-------
Table 6-5a.  Vertical Wind Direction Turbulence Criteria for Initial
             Estimate of Pasquill Stability Category.  Use with Table 6-5b,


     Initial estimate of             Standard deviation of vertical wind
 Pasquill stability category       direction fluctuations, OE, in degrees
A
B
C
D
E
F
11.5
10.0
7.8
5.0
2.4

1°E

T ac <
1 °E «
< 
-------
                          If the site  roughness  length  is  other  than  15  cm,
the category boundaries listed in Table 6-5a may need adjustment.  As  an
initial adjustment, multiply the values listed by,
                          (z0/15) °'2,
where ZQ is the site roughness in centimeters.   This factor, while theoret-
ically sound, has not had widespread testing.   It is likely to be a useful
adjustment for cases when z0 is greater than 15  cm.  It is yet problematical
whether the adjustment is as useful  for cases when  z0 is less than 15  cm.
                          If the measurement height  is  other than 10m, the
category boundaries listed in Table  6-5a will need  adjustment.   As an
initial adjustment, multiply the lower bound values  listed by,
                          (Z/10)Pe,
where Z is the measurement height in meters.  The exponent pe varies  as  a
function of stability category as,
                   To determine
wer bound
ategory
A
B
C
D
E
Value of
pe
0.02
0.04
0.01
-0.14
-0.31
                          The above suggestions summarize the results of sev-
eral studies conducted in fairly ideal  circumstances.   It is anticipated that
readers of this document are often faced with conducting analyses in less than
ideal circumstances.  Therefore, before trusting the Pasquill category esti-
mates, the results should be spot checked.  This can easily be accomplished.
Choose cloudless days.  In midafternoon during a sunny day, categories A and
B should occur.  During the few hours just before sunrise, categories E and
                                    6-26
-------
F should occur.  The bias, If any,  in the turbulence criteria  will  quickly
be revealed through such comparisons.  Minor adjustments  to  the  category
boundaries may tailor the turbulence criteria to the particular  site  charac-
teristics, but should be made only  in consultation  with the  reviewing agency.

                 6.4.4.3  Lateral turbulence (o^) and wind speed method
                          The following discussion  describes a method for
estimating Pasquill stability categories in terms of the  standard deviation
of the horizontal wind direction fluctuations, o/\,  and the scalar mean
wind speed, US.  The reader should  note that the method and  parameters
specified in this subsection are identical  with those in  the Guideline on
Air Quality Models (Revised).4  However, several  refinements are added that
provide for wider applicability and for less ambiguous distinctions between
stability classes.
                          The criteria in Table 6-6a and  Table 6-6b are for
data collected at 10m and the roughness length is 15 cm.  For  use in  Table
6-6b, nighttime is the period from  one hour before  sunset to one hour after
sunrise.  Wind speed and direction  data collected within  the height range
from 20z0 to 100z0 should be used.   For sites with  very low  roughness, these
criteria are slightly modified. The lower bound of measurement  height should
never be less than 1.0m.  The upper bound should never be less than 10m.  To
obtain 1-hour averages, the recommended sampling duration is 15  minutes, but
it should be at least 3 minutes and may be as long  as 60  minutes.   The re-
lationships employed in the estimation methods assume conditions are  steady
state.  This is more easily achieved if the sampling duration  is less than
30 minutes.
                                    6-27
-------
Table 6-6a.   Lateral Wind Direction Turbulence  Criteria  for Initial
	Estimate of Pasquill Stability Category.  Use  with Table 6-6b.

        Initial estimate of          Standard  deviation of  horizontal  wind
    Pasquill stability category     direction  fluctuations,  o/\,  in degrees
A
B
C
D
E
F
22.5
17.5
12.5
7.5
3.8

o^A
^ O A *
<" OA <
^ O A *
1 °A '
OA *

c 22.5
c 17.5
c 12.5
c 7.5
c 3.8
Table 6-6b.   Wind Speed Adjustments  for  Determining  Final  Estimate of
	Pasquill Stability Category from OA.  Use with Table 6-6a.
            Initial estimated10m  scalar  windFinal  estimate of
               category          speed  (US)  (m/s)         stability category

 Daytime          A                     US < 3                   A
                                    3 _< US < 4                   B
                                    4 < US < 6                   C
                                    6 £ US                        D

                  B                     US < 4                   B
                                    4 <_ US < 6                   C
                                    6 _< US                        D

                  C                     US < 6                   C
                                    6 <_ US                        D

               D,E or F                 ANY                        D

 Nighttime        A                     US < 2.9                  F
                                   2.9 _< US < 3.6                  E
                                   3.6 _< US                        D

                  B                     US < 2.4                  F
                                   2.4 _< US < 3.0                  E
                                   3.0 _< US                        D

                  C                     US < 2.4                  E
                                   2.4 _< US                        D

                  D                     ANY                        D

                  E                     US < 5.0                  E
                                   5.0 £ US                        D

                  F                     US < 3.0                  F
                                   3.0 _< US < 5.0                  E
                                   5.0 < US                        D
                                     6-28
-------
                          If the site roughness length 1s  other  than  15  cm,
the category boundaries listed in Table 6-6a may need  adjustment.   As an
initial adjustment, multiply the values listed by,
where z0 is the site roughness in centimeters.   This factor,  while  theoret-
ically sound, has not had widespread testing.   It is likely to  be a useful
adjustment for cases when z0 is greater than 15 cm.   It is yet  problematical
whether the adjustment is as useful  for cases  when z0 is less than  15 cm.
                          If the measurement height  is other  than 10m, the
category boundaries listed in Table  6-6a will  need adjustment.   As  an
initial adjustment, multiply the lower bound values  listed by,
                          (Z/10)Pa,

where Z is the measurement height in meters. The exponent pa varies as a
function of stability category as,
                           To determine
ter bound
itegory
A
B
C
D
E
Value of
pa
-0.06
-0.15
-0.17
-0.23
-0.38
                          The above suggestions  summarize  the  results  of
several studies conducted in fairly ideal  circumstances.   It is  anticipated
that readers of this document are often faced with  conducting  analyses in
less than ideal circumstances.  Therefore, before trusting the Pasquill
                                    6-29
-------
category estimates, the results should be  spot  checked.  This  can  easily
be accomplished.  Choose cloudless days.   In  midafternoon  during a sunny  day,
categories A and 8 should occur.  During the  few hours just  before sunrise,
categories E and F should occur.  The bias,  if  any,  in the turbulence criteria
will quickly be revealed through such comparisons.   Minor  adjustments to  the
category boundaries may tailor the turbulence criteria to  the  particular  site
characteristics, but should be made only in  consultation with  the  reviewing
agency.

                 6.4.4.4  Accuracy of stability category estimates
                          Results are not  available  comparing  the  performance
of the methods outlined above in this section.   There are  comparison results
for similar methods.  From these studies,  it  is concluded  that the methods
will estimate the same stability category  about 50%  of the time.   They will
estimate within one category of each other about 90% of  the  time.   Adjustment
of the turbulence criteria resulting from  spot  checks is necessary to achieve
this performance.

          6.4.5  Other Stability Measures
                 6.4.5.1  Flux Richardson  number
                          Buoyancy forces  may act to enhance or suppress
turbulent wind fluctuation motions.  A very useful  measure in  this regard
is the flux Richardson number, Rf,

                         Production of turbulent thermal kinetic energy
                 Rf -  -
                         Production of turbulent mechanical kinetic energy
                                    6-30
-------
The denominator is always positive near the surface.   Rf is  negative when
buoyant forces tend to enhance turbulent motions in the vertical.   It is
positive when buoyant forces tend to suppress turbulent motions  in  the
vertical.  Stable conditions exist when Rf is positive.  When Rf is near zero,
stability conditions are neutral.  During such times, the wind speed profile
often varies linearly with the logarithm of height.  When Rf is  negative,
stability conditions are unstable.

                 6.4.5.2  Monin-Obukhov length
                          A more easily estimated stability  measure, related
to Rf, is the Monin-Obukhov length, L,
                 Rf = Z/L.
                          A variety of methods are available for estimating
L.  The choice of method is dependent upon the type of meteorological data
available.  In all the estimation methods, use the scalar mean wind speed.
Only wind speed and temperature data collected within the height range from
20z0 to 100z0 are used.  For sites with very low roughness,  these criteria
are slightly modified.  The lower bound of measurement height should never
be less than 1.0m.  The upper bound should never be less than 10m.   To
obtain 1-hour averages, the sampling duration should  be at least 3  minutes
and may be as long as 60 minutes.  The relationships  employed in the estima-
tion methods assume conditions are steady state.  This is more easily achieved
if the sampling duration is less than 30 minutes.
                          When temperature and wind speed are available at
three or more heights, use of the procedure presented by Nieuwstadt^ is
recommended.  Wind speed at one level and direct measurements of temperature
                                    6-31
-------
difference in the vertical  may be available.   For these  cases the  procedures
outlined by Irwin and Binkowski  should be used35.  When  only the routine
weather observations are available,  L should  be  estimated  with  the procedure
outlined in the appendix to the article by Holtslag36.   The latter two  pro-
cedures are incorporated into the meteorological  processor, MPDA-1.^8
                          The uncertainty of  the empirical constants  used
in the estimation methods means that at best  L estimates have two  significant
figures accuracy.  Often, especially for the  cases using the  routine  weather
observations, the estimate has only one significant figure accuracy.

     6.5  Model Inputs
          The majority of point source models recommended in  EPA's Guideline
on Air Quality Models (Revised)^ require that hourly meteorological data
be input in a format that has been standardized  by EPA's meteorological
preprocessor program.^  EPA desires to maintain this consistency  and extend
it to on-site meteorological data sets.  EPA  is  developing a  meteorological
processor for regulatory applications (MPRA)  that will  provide  this consistency
when available.
          6.5.1  Formats
                 As noted above, the input data  format for EPA  short-term
regulatory models has been standardized by the meteorological  preprocessor,
RAMMET, as described in Reference 12.  A consistent format for  model  input
should be used when processing on-site meteorological data.   Since on-site
wind direction data are reported to the nearest  degree, the  actual observed
winds should be repeated in the field reserved for the randomized  flow
vector generated for National Weather Service (NWS) data. The  input  format
                                    6-32
-------
for the EPA long-term models should be of the stability wind rose (STAR)
variety generated for NWS stations by the National  Climatic Data Center.
Individual model user's guides should be referred to for additional  details
on input data formats.

          6.5.2  Treatment of Calms
                 EPA's policy is to disregard calms until  such time as an
appropriate analytical approach is available.  The recommended EPA models
contain a routine that eliminates the effect of the calms  by nullifying
concentrations during calm hours and recalculating short-term and annual
average concentrations.  Certain models lacking this built-in feature can
have their output processed by EPA's CALMPRO program^ to  achieve the same
effect.  Because the adjustments to the concentrations for calms are made
by either the models or by postprocessor, actual measured  on-site wind
speeds should always be input to the preprocessor.   These  actual wind speeds
should then be adjusted as appropriate under the current EPA guidance^ by
the preprocessor.
                 Measured on-site wind speeds of less than 1.0 m/s,  but above
the instrument threshold, should be set equal to 1.0 m/s by the preprocessor
when used as input to Gaussian models.  Wind speeds below  the starting thres-
hold of the anemometer or vane, whichever is greater, should be considered
calm.  Calms are identified in the preprocessed data file  by a wind  speed
of 1.0 m/s and a wind direction equal to the previous hour.

          6.5.3  Treatment of Missing Data
                 Missing data refers to those hours for which no data are
available from the primary on-site source for the variable in question.
                                    6-33
-------
In order for the regulatory models  to function  properly,  there  must  be  a  data
value in each input field.   When missing  values arise,  they  should be
handled in one of the ways  listed below,  in  the following order of preference,
                 (1)  If there are  other  on-site data,  such  as  measurements
at another height, they may be used when  the primary data are missing.   If
the height differences are  significant, corrections  based on established
vertical profiles should be made.  Site-specific vertical profiles based  on
historical on-site data may also be appropriate to use  if their determination
is approved by the reviewing authority (see  Section  6.1.3).  If there  is
question as to the representativeness of  the other on-site data, they  should
not be used.
                 (2)  If there are  only one  or  two missing hours, then
linear interpolation of missing data may  be  acceptable, however, caution
should be used when the missing hour(s) occur(s) during day/night transition
periods.
                 (3)  If representative off-site data exist, they may  be
used.  In many cases this approach  may be acceptable for cloud  cover,
ceiling height, mixing height and temperature.   This approach will  rarely be
acceptable for wind speed and direction.   The representativeness of  off-site
data should be discussed and agreed upon  in  advance  with the reviewing
authority (see Section 6.6).
                 (4)  Falling any of the  above, the  data field  should  be
coded as a field of nines.   This value will  act as  a missing data flag in
any further use of the data set.
                 At the present time, the short term regulatory models con-
tain no mechanism for handling missing data  in  the  sequential  input  file.
                                    6-34
-------
Therefore, in order to run these models a complete data set, including
substitutions, is required.  Substitutions for missing data should only be
made in order to complete the data set for modeling applications,  and should
not be used to attain the 90% data retrieval  recommended in Section 5.0.

     6.6  Use of Off-Site Data

          6.6.1  Representativeness of Meteorological  Data
                 Evaluations of the atmospheric dispersion characteristics
of the site of a pollutant source, make it necessary to determine  if available
meteorological data can be used to adequately characterize the atmospheric
dispersion conditions.
                 Such determinations are required when the available meteoro-
logical data are acquired at a location other than that of the proposed
source.  In some instances, even though meteorological data are acquired at
the location of the pollutant source, they still may not correctly characterize
the important atmospheric dispersion conditions.
                 Considerations of representativeness  are always made with the
meteorological data sets used in atmospheric dispersion modeling whether the
data base is "on-site" or "off-site."  These considerations call for the
judgment of a meteorologist or an equivalent professional with expertise
in atmospheric dispersion modeling.
                 Representativeness has been defined in the Workshop on the
Representativeness of Meteorological Observations*1 as "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."  Any judgments of the representativeness  of

                                    6-35
-------
meteorological  data should necessarily factor  in  considerations  of spatial
and temporal  dependence.
                 6.6.1.1  Spatial  dependence
                          The location where the  data  base  was acquired
should be compared to the source location for  similarity  of terrain features.
For example, in complex terrain, the following considerations  should be
addressed:

                          1.  Aspect ratio of  terrain, i.e., ratio of:
                              a.  Height of valley walls  to width  of valley;
                              b.  Height of ridge to length of ridge; and
                              c.  Height of isolated hill to width of hill
                                  at base.
                          2.  Slope of terrain
                          3.  Ratio of terrain height  to  stack/plume
                              height.
                          4.  Distance of source  from terrain, i.e. how close
                              to valley wall,  ridge, isolated hill.
                          5.  Correlation of terrain feature to  prevailing
                              meteorological conditions.
Likewise, if the source is to be located on a  plateau  or  plain,  the source
of meteorological data should be from a similar plateau or  plain.
                          Judgments of representativeness should be made
only when sites are climatologically similar.   Sites in nearby but different
air sheds often exhibit different weather patterns.  For  instance, meteoro-
logical data acquired along a shoreline are not normally  representative of
inland sites and vice versa.
                                    6-36
-------
                          Meteorological  data collected need to be examined
to determine if drainage, transition,  and synoptic flow patterns are  charac-
teristics of the source, especially those critical to the regulatory  applica-
tion.  Consideration of orientation, temperature, and ground cover should
be included in the review.
                          An important aspect of space dependence is
elevation above the ground.  Where practical, meteorological  data should  be
acquired at the release elevation, as  well as above or below, depending on
the buoyancy of the source's emissions.
                 6.6.1.2  Temporal dependence
                          To be representative, a meteorological data base
must be of sufficient duration to define the range of sequential atmospheric
conditions anticipated at a site.  As  a minimum, one year of on-site  meteoro-
logical data covering the four seasons is necessary to prescribe this time
series.  Multiple years of data are used to describe variations in annual,
and short term impacts.  In general, the climatic period of five years is
adequate to represent these yearly variations.  The length of the required
data period relates to the standard being addressed.  In general, the
longer the time period of the ambient  air quality standard, the longer the
period of meteorological data required to demonstrate compliance with that
standard.

                 6.6.1.3  Further considerations
                          It must also be recognized that consideration of
alternative data sets extends beyond space and time representation.   The
data from the onset must be compatible with the impact analysis requirements
                                    6-37
-------
as set forth in the source's modeling protocol.   If a  meteorological  data
set were acquired in an incompatible form,  it may be considered  inadequate
and, therefore, "not representative."  Also,  consideration must  be given to
the response characteristics of the instruments  and their ability to correctly
describe the atmospheric dispersion processes.  If these response character-
istics restrict the instrument from sensing the  most critical  atmospheric
processes (those resulting in the highest impacts), they may not be represen-
tative from an atmospheric dispersion standpoint.
                          It may be necessary to recognize the non-homogeneity
of meteorological variables in the air mass in which pollutants  disperse.
This non-homogeneity may be essential in correctly describing the dispersion
phenomena.  Therefore, measurements of meteorological  variables  at multiple
locations and elevations may be required to correctly  represent  these
meteorological fields.  Such measurements are generally required in complex
terrain or near large land-water body interfaces.
                          It is important to recognize that, although
certain meteorological variables may be considered unrepresentative of
another site  (for instance, wind direction or wind speed), other variables
may be representative (such as temperature, dew point, cloud cover).
Exclusion of one variable does not necessarily exclude all.
                          Other factors affecting representativeness include
change in surface roughness, topography and atmospheric stability.
                          Currently there are no established analytical or
statistical techniques to determine representativeness of meteorological
data.  As implied above, any criteria would be variable-specific and involve
a judgment based on case-by-case considerations.  Even if such criteria

                                    6-38
-------
could be established, they would require the acquisition of some on-s1te
data.  The establishment and maintenance of such an on-s1te data collection
program generally fulfills the requirement for "representative"  data.

          6.6.2  Alternative Meteorological Data Sources
                 It is necessary in the consideration of most air pollution
problems to obtain information on site-specific atmospheric dispersion.
Frequently, an on-site measurement program must be initiated. As discussed
in Section 6.5.3, representative off-site data may be used to substitute
for missing periods of on-site data.  There are also situations  where
current or past meteorological records from a National  Weather Service
station may suffice.  The following outline provides a brief insight into
the types of observations taken at Weather Stations and some of  the summaries
compiled from this data.

                 6.6.2.1  National Weather Service (NWS)
                          (a)  First Order Stations
                               There are about 200 National Weather Service
(NWS) stations where 24 hourly observations are taken daily.  Among the
measurements taken are:  dry bulb temperature and wet bulb temperature
(from which dew point temperature and relative humidity are calculated),
pressure, wind direction and speed, cloud cover and visibility.   The
National Climatic Data Center (NCDC) in Asheville, North Carolina, main-
tains records of these observations.
                          (b)  Second Order Stations
                               These stations usually take hourly observations
simi-lar to the first order stations above, but not throughout the entire day.

                                    6-39
-------
                 6.6.2.2  Military observations
                          Many military installations,  especially  Air  Force
Bases, take hourly observations.   These are  transmitted on military teletype
circuits and therefore are not usually available  for  general  use.  No  routine
publication of these data is done.  Records  of observations  are  sent to  NCDC
where special summaries can be made.

                 6.6.2.3  Supplementary Airways Reporting Stations
                          These stations are at smaller airports.  The
observations are not at regular intervals,  usually being  taken according to
airline schedules at the airport.   These observations are not published  and
are not usually digitized.  Original records are  sent to  NCUC, however.

                 6.6.2.4  Upper air
                          There are between  60 and 70 stations in  the  contig-
uous United States where upper air observations are taken twice  daily  (at
0000 GMT and 1200 GMT) by radiosonde balloon and  radio direction-finding
equipment.  The measurements made are temperature, pressure, and relative
humidity with height and wind speed and direction. These data are obtained
primarily for knowledge of the large scale meteorological pattern  and  have
relatively little refinement in the lower 500 to  1000 meters of the  atmos-
phere.  These observations are transmitted by teletype and  original  records
sent to NCDC where these data are published.  These data  form the  basis  for
most determinations of mixing height input to regulatory  air quality models.

                 6.6.2.5  Evaluation of NWS  and military  data sources
                          If these NWS and military meteorological data
                                    6-40
-------
sources are to be used in making atmospheric dispersion estimates of a
source, a judgment as to the representativeness of these data sources
should be made using the considerations provided in Section 6.6.1 above.
                          In addition, it must be recognized that these
data sources have the following limitations:
                          (a)  Human error
                                      4
                               The observational data are a result of human
interpretation and as such are then subject to individual bias and variation
in reported data.  Such observational bias is sometimes apparent upon review
of the data.  For instance, some observers will report wind directions to
the nearest 20 degrees, resulting in a higher frequency of occurrence of
even numbered wind directions.  This is apparent from a casual observation
of the wind rose constructed on such a biased data set.  It is important  that
all relevant NWS meteorological observational data be reviewed for human
bias.
                          (b)  Accuracy of the wind direction observation
                               Wind directions are only reported to the nearest
10 degrees, with no attempt to electronically average the data.  Dispersion
modeling estimates for short term impacts have traditionally relied upon
directions specified to the nearest degree.  In order to achieve that level
of specificity and consistency, EPA has generated a random number string  to
be applied to the data set.
                          (c)  Time period of observation
                               While on-site meteorological  data are generally
of a continuous nature, NWS and military observations are constrained to  a
short time period preceding the hour.  Gradual shifts in the data over that
                                    6-41
-------
time period are generally unreported.   Other* significant  shifts  in  observa-
tions, although observed and reported  are  not handled  by  the  meteorological
data preprocessor.  These shortcomings are known  to  be inherent  In  such
data yet, historically, these observations have provided  acceptable data
for regulatory applications.

                 6.6.2.6  Meteorological data from private networks
                          As with NWS  and  military data sources, meteoro-
logical data acquired from private monitoring networks may be used  in making
atmospheric dispersion estimates of a  source if judged to be  representative
by the criteria provided in Section 6.6.1  above.
                          Data from such sources  may not  be accompanied
by the problems associated with NWS and military  data  as  noted above. How-
ever, such meteorological data sets are not generally  subject to the same
level of public dissemination and review.   Therefore,  any use of such data
sets should be accompanied by a review of  the quality  assurance plans for
these data acquisition systems.  Such  meteorological data should be collected
in accordance with the guidance on quality assurance and  maintenance contained
in Section 8.0 of this document.
     6.7  Recommendations
          It is recommended that for hourly mean  wind statistics in straight-
line Gaussian dispersion models, scalar wind speed and scalar wind  direction
processing be used.  For microprocessor-based digital  systems, the  unit
vector mean wind direction is also acceptable. The standard  deviation of
the wind direction fluctuations should be  calculated about the scalar or
unit vector mean direction, or may be  estimated  using the techniques of
Mardia^4 or Yamartino^6.  These hourly values may be obtained by averaging
samples over an entire hour or by averaging a group of shorter period aver-
ages.  If shorter period averages are  used, it is recommended that  wind
statistics be computed over intervals  of  15 minutes, and  that at least two
valid 15-minute periods be averaged to represent  the hour.  A minimum of
                                    6-42
-------
360 data samples should be used to calculate the standard deviation and at
least 60 samples should be used to calculate the mean,  regardless  of the
averaging period.  Thus, to calculate the standard deviation for a 15-minute
sampling duration, the data should be sampled at least  once every  2.5
seconds, and if the data are only averaged every hour,  then the data should
be sampled at least once every ten seconds.  If the single-pass processor
described by Equations 6.1.4 and 6.1.5 in Section 6.1.2 is used for wind
direction, it is recommended that the data be sampled at least  once per
second, to assure that the difference between consecutive values is less
than 180°.

          The hourly vertical temperature gradient may  be determined by
averaging samples over the entire hour or by averaging  a group  of  shorter
period averages.  If shorter period averages are used,  it is recommended
that four 15-minute averages be used with at least 60 samples for  each
15-minute period.  For other primary variables, including temperature, dew
point, pressure and radiation, four 15-minute averages  of digital  data are
recommended, but one-hour point or one-hour average analog values  may be
acceptable.  Precipitation data should be processed to  obtain a total  for
every hour.

          It is recommended that effective roughness length be  determined
from equation 6.4.2

          The atmospheric stability category should be  determined  from one
of the following schemes, following the order of preference given  in the
Guideline on Air Quality Models (Revised):4

                 (1)  Turner's 1964 method-*8 using site-specific data which
include cloud cover, ceiling height and surface (~10m)  wind speeds;

                 (2)  OE from site-specific measurements modified  by
wind speed (o£ may be determined from elevation angle measurements or
may be estimated from measurements of ay according to the transform:
°E = 
-------
instrument threshold should be set equal  to 1.0  m/s  by  the  preprocessor  when
used as input to Gaussian models.   Wind  speeds below the  instrument  threshold
of the cup or vane, whichever is greater, should be  considered  calm, and are
identified in the preprocessed data file by a wind speed  of 1.0 m/s  and  a
wind direction equal to the previous hour.

          If data are missing from the primary source,  they should be handled
as follows, in order of preference: (1)  substitution of other representative
on-site data; (2) linear interpolation of one or two missing hours;  (3)
substitution of representative off-site  data; or (4) coding as  a field of
nines, according to the discussions in Section 6.5.3 and  6.6.   However,  in
order to run existing short-term regulatory models,  a complete  data  set,
including substitutions, is required.

          If the data processing recommendations in  this  section cannot  be
achieved, then alternative approaches should be  developed in conjunction with
the Regional Office.
                                    6-44
-------
7.0  DATA REPORTING AND ARCHIVING
     Because of the different data requirements for different types of
analyses, there is no fixed format that applies to all  data sets.   However,
a generalization can be made.  All on-site meteorological  data should be col-
lated in chronological order and tabulated according to the observation time.
Observation time should be defined as the time at the beginning of the averag-
ing period, e.g., 0100 refers to the period from 0100 to 0200.   Note that NWS
data is based on a somewhat different recording scheme and cannot  be interpreted
in the same manner.  If an EPA regulatory decision is involved, the on-site
data must be furnished to the reviewing agency upon request.

     7.1  Reporting Formats
          When data are requested by the reviewing agency, two types of
reports will generally be required.  The first will be a written summary
report which should include a discussion of the overall monitoring program
followed by details on data sources, data quality, completeness, data
handling procedures and computational methods.  The second report  will
include the actual data.   Different forms of actual data reporting are
discussed briefly below.
          7.1.1  Preprocessed Data
                 In most  cases, the reviewing agency will  request  a copy of
the preprocessor output in tape and hardcopy form.
          7.1.2  SAROAD/AIRS
                 In some  cases, the reviewing agency will  require  that
validated measured data be reported to EPA's ambient monitoring data base
                                    7-1
-------
system (SAROAD/AIRS) on a quarterly basis.   In  these instances,  all  variables
that have a SAROAD/AIRS parameter code should be  submitted  in SAROAD/AIRS
format on a quarterly basis.   In some cases, both preprocessor output  and
SAROAD/AIRS format data may be required.

     7.2  Archiving
          While there are currently no EPA  regulatory requirements  for
meteorological  data archiving, it is considered prudent  practice for
collectors of such data to establish an archiving program.   When the data
are being collected for use in a regulatory setting, they must be made avail-
able to the reviewing agency upon request.   Thus, until  a particular regulatory
action is complete, all data must be available.  Since a particular data set
may have applicability in more than one regulatory action,  or  since litigation
may follow a regulatory action, the need  for the  raw data set may extend well
beyond its original application.  EPA suggests  the following considerations
in designing an archiving program.

          7.2.1  Raw Data
                 The raw data records are the most basic data  elements and
should be given the highest priority in archiving.  The  raw data may include
variables that, although not currently used by  recommended  models,  might
be used in future models.  Therefore, comprehensive archiving  is recommended.
Hourly averaged data should be stored in machine-readable form,  e.g.,  magnetic
tape, for convenience and easy access.  However,  magnetic tapes  need to be
copied periodically to insure integrity, and care should be taken to select
a format for encoding the data that will  be as  compatible as possible  with
                                     7-2
-------
other computer systems.  Where data were originally reduced from strip chart
records, the charts should also be archived.

          7.2.2  Preprocessed Data
                 Since, in theory, all preprocessed data can be recreated
from the raw data, the preprocessor data should be given a lower priority.
However, the ready-to-use nature of the preprocessor output and the cost of
preprocessing raw data argue strongly for archiving the preprocessed data
as well.

          7.2.3  Retention Time
                 Experience shows that good data sets have long, useful
lives and thus should be archived as long as possible.   When evaluating
whether an old data set remains useful, primary consideration should be
given to a comparison of the actual collection program with the most cur-
rent guidance.  As long as the instrumentation, siting, quality assurance
and completeness criteria are still satisfied, it is recommended that the
data be retained indefinitely in machine-readable form.  Original strip chart
records should be retained for a minimum of five years.  If an archive is to
be eliminated, an attempt should be made beforehand to contact other modelers
who may wish to receive the data.
     7.3  Recommendations
          In general, the data reporting and archiving requirements will  be
worked out in consultation with the reviewing agency.   An agency may request
meteorological data in either a preprocessed form, or in the SAROAD/AIRS
data base format, or both.  All meteorological  data must be available to
the reviewing agency until a regulatory action  is completed.  However, the
need for a data set may extend beyond its original application due to liti-
gation, or due to its applicability to another regulatory action.   Therefore,
                                    7-3
-------
it is recommended that data be retained  indefinitely,  provided  that  the
guidance criteria for on-site meteorological monitoring  are still  satisfied.
It is recommended that the observation time  reported  refer to the  time at
the beginning of the averaging period.
                                    7-4
-------
8.0  QUALITY ASSURANCE AND MAINTENANCE
     The purpose of quality assurance and maintenance is the generation of
a representative amount (90% of hourly values for a year, Section 5.3.2) of
valid data (Sections 5.1 and 8.6).  Maintenance may be considered the
physical activity necessary to keep the measurement system operating as it
should.  Quality assurance is the management effort to achieve the goal of
valid data through plans of action and documentation of compliance with the
plans.
     Quality assurance (QA) will be most effective when following a QA Plan
which has been signed-off by appropriate project or organizational authority.
The QA Plan should contain the following information (paraphrased and
particularized to meteorology from Lockhart^):
     1.  Project description - how meteorology is to be used
     2.  Project organization - how data validity is supported
     3.  QA objective - how QA will document validity claims
     4.  Calibration method and frequency - for meteorology
     5.  Data flow - from samples to archived valid values
     6.  Validation and reporting methods - for meteorology
     7.  Audits - performance and system
     8.  Preventive maintenance
     9.  Procedures to implement QA objectives - details
     10.  Management support - corrective action and reports
     It is important for the person providing the quality assurance (QA)
function to be independent of the organization responsible for the collection
of the data and the maintenance of the measurement systems.  Ideally, the QA
auditor works for a separate company.  There should not be any lines of

                                    8-1
-------
intimidation available to the operators which  might  be  used  to  influence
the QA audit report and actions.
     With identical goals of valid data, the QA person  should encourage the
operator to use the same methods  the QA person uses  (presumably these  are
the most comprehensive methods) when challenging the measurement system
during a performance audit.   When this is done, the  QA  task  reduces  to spot
checks of performance and examination of records thus providing the  best
data with the best documentation  at the least  cost.
     The subsections will be specific to the variable to be  measured.   Wind
speed will refer to those common  mechanical  anemometers (cups and vane-
oriented propellers) which use the pressure  force of the air passing the
aerodynamic shape of the anemometer to turn  a  shaft. Except for Doppler
SODARS (see Section 9.0), the more complicated indirect or remote measuring
systems, such as sonic anemometers, hot wire or hot  film anemometers,  laser
anemometers and the like, are not commonly used for  routine  monitoring and
are beyond the scope of this guide.
     Wind direction will refer to common wind  vanes  which provide a
relative direction with respect  to the orientation of the direction  sensor.
There are three parts of the direction measurement which must  be considered
in quality assurance.  These are  (1) the relative accuracy of the vane per-
formance in converting position to output, (2) the orientation  accuracy in
aligning the sensor to TRUE NORTH and vertical, with respect to a level
plane, and (3) the dynamics of the vane and  conditioning circuit response
to turbulence for calculation of  sigma theta.
                                    8-2
-------
     Temperature and temperature difference require QA focused on the appli-
cation of the data.  Dew point temperature, precipitation, atmospheric
pressure and radiation are also addressed.
     8.1  Instrument Procurement
          The specifications required for the applications for which the data
will be used (see Sections 5.0 and 6.0) along with the test method to be used
to determine conformance with the specification should be a part of the pro-
curement document.  A good QA Plan will require a QA sign-off of the procure-
ment document for an instrument system containing critical requirements.  An
instrument should not be selected solely on the basis of price and a vague
description, without detailed documentation of sensor performance.
          8.1.1  Wind Speed
                 The performance specification for an anemometer might read:
                 Range                             0.5 m/s to 50 m/s
                 Threshold (1)                   £0.5 m/s
                 Accuracy (error)(1)(2)          <_ (0.2 m/s +5% of observed)
                 Distance Constant (1)           _< 5 m at 1.2 kg/m3 (standard
                                                   sea-level  density)
                 (1) as determined by wind tunnel  tests conducted on pro-
                 duction ;
                 methods.'
duction samples in accordance with ASTM D-22.11 test
                 (2) aerodynamic shape (cup or propeller) with permanent
                 serial  number to be accompanied by test report, trace-
                 able to NBS, showing rate of rotation vs. wind speed at
                 10 speeds.
                 The procurement document should ask for (1) the starting
torque of the anemometer shaft (with cup or propeller removed) which repre-
                                    8-3
-------
sents a new bearing condition,  and  (2)  the  starting  torque which  represents

the threshold speed, above which  the  anemometer will  be out of specification.

The latter value is a flag requiring  the  action of bearing or sensor  re-

placement.

                 The ASTM test  cited  above  includes  a measurement of  off-axis

response.  Some anemometer designs  exhibit  errors greater than the accuracy

specification with off-axis angles  of as  little as 10 degrees.  However,

there is no performance specification for this type  of error at this  time,

due to a lack of sufficient data  to define  what the  specification should  be.


          8.1.2  Wind Direction

                 The performance  specification for the wind vane  might read:

                 Range                          001  to 360 degrees or
                                                001  to 540 degrees

                 Threshold (1)                  <0.5 m/s

                 Accuracy (error)(l)            <3 degrees  relative to the
                                                     sensor mount or  index
                                                     (4> degrees  absolute
                                                     error  for  installed
                                                     system)

                 Delay Distance (1)            <5 m at  1.2 kg/m3 (standard
                                                     sea-level density)

                 Damping Ratio (1)              >0.4 at  1.2 kg/m3 or
                 Overshoot (1)                   <25% at  1.2 kg/rn3

                 (1) as determined  by wind  tunnel tests  conducted on  pro-
                 duction samples  in accordance with  ASTM  D-22.1I  test methods.


                 The procurement  document should  ask for  (1) the  starting

torque of the vane  shaft (with the vane removed)  which represents a new

bearing  (and potentiometer) condition, and  (2) the  starting torque which
                                    8-4
-------
represents the threshold speed, above which the vane will be out of specifi-
cation.  The latter value is a flag requiring the action of bearing or
sensor replacement.
                 The range of 001 to 540 degrees was originally conceived
to minimize strip chart "painting" when the direction varied around 360
degrees.  It also minimizes errors (but does not eliminate them) when
automatic sigma meters are used.  It may also provide a means of avoiding
some of the "dead band" errors from a single potentiometer.  In these days
of "smart" data loggers, it is possible to use a single potentiometer
(001 to 360 degree) system without excessive errors for either average
direction or sigma theta.
                 If the wind direction samples are to be used for the cal-
culation of sigma theta, the specification should also include a time
constant requirement for the signal conditioner.  Direction samples should
be effectively instantaneous.  At 5 m/s, a 1m delay distance represents
0.2 seconds.  A signal conditioner specification of a time constant of <0.2
seconds would insure that the sigma theta value was not attenuated by an
averaging circuit provided for another purpose.

          8.1.3  Temperature and Temperature Difference
                 When both temperature and differential temperature are
required, it is important to specify both accuracy and relative accuracy
(not to be confused with precision or resolution).  Accuracy is performance
compared to truth, usually provided by some standard instrument in a con-
trolled environment.  Relative accuracy is the performance of two or more
sensors, with respect to one of the sensors or the average of all sensors,
                                    8-5
-------
in various controlled environments.   A temperature  sensor  specification
might read:
                 Range                           -40 to +60 degrees C.
                 Accuracy (error)                 5 0.5 degree C.
A temperature difference specification might read:
                 Range                           -5 to +15 degrees C.
                 Relative accuracy (error)        £0.1 degrees  C.
                 While calibrations  and audits  of both accuracy  and relative
accuracy are usually conducted in  controlled environments, the measurement
is made in the atmosphere.   The greatest source of  error is usually solar
radiation.  Solar radiation shield specification is therefore an important
part of the system specification.   Motor aspirated  radiation shields  (and
possibly high performance naturally ventilated  shields) will satisfy the
less critical temperature measurement.  For temperature difference, it  is
critical that the same design motor aspirated shield be used for both
sensors.  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.

          8.1.4  Dew Point Temperature
                 Sensors for measuring dew  point temperature can be
particularly susceptible to precipitation,  wind, and radiation effects.
                                    8-6
-------
Therefore, care should be taken in obtaining proper (manufacturer-recommended)
shielding and aspiration equipment for the sensors.  If both  temperature
and dew point are to be measured, aspirators can be purchased which will
house both sensors.  If measurements will  be taken in polluted atmospheres,
gold wire electrodes will minimize corrosion problems.  For cooled  mirror
sensors consideration should be given to the susceptibility of the  mirror
surface to contamination.

          8.1.5  Precipitation
                 For areas where precipitation falls in a frozen form,
consideration should be given to ordering an electrically heated rain and
snow gage.  AC power must be available to the precipitation measurement
site.  For remote sites where AC power is not available, propane-heated
gages can be ordered.  However, if air quality measurements are being made
at the same location, consideration should be given to the air pollutant
emissions in the propane burner exhaust.
                 Ai r movement across the top of a gage can affect the amount
of catch.  For example, Weiss4-* reports that at a wind speed  of 5 mph,  the
collection efficiency of an unshielded gage decreased by 25%, and at 10 mph,
the efficiency of the gage decreased by 40%.  Therefore, it is recommended
that all precipitation gages be installed with an Alter-type wind screen,
except in locations where frozen precipitation does not occur.
                 Exposure is very important for precipitation gages; the
distance to nearby structures should be at least two to four  times  the
height of the structures (see Section 3.1.3).  Adequate lengths of  cabling
                                    8-7
-------
must be ordered to span the separation distance of the gage from the data
acquisition system.
                 If a weighing gage will  be employed,  a set of calibration
weights should be obtained.

          8.1.6  Pressure
                 The barometric pressure sensor should normally have a
proportional and linear electrical  output signal  for data recording.
Alternately, a microbarograph can be used with a mechanical recording
system.  Some barometers operate only within certain pressure ranges; for
these, care should be taken that the pressure range is appropriate for the
elevation of the site where measurements will  be taken.

          8.1.7  Radiation
                 Radiation instruments should be selected from commer-
cially available and field-proven systems.  These sensors generally have a
low output signal, so that they should be carefully matched with the signal
conditioner and data acquisition system.  Another consideration in the
selection of data recording equipment is the fact that net radiometers have
both positive and negative voltage output signals.

     8.2  Acceptance Testing
          It is common for acceptance tests to be just checking the shipment
part numbers against the packing slip.  Lacking more detailed instructions,
it is  all a receiving department can do.  Such a test does not provide any
technical information.
                                    8-8
-------
          8.2.1  Wind Speed
                 A technical acceptance test may serve two purposes.   First,
it can verify that the instrument performs as the manufacturer claims,
assuming the threshold, distance constant and transfer function (rate of
rotation vs.  wind speed) are correct.   This test catches shipping damage,
incorrect circuit adjustments, poor workmanship, or poor QA by the manufac-
turer.  This level of testing should be equivalent to a field performance
audit.  The measurement system is challenged with various rates of rotation
on the anemometer shaft to test the performance from the transducer in the
sensor to the output.  The starting torque of the bearing assembly is
measured and compared to the range of values provided by the manufacturer
(new and replacement).
                 The other purpose of a technical acceptance test is  to deter-
mine if the manufacturer really has an  instrument which will  meet the specifi-
cation.  This action requires a wind tunnel test.  The results would  be used
to reject the instrument if the tests showed failure to comply.  An independent
test laboratory is recommended for conducting the ASTM method test.
                 The specification most likely to fail for a low cost
anemometer is threshold, if bushings are used rather than quality bearings.
A bushing design may degrade in time faster than a well designed bearing
assembly and the consequence of a failed bushing may be the replacement of
the whole anemometer rather than replacement of a bearing for a higher
quality sensor.  A receiving inspection cannot protect against this problem.
A mean-time-between-failure specification tied to a starting threshold
torque test is the only reasonable way  to assure quality instruments  if
quality brand names and model numbers cannot be required.

                                    8-9
-------
          8.2.2  Wind Direction
                 A technical  acceptance test  can  verify  the  relative
direction accuracy of the wind  vane  by employing  either  simple fixtures or
targets within a room established by sighting along a  30-60-90 triangle.
There is no acceptance test for sighting or orientation,  unless  the manu-
facturer supplies an orientation fixture and  claims that the sensor is  set
at the factory to a particular  angle (180 degrees for  example) with respect
to the fixture.  This could be  verified.
                 If sigma theta is to be calculated from direction output
samples, the time constant of the output to an instantaneous change should
be estimated.  If the direction output does not change as fast  as a test
meter on the output can react,  the time constant  is too  long.
                 If sigma theta is calculated by  the system, a  receiving
test should be devised to check its  performance.   The  manual for the  system
should describe tests suitable  for this challenge.

          8.2.3  Temperature and Temperature  Difference
                 The simplest acceptance test for temperature and temperature
difference would be a two point test, room temperature and a stirred  ice
slurry.  A reasonably good mercury-in-glass thermometer with some calibration
pedigree can be used to verify  agreement to within 1 degree C.   It is impor-
tant to stir the liquid to avoid local gradients.  It  should not be assumed
that a temperature difference pair will read  zero when being aspirated  in  a
room.  If care is taken that the air drawn into each of the shields comes
from the same well mixed source, a zero reading might  be expected.
                                    8-10
-------
                 A second benefit of removing the transducers  from the
shields for an acceptance test comes to the field calibrator and auditor.
Some designs are hard to remove and have short leads.   These conditions  can
be either corrected or noted when the attempt is first made in the less
hostile environment of a receiving space.
          8.2.4  Dew Point Temperature
                 A dew point temperature acceptance test  at one point  inside
a building, where the rest of the system is being tested,  will  provide assur-
ance that connections are correct and that the operating  circuits  are  func-
tioning.  The dew point temperature for this test should  be measured with  a
wet-dry psychrometer (Assman type if possible) or some other device in which
some measure of accuracy is documented.  If it is convenient to get a second
point outside the building, assuming that the dew point temperature is dif-
ferent outside (usually true if the building is air conditioned with water
removed or added), further confidence in the performance  is possible.  Of
course, the manufacturer's methods for checking parts  of  the system (see the
manual) should also be exercised.

          8.2.5  Precipitation
                 The receiving inspection for a precipitation  gage is straight-
forward.  With the sensor connected to the system, check  its response to water
(or equivalent weight for weighing gages) being introduced into the collector.
For tipping bucket types, be sure that the rate is less than the equivalent of
one inch (2bmm) per hour if the accuracy check is being recorded.   See the
section on calibration (8.3) for further guidance.
                                     8-11
-------
          8.2.6  Pressure
                 A check inside the building  is  adequate  for  an  acceptance
test of atmospheric pressure.   An aneroid  barometer which has been  set  to
agree with the National  Weather Service (NWS)  equivalent  sea-level  pressure
can be used for comparison.   If station pressure is to  be recorded  by the
pressure sensor, be sure that  the aneroid  is  set to agree with the  NWS
station pressure and not the pressure broadcast  on radio  or television. A
trip to the NWS office may be  necessary to set the aneroid for this agreement
since the station pressure is  sensitive to elevation and  the  NWS office may
be at a different elevation than the receiving location.

          8.2.7  Radiation
                 A simple functional test  of  a pyranometer or solarimeter
can be conducted with an electrical light  bulb.   With the sensor connected
to the system as it will be in the field,  cover  it completely with  a box
with all cracks taped with an  opaque tape. Any  light can bias a "zero"
check.  The output should be zero.  Do not make  any adjustments without
being absolutely sure the box  shields the  sensor from any direct, reflected,
or diffuse light.  Once the zero is recorded, remove the  box  and bring  a
bulb (100 watt or similar) near the sensor.  Note the output  change. This
only proves that the wires are connected properly and the sensor is sensi-
tive to light.
                 If a net radiometer is being checked,  the bulb on  the  bottom
should induce a negative output and on the top a positive output.  A "zero"
for a net radiometer is much harder to simulate.  The sensor will (or may)
detect correctly a colder temperature on the  bottom of  the shielding box than
                                    8-12
-------
the top, which may be heated by the light fixtures  in the room.   Check  the
manufacturer's manual for guidance.
     8.3  Routine Calibrations
          It is not possible to generalize a routine calibration.  One
system design might require "routine calibrations"  quarterly  while  another
might require them daily.  This section will address what the calibration
should be and how the required period might be determined.   For  this  section,
all variables will be considered under each category.
          8.3.1  Sensor Check
                 There are three types of action which can be considered a
sensor check.  First, one can look at and perform "housekeeping" services
for the sensors.  Secondly, one can measure some attribute of the sensor to
detect deterioration in anticipation of preventative maintenance.  Thirdly,
the sensor can be subjected to a known condition whose consequence  is pre-
dictable through the entire measurement system, including the sensor
transducer.  Each of these will be addressed for each variable,  where appro-
priate, within the divisions of physical  inspection and measurement and
accuracy check with known input.
                 8.3.1.1  Physical inspection
                          The first level of inspection is visual.  The
anemometer and vane can be looked at, either directly or through binoculars
or a telescope, to check for physical damage or signs of erratic behavior.
Temperature shields can be checked for cleanliness.   Precipitation  gages
can be inspected for foreign matter which might effect performance.  The
                                    8-13
-------
static port for the atmospheric  pressure  system  also  can  be examined  for
foreign matter.  Solar radiation sensors  should  be wiped  clean at every
opportunity.
                          A better level  of physical  inspection  is  a  "hands
on" check.   An experienced technician  can feel the condition  of  the anemometer
bearing assembly and know whether or not  they are in  good condition.   This
is best done with the aerodynamic shape (cup wheel, propeller, or vane)
removed.  Caution:  Damage to anemometers and vanes is  more likely  to result
from human handling than from the forces  of the  wind, especially during
removal or installation and transport  up  and down a tower.  The  proper
level of aspiration through a forced aspiration  shield  can be felt  and
heard under calm condition.
                          The best level  of sensor check  is a measurement.
The anemometer and wind vane sensors have bearings which  will certainly
degrade in time.  The goal is to change the bearings  or the sensors before
the instrument falls below operating specifications.  Measurements  of
starting torque will provide the objective data  upon  which maintenance
decisions can be made and defended.   The  presence, in routine calibration
reports, of starting torque measurements  will support the claim  for valid
data, if the values are less than the replacement torques.
                          The anemometer, identified  by the serial  number of
the aerodynamic shape, should have a wind tunnel calibration  report (see
Section 8.1) in a permanent record folder.  This is the authority  for the
transfer function (rate of rotation to wind speed) to be  used in the  next
section.  The temperature transducers, identified by  serial  number, should
have calibration reports showing their conformity for at  least three  points

                                    8-14
-------
to their generic transfer function (resistance to temperature, usually).
These reports should specify the instruments used for the calibration and
the method by which the instruments are tied to national  standards (NBS).
The less important sensors for solar radiation and atmospheric pressure can
be qualified during an audit for accuracy.

                 8.3.1.2  Accuracy check with known input
                          Two simple tests will determine the condition of
the anemometer (assuming no damage is found by the physical  inspection).
The aerodynamic shape must be removed.  The shaft is driven  at three known
rates of rotation.  The rates are known by independently  counting shaft
revolutions over a measured period of time in synchronization with the
measurement system timing.  The rates should be meaningful such as the
equivalent of 2 m/s, 5 m/s and 10 m/s.  Conversion of rates  of rotation to
wind speed is done with the manufacturer's transfer function or wind tunnel
data.  For example, if the transfer function is m/s = 1.412  r/s + 0.223,
then rates of rotation of 1.3, 3.4 and 6.9 revolutions per second (r/s) would
be equivalent to about 2, 5 and 10 m/s.  All that is being tested is the
implementation of the transfer function by the measuring  system.   The
output should agree within one increment of resolution (probably 0.1  m/s).
If problems are found, they might be in the transducer, although failures
there are usually catastrophic.  The likely source of trouble is the measure-
ment system (signal conditioner, transmitting system, averaging system and
recording system).
                          The second test is for starting torque.  This
test requires a torque watch or similar device capable of measuring in the
                                    8-15
-------
range of 0.1  to 10 gm-cm depending upon  the  specifications  provided  by the
manufacturer.
                          A successful  response to  these two tests will
document the fact that the anemometer is operating  as well  as it did  at
receiving inspection, having verified threshold and accuracy.  Changes in
distance constant are not likely unless  the  anemometer design has changed.
If a plastic cup is replaced by a stainless  steel cup, for  example,  both
the transfer function and the distance constant will likely be different.
The distance constant will vary as the inverse of the air density.   If a
sea-level distance constant is 3.0m,  it  may  increase to 3.5m in Denver
and 4.3m at the mountain passes in the Rockies.
                          For wind direction, a fixture holding the  vane,
or vane substitute, in positions with a  known angle change  is a fundamental
challenge to the relative accuracy of the wind vane.  With  this method,
applying the appropriate strategy for 360 or 540 degree systems, the accuracy
of the sensor can be documented.  The accuracy of the wind  direction measure-
ment, however, also depends on the orientation of the sensor with  respect
to true north.
                          The bearing to distant objects may be determined
by several methods.  The recommended  method  employs a solar observation
(see Reference 3, p.11) to find the true north-south line where it passes
through the sensor mounting location.  Simple azimuth sighting devices can
be used to find the bearing of some distant  object  with  respect to the
north-south line.  The "as found" and "as left" orientation readings should
report the direction to or from that  distant object. The object should  be
                                    8-16
-------
one toward which the vane can be easily aimed and not likely to become
hidden by vegetation or construction.
                          There are two parts of most direction vanes which
wear out.  One part is the bearing assembly and the other is the transducer,
usually a potentiometer.  Both contribute to the starting torque and hence
the threshold of the sensor.  A starting torque measurement will  document
the degradation of the threshold and flag the need for preventive mainte-
nance.  An analog voltmeter or oscilloscope is required to see the noise
level of a potentiometer.  Transducer noise may not be a serious problem
with average values but it is likely to have a profound effect on sigma
theta.
                          The dynamic performance characteristics of a wind
vane are best measured with a wind tunnel test.  A generic test of a design
sample is adequate.  As with the anemometer, the dynamic response character-
istics (threshold, delay distance and damping ratio) are density dependent.
                          Temperature transducers are reasonably stable, but
they may drift with time.  The known input for a temperature transducer is
a stable thermal mass whose temperature is known by a standard transducer.
The ideal thermal mass is one with a time constant on the order of an hour
in which there are no thermal sources or sinks to establish local gradients
within the mass.  It is far more important to know what a mass temperature
is than to be able to set a mass to a particular temperature.
                          For temperature difference systems, the immersion
of all transducers in a single mass as described above will  provide a
zero-difference challenge accurate to about 0.01  degrees C.   When this test
is repeated with the mass at two more temperatures, the transducers will

                                    8-17
-------
have been challenged with respect to how well  they  are  matched  and  how  well
they follow the generic transfer function.   Mass  temperatures in  the  ranges
of 0 to 10 degrees C, 15 to 25 degrees C, and  30  to 40  degrees  C  are  recom-
mended.  A maximum difference among the three  temperatures  (i.e.,, 0,  20,  and
40 degrees C) is optimum.  Once the match has  been  verified,  known  resis-
tances can be substituted for the transducers  representing  temperatures,
according to the generic transfer function,  selected to produce known
temperature difference signals to the signal conditioning circuitry.  This
known input will challenge the circuitry for the  differential measurement.
                          Precipitation sensors can be  challenged by  insert-
ing a measured amount of water, at various reasonable rainfall  rates  such
as 25 mm or less per hour.  The area of the  collector can be  measured to
calculate the amount of equivalent rainfall  which was inserted.  The  total
challenge should be sufficient to verify a 10% accuracy in  measurement  of
water.  This does not provide information about errors  from siting  problems
or wind effects.
                          Dew point temperature (or relative humidity),
atmospheric pressure and radiation are most  simply  challenged in  an ambient
condition with a collocated transfer standard.  An  Assmann  psychrometer may
be used for dew point.  An aneroid barometer checked against  a  local  National
Weather Service instrument is recommended for atmospheric  pressure.  Another
radiation sensor with some pedigree or manufacturer's certification may be
used for pyranometers and net radiometers.  A complete opaque cover will
provide a zero check.
                                    8-18
-------
          8.3.2  Signal  Conditioner and Recorder Check
                 For routine calibration of measurement  circuits  and
recorders, use the manufacturer's recommendations.   The  outputs required  by
the test described in 8.3.1.2 must be reflected  in  the recorded values.
Wind speed is used as an example in this section.  Other variables will
have different units and different sensitivities but the principle is  the
same.  For sub-system checks, use the manual for specific guidance.

                 8.3.2.1  Analog system
                          Some systems contain "calibration"  switches  which
are designed to test the stability of the circuits  and to provide a basis
for adjustment if changes occur.  These should certainly be exercised  during
routine calibrations when data loss is expected  because  of calibration.   In
the hierarchy of calibrations, wind tunnel is first, known rate of rotation
is second, substitute frequency is third and substitute  voltage is fourth.
The "calibration" switch is either third or fourth.
                          If analog strip chart  recorders are used, they
should be treated as separate but vital parts of the measurement  system.
They simply convert voltage or current to a mark on a time scale1printed  on
a continuous strip of paper or composite material.   The  output voltage or
current of the signal conditioner must be measured  with  a calibrated meter
during the rate of rotation challenge.  A simple transfer function, such  as
10 m/s per volt, will provide verification of the measurement circuit  at
the output voltage position.  The recorder can be challenged  separately by
inputting known voltages and reading the mark on the scale, or by noting
                                    8-19
-------
the mark position when the rate of rotation and output voltage are both  known.
See the recorder manual for recommendations should  problems  arise.
                          This special  concern with recorders results from
the variety of problems which analog recorders can  introduce.   A good measure-
ment system can be degraded by an inappropriate recorder selection.   If
resolution is inadequate to distinguish between 1.3 m/s and  1.5 m/s, a 0.2
m/s accuracy is impossible.  If enough  resolution is just barely there,
changes in paper as a function of relative humidity and changes in paper
position as it passes the marking pen and excessive pen weight on the paper
can be the limit of accuracy in the measurement.  If the strip chart recorder
is used only as a monitor and not as a  backup for the primary system, its
accuracy is of much less importance. The recorder from which data are re-
covered for archiving is the only recorder subject to measurement accuracy
specifications.

                 8.3.2.2  Digital system
                          A digital system may also present  a variety of con-
cerns to the calibration method.  One extreme is the digital system which
counts revolutions or pulses directly from the sensor.  No signal condition-
ing is used.  All that happens is controlled by the software of the digital
system and the capability of its input  hardware to detect sensor pulses  and
only sensor pulses.  The same challenge as described in 8.3.1.2 is used.  The
transfer function used to change rate of rotation to m/s should be found in
the digital software and found to be the same as specified by the manufacturer
or wind tunnel test.   If any difference is found between the speed calculated
from the known number of revolutions in the synchronous time period and  the
                                    8-20
-------
speed recorded In the digital recorder, a pulse detection problem is certain.
A receiving inspection test may not uncover interference pulses  which exist
at the measurement site.  For solution of this type of problem,  see the
digital  recorder manufacturer's manual or recommendations.
                          A digital data logger may present different con-
cerns.  It may be a device which samples voltages,  averages them, and trans-
fers the average to a memory peripheral, either at  the site or at the end
of a communication link.  Conversion to engineering units may occur at
almost any point.  The routine calibration should look at the output voltage
of a signal conditioner as a primary point to assess accuracy of measurement.
Analog to digital conversion, averaging and transmission and storage would
be expected to degrade the measurement accuracy very little.  Such functions
should contribute less than 0.05 m/s uncertainty from a voltage  input to a
stored average value.  If greater errors are found  when comparing known rates
of rotation and known signal conditioning output voltages to stored average
wind speed values, check the data logger manual  for specifications and
trouble-shooting recommendations.

          8.3.3  Calibration Data Logs
                 Site log books must record at least the following:
                 A.   Date and time of the calibration period (no valid data)
                 B.   Name of calibration person or  team members
                 C.   Calibration method used (this  should identify SOP number
                     and data sheet used)
                 D.   Where the data sheet or sheets can be found on site
                 E.   Action taken and/or recommended
                                    8-21
-------
                 The data sheet should contain this same information along
with the measurement values found and observations made.   Model  and serial
numbers of equipment tested and used for testing must appear.   The original
report should always be found at the site location and a copy  can be used
for reports to management (a single-copy carbon form could be  used).  The
truism that "it is impossible to have too many field notes" should be under-
scored in all training classes for operators and auditors.

          8.3.4  Calibration Report
                 The calibration report may be as simple as copies of the
calibration forms with a cover page, summary and recommendations.
                 While the calibration forms kept at the site  provide the
basis for the operator or the auditor to trace the performance of the instru-
ment system, the copies which become a part of the calibration report pro-
vide the basis for management action should such be necessary.  The cali-
bration report should travel from the person making out the report through
the meteorologist responsible for the determination of data validity to the
management person responsible for the project.  Any problem should be high-
lighted with an action recommendation and a schedule for correction.  As soon
as the responsible management person sees this report the responsibility
for correction moves to management, where budget control usually resides.  A
signature block should be used to document the flow of this information.

          8.3.b  Calibration Schedule/Frequency
                 Frequency of calibration may be determined by an iterative
process; the minimum period may be fixed by regulation.  Whenever a calibration
                                    8-22
-------
of the type described in 8.3.1.2 is conducted, monitored data are lost.   The
first field calibration should be just after installation is completed.   The
second might be a week later.  If problems are found and corrected, the  one
week period should be repeated.  When no problems are found, the next calibra-
tion might be a month later.  If no problems are found at one month, the
next calibration might be three months later.   If the next calibration is
another three months later and shows no problems, try six months.  The
system should be calibrated at least every six months.
                 It must be clearly understood that the risk of the loss of
large amounts of data increases when long periods of time are allowed to
pass without any attention paid to the data or the instrument.  The method
of establishing the frequency of calibrations  presumes the existence of
operational checks and preventive maintenance  as described below.  The most
important function to avoid loss of large amounts of data is the routine
(daily or at least weekly) quality control (QC) inspection of the data by an
experienced meteorologist.  The data themselves will usually expose failures
of the measurement system.  The lack of problems reported from progressively
less frequent calibration and the experience gained from weekly assessment
of data validity is the most cost effective method for archiving the most
valid data.  A carefully followed program of preventive maintenance will
lower the risk of large blocks of invalid data.
          8.3.6  Data Correction Based on Calibration Results
                 Corrections to the raw data are to be avoided.  A thorough
documentation of an error clearly defined may  result in the correction of
data (permanently flagged as corrected).  For  example, if an operator
                                    8-23
-------
changes the transfer function in a digital  logger  program  and  it  1s  subtle
enough not to be detected in the quality  control Inspection  of the data
stream, but is found at the next calibration,  the  data  may be  corrected.
The correction can be calculated from the erroneous  transfer function  and
applied to the period starting when the logger program  was changed  (determined
by some objective method such as a log entry)  and  ending when  the error was
found and corrected.
                 Another example might be a damaged  anemometer cup  or  pro-
peller.  If an analysis of the data points  to  the  time  when  the damage
occurred, a correction period can be determined.   A  wind tunnel test will
be required to find a new transfer function for the  damaged  cup or  propeller
assembly.  With the new transfer function defining the  true  speed respon-
sible for a rate of rotation, and with the assumption that the average
period is correctly represented by a steady rate of  rotation,  a correction
can be made and flagged.  This is a more  risky example  and judgment  is
required since the new transfer function  may be grossly different and
perhaps non-linear.
     8.4  Audits
          The system audit (see Ref. 44)  is intended to provide an  independent
assessment of the QA Plan, how it is being implemented, and  how the evidence
of the operator's actions is kept.  Given the  joint  goal of  the auditor  and
the operator to achieve valid data with defendable documentation, the  audit
becomes a training tool.  Whichever is the most experienced  will  teach the
other for the good of the joint goal.
                                    8-24
-------
          When the period of time between calibrations  or performance  audits
is three months or longer, it is critical to examine the methods  by which  the
experienced quality control meteorologist determines the validity of the data
on a routine schedule.   It is also important to assure  the proper documentation
of the data inspection process where changes or selective deletions are allowed.
          The performance audit is a direct challenge to the performance of
the measurement system.  The recommended methods described in 8.3.1.2  are
the same as would be used in a performance audit for the reason mentioned
in Section 8.0.
          The use of a collocated transfer standard is  an additional chal-
lenge to be considered.  This is accomplished by locating a like  instru-
ment as close as practical to the instrument being audited to serve as a
standard for comparison of the transfer function.   If a good exposure  is
possible for a collocated instrument, such a test can be considered a
substitute for a wind tunnel test of the transfer function.  The  wind
tunnel will always be superior for controlled testing in laminar  flow.  The
data taken in Boulder,  Colorado and partially reported  in Kaimal  et al.^5
suggest a collocated instrument can provide an opportunity to assess the
absolute accuracy of a monitoring system within the accuracy specifications
listed in Section 8.1.   If a suitable data sample size  is achieved over a
reasonable range of wind speeds (usually found in a few diurnal cycles),
the average difference can be considered the accuracy error and the root-
mean-square of the difference can qualify the test period and relative
siting as acceptable or not.  An experienced assessment of exposure is
critical to the proper use of this method.
                                    8-25
-------
          Somewhere between the system audit and the performance audit is
found the independent technical appraisal  of such things as the suitability
of the deployment of sensors with respect  to the intended data application
(sensor siting, Section 3.0), the sample summarization method (Section 6.1.2),
and model suitability (Section 6.b).  The  value of this type of appraisal is
proportional to the qualifications of the  auditor, but the fact that these
questions are addressed at all will  help focus the thinking to these impor-
tant considerations.  As a consensus develops on these operational  design
considerations, objective guidance will follow.

          8.4.1  Schedule
                 Audits are most effective in the initial phases of monitor-
ing programs.  It would be useful to have  an audit concurrent with the
initial field calibration.  The audit methods might be carried out by the
operators with the auditor assisting and making an independent report of
the findings.
                 The optimum frequency of  an audit is dependent upon the
findings as they affect data validity.  When the effort of the operating
organization provides all the technical oversight to assure data reliability
and validity, the audit becomes simply an  independent statement to that
effect.  When the operating organization falls short of that goal, the
audit becomes a motivation for improvement.  A six month frequency should
be adequate for audits.  This provides a beginning, a mid-point check and a
final check for a one year monitoring program.  The audits will comment on
the calibration performance and coupled with experienced data quality control,
                                    8-26
-------
become a basis for legal  claim to data validity.   The independence of the
auditor is critical to the legal  claim of validity without operational  bias.

           8.4.2  Scope
                  The scope of the audit is discussed above in Section 8.4.
An audit must begin with a briefing which states  the goals of the audit,
the methods to be employed, and the work required from the operator in
assistance to the auditor.  This  should include a specific requirement for
the operator to remove the anemometer, wind vane  and temperature instruments
from their mounts, after as-found observations are made,  and connect them
back to the system in a sheltered work place.   A field audit (or calibration,
for that matter) should be as close to a laboratory test  as conditions
allow.  It is not acceptable to merely audit at the top of a 10 meter mast
or 60 meter tower.  When the audit is completed,  an exit  interview is
required.  Management level people should be present at both the initial
briefing and the exit interview.

          8.4.3  Audit Report
                 The audit report is the evidence of the  audit.  It must  be
complete and submitted in a timely manner, within 3D days of the audit
performance.  The findings should be as objective as possible but subjective
judgments are valuable, particularly in those  areas mentioned above that
fall between the purview ot the system audit and  the performance audit.
Where possible the audit report should contain copies of  the forms used in
the audit rather than, or in addition to, summarizations  of the findings.
                                    8-27
-------
          8.4.4  Audit Responses or Corrective Action
                 An audit is not worth  the  cost if  there  is  not  the  support
from the management of the operators to react  promptly  to required corrective
action.  The highest priority must  rest with the performance audit findings
where questionable data are being collected.   Immediate corrective action is
required before the collection of data  can  be  considered  useful.

     8.5  Operational  Checks and Preventive Maintenance
          There may be little difference between operational checks  and
calibration checks.  If the same person performs both  functions, they may
both be considered calibration checks.   As  such they deserve high credibility
with respect to data validity.  It may  be the  case, as  it often is,  that
other measurements (such as air chemistry)  are made at  the same station.
These instruments usually require more  frequent attention than do the meteoro-
logical instruments.  As long as the visit  takes place, some attention to the
meteorological instrument is advisable.  The  following  sub-sections  will
assume that the frequent visitor to the station is  a different person from
the one who calibrates the meteorological instrument.   The checker  requires
training to properly check the meteorological  system.

          8.5.1  Visual Inspection
                 A look at the anemometer and  vane, probably through field
glasses is desirable.   Look for any evidence  of physical  damage or abnormal
condition.  For example, if icicles are hanging from the  cups or vane, it
should be communicated to the operator  and noted in the log.
                 A diagram showing switch positions for normal operation
should be posted near the system electronics.   The person visiting  for

                                    8-28
-------
other reasons should check to see that the switches are in the correct
positions.  If not, contact should be made with a knowledgeable operator
before changes are made.  The observation, consulting information  and
consequent action must be entered in the log.

          8.5.2  Manual Inspection
                 There should not be any manual (hands-on) inspection of
the meteorological instrument by persons not qualified to perform  calibra-
tions.

          8.5.3  Recorder Inspection
                 If the system has an analog recorder, the person  visiting
for other purposes should check the recorded data for signs of malfunction.
If problems are found, contact the operator and decide what the appropriate
action might be.
                 Unwind the strip chart so that the previous 24 hours can
be seen.  Look at the range of values recorded.  Does it look reasonable or
does there seem to be a limit on the high or low end of the trace?  Check
the nature of the speed and direction fluctuations.   During the day  there
should be more wiggles (more turbulence) than at night.  If the trace is
always steady it might be a sign of excessive pen weight or a defective
sensor.  Check to see that the marking method (inking, for example)  is
working reliably and that supplies are sufficient to last until  the  next
scheduled service visit.  Check the paper drive to be sure that the  chart
is moving accurately with time and that the sprocket pins are engaged in
the paper drive holes.  Check the time marks on the chart to be sure they
are correct.   Mark on the chart a note indicating who and when this  check
                                    8-29
-------
was made.  Also note in the log book that  the check  was  made.   Rewind  the
strip chart and make sure that it  is moving  correctly  before leaving.
                 If there is another indication  of wind  meteorological  out-
puts in the system, a meter or a digital  readout for example,  note  the
values on the strip chart.  They may or may  not  agree  exactly  because  of
averaging time constants, but they should  agree  generally.   If they do not,
watch the meter for a few minutes  and note a few values  on  the chart paper.
If there is still  not agreement, call  the  operator and report  the finding
and note it in the log.  A visual  examination of the direction the  wind
vane is pointing may also be used  to independently check the recorder  out-
put, provided that the wind direction is fairly  steady.   This  check will
detect slippage in the alignment of the wind direction sensor  due to a
loose collar.
          8.5.4  Preventive Maintenance
                 8.5.4.1   Wind Speed
                          The anemometer has just one  mechanical  system
which will benefit from preventive maintenance.   That  is the bearing as-
sembly.  There are two strategies  from which to  choose.   One is to  change
the bearings (or the entire instrument if  a  spare is kept for  that  purpose)
on a scheduled basis and the other is to make the change when  torque measure-
ments suggest change is in order.   The former is most  conservative  with
respect to data quality assuming that any  time a torque measurement indicates
a bearing problem, the bearing will be changed as a  corrective maintenance
action.
                          As routine calibrations become less  frequent
(8.3.5), the probability increases that a  starting torque measurement  will

                                    8-30
-------
be made which indicates the anemometer is outside its performance specifica-
tion.  This will  effect both the threshold (by increasing it)  and the  trans-
fer function (by moving the non-linear threshold toward high speeds).   It
is unlikely that corrections can be properly made to the data  in  this  case.
The consequence might be the loss of a half-year's data, if that  is  the
period for routine calibration.   If experience indicates that  the anemometer
bearing assembly shows serious wear at the end of one year or  two years
(based on torque measurements),  a routine change of bearings at that frequency
is recommended.

                 8.5.4.2  Wind Direction
                          The wind vane usually has two mechanical  systems
which will benefit from preventive maintenance.   The bearing assembly  is
one and can be considered in the same way as the anemometer bearing  assembly
described above.   The other is the potentiometer which will  certainly  "wear
out" in time.  The usual mode of failure for a potentiometer is to become
noisy for certain directions and then inoperative.  The noisy  stage  may not
be apparent in the average direction data.  If sigma theta is  calculated,
the noise will  bias the sigma value toward a higher value.   It will  probably
not be possible to see early appearance of noise in the sigma  data.  When
it becomes obvious that the sigma is too high, some biased  data may  already
have been validated and archived.  Systems with time constant  circuits built
into the direction output will both mask the noise from the potentiometer
(adding to the apparent potentiometer life) and bias the sigma theta toward
a lower value.   Such circuits should not be used if they influence the actual
output capability of the sensor.
                                    8-31
-------
                          Each manufacturer may  be different  in  their selec-
tion of a source and specifications  used  in buying potentiometers.   The  oper-
ator needs to get an expected life for the potentiometer from the manufacturer
and monitor the real life with a noise sensitive test.   An  oscilloscope  is
best and can be used without disrupting the measurement. When potentiometer
life expectations have been established,  a preventive maintenance replacement
on a conservative time basis is recommended.

                 8.5.4.3  Temperature and Temperature Difference
                          Aspirated  radiation shields use fans which will
also fail in time.  The period of this failure should be several  years.   The
temperature error resulting from this failure will be easily  detected by a  QC
meteorologist inspecting the data.  Some aspirated radiation  shields include
an air flow monitoring device or a current check which  will  immediately
signal a disruption in aspiration.  Preventive maintenance  is not required
but spare fans should be on the shelf so that a change  can  be made  quickly
when failure does occur.

                 8.5.4.4  Dew Point  Temperature
                          Field calibration checks of the dew point tempera-
ture measurement system can be made with a high-quality Assmann-type or  por-
table, motor-aspirated psychrometer.  Sling psychrometers should not be  used.
Several  readings should be taken at the intake of the aspirator or  shield at
night or under cloudy conditions during the day.  These field checks should
be made  at least monthly, or in accordance with manufacturer's suggestions,
and should cover a  range of relative humidity values.
                                    8-32
-------
                          Periodically (at least  quarterly)  the lithium chlo-
ride in dew cells should be removed and recharged with  a fresh  solution.   The
sensor should be field-checked as described above before and at least  an hour
after the lithium chloride solution replacement.
                          If cooled-mirror type dew point systems  are  used,
follow the manufacturer's service suggestions  initially.  The quality  of the
data from this method of measurement is dependent upon  the mirror  being kept
clean.  The frequency of service required to keep the mirror clean is  a func-
tion of the environment in which the sensor is installed. That environment
may vary with seasons or external weather conditions.   If changes  in dew
point temperature of a magnitude larger than can  be tolerated are  found after
service scheduled according to the manufacturer's suggestion, increase the
service frequency until the cleaning becomes preventive maintenance rather
than corrective service.  This period will vary and can be defined only by
experience.  Station log data must include the "as found" and the  "as  left"
measurements.  Dew point temperature does not  change rapidly (in the absence
of local sources of water) and the difference  between the two measurements
will usually be the instrument error due to a  dirty mirror.

                 8.5.4.5  Precipitation
                          The gage should be inspected  at regular  intervals
using a bubble level  to see that the instrument base is mounted level.
Also, the bubble level  should be placed across the funnel orifice  to see
that it is level.  The wind screen should also be checked to see that  it is
level, and that it is located 1/2 inch above the  level  of the orifice, with
the orifice centered  within the screen.
                                    8-33
-------
                 8.5.4.6  Pressure
                          The output  of  the  pressure  sensor  should  be
regularly checked against a collocated instrument.  A precision  aneroid
barometer can be used for this check.  The collocated barometer  should be
occasionally checked against a mercurial  barometer  reading at  a  nearby NWS
station.

                 8.5.4.7  Radiation
                          The optical  hemispheres on  pyranometers  and  net
radiometers should be cleaned frequently (preferably  daily)  with a  soft,
lint-free cloth.  The surfaces of the hemispheres should be  regularly
inspected for scratches or cracks.   The  detectors should be  regularly
inspected for any discoloration or deformation.   The  instruments should  be
inspected during cool temperatures for any  condensation which  may  form on
the interior of the optical surfaces.
                          While calibrations must be  done by the manufac-
turer, radiation can be field-checked using  a recently-calibrated,  collocated
instrument.  Since signal processing is  particularly  critical  tor these
sensors, the collocated instrument should also use  its own signal  conditioner
and data recording system for the check.  This kind of field check  should
be done every six months.

                 It is mandatory to log "as  found"  and "as left" information
about the parts of the system which seem to require work.  Without this
information it becomes difficult, if not impossible,  to assess what data are
usable and what are not.
                                    8-34
-------
     8.6  Data Validation
          The data collected by an on-site meteorological  monitoring program
must be validated prior to their use in air quality modeling analyses.   The
data validation process should consist  of a review of the  data by  experienced
personnel, a screening of the data to identify possible incorrect  values,  and
a comparison of randomly selected data  with other available data.   These
procedures, if followed, will help to identify problems within the monitoring
program which escape detection by other quality assurance  checks.

          8.6.1  Manual Data Review
                 Soon after the meteorological data have been collected, a
hard copy of the 15-minute or 1-hour averaged values should be reviewed  by
experienced personnel.  The data should be scanned to determine if the
reported values are reasonable and in the proper format.   Periods  of missing
data should be noted and investigated as to the causes.

          8.6.2  Data Screening Tests
                 The data should then be run through a screening program.
This involves comparing the measured value with some expected value or
range of values.  The range test, in which data are checked to see if they
fall within specified limits, is the most common and simplest test.   The
limits are set usually based upon historical data or physically realistic
values.  In a similar test, the rate of change test, the difference between
the current measured value and the value from the previous time period is
compared with physically realistic values.   Suggested screening criteria
are listed in Table 8-1.  Other values  may be more appropriate for a given
location, therefore site-specific screening values should  be developed by

                                    8-35
-------
                                Table 8-1
                     Suggested  Data Screening Criteria'
 Meteorological
    Variable
 Wind Speed
 Wind Direction
 Temperature
 Temperature
 Difference
 Dew Point
 Temperature
 Precipitation
 Pressure
 Radiation
   Sc reen i ng Criteria

Flag the data if the value:

- is less than zero or greater than  25 m/s
- does not vary by more than  0.1  m/s for  3  consecutive
  hours
- does not vary by more than  0.5  m/s for  12 consecutive
  hours

- is less than zero or greater than  360 degrees
- does not vary by more than  1 degree for more
  than 3 consecutive hours
- does not vary by more than  10 degrees for 18 consecutive
  hours

- is greater than the  local  record high
- is less than the local  record low
  (The above limits could be  applied on a monthly  basis.)
- is greater than a 5°C change from  the previous hour
- does not vary by more than  0.5°C for  12 consecutive hours

- is greater than 0.1°C/m during  the daytime
- is less than -0.1°C/m during the night  time
- is greater than 5.0°C/m or  less than  -3.U°C/m

- is greater than the  ambient temperature for the  given
  time period
- is greater than a 5°C change from  the previous hour
- does not vary by more than  0.5°C for  12 consecutive hours
- equals the ambient temperature  for 12 consecutive hours

- is greater than 25 mm in one hour
- is greater than 100 mm in 24 hours
- is less than 50 mm in three months
  (The above values can be adjusted  based on local  climate.)

- is greater than 1060 mb (sea level)
- is less than 940 mb  (sea level)
  (The above values should be adjusted  for
  elevations other than sea level.)
- changes by more than 6 mb in three hours

- is greater than zero at night
- is greater than the maximum possible  for
  the date and latitude
*Some criteria may have to be changed for  a  given  location.

                                    8-36
-------
an experienced meteorologist.  It" the data do not fall  within the screening
criteria, the data should be flagged for further investigation.   Relation-
ships between different variables should be considered in evaluating flagged
data.  Conditional flags may also be developed to account for these relation-
ships in the screening program, e.g., comparing temperature and  dew point
during precipitation events, or checking for low wind speeds during highly
variable wind directions.

          8.6.3  Comparison Program
                 After the data have passed through the screening program,
they should be evaluated in a comparison program.  Randomly selected values
should be manually compared with other available, reliable data  (such as,
data obtained from the nearest National Weather Service observing station).
At least one hour out of every 10 days should be randomly selected.  To ac-
count for hour-to-hour variability and the spatial displacement  of the NWS
station, a block of several hours may be more desirable.  All data selected
should be checked against corresponding measurements at the nearby station(s).
In addition, monthly average values should be compared with climatological
normals, as determined by the National Weather Service from records over a
30-year period.  If discrepancies are found which can not be explained by
the geographic difference in the measurement locations or by regional cli-
matic variations, the data should be flagged as questionable.

          8.6.4  Further Evaluations
                 Any data which are flagged by the screening program or the
comparison program should be evaluated by personnel with meteorological
expertise.  Decisions must be made to either accept the flagged  data, or

                                    8-37
-------
discard and replace it with back-up or interpolated  data,  or data from a

nearby representative monitoring station (see  Section  6.5.3).  Any changes

in the data due to the validation process should  be  documented as to  the

reasons for the change.  If problems in the monitoring system are identified,

corrective actions should also be documented.   Any edited  data should continue

to be flagged so that its reliability can be considered in the interpretation

of the results of any modeling analysis which  employs  the  data.


     8.7  Recommendations

          It is recommended that the quality assurance (QA) program for an
on-site meteorological monitoring system should follow a QA plan  that has
been approved by appropriate project or organizational authority.  The QA
function should be independent of the organization responsible for the
collection of the data and the maintenance of  the measurement systems.

          To insure that instrumentation of proper accuracy and  response
characteristics are purchased, procurement documents for meteorological
monitoring systems should include the specifications required for the
applications of the data (see Section 5.0), along with the test  method by
which conformance with the specification will  be  determined.  The procurer
should review the manufacturer's documentation of the  tests used  to demon-
strate an instrument's conformance to specifications.   An  instrument  should
undergo an acceptance test to verify that it performs  as the manufacturer
claims, assuming that the specifications are correct.   These acceptance
tests should be similar in scope to a field calibration.

          Routine system calibrations and system  audits should be performed
at the initiation of a monitoring program and  at  least every six  months there-
after.  More frequent calibrations and audits  may be needed in the early
stages of the program if problems are encountered, or if valid data retrieval
rates are unacceptably low.

          Regular and frequent routine operational checks  of the monitoring
system are essential to ensuring high data retrieval rates.  These should
include visual inspections of the instruments  for signs of damage or wear,
inspections of recording devices to ensure correct operation and  reasonable-
ness of data and periodic preventive maintenance measures.  The  latter
should include periodic checks of wind speed and  direction bearing assemblies,
cleaning of aspirated shield screens in temperature  systems, removal  and
recharging (at least quarterly) of lithium chloride  dew cells, cleaning the
mirror in cooled mirror dew cells, clearing the precipitation gage funnel
of obstructing debris, and frequent (preferably daily) cleaning  of the
optical surface of a pyranometer or net radiometer.
                                    8-38
-------
          Also crucial to achieving acceptable valid data retrieval  rates
is the regular review of the data by an experienced meteorologist.   This
review should include a visual scanning of the data for reasonableness, and
automated screening and comparison checks to flag out-of-range or unusual
values.  This review should be performed at least weekly, and preferably
on a daily basis.
                                   8-39
-------
9.0  REMOTE SENSING - DOPPLER SODARS
     In recent years, Doppler SODAR (an acronym for Sound Detection And
Ranging) systems have gained recognition as effective tools for remote
measurement of meteorological variables at heights up to several  hundred
meters above the surface.  There has been an increased interest in using
SODARs to develop the meteorological data bases required as input to dis-
persion models.  While SODARs in rare cases have been approved and used for
this purpose, there is a distinct void in terms of the guidance needed to
help potential users and the regulatory community alike develop acceptable
on-site meteorological measurement programs with SODARs.  The purpose of this
section of the document is to provide a first attempt at filling this void.
     Two intercomparison experiments, carried out in 1979 and 1982, compared
winds measured by Doppler SODAR systems manufactured by four different companies
against tower measurements at the Boulder Atmospheric Observatory (BAO).^5,46
The results of the intercomparison experiments were quite encouraging for mean
winds.  All four systems demonstrated virtually no bias for wind speed and
direction, and scatter was in a range that might be expected, given that the
SODAR systems were measuring winds in volumes of air displaced in space and
time as opposed to the single-point tower measurements.  Turbulence measure-
ments were not as encouraging (see the discussion on this topic later in
this section) although they do hold some promise.
     While encouraging, the BAO intercomparison results should not be
regarded as an unqualified endorsement of SODAR technology.  Meteorological
conditions during the test and characteristics of the BAO site were close
to ideal for optimal  SODAR performance.  Furthermore, manufacturers operated
their own systems and were given the opportunity to submit only data that

                                    9-1
-------
they believed were valid.  Many real-world applications  involve conditions
that may produce return spectra that are interpreted as  valid  when  in  fact
they are not, such as high background noise,  electrical  interference,  and
ground clutter.  Careful  attention to siting  requirements  and  data  validation
procedures is necessary to successfully overcome these real-world problems.
     Doppler SODARs operate on a fundamentally simple principle, yet the
systems that control their operation are quite complex.  Thanks to  diligent
work on the part of SODAR manufacturers, systems have been engineered  to
operate reliably and with relatively little operator interface.  However,
the potential user should be aware that unattended and/or  careless  operation
of a SODAR could result in the collection of  erroneous data.   Diligence and
close scrutiny of the data on a regular basis, by someone  experienced  in
meteorology and trained to recognize instrument problems,  is  a necessity
(this is true for any meteorological measurement system, but  particularly
so for SODARs).
     It should be noted that SODAR systems made by different  manufacturers
differ greatly in the generation of transmit  pulses and in analyzing and
processing return echoes from the atmosphere.  It is not yet  possible to make
definitive recommendations as to which system works best in specific applica-
tions.  Because of these differences (and because of the unique nature of
Doppler SODARs), guidance provided herein is  more generic  than in previous
sections.  Specific operating procedures and quality assurance plans prepared
based on this guidance and on other case-specific factors  should provide
feedback so that the guidance can be expanded and improved based on experi-
ence.  The guidance may also be expanded or modified based on further con-
trolled tests of Doppler SODARs that may be conducted at BAO  in the future.

                                    9-2
-------
Such tests are anticipated to evaluate developments in SODAR  technology
designed to yield better turbulence measurements and the performance of
automated data validation routines.
     The development in recent years of "mini-SODAR" technology47 represents
a somewhat different approach to remote acoustic sounding, involving a phased
array of speakers in place of a large transducer and antenna, and operating
at much higher frequencies than more conventional  SODARs.  This technology
allows measurements to be taken much closer to the surface than with more
conventional SODARs, but is considered to be a research tool  at this time.
The information presented in the rest of this section is applicable primarily
to more conventional SODARs, represented by the types of instruments tested
in the BAO intercomparison experiments.
     9.1  SQDAR Fundamentals
          The requirements for installing and operating a SODAR and for
developing a modeling data base flow directly from the requirements for
obtaining a good single pulse return from one antenna.  This  section discusses
the SODAR fundamentals involved with getting a good signal return.  An under-
standing of these fundamentals will help in understanding what needs to be
done to develop an acceptable data base.
          A SODAR transmits a strong (typically 100-300 watts) acoustic pulse
into the atmosphere and listens for that portion of the transmitted pulse
that is scattered and returned.  A monostatic system uses the same acoustic
driver both to transmit the pulse (driver acting as a powerful speaker) and
to receive the return signal (driver acting as a sensitive microphone).  A
bistatic system uses different antennas to transmit and receive.   Monostatic
                                    9-3
-------
systems generally have collocated  antennas while  a bistatic configuration


generally requires that the antennas  be  separated by a distance  (typically


several hundred meters) that is  determined by  the height  at which measure-


ments are desired.  This section is concerned  primarily with the most


common Doppler SODAR configuration, namely a collocated monostatic  system.


          A volume of air will  scatter incident acoustic  energy.  Most  of


the scattering occurs in the direction of propagation, but a small  percentage


of the energy is scattered back  to the source.  Scattering is  due to wind


speed and temperature discontinuities in the volume of air.  An  equation


has been developed48 that expresses the  amount of scattering as  a function


of the angle measured to the direction of propagation of  the transmitted


pulse, and the velocity and thermal structure  functions,  Cv2 and Cj .   The


structure functions can be interpreted as expressing the  degree  of  instan-


taneous velocity or temperature difference between  points a unit distance


apart.  If the direction of propagation  is 0°  and scattering directly  back


to the source is 180°, the following  generalizations can  be made based  on


the scattering equation:


          1.  There is no scattering  at  90°  or 270°  (right angles);


          2.  Scattering at 180° is due  to Cy2 only, where Cy   scattering


is a maximum;


          3.  Scattering at intermediate angles is  due to both Cy   and  Cy2;

                        y
the contribution from Cv  reaches a maximum  at 135°.


          Return signal strength for  a bistatic system thus depends on  both


Cj2 and Cv  , while the strength of the returned signal for a monostatic system

                  P
depends only on Cy .  Scattering is  accomplished  by temperature  variations  on


a spatial scale of one-half of the wavelength  of  the transmitted sound, approx



                                    9-4
-------
imately 10 cm for a SOOAR operating at 1500 Hz.49  The return signal  is  scat-
tered from many of these small  "targets" in the atmosphere.
          The existence of atmospheric targets for a monostatic SODAR depends
on the presence of a temperature gradient and small-scale turbulence  that
creates local instantaneous temperature differences much greater than the  mean
temperature gradient.  A strong return signal can be produced either  with  an
unstable potential temperature gradient and little wind shear (in a convective
boundary layer) or with a stable potential  temperature gradient and large
wind shear (in a stable boundary layer).  Fortunately for the science of
                  o
doppler SODARs, Cj  never disappears entirely and, although  a diurnal  pattern
of signal strength does occur, adequate targets are available most of the
time.
          Although a strong signal  return indicates the presence of many
atmospheric targets, it does not by itself signify that mixing is occurring
on a scale that would diffuse the plume from a pollutant source.  It  is
through the analysis of time-height patterns of signal  strength, generally
displayed on an analog facsimile chart, that mixing height information is
inferred (see the later discussion  on mixing heights).
          The real strength of a SOOAR system (for developing modeling data
bases) lies in its ability to detect shifts in the frequency of the trans-
mitted acoustic pulse.  Frequency shifts are caused by the Doppler effect
and are directly proportional to the speed of an air parcel  moving away  from
(lower frequency) or towards (higher frequency) the transmitting antenna.
If the antenna is tilted away from  the vertical, simple trigonometry  can be
used to calculate the horizontal component of the motion of  a parcel  of  air.
If the return pulse is analyzed at  different times following pulse transmis-

                                     9-5
-------
si on, speeds can be assigned to different  heights  above  the  surface  based  on
trigonometry and the speed of sound.   Many pulses  can  be averaged  at each
height to get an average speed for a  time  interval  as  a  function of  height.
A second tilted antenna will produce  a second (orthogonal) component.  Vector
wind direction and speed can be calculated from the two  components at each
level.  Wind fluctuation statistics can then  be calculated from the  two com-
ponents at each level along with the  mean  values.
          Most monostatic Doppler SODAR systems (referred to henceforth in
this section simply as SODARs) include a third, vertically-pointing  antenna
that measures vertical motion (mean and standard deviation)  and also pro-
duces a time-height display of signal strength on  a facsimile chart.  Maximum
heights and averaging intervals are generally user-selectable and  typically
range up to 1500 meters and from 2 to 60 minutes.   The minimum SODAR wind
level is 30-50 meters.  Lower heights are  not possible because of  the time re-
quired for the diaphragm in the acoustic driver to come  to  rest and  for the
driver to be switched from the transmit to the receive mode.
          The three antennas generally are not pulsed  simultaneously.  If
they were, they would be listening to each other's return signals, and there-
fore they are pulsed sequentially.  Furthermore, since an antenna  must con-
tinue listening until it receives a return signal  from the maximum height,
setting the SODAR to higher heights reduces the effective sampling rate.
At 600 meters, the effective sampling rate for each antenna  is approximately
once every 13 seconds.
          A conclusion that can be drawn from the foregoing  discussion is
that the success of a SODAR hinges primarily on its ability  to extract a
peak frequency (single or double) from the return  signal, as well  as its

                                    9-6
-------
ability to transmit a pulse with a sharply defined and precisely  known  peak
frequency.  Figure 9-1 illustrates what an ideal  signal  return  might  look
like for one single-peak pulse.  Frequency is plotted on the abscissa,
amplitude on the ordinate.  The graphs represent  a "snap-shot"  of the return
spectrum at times following pulse transmission corresponding to 60 and  600
meters above the surface.  What is shown is a sharp peak in the spectrum,  a
high signal-to-noise ratio, and no other interfering peaks.  The  attenuation
of the return signal with height is also shown.
          Acceptability of the return pulse depends in part on  a  strong,
clear, concentrated transmit pulse.  The pulse is created by a  heavy-duty
acoustic driver that is mounted above the parabolic dish.  The  antenna  dish
focuses the pulse and gives it its direction and  inclination.  A  sound-dead-
ening enclosure for the dish is required to reduce side-lobe effects,
prevent ambient noise from reaching the microphone when  the driver is in
the receive mode, and to reduce the amount of nuisance created  by the
transmit pulse.
          Given a good transmit pulse, there are  still other sources  of
interference that can influence the quality of the data  extracted from
return spectra.  The unique nature of SODARs for  measuring meteorological
variables lies in the fact that a SODAR is a remote measurement device  that
probes the medium (the atmosphere) and actually measures only the response
to that probe.  Data quality therefore is related to the probe  itself and
to the fact that the nature of the probe (acoustic energy) is such that there
can be many sources of interference.  This can be contrasted to a wind  vane
which is located in the medium that it is measuring.  Sources of  a voltage
that could be interpreted as an erroneous wind direction signal from  a  wind

                                    9-7
-------
Amplitude  (Arbitrary Units)
uv
50
40
30
20
10
n

—
—
—
—
— /
	 -I--"""""
f»
\ 60 Meters
!
•
•
•
I "--t 	 ••"• I ""•""
           1540     1580     1620
                    Frequency  (Hz)

Amplitude  (Arbitrary Units)
  60   r-
                                      1660
  50

  40

  30

  20

  10
                               600  Meters
                      	••-•-••••.•••r--	-
           1540     1580      1620
                    Frequency (Hz)
                                      1660
    Figure 9-1.  Example SODAR Return Spectra
                        9-8
-------
vane are much fewer than potential sources of interference for a SODAR.
          A successful SODAR-based measurement program depends on maximizing
the occurrence of "ideal" spectra such as discussed above, minimizing the
number of times when data is lost due to high background noise (low signal-to-
noise ratios), minimizing the number of times when interfering signals are
interpreted as atmospheric returns (thereby producing erroneous data), and
validating the data to ensure that erroneous data do not enter the data  base.
The rest of this section presents guidance on how to develop an operational
plan to achieve these ends.  The operational plan addresses siting and ex-
exposure, operation and maintenance, quality control, quality assurance,
data validation, data management, and data use.
          It is important to again note that different SODAR manufacturers
have designed their systems with different techniques for producing transmit
pulses and for extracting the atmospheric signal  from return spectra. There-
fore, different systems have different means of discriminating acceptable
spectra.  The techniques described herein for maximizing valid data capture
will have a different emphasis based on the system chosen.  An operational
plan, including Standard Operating Procedures and a Quality Assurance Plan,
can therefore differ between systems.  The manufacturer may already have
developed most of the information required for the plan.  Nonetheless, each
of the aspects of this plan, as discussed in this document, should be ad-
dressed in some fashion and agreed to between applicant and regulatory
agency, prior to the start of data collection.

     9.2  Siting and Exposure
          The fundamental requirement of a return signal with a sharply
                                    9-9
-------
defined atmospheric peak frequency places  special  requirements  on  the
siting of a SODAR.  Siting criteria described  elsewhere  in  this document
should be followed in general;  in addition,  the  other  factors discussed
here that are unique to SODARs  must be assessed.
          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't
find a peak or because the signal-to-noise ratio is too  low.  The  net
effect is not to produce erroneous data but  to lower the effective sampling
rate due to the loss of many of the pulses.  The manufacturer should be
consulted as to what noise level would be  acceptable.  A qualitative survey
should be conducted to identify potential  noise  sources, and a  quantitative
noise survey may be necessary to determine if  noise levels  are  within  the
manufacturer's minimum requirements.
          Examples of active, broad-band noise sources include  highways,
industrial facilities or power  plants, and heavy machinery  operating near
the SODAR.  Some of these noise sources have a pronounced diurnal, weekly
or even seasonal pattern (farm machinery,  for  example).   The noise survey
should at least cover diurnal and weekly patterns.  Examination of land-use
patterns and other sources of information  may  have to  be relied on to  deter-
mine if any seasonal activities would be a problem.  A noise  survey will
not cover all bases, but a carefully designed  survey  should help decide  if
a site is suitable.
          Examples of active, fixed-frequency  noise sources include rotating
fans, the back-up beeper on a piece of heavy equipment,  and birds  and in-
                                    9-10
-------
sects.  If these noise sources have a frequency component in the SOOAR
operating range, they may be picked up as good data by the SODAR.  Some of
these sources can be identified during the site selection process.   Problems
can be avoided by taking precautions such as pointing the antennas  away from
the instrument shelter (where the sound of an operating air conditioner might
be picked up).  Wind blowing over the enclosures and rain impacting on  the
horn or enclosure also represent noise sources that may affect data capture.
          One approach to reducing the problem of fixed frequency,  narrow-
band 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 the ground or elevated
(such as tall towers, electric power transmission lines, buildings  and  trees)
that can reflect a transmitted pulse back to the antenna.  While most of the
acoustic energy is focused in a narrow beam, side-lobes do exist and are of
particular concern when antenna enclosures have degraded substantially.
Side-lobes reflecting off of 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 pre-
cisely which objects may be a problem.  Anything in the same general direc-
tion that the antenna is pointing and higher than 5 to 10 meters is a poten-
tial reflector.  It is therefore important to construct an "obstacle vista
diagram" prior to SODAR installation that identifies potential  reflectors
and their height as a function of direction from the antenna.  This diagram
can be used after some data have been collected to assess whether or not

                                     9-11
-------
reflections are of concern at some SODAR  height  ranges.   It  should  be  noted
that reflections from an object  at distance  X  from  an  antenna will  show  up
at a height X cos(theta) where theta is the  tilt angle of the antenna  from
the vertical.
          An approach to dealing with the problem caused  by  fixed echoes
is to utilize software that eliminates signal  returns  where  the  peak fre-
quency is the same as the transmit frequency.  This technique can also rec-
ognize a zero Doppler shift caused by antenna  "ringing",  where the  speaker
diaphragm, or driver mounting hardware continues to vibrate  after the  driver
has been switched to the receive mode. The  potential  for rejecting valid zero
Doppler shift returns would have to be addressed when  utilizing  this type
of software.
          The mobility of trailer-mounted SODARs allows  them to  be  set up and
operated in a temporary mode with very little  site  preparation.   For installa-
tions where a long-term data base is desired,  the SODAR  should be  installed
on a more permanent base such as a concrete  pad.
          The two horizontal 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 be  established  based on one of  the  techniques
described in the Quality Assurance Handbook  for  Air Pollution Measurement
Systems:  Volume IV, Meteorological Measurements.3   Orientation  of  the
SODAR antennas should be based on the axis of the parabolic  dish that
focuses the sound pulse.  Since the dishes are hidden from view  by  the
antenna enclosures, orientation  is commonly  accomplished with  reference to
the trailer or the enclosure sides.  This is acceptable  as a quick  check,
as long as the measurement that  is taken  on  the  trailer  or enclosure  side

                                    9-12
-------
is related to the measurement that is required (relative to the antenna
dish) on a periodic basis.
          Another siting concern that is unique to SODARs relates  to  the
fact that wind measurements are a composite of two independent  measurements
of air parcels separated in space.  For typical height ranges the  parcels
may be separated by several hundred meters, depending on the antenna  tilt
angle and the measuring height.  In complex terrain, the different parcels
may be in different flow regimes.  A topographic map should be  used to  "plot"
air parcels based on antenna geometry, and the location of the  parcels  rel-
ative to terrain should be evaluated.
          One last item that should be considered in a SOOAR siting decision
is the effect of the instrument on its surroundings.  The sound pulse is quite
audible and could create a disturbance if antennas are located  too close to
residences.

     9.3  Operation and Maintenance; Quality Control
          Detailed operation and maintenance (O&M) procedures are  specific to
each manufacturer's instrument.  This section discusses O&M procedures  in
general and recommends elements that should be addressed in any SODAR O&M
plan.
           When setting up a SODAR for operation in the field,  it  is  important
to consider several factors when selecting the averaging interval  and height
range.  Predicted plume heights of sources to be modeled is one factor.  The
effective sampling rate is another factor that should be considered (higher
heights result in fewer transmit pulses).  The height and averaging interval
settings should initially be fixed at some nominal  values,  such as 600 meters
                                    9-13
-------
and 15 minutes.  A different height can be specified,  but it is suggested
that 300 meters be the minimum height.
          The Quality Control (QC) function is closely related to operating
procedures which should provide for data review as well  as site visits.
The procedures developed for a specific instrument at  a  specific site
should be written up in a standard operating procedures  document (SOP)  that
can help ensure that all important aspects of SODAR operation are checked
at regular intervals, and that other procedures for data review and manage-
ment are being followed.  There are not many example SOPs available.  As
more SOPs are developed, a greater body of knowledge will be available  to
build on.  Manufacturers can also provide a great deal of information that
can be incorporated into a site-specific SOP.
          The purpose of an SOP is to spell out operating and QC procedures
with the ultimate goal of maximizing valid data capture.  The keys to a
successful SOOAR QC program, based on the experience of many users, are
(1) timely data review by an individual with meteorological expertise and
SODAR experience and (2) diligence in regular checking of all aspects of
SODAR operation under the direction of highly qualified  electronics personnel
          It is helpful here to recall the fundamentals  of reliable SODAR
operation; a clear, sharp transmit pulse with sharp frequency peak(s),  and
return spectra with low background noise and well-defined frequency peak(s)
due to atmospheric echoes.  Departures from this ideal can produce either
erroneous data or a severe loss of data.  Some departures from the ideal
will occur in any SODAR data base; a later section will  discuss refining
and validating that data base.  Timely data review and regular site checks
will serve both to identify and fix "fatal flaws", and to minimize to the

                                     9-14
-------
greatest degree possible, the amount of data that has to be "weeded out".
The type of system that is used also affects the degree to which data must
be validated.
          A "fatal flaw" can include an instrument failure which is the
most obvious problem to identify (i.e., no data are being produced).  Another
fatal flaw might be the complete or partial failure of one of the acoustic
drivers.  Data would still be collected if this occurred but with one
component missing.  If this was a horizontal component, the data would be
virtually useless.  Data capture from one antenna might degrade to the
point where it is almost entirely missing, if the diaphragm in that driver
is on the verge of failure or if snow and/or ice has built up to a significant
degree in the antenna dish (remember that the parabolic dish shapes and
focuses the transmit pulse - snow and ice build-up will distort the pulse).
An antenna dish heater is recommended to reduce this problem in locations
where frozen precipitation can occur.  Mechanical relays that switch drivers
from the transmit to receive mode can also fail causing a loss of data.
          Timely data review and regular site checks can also serve to iden-
tify "non-fatal" flaws.  Non-fatal  flaws generally are data anomalies that
would cause some levels of data to be invalidated but not enough to consider
the period "missing".  Echoes that occur intermittently should be noted.
Antenna ringing, caused by continued vibration of a component in the driver
or on the driver mounting hardware after the driver has been switched to the
receiver mode, will show up as zero's in the lower levels of the data.
Periods of data loss that are not otherwise explainable may help identify
noise sources not previously identified (farm machinery operating near the
site, for example).

                                    9-15
-------
          Some "non-fatal" flaws  can  be  fixed,  others  cannot.   Flaws  that
can't be fixed should be noted for the final  validation  process.   Problems
that are persistent should be tracked down,  although  sometimes  this  is  not
possible because the problem doesn't  occur when help  is  available  to  track
it down.  The main objective of the timely data review/regular  site  check
process is to keep the non-fatal  flaws from  becoming  fatal  flaws which
would translate into substantial  data loss.
          An SOP should be tailored to a particular instrument  at  a  partic-
ular site.  What follows is a description of major elements of  data  review
and site procedures that should be addressed in any SOP.

          Data Review
          0 Ideally the data should be spot-checked on a daily  basis  (this
            is generally possible only for sites with  a  remote  interrogation
            capability);
          0 A more complete data review  should  be conducted on  a weekly basis,
            The following types of data  reports have  been found to be useful:
            - component-specific reports that display  time-series  of the data
              profiles for each component (mean and standard deviation);
            - printouts that group many  averaging periods on the complete
              data set on one page;
            - hourly averaged data displayed in manner that will highlight
              diurnal patterns; and
            - summaries of raw frequency data analyses.
          0 On a monthly basis preliminary data capture summaries  should be
            prepared on a component-specific basis and for resultant data.
                                    9-16
-------
0 A tower (at a minimum of 10 meters)  should be installed  at  the
  SODAR site.  A tower would generally be required  to  provide sur-
  face data as input to stability determinations, but  can  also be
  valuable in the QC process.  A measurement system capable of
  providing u and v components at the  same time as  the SODAR
  data is preferred.  Some manufacturers  offer a 10-meter  tower
  as an integral part of their SODAR systems.  In complex  terrain,
  siting of the tower may be problematical and its  usefulness may
  be limited as a result.

Site Visits
0 Perform instrument diagnostics as specified/recommended  by  the
  manufacturer.
0 Obtain printouts of data collected during site visit and provide
  qualitative description of how well  actual site conditions  are
  reflected by the data.  NOTE: This could include  making  observa-
  tions of stack plume direction and amount of plume rise, compar-
  ison of SODAR data to tower data, etc.
0 Check operation of facsimile chart recorder; provide description
  of how well actual site conditions are  reflected  in  the  data -
  primarily cloud cover, time of day,  wind speeds.
0 Inspect all antennas for accumulation of snow (which may indicate
  faulty heater cables), and birds or  insects present  inside  enclo-
  sure.  Listen to several pulses from each antenna to verify that
  the driver is in good shape.
0 Collect raw frequency data, if done  as  part of the QC process.
                          9-17
-------
          0 Remove and replace  magnetic tape,  if being utilized.
          0 Site visits should  be  made frequently enough that data capture
            objectives can be met.  The frequency of visits may depend on
            how much information on SODAR operations can be obtained by
            remote interrogation.

     9.4  Quality Assurance
          Major elements of a SODAR Quality Assurance  (QA) plan are:  QC
procedures, periodic audits, and data validation.  QC  procedures are dis-
cussed in the previous section  in  the context  of an SOP.  Data validation
is discussed in the next section on data use,  and audits are discussed here.
It is quite important for all three elements to be present.  An audit by it-
self can ensure that the instrument is operating correctly at the time that
the audit is conducted.  Comprehensive QC procedures  (carried out through
site visits and data review) are necessary to  ensure that good data are
collected between audits, and data validation  is necessary to ensure that
anomalous data do not enter into a final data  base used for modeling.
          SODAR audits should be conducted when the system first begins on-
site operation and every six months thereafter, although some elements do
not have to be repeated at each audit.  Specific procedures will vary among
manufacturers, but the four main elements are  as follows:  site evaluation,
internal and external instrument checks, a system audit and a performance
audit.  These terms are somewhat loosely defined here; some overlap is
possible in the elements as stated.
          Site Evaluation:  The SODAR  site characteristics in terms of noise
potential, both active and passive,  should be  evaluated and documented  (refer
                                    9-18
-------
to the previous discussion on siting and installation).
          Internal and External Instrument Checks;   Some of these checks
should mirror the checks made on a routine basis, and some are quite specific
to each instrument.  Some of the checks that can be made are for electronic
noise, local oscillator frequency, ramp and amplifier gain circuits, and
automatic gain control circuits.  An effort should  be made to check the
circuits that control the transmit pulse frequency, particularly if that
frequency is adjusted from one period to the next.   Accurate transmit
frequency is directly related to data accuracy, since speed computations
are based on the frequency shift of the measured return  peak frequencies
where the transmit frequency has to be assumed.
          External checks should also be carried out and should also mirror
to some extent the routine checks.  Each antenna should  be examined, the
enclosure lining material checked, and the tilt and orientation measured.
Transmit pulses from each antenna should be listened to, to determine if
the acoustic drivers are functioning properly.
          Facsimile chart records, if collected, should  be examined to deter-
mine if conditions recorded on the charts reflect actual conditions for the
day.  Charts should be reviewed for some time period prior to the audit to
identify potential large periods of missing or invalid data.
          Acoustic pulses of known frequencies may  be used to determine if
the SODAR correctly detects and interprets frequency shifts in the return
signal.  This technique, known as static calibration, tests portions of the
SODAR's electronic circuitry, but does not test a system's ability to extract
a valid Doppler shift from a return signal that contains background noise
or to identify the presence of fixed echoes or electronic interference.

                                    9-19
-------
          System Audit:   This should include a  review of  data  handling
procedures and conformance to site inspection and  data review  procedures.
Since what happens in between audits is a critical  element  of  a  successful
SOUAR program, the audit itself provides a good opportunity to critically
review conformance to the data review and site  inspection requirements  of
the SOP.  As part of a system audit, data should be produced and reviewed
in the same manner as for the QC checks.
          Performance Audit:  The site evaluation, internal and  external
intrument checks, and system audit ensure that  the SODAR  is being operated
correctly.  A performance audit compares SODAR  wind measurements with an
independent measurement.  SODAR performance audits should consist of com-
paring data on a component-specific basis, as well  as comparing  resultant
speed and direction.  Any one of the following  approaches to testing SODAR
performance may be considered:
          1.  Use of a temporary measurement system such  as a  tethersonde  or
kite anemometer.  Data from this test should cover as many meteorological
conditions as possible.  A sample size of 120 15-minute samples  would gener-
ally be considered adequate.  The independent measuring technique should  be
used to collect data for a full averaging period at one height,  rather than
measuring at several heights during the period.  Samples  should  be taken  at
several heights during the course of the audit.
          2.  Use of a fixed tower measuring data at an elevation corres-
ponding to an elevation measured by the SODAR.   A tower that utilizes terrain
to achieve part of the elevation may be acceptable in some situations (refer
to Section 3.2 for a discussion of this issue).  Since a tower provides a
continuous measurement, the data produced can actually serve two purposes.

                                    9-20
-------
First, the data can be used in the performance audit by comparing  SODAR  to
tower measurements for a period of time corresponding to the  audit (nominally
one week of continuous data), and also for the period of time since the
last audit.  Second, the data can provide a valuable input  to the  U.C process,
as a continuous check on SODAR performance.
          3.  Use of a second SODAR operating at a different  transmit fre-
quency.  Not many tests of this type have been carried out.  The advantages
include being able to provide comparisons of complete profiles and being able
to provide comparisons continuously for the period of the test. A nominal
testing period of one week of continuous data is suggested  as a minimum.
          The following factors should be considered when conducting a per-
formance audit:
          0 Good comparisons between SODAR and tethersonde/kite anemometer
            systems give confidence that both systems are working  well.   Bad
            comparisons, on the other hand, do not necessarily mean that the
            SODAR is faulty, rather, it could mean that the alternate measure-
            ment technique is faulty or that the difference in measurement
            techniques simply produce different values for  the conditions
            measured.  The usefulness of such a test is therefore  limited by
            the potential to produce results that are not meaningful.
          0 Tethersondes and kite anemometers are limited to  daytime use.
            For applications where nighttime, stable conditions are important,
            a performance test such as this is not useful for determining
            whether these conditions are adequately measured.
          0 The continuous one-level comparison provided by the 10m tower
            can provide a means of continuous comparison with an independent

                                    9-21
-------
            measurement.   It  is  important  to  understand that the tower  is
            not measuring the same  thing as the  first acoustic  level, and
            therefore cannot  replace the performance audit.  However, evalu-
            ating the complete profile  on  both a resultant and  component-
            specific basis can contribute  to  an  assessment of the  accuracy
            of the acoustic portion of  the data. This assessment  is  parti-
            cularly useful when  evaluating profiles measured in well-mixed,
            neutral atmospheric  conditions.   Severe terrain in  the immediate
            vicinity of the SODAR site  will limit the usefulness of this
            comparison.
          A performance audit should be performed at the  beginning of a SODAR
measurement program, and  at least annually thereafter.  As stated  above, other
portions of the audit should  be  conducted  at  six month intervals.

     9.5  Data Validation, Data  Management and Data Use

          9.5.1  Data Validation
                 A carefully  sited, well-maintained SODAR will  produce  high
quality data most of the  time.  Since the  SODAR  can occasionally misinterpret
interfering signals and assign "valid"  codes  to  the resulting data, validation
is an important step in developing a modeling data base.  The degree to which
validation and post-processing is necessary depends partially on the site  but
also on the system being used -  some SODARs are  more  selective  than others  in
                                                                          *
accepting return pulses,  and  some SODARs  are  being  introduced with built-in
validation software.
                 Section  9.1  describes  the types of anomalous data that can
occur.  Final validation  should  not occur  until  after at  least  one complete

                                     9-22
-------
audit has been conducted, although "fatal  flaws" (which would invalidate an
entire data period) should be removed from the data base shortly  after they
are discovered.
                 It is not possible to provide specific guidance  on SOOAR
data validation procedures at this time.  The following are suggested steps
that would need to be enhanced (and could be modified)  for a particular
system and a specific application.
                 1.  Data should be reviewed by a meteorologist familiar
with SODAR operation soon after they are collected, on  at least a weekly
basis.  Fatal flaws should be identified and removed.
                 2.  A screening program should be developed that produces
flags for each level on each antenna.  The flags could  be assigned based
on the amount of shear between levels, the value of the radial  standard
deviation, and other values that characterize anomalous data (refer to
Section 8.6).  The flags should be numeric (possibly 0-9) with  values as-
signed on a sliding scale.  For example, a value of 1 might be  assigned to
a difference between 2 levels of 2 meters/second, a value of 9  to a differ-
ence of 10 m/s.  Likewise, a value of 1 might be assigned to a  standard
deviation of 1.5, a value of 9 assigned to a standard deviation of 3.0.
Since perfect data may be equally suspect, a value of 9 might be  assigned
to a standard deviation of 0.0.
                 3. When the data with flags are reviewed (again  by a meteor-
ologist familiar with SODAR operation) the flags may be manually  changed if
the reviewer feels that the screening flags are inappropriate.  This addi-
tional review is important, since the reviewer can rely on an assessment
of the entire profile - something which is difficult to accomplish with a

                                    9-23
-------
computer program.   It is also important to thoroughly  document  the  changes
and the rationale  for the changes,  such that  an  independent  reviewer  can
distinguish between manual  and automatic flags.
                 4.  A final  data base should be created  by  automated means,
based on a test of the flags.  The  entire data base should be examined to
determine what level should be accepted - a value of 2 or less  might  be
accepted, for example, while a value of 3 or  greater rejected.
                 5.  Reserving final data validation until a full year of
data has been collected will  allow  statistical  and climatological summaries
of the data to be  prepared and further data checks to  be  made against other
data sources (e.g., nearby NWS upper-air stations or nearby  towers).   This
additional information can help in  the validation process by providing a  ref-
erence against which individual data points can  be evaluated (for example,
a profile initially thought to be an anomaly  may occur several  times  and  be
traced to a real meteorological phenomenon).

          9.b.2  Data Management
                 A SODAR produces a prodigious amount  of  information.  If
set at 600 meters, 15 minute averages, 30 meter increments with one tower
level, several variables are produced and recorded at twenty levels.   It
is important to plan for managing these data  prior to the start of  the
measurement program.  The data management scheme should accommodate the
following:
                 1.  Initial  checks to ensure that the data  have been trans-
ferred correctly (i.e., that magnetic tapes can be read or data sets  trans-
ferred by phone link are intact);
                                    9-24
-------
                 2.  Quick data turn-around in a format that can be reviewed
to identify fatal flaws and instrument problems that can be fixed.   This is
not a trivial task, and should include the following (as input to the QC
procedures):
                     a.  Reports that summarize profile data from each
antenna on one line for each time period;
                     b.  Reports that present a significant portion of the
data from each time period (to cut down on the amount of paper produced,
several time periods can be placed on one page);
                     c.  Reports that present hourly averages in a  format
where diurnal patterns can be examined; and
                     d.  Reports that summarize raw frequency data  analyses.
                 3.  A provision for editing the data if errors occur or as
a result of the data validation process.  All editing functions should be
carefully controlled and documented; and
                 4.  Methods for archiving the data.

          9.5.3  Data Use
                 Several types of data are produced by a SODAR; furthermore,
data availability can vary with height as a function of atmospheric condi-
tions (the existence of suitable "targets") and ambient noise (more noise,
less data).  Three important questions that will be addressed in this
section are:  1) which data types can be used in regulatory modeling; 2)
what level(s) are appropriate to use in a dispersion model, and how are
they to be used; and 3) how should data availability be defined (and what
percentage of data capture is required).
                                    9-25
-------
                 9.b.3.1  Data Types
                          Mean Wind Values:   Wind  speed  and  wind  direction
values are reported for many heights.   Based  in  part  on  the  results  of  the
BAO intercomparison results, the mean  values  are appropriate for  use in
regulatory modeling if the SODAR system is  subject to an approved QA plan
and the data are validated prior to use. Treatment of low wind speeds  is  an
important consideration since the SODAR produces a vector-averaged speed.
Mean vertical  wind speed, a variable that  is  also  reported by SODAR  systems,
is not yet used in regulatory modeling although  the reported values  may pro-
vide some meteorological insights.
                          Wind Fluctuation  Values: Most SODAR systems
report the standard deviations of horizontal  wind  direction  (o/\)  and of
vertical wind speed (ay).  Values of o/\ from  SODAR are usually much  larger
than values recorded by a wind vane, although the  overestimation  appears to
lessen with higher wind speeds.  A fundamental  problem is that SODAR winds
are composed of samples taken from different  volumes  of  air  at different times.
Wind direction fluctuations cannot be  calculated directly, and the estimation
techniques tend to over-estimate the amount of fluctuation.
                          As a result  of these concerns, o/\  data  from SODARs
are not being recommended for modeling use  at this time.  Some work  has
been done to develop corrections to SODAR  a^  data.  '    Furthermore, some
manufacturers are exploring ways of designing the  system to  avoid the funda-
mental problem (e.g., using a configuration that that points to monostatic
antennas at the same volume of air, pulsed  at the  same time  but at a different
frequency so that the signals do not  interfere with each other).
                          The BAO results  indicate that  cty values do not

                                    9-26
-------
        with tower measurements as well  as  wind speed  or  direction,  although
daytime (convective) values show better  agreement  than nighttime  (stable)
values.  In order to relate o^ to diffusion,  a transformation  to  oj:  (stan-
dard deviation of elevation angle fluctuations) must be made by dividing  oy
by wind speed (see Section 6.4.1).  Since SODAR wind speed  is  a vector  aver-
age, overprediction of OE is likely to occur  under low wind speed condi-
tions.  Use of ON data from SODARs is also  not recommended  for regulatory
modeling at this time.
                          An obvious point  to make is  that  no  model  currently
in Appendix A of the Guideline on Air Quality Models  (Revised)4 is capable
of utilizing direct turbulence measurements.   The  purpose of including  this
discussion is that this guidance is also intended  for  applications where
nonguideline models are being evaluated  and there  may  be  some  nonguideline
models that can utilize the turbulence data.   Furthermore,  models under devel-
opment by EPA that utilize turbulence data  may eventually be included in
the guideline.  This discussion is not meant  to categorically  deny the  use
of turbulence data from a SODAR.  If an  applicant  wishes  to use the  data,
it is up to the applicant to overcome the concerns expressed here.  Further
improvements in processing techniques, correction  factors,  or  improvements
in equipment may make SODAR turbulence data acceptable for  regulatory
modeling.
                          Mixing heights:  The facsimile  chart produced by
a SODAR can be analyzed to estimate mixing  heights.  Mixing heights  estimated
in this manner are not recommended for routine modeling use, primarily
because of height limitations.  A typical convective boundary  layer  appears
on the facsimile chart as a series of spikes  ("thermal plumes").   Occasionally

                                    9-27
-------
a limiting stable layer can be observed  by  a  skilled  analyst that can  properly
be interpreted as a limit to the vertical extent  of mixing.  More commonly,
the elevated stable layer is not strong  enough  to produce  an unambiguous
trace or is out of range of the instrument  (facsimile charts are generally
set at 500 or 1000 meters).  In this case the top of  the visible thermal
plumes does not necessarily indicate the vertical  extent of mixing,  just
that the atmospheric targets are not strong enough to produce  a visible
trace at that height.  (It should be noted  that the dynamic response charac-
teristics of the facsimile chart recorder are different from the part  of  the
SODAR that interprets frequency shifts.  Therefore wind data can be  derived
at heights well above the end of the visible  trace on the  chart recorder.)
                          If mixing heights are thus  underestimated, their
use in a model may lead to under- or over-predictions. This is because most
EPA models employ the assumptions that ground-level concentrations are zero
when a plume is above the mixing height, and  that complete reflection  of
the plume occurs if the plume is below the  mixing height.
                          As in the case of turbulence values, an  applicant
has the opportunity to use SODAR mixing heights if the concern expressed
here is overcome.  Use of the Holzworth interpolation11  scheme with  some  of
the facsimile information may have some promise.   Manufacturers have recently
begun to offer automatic mixing height detection  routines. These  routines
should be examined carefully prior to approving their use.
                          SOOAR facsimile charts  can, on  the other hand,
provide valuable information on the condition of  the  atmosphere.   Although
translating that information into data usable in  a regulatory  context  is
problematical, the information could be used  in a diagnostic sense when

                                    9-28
-------
conducting a model evaluation study.  Users are encouraged to develop schemes
for using the data, although it should also be noted that facsimile charts
are not easy to handle.

                 9.5.3.2  Levels for model input
                          Wind speed and direction data from many levels
are available from a SODAR, and data are generally available well  above the
100m level that is considered a practical limit for tower heights.  A scheme
for utilizing SODAR data for regulatory model input is recommended below.
Other schemes may be approved on a case-by-case basis.
                          1.  Wind data at stack top or at plume height may be
used as input to regulatory models.  Wind speed is generally used for plume
rise and dilution calculations, and wind direction is used to determine plume
transport direction.  Selecting a single measurement height representative  of
average plume height under critical meteorological conditions is acceptable.
                          2.  A SODAR measurement is derived from signal  re-
turns from a layer of the atmosphere, rather than a single level.   The speed
or direction values at one level are essentially averages across the layer.
If the elevation of the measurement height selected for model input (stack
top or plume height) is close to the elevation of the center of a SODAR
range gate, then the data from that level should be used.  If the height
selected for model input is close to the upper or lower end of a range gate,
then the speed and direction data should be interpolated between the two
adjacent range gates.
                          3.  If data are not available at the height selected
for model  input but the data period is considered valid as defined below in
                                    9-29
-------
Section 9.5.3.3, substitutions should be handled as  follows.   The wind speed
at model input height may be determined from a  logarithmic  profile based  on
available data from at least three levels.   Wind direction  from  the closest
level with valid data may be substituted, as long as that level  is at  least
at 100m.  If the data are not available for these substitutions, or if the
averaging period is not considered valid, refer to Section  6.5.3 for guidance
on treatment of missing data.
                          4.  An upper bound should  be established for selec-
tion of a measurement height for model input.   This  is because data capture
becomes more erratic at greater heights, and also because  return signals  are
more saturated with noise at greater heights and erroneous  data  are more
likely to occur.  It is recommended that the cut-off level  for model input be
the highest height with data capture of at  least 80%.  See  Section 9.5.3.3
below for a more complete discussion of data capture requirements.

                 9.5.3.3  Data capture requirements  and definition
                          Data capture for a SODAR data base  must be defined
somewhat differently than for more conventional instruments.   Data capture
for SODARs is a strong function of height.   A valid  data period  should not
be defined in terms of a specific height because of  the possibility that  data
at that height might be invalidated due to intermittent echoes.   The following
definitions and requirements should apply to SODAR data bases:
                          1.  A SODAR averaging period will be considered
valid if there are at least three complete (both components), valid levels
for the period (independent of height).  "Valid level" refers to data that
have gone through final validation.
                                     9-30
-------
                          2.  An hour will  be considered  valid  if  at  least

30 minutes are valid (i.e., 2 out of 4 15-minute periods);  and

                          3.  Valid SODAR data as defined in (1) and  (2)

should be available at least 90% of the time on an annual  basis.


     9.6  Recommendations

          Doppler SODARs can be used to provide mean wind speed and direc-
tion at heights not readily achievable by towers, and in  some cases mixing
heights, for on-site meteorological measurement programs.   The  turbulence
data available from most SOOAR systems are currently not  recommended  for
routine use.

          A proposal to utilize Doppler SODAR in an on-site program should
be closely coordinated with the reviewing agency.  An overall operational
plan, including Quality Assurance procedures, should be prepared prior to
data use and preferably prior to the start of data collection.  The details
of the operational plan will change with the specific instrument manufac-
turer.  The following topics and recommendations should be addressed  in the
operational plan.  The text of previous sections contains more  detailed
discussion on these topics.

          Siting and Installation

          0  Noise survey:  qualitative followed by quantitative if necessary
          0  Identification of potential reflection targets
          0  Disturbance potential
          0  Analysis of flow regime being measured
          0  Initial alignment

          Operation and Maintenance; Quality Control (QC)
          0  Many aspects of O&M specific to manufacturer
          0  Initial  settings of 15 minutes for averaging period and at  least
               300m for height.
          0  Collocated tower (at a minimum of 10 meters)
          0  Standard Operating Procedures:

               Timely and thorough data review:  daily,  weekly,  monthly
                 procedures
               Regular instrument checks (frequency  based on  degree  of remote
                 interrogation available)

          Quality Assurance Plan

          0  Major elements are QC procedures, periodic  audits,  and  data
               validation
                                    9-31
-------
0  Audits should be conducted at six month  intervals  and should
     include:

     Site elevation
     Internal and external  instrument checks
     System audit
     Performance audit:   when instrument  is placed  in service
       and at least annually thereafter

Data Validation

0  Should be carried out, on a component-specific basis, prior to
     using data in a model  for regulatory purposes
0  Procedures should be  manufacturer-specific

Data Management

0  Prodigious amount of  information necessitates careful planning
0  Management plan should incorporate timely  review and  archiving
     of data

Data Use

0  Wind speed and direction recommended for use
0  Wind speed and direction at stack top  or at  plume  height for
      model input
0  An upper bound should be established,  where  data capture is at
     least 80%, for developing model inputs
0  Mixing height may be  acceptable on a case-by-case  basis

Data Capture Requirements

0  Valid hours must be available 90% of the time
0  Valid hour defined as at least three complete valid levels for
     30 minutes out of an hour (two 15-minute values)
                          9-32
-------
 10.0  REFERENCES

 1.   Environmental  Protection Agency,  1987.   Ambient  Monitoring  Guidelines
     for Prevention of Significant Deterioration (PSD).   EPA-450/4-87-007.
     U. S.  Environmental  Protection Agency,  Research  Trianyle Park,  NC.

 2-   Federal  Register  45:52676-52748.   August 7,  1980.

 3.   Finkelstein, P. L.,  D. L. Mazzarella,  T. J. Lockhart, W. J. King and
     J. H.  White, 1983.  Quality Assurance  Handbook for  Air Pollution
     Measurement Systems, Volume IV:  Meteorological  Measurements.   EPA-
     600/4-82-060.   U. S. Environmental  Protection Agency, Research  Triangle
     Park,  NC.

 4.   Environmental  Protection Agency,  1986.   Guideline on Air Quality Models
     (Revised).  EPA-450/2-78-027R.  U.  S.  Environmental  Protection  Agency,
     Research Triangle Park, NC.

 5.   Strimaitis, D., G. Hoffnagle and  A. Bass, 1981.   On-Site Meteorological
     Instrumentation Requirements to Characterize Diffusion from Point
     Sources-Workshop Report.  EPA-600/9-81-020.  U.  S.  Environmental Protec-
     tion Agency, Research Triangle Park, NC.

 6.   American Nuclear Society, 1984.  Standard for Determining Meteorological
     Information at Nuclear Power Sites.  ANSI/ANS-2.5-1984.   American Nuclear
     Society, La Grange Park, IL.

 7.   Mason, C.  J. and H.  Moses, 1984.   Meteorological Instrumentation.  In:
     Atmospheric Science  and Power Production, D.  Randerson (ed.).   DOE/TIC-
     27601.U. S.  Department of Energy, Oak Ridge, TN.

 8.   Great  Britain  Meteorological Office, 1956.   Handbook of  Meteorological
     Instruments, Part I.  Her Majesty's Stationery Office, London,  England.

 9.   Middleton, W.  E. K.  and A. F. Spilhaus, 1953.  Meteorological  Instruments,
     3rd ed.   University  of Toronto Press,  Toronto, Canada.

10.   Wang,  J. Y. and C. M. M. Felton,  1983.   Instruments  for  Physical Environ-
     menta1 Meas urement s, 2nd ed.  Kendall/Hunt  Publishing Company,  Dubuque, IA.

11.   Holzworth, G.  C. 1972.  Mixing Heights, Wind Speeds, and Potential  for
     Urban  Air  Pollution  Throughout the  Contiguous United States.  AP-101.
     U.S. Environmental Protection Agency,  Research Triangle  Park, NC.

12.   Environmental  Proteciton Agency,  1977.   User's Manual for Single-Source
     (CRSTER) Model.  EPA-450/2-77-013.   U.S. Environmental Protection
     Agency,  Research Triangle Park, NC.

13.   World  Meteorological Organization,  1971.  Guide  to  Meteorological Instru-
     ments  and  Observing  Practices. WMO No. 8TP3, 4th ed., Secretariat  of
     WMO, Geneva, Switzerland.
                                     10-1
-------
14.   Environmental  Protection Agency,  1985.  Guideline  for Determination  of
     Good Engineering Practice Stack Height  (Technical  Support Document for
     the Stack Height Regulations)  - Revised.   EPA 450/4-80-023R.  U. S.
     Environmental  Protection Agency,  Research  Triangle Park, NC.

15.   Hanna, S. R.,  G. A.  Briggs,  and R.  P. Hosker, 1982.  Handbook on Atmospheric
     Diffusion.  DOE/TIC-11223.   U. S. Department of  Energy, Oak Ridge, TN.

16.   U. S. Department of  Commerce,  1972.   National Weather Service Observing
     Handbook No.2, Substation Observations.  Environmental Science  Services,
     Office of Meteorological  Operations,  Silver Springs, MD.

17.   Raynor, G. S,  P. Michael, and  S.  SethuRaman, 1979.  Recommendations
     for Meteorological Measurement Programs and Atmospheric Diffusion Pre-
     diction Methods for  Use at Coastal  Nuclear Reactor Sites.  NUREG/CR-0936.
     U.S. Nuclear Regulatory Commisison,  Washington,  DC.

18.   Brooks, C. E.  P. and N. Carruthers,  1953.  Handbook of Statistical Methods
     in Meteorology, M.0.538.   Her  Majesty's Stationery Office, London, England.

19.   Mazzarella, D. A., 1972.  An Inventory  of  Specifications for Wind
     Measuring Instruments.  Bull.  Amer.  Meteor. Soc.,  Vol. 53, Mo.  9.
     pp.860-871, American Meteorological  Society, Boston, MA.

20.   Gill, G. C., 1967.   On the Dynamic  Response of Meteorological Sensors
     and Recorders.  Proceedings of the  First Canadian  Conference on Micro-
     meteorology, Part I, Meteorological  Service of Canada, Toronto, Canada.

21.   American Society for Testing and  Materials, 1985.  Standard Method for
     Measuring Surface Wind by Means of  Wind Vanes and  Rotating Anemometers.
     ASTM D4480-85, American Society  for Testing and  Materials, Philadelphia,
     PA.

22.   Mazzarella, D. A., 1978.  Meteorological  Instruments for Use Near the
     Ground: Their Selection and Use in  Air  Pollution Studies.  In:  Air
     Quality Meteorology  and Atmospheric Ozone, Morris  and Barris  (eds.),
     American Society for Testing and  Materials, Philadelphia, PA.

23.   Snow, J. T., 1985.  Meeting of ASTM Subcommittee D-22.11:  Meteorology.
     Bull. Amer. Meteoro. Soc., Vol. 66,  No.  11, p.1432, American Meteorological
     Society, Boston, MA.

24.   Mardia, K. V., 1972.  Statistics  of Directional  Data.  Academic Press,
     New  York, NY.

25.   Turner, D. B., 1986.  Comparison  of Three  Methods  for Calculating the
     Standard Deviation of the Wind Direction.  J. Climate Appl.Meteor..
     Vol. 25, pp 703-707.
                                    10-2
-------
26.  Yamartino, R. J., 1984.  A Comparison of Several  "Single-pass" Estimators
     of the Standard Deviation of Wind Direction.  J.  Climate Appl. Meteor.,
     Vol. 23. pp.1362-1366.

27.  Irwin, J.S., 1979.   A Theoretical Variation of the Wind Profile Power-
     Law Exponent as a Function of Surface Roughness and Stability.  Atmos.
     Env..  Vol. 13. pp.  191-194.

28.  Paumier, J., D. Stinson, T. Kelly, C. Bellinger,  ana J. S. Irwin, 1986.
     MPDA-1:  A Meteorological Processor for Diffusion Analysis - User's
     Guide.  EPA 600/8-86-011.  U. S. Environmental Protection Agency,
     Research Triangle Park, NC.

29.  Weber, A., J. S. Irwin, J. P. Kahler, and W. B. Petersen, 1975.
     Atmospheric Turbulence Properties in the Lowest 300 Meters.
     EPA 600/4-75-004.  Environmental Protection Agency, Research Triangle,
     Park,  NC.

30.  Deihl, B. J., 1984.  Vertical Wind Direction Standard Deviation (o^);
     Variation With Height and Comparison of Measurement Systems.  Public
     Service Company of  New Mexico, Albequerque, NM«

31.  Wieringa, J., 1976.  An Objective Exposure Correction Method for Average
     Wind Speeds Measured at a Sheltered Location.  Quart. J. Roy. Meteorol.
     Soc..  Vol. 102. pp.241-253.

32.  Wieringa, J., 1980:  Representativeness of Wind Observations at Airports.
     Bull.  Amer. Meteor. Soc., Vol. 61. pp.962-971.

33.  Irwin, J. S., T. M. Asbury, and W. B. Petersen, 1986.  Description of the
     Savannah River Laboratory Meteorological Data Base for 1975 to 1979.  U. S.
     Environmental Protection Agency, Research Triangle Park, NC.

34.  Nieuwstadt, R. T. M., 1978.  The Computation of the Friction Velocity u*
     arid the Temperature Scale T* from Temperature and Wind Velocity Profiles
     by Least-Square Methods.  Bound.-Layer Meteorol., Vol. 14, pp.235-246.

35.  Irwin, J. S. and F. S. Binkowski, 1981.  Estimation of the Monin-Obukhov
     Scaling Length using On-Site Instrumentation.  Atmos. Environ., Vol. 15,
     pp.1091-1094.  (Erratum, 1982, Atmos. Environ., Jj>, "887).

36.  Holtslag, A. A. M., 1984.  Estimates of Diabatic  Wind Speed Profiles
     From Near Surface Weather Observations.  Bound-Layer Meteor., Vol. 29,
     pp.225-250.                                ~'

37.  Pasquill, F., 1961.  The Estimation of the Dispersion of Windborne
     Material, Meteorol. Mag., Vol_. 90, pp.33-49.

38.  Turner, D. B., 1964.  A Diffusion Model for an Urban Area.  J. Appl.
     Meteor., Vol. 3, pp.83-91.
                                    10-3
-------
39.  List, R. J. 1966.   Smithsonian  Meteorological Tables, Sixth Revised
     Edition (Third Reprint).Smithsonian  Institution, Washington, D. C.

40.  Environmental  Protection  Agency,  1984.   Calms Processor  (CALMPRO) User's
     Guide.  EPA-9U1/9-84-001.  U. S.  Environmental  Protection Agency, Region  I,
     Boston, MA.

41.  Nappo, C. J. et al, 1982.  The  Workshop on  the  Representativeness of
     Meteorological Observations,  June 1981, Boulder,  CO.  Bull. Amer. Meteor.
     Soc., Vol. 63, No.  7, pp.761-764.  American Meteorological Society,
     Boston, MA.

42.  Lockhart, T. J., 1985.  Quality Assurance of Meteorological Measurements.
     In:  Quality Assurance for Environmental Measurements, ASTM STP  867.
     J. K. Taylor and T. W. Stanley  (eds).American Society  for Testing and
     Materials, Philadelphia,  PA.

43.  Weiss, L. L., 1961.  Relative Catches  of Snow in Shielded  and Unshielded
     Gauges at Different Wind  Speeds.   Monthly Weather^ Review, Vol. 89.

44.  Environmental Protection  Agency,  1976.   Quality Assurance  Handbook for
     Air Pollution Measurement Systems:  Volume  I -  Principles.  EPA  600/9-76-
     005.  U. S. Environmental Protection Agency, Research Triangle Park,  NC.

45.  Kaimal, J. C., J. E. Gaynor,  P. L. Finkelstein, M.  E. Graves, and
     T. J. Lockhart, 1984.  A  Field  Comparison of In-Situ Meteorological Sensors,
     BAO Report No. 6, NOAA/ERL Wave Propagation Laboratory,  Boulder, CO.

46.  Kaimal, J. C., J. E. Gaynor,  and  W. M.  Baynton  (Ed.), 1980.   The Boulder
     Low-Level  Intercomparison Experiment -- Preprint of WMO  Report,  BAO Report
     No. 2.  NOAA/ERL Wave Propagation Laboratory, Boulder, CO.

47.  Coulter, R. L., and T. J. Martin, 1987. Measurement of  High  Resolution
     Wind Profiles with a Minisodar.  Sixth Symposium on Meteorological
     Observations and Instrumentation  - Extended Abstracts, New Orleans, LA,
     January 12-16, 1987.  American  Meteorological Society, Boston, MA.

48.  Underwood, K. H., 1981.  SODAR  Signal  Processing Methods and  the Riso-78
     Experiment.  PhD Thesis,  Penn State University, State College, PA.

49.  Tombach,  I., R. Baxter, and R.  Drake, 1983. Automatic Determination  of
     Atmospheric Mixing Depth  and Inversion Heights:  Phase  I.  AV-FR-83/536.
     Aerovironment Corporation, Pasadena, CA.

50.  Kristensen, L. and J. E.  Gaynor,  1986.  Errors  in Second Moments Estimated
     from Monostatic Doppler Sodar Winds.  Part  I:   Theoretical Description.
     J. Atmos.  Oceanic Tech.,  Vol. 3,  pp.523-528.

51.  Gaynor, J. E. and L. Kristensen,  1986.  Errors  in Second Moments Estimated
     from Monostatic Doppler Sodar Winds.  Part  II:   Application  to Field
     Measurements.  J. Atmos.  Oceanic Tech., Vol. 3, pp.529-534.
                                     10-4
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing}
1. REPORT NO.
EPA-45G/4-87-013
4. TITLE AND SUBTITLE
On-site Meteorological
Guidance for Regulatory
Modeling Applications
r
Program
7. AUTHOR(S)
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Office of Air Quality Planning and Standards
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
12. SPONSORING AGENCY NAME AND ADDRESS


3. RECIPIENT'S ACCESSION NO.
5. REPORT DATE
June 1987
6. PERFORMING ORGANIZATION CODE
8. PERFORMING ORGANIZATION REPORT NO.
10. PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO.
13. TYPE OF REPORT AND PERIOD COVERED
14. SPONSORING AGENCY CODE
15. SUPPLEMENTARY NOTES
16. ABSTRACT
 This document provides  EPA's guidance on the collection and use of on-site meteor-
 ological data for regulatory modeling applications.  It will form the basis  for  the
 regulatory review of proposed meteorological monitoring plans by the EPA Regional
 Offices and States.  The  document contains comprehensive and detailed guidance for
 on-site meteorological  measurement programs, covering initial design and siting
 of a system, through data recording and processing, up to air quality model  input.
17. KEY WORDS AND DOCUMENT ANALYSIS
a. DESCRIPTORS
Air Pollution
Atmospheric Models
Meteorological Instrumentation
Meteorological Monitoring
Meteorology
18. DISTRIBUTION STATEMENT
Unlimited
b. IDENTIFIERS/OPEN ENDED TERMS

19. SECURITY CLASS (This Report)
Non
-------
r*«
-------