United States
Environmental Protection
Agency
Atmospheric Sciences
Research Laboratory
Research Triangle Park NC 2771 1
Research and Development
EPA-600/S3-84-111 Jan 1985
Project Summary
An  Evaluation  of Wind
Measurements  by
Four  Doppler  Sodars

J C. Kaimal, J.E. Gaynor, P L. Fmkelstem, M.E. Graves, and T.J. Lockhart
  Measurements by four Doppler sodars
of wind speed, wind direction, and the
vertical component of turbulence were
compared  to  similar  measurements
made on the 300-m instrumented
tower at the  National Oceanic and
Atmospheric Administration's Boulder
Atmospheric Observatory. The sodars
were manufactured and were operated
during the test by Aerovironment Inc.,
Radian Inc., Remtech et Cie (formerly
Berlin), and Xontech Inc. Sodar measure-
ments were compared to measurements
made by fast  response sonic anemo-
meters at 100, 200, and 300  m. The
comparison ran continuously for 21 d in
September, 1982. Results  of  the
experiment indicated that all  sodars
measured wind speed and direction
quite accurately and with  reasonably
high precision. Comparison  of  the
measurements of the vertical compo-
nent of  turbulence indicated that the
sodars tended to overestimate  the
standard deviation of vertical velocity
at night and to underestimate it during
strongly convective situations. The
precision of measurement for  the
vertical component of turbulence was
considerably poorer than for average
wind speed and direction. Comparison
also indicated some differences among
the various types of sodars. Analysis of
the  spectra measured by the  sodars
indicated that measurement inaccurac-
ies may have been due to a combination
of sampling volume  and aliasing prob-
lems.
  This Project Summary was developed
by EPA's A tmospheric Sciences Research
Laboratory. Research Triangle Park.
NC, to announce key findings of the
research project that is fully documented
in a separate report of the same title (see
Project Report ordering information at
back),

Introduction
  During the first three weeks of Septem-
ber, 1982, an experiment was conducted
at the Boulder Atmospheric Observatory
(BAO),  under the sponsorship of the
Environmental Protection Agency (EPA),
to assess the ability of tn situ and remote
sensors to measure the mean and
turbulent properties of the lower atmo-
sphere The experiment was conducted in
response to the  need for comparative
data from which scientists could evaluate
the accuracy, field precision, and general
performance of  some of the more
commonly  used meteorological instru-
ments that measure atmospheric turbu-
lence
  Recent advances in the modeling of
transport  and diffusion of pollutants,
achieved  largely through theoretical
insights gained from field experiments,
point to the site-specific nature  of
turbulence  Attention is, therefore, being
directed to better on-site characterization
of turbulence and to the development of
techniques for measuring the mean and
turbulent wind variables needed for input
into the models  This experiment was
designed to provide information needed
to formulate a monitoring strategy for the
development of site-specific dispersion
meteorology
  The BAO was chosen as the site for the
experiment because of the availability of
precise profile and turbulence data  from
sensors on  a 300-m tower Facilities for
launching rawmsondes and for processing

-------
the data received from them were added
benefits  Four  commercially-available
Doppler sodars with the capability to
measure variances in vertical wind
component in addition to the mean three-
dimensional  wind field  were used The
question addressed here is whether the
sodars can measure  the  mean and
turbulent properties of the flow at heights
100 to 300 m above the ground

Description  of Instrumentation
  Four Dopplersodar manufacturers who
currently  market their  products m  the
United States  were invited by EPA to
participate  m  the  experiment  Under
arrangements  made through EPA's
principal contractor, Meteorology Research
Inc , the sodars were installed and
operated by personnel from the participat-
mg firms  The four systems differ
significantly in their physical configuration
and approach to signal processing Two of
the systems used three-axis monostatic
arrays, one used a bistatic array and the
other used a colocated monostatic/bistatic
array  The following four systems were
deployed

AeroVironment Three-Axis
Monostatic  System (A V)
  The system consisted of three acoustic
transceivers  mounted on  a  trailer
Doppler shifts m  the backscattered
signals  received  on each  axis were
interpreted as  wind components m the
radial directions Wind components thus
measured were transformed into compon-
ents along  the N-S,  E-W, and vertical
directions Because m this configuration
sampling volumes are separated by large
distances, the  assumption of horizontal
homogeneity m the mean wind field is
essential to justify  the  use of wind
components measured along the different
axes for the coordinate transformation
  The AV  system transmitted a sound
pulse (1 50-200 W) at a frequency of 1500
Hz  (duration 0-18  s) sequentially from
each  of three  adjacent pencil-beam
antennas One tilts south 30° from the
vertical to be sensitive  to  the N-S
component,  one tilts  west 30° from
vertical to be sensitive to the  E-W
component, and one points straight up to
be  sensitive  to  the vertical  component
The receiver echo  is heterodyned and
then passed through an electronic comb
filter  with  31   teeth to yield  the  full
spectral distribution m the return signal
For each 33 3-m altitude range gate, the
spectrum is  examined  according to
several criteria to obtain a best estimate
of Doppler shift, along with an estimated
reliability factor The pulse repetition
interval was 8 s

Remtech Three-Axis
Monostatic System (REM)
  The REM system (developed originally
at Bertm  et Cie) also uses a  trailer-
mounted array of  three transceivers
They  are operated  m sequence as
monostatic systems, the same assumption
of horizontal homogeneity is invoked for
wind measurements In this system, the
horizontal wind  sensing  antennas  are
tilted 18° from the vertical in the same
directions as are the AV's, the transmitted
pulse is  a 1600-Hz signal of  0 08-s
duration  The  received  signals are
digitized after  appropriate bandpass
filtering, and the  Doppler frequency shift
is extracted using Fast FounerTransforms
(FFT) techniques  The  pulse repetition
interval was 5 s

Radian  Corporation Colocated
Monostatic/Bistatic System
(RAD)
  The RAD's antenna  configuration
permitted  both monostatic and bistatic
operation  Both systems shared the
central, vertically pointing, pencil-beam
transceiver  The two  tilted (18° from
vertical) monostatic transceivers were
not located close to the vertical transceiver
as on the AV and REM systems,  but
were aimed to intersect at a height of
150-m  In the  bistatic  mode, two fan-
beam transmitters (located 250 m to the
south and to the west) illuminated the
vertical beam of the central transceiver
The movement of the sound pulse up the
vertical beam was followed by time gating
of the receiver signal. Doppler frequency
shifts  in  each gated  segment were
converted to wind velocity components to
produce a wind profile
  In  both configurations, the  vertical
transceiver was operated in the monostatic
mode to measure the vertical wind
component Because the three monostatic
beams were not divergent, the assumption
of homogeneity is not as critical here The
RAD system transmitted 120-W pulses at
20 kHz (0 1-s duration) and computed
Doppler shifts using the Complex Covanance
method  RAD operated in three  modes
monostatic, bistatic, and multistatic(alter-
natmg one series of monostatic and one
series of bistatic pulses) Pulse repetition
interval for all systems was 5 s

Xontech Three-Axis Bistatic
System (XON)
  This  bistatic  system consisted of a
vertical pencil-beam transceiver and two
fan-beam receivers aimed at a central
vertical common volume Therefore, the
geometry m this system was exactly the
reverse of the RAD bistatic system  The
transmitted frequency was 2 OkHz(0 08-
and  016-s duration under computer
control)   Its bistatic baseline was 350-m
long  The fan-beam  antennas  received
signals  scattered from  the  vertical
transceiver beam, so the winds were
computed along a vertical column above
the transceiver as in the  RAD bistatic
system A microcomputer determined the
Doppler  frequency shift with an  FFT
detection scheme  powerful  enough to
sense small frequency variations in the
presence of high ambient noise levels
The wind data are, therefore, presented
without   qualifiers, but if  the  program
could detect a consistent signal for the
entire averaging  period, no  data were
printed for the height range The pulse
repetition interval was 5 s

BAO Instrumented Tower
  The 300-m tower at the  BAO is
instrumented at 8 levels 10, 22, 50, 100,
150,  200, 250,  and  300  m  Sonic
anemometers installed at each  of these
levels measured the  three-dimensional
wind  field  R M  Young propeller-vane
anemometers are mounted on the side of
the tower opposite from  the  sonic
anemometers to serve as back  up wind
sensors when the tower is shadowing the
sonic anemometers For this experiment,
the sonic anemometers were mounted
on the booms  pointing SSW  and  the
propeller-vane  anemometers  were
mounted  on booms pointing NNE. These
booms also supported sensors for meas-
uring mean and fluctuating air tempera-
tures and the dewpoint temperature. Data
from the  sonic anemometers and other
fast-response sensors were sampled ten
times per second, while the propeller-
vane  anemometers, like  other  slow-
response sensors, were sampled only
once per  second
  Data from the BAO are recorded in one
of two modes  In the "regular" mode, only
the 10-s  averaged data points and 10-s
grab samples (last point m a 10-s data
block) of  the time series are retained In
the "raw  data" mode, all data points are
recorded  In both modes, the software
computes and lists once every 20 mm the
means, variances, and fluxes for  the
preceedmg 20-min period These listings
become  the  common  reference  for
comparing the  performance  of  the
different  sodars. The raw data  mode is
employed only when the full time series is
needed for special analyses or for the

-------
details in the structure of the flow Such
recordings were made for three relatively
brief periods during this experiment for
the purpose  of comparing the spectral
response of the sodars and examining the
limitations imposed by sampling volumes
and sampling rates


Procedures
  The sodar measurements were centered
over an area 05x03 km, about 0 65 km
southwest of the BAO tower  The bistatic
arrays were laid out to provide a height
range  of  at  least 300  m,  hence  the
requirement for such  a  large test area
The  electronic equipment  associated
with the Doppler systems were housed in
trailers located within the visitors area A
larger trailer in the same area served as
the control center for the experiment
  The terrain in the vicinity of the tower,
is reasonably flat Except for the trailers
and the fence  surrounding  the  visitors
area,   the site is free  of small-scale
surface obstructions
  Procedures for data  collecting and
reporting were established to insure
against  unfair  bias for  any  of  the
participants. All systems were assumed
to be  capable of unattended continuous
operation  All systems provided  data in
the form of wind speeds, wind directions,
vertical wind components, and standard
deviations of vertical wind speed averaged
over 20-min periods coincident with the
BAO averaging periods  The three compari-
son levels were 100, 200, and 300 m The
data collected  over the previous 24-h
period were submitted to EPA personnel
directing  the experiment every morning
at 0800-h MST in exchange for tower
data covering the same period
  Concurrent operation  of some of the
sodars was  considered at one  time, but
quickly ruled  out because of  cross-
contamination,  even  between systems
operating  at  different frequencies The
sodars were, therefore, operated m
sequence, with the switchover from one
system to  another controlled by a central
timer switch  The assigned observing
interval was normally one 20-mm period
each cycle The experiment covered the
period  from September 1 to 21, 1982
  The AV, REM, and RAD computed the
standard deviation of w (the vertical wind
component) from their  time  series
Missing data points were not filled in by
interpolation, but the  number  of points
missed (or accepted) was displayed The
REM used 4-pomt block averages  instead
of the original time  series.  The XON
computed  its standard  deviation from the
width of its  2-mm w spectra, estimated
for each level Successive 2-mm standard
deviations were averaged to obtain the 20-
min values Each spectrum was automati-
cally  examined  for level and  shape of
background  noise and steps were taken
to remove their effects
  No  attempt  is made m this report to
present  the results of our  standard
deviation of wind direction comparisons
The azimuth direction standard deviations
showed very large scatter  The data are
withheld pending a better understanding
of the reasons for the scatter Meanwhile,
we can only suggest caution in using
standard deviation of wind  direction for
diffusion predictions
  The AV, REM, and XON maintained a
consistent operating pattern throughout
the experiment  However, RAD changed
its operating mode every 24 h, switching
from multistage to bistatic and monostatic,
then back to multistatic, and so on

Results

Measurement of the Standard
Deviation of  Vertical Wind
Velocity
  Because the sonic calculations of the
standard deviation  of vertical wind speed
provide  reference  values,  the  accuracy
and precision of each sodar system can
be determined from the collection of 20-
mm average differences  The two input
variables for these computations  are the
standard deviation  of vertical wind
velocity calculated from sodar  measure-
ments and the standard deviation of the
sonic vertical wind velocity  The compara-
tive statistics used to estimate accuracy
and precision then become the average
difference (sample bias) and the standard
deviation of the differences  In addition,
the root mean square difference, or com-
parability,  is computed,  this statistic
characterizes  the repeatability of  a
system  Finally, the precision is also
represented as a  percentage of  the
average value, a coefficient of variation
  Values for these statistics are presented
in Table 1 for the combined  sodar
observations at  each of three heights, as
well  as  for the individual  vendor data
subsets The sample bias showed a large
range of values around nearly constant
composite values of 0 08 At 100  m the
spread  was  greatest,  with  the  REM
having  the only negative  sample bias
value (i e ,  sodar < sonic) and the XON
having a sizable 0 23 m/s  sample bias
The AV was  well  below  the  composite
sample bias value, the RAD was above it
At 200 m  the  REM value  was slightly
negative,  the AV  remained small,  the
XON was  the  same as the  composite
value, but the RAD was in excess  of 0 2
m/s At 300 m the RAD  sample bias
continued to be relatively  large, but the
other vendors  were grouped  between
003 and 007 m/s
  From Table 1 it  is clear that there is
much scatter m the comparability  and
percentage of the  average value  about
the true value in all systems There was
no statistical  difference  between  the
standard deviation of the differences and
Table 1.    Standard Deviation of Sodar Vertical Wind Speed Compared to the Standard Deviation
          of Sonic Vertical Wind Speed

                                                     s3         s4       AT
                                                   jVn/s;        (%)

                                                    0 22
                                                    0 16
                                                    0 21
                                                    0 17
                                                    0 24

                                                    0 26
                                                    020
                                                    0 32
                                                    0 19
                                                    0 24

                                                    026
                                                    0 25
                                                    030
                                                    0 22
                                                    0 18

'b = sample bias (sodar measurement-sonic measurement) Estimates accuracy
2c - comparability
3s - standard deviation of differences Estimates precision
4s' - s expressed as a percentage of the average value of the sonic standard deviation
5N = number of observations
Height
(m)
WO




200




300




Vendor
Composite
AV
Ft AD
REM
XON
Composite
AV
RAD
REM
XON
Composite
AV
RAD
REM
XON
(m/s)
008
001
0 12
-005
0 23
008
003
022
-000
008
009
004
023
007
004
(m/s)
0 24
0 16
0 25
0 18
034
0 27
020
039
0 19
0 25
0 27
025
038
0 23
0 19
50
35
47
38
53
54
43
65
39
51
54
53
62
47
38
678
190
178
139
171
576
167
144
119
146
665
214
158
136
157

-------
the percentage of the average value for
the AV and the REM at any of the levels

Measurement of Wind Speed
  To examine the accuracy and precision
of the sodars, simultaneous observations
of wind speed were recorded from sonic
anemometers at the same three heights
on a tower about 600 m from the sodar
systems  The sonic systems had  a
sampling rate of 10 Hz, and they were
regarded as the reference instruments m
the evaluation However, due to a wind
shadow zone created  by  the tower,
extending  +40° from north for the sonic
instruments, reference data m this sector
were obtained by the propeller vane at the
BAO tower  A comparison of sonic and
propeller wind speed measurements on
the tower  showed that the instruments
were approximately equivalent
  Values of  sample bias, comparability,
standard  deviation for  the differences
between sodar and reference values are
presented  m Table 2 for combined sodar
observations at each height as well asfor
the sodar record of each vendor Propeller
wind speeds were excluded when the wind
speed was less than 1 m/s
  The estimates of sample bias m Table 2
show mostly negative values at 100 m,
with a composite value  near -04 m/s
Because the difference was sodar minus
reference,  this means  that the sodar
systems tended to register too low  An
exception  was the RAD, which did not
have a significant sample bias at 100 m
  At 200 m, the vendors all recorded too
high At 300 m the RAD and XON again
recorded too high, whereas, the AV and
REM were unbiased  Sample biases were
also computed for day (0600-1800 h) and
night (1800-0600 h) values Most differences
between day and night were insignificant
  The comparability of sodar wind speeds
with reference  values  is also  given m
Table 2 Precision  is  represented by
standard deviation and percentage
deviations  The percentage deviation
values ranged from  about 15% to 35%
around composite values near 25%


Measurement of Wind
Direction
  An investigation of the propeller-vane
data indicated that  they could not be
substituted as reference  values  There
were  unexplained disparities between
the propeller and sonic data that are still
under investigation  Because sonic data
showed more consistent behavior and we
believe them to be more reliable, only the
sonic  wind direction measurements are
used as reference data
  Values of sample  bias, comparability,
and standard deviation for the differences
between  sodar and sonic  reference
values  are presented  m Table  3 for
combined sodar observations at each
height as well as for the sodar record for
each vendor
  The values of bias in  Table 3 show
negative values at 100  m and 200 m for
all vendors, but positive values at  300 m
for all vendors except for RAD  However,
most of these values were not significantly
different  from zero, considering the
Table 2    Sodar Wind Speed Compared With Reference Wind Speed
Height
(m)
100




200




300




Vendor
Composite
AV
RAD
REM
XON
Composite
AV
RAD
REM
XON
Composite
AV
RAD
REM
XON
b'
(m/s)
-042
050
002
-012
-1 04
0 14
005
031
0 12
009
0 16
-0 10
029
002
0 44
c?
(m/s)
1 28
1 03
1 18
062
1 88
098
072
1 00
0 73
071
1 24
1 15
1 17
0 74
1 20
s3
(m/s)
1 21
090
1 18
060
1 56
096
0 72
1 47
072
0 70
1 23
1 15
1 69
0 74
1 12
s4
(%)
28
21
28
14
37
23
17
35
17
17
27
25
37
17
25
/V-"
1179
327
315
236
301
1019
298
258
194
269
1005
328
198
183
296
variability of the wind direction and the
number of cases included
  The  comparability  of  sodar wind
direction with sonic reference values is
also  given  in Table 3  Considering the
scatter in data, it can be assumed thatthe
vendors' measurements of wind directions
were equivalent to each other
  The Project Report also gives informa-
tion on the sodar performance above 300
m It also  compares vertical velocity
spectra determined  from  the  sodar
measurements and  from  the  sonic
anemometer measurements

Conclusions and
Recommendations
  The wind measurements analyzed  in
this report represent the state-of-the-art
in Doppler sodar wind sensing Consider-
ing the requirements that the  sodar
operation   be  unattended,  except  for
maintenance and repair,  and that the
data  be subjected to no editing  by the
vendors, the results obtained were
reassuringly good The scatter inthe wind
speed and direction data compared very
well  with  scatter in past experiments
when some of the same sodar systems
were  compared  under more  controlled
conditions  In these data, some vendors
showed more scatter than others, but
much of that could be attributed to factors
such as ram, high winds, cable damage,
and  transducer failures  If data from
these suspected periods are  eliminated,
the scatter for  the different systems
would appear more  similar
  The measurement  of standard deviation
of vertical  wind velocity  with sodars
seems to  show promise, at least for
daytime conditions  Here too, vendor
performance showed variations that
could be attributed  in some cases  to
weather and equipment failure, but
individual differences  in processing the
data  might  also  have been  a factor here
The predictable behavior of sodar vertical
velocity spectra m the convective boun-
dary  layer leads one to believe that verti-
cal velocity variance measurements can
be made to within 10% of heights above
100  m  However, the nighttime results
are not so encouraging  More work  is
needed  to ascertain the reasons for the
large discrepancies in the vertical velocity
measurements at night
'b = sample bias (sodar measurement-sonic measurement) Estimates accuracy
2c - comparability
3s = standard deviation of differences Estimates precis/on
4s' = s expressed as a percentage of the average value of the sonic standard deviation
5/V = number of observations

-------
Table 3.   Sodar Wind Direction Compared With Sonic Wind Direction
Height
fm)
WO




200




300




Vendor
Composite
AV
RAD
FtEM
XON
Composite
AV
RAD
REM
XON
Composite
AV
RAD
REM
XON
b'
(deg)
-441
-387
-676
-203
-449
-343
-079
-786
-389
-1 85
075
026
-3 25
062
405
c2
(deg)
2859
26 70
2706
1851
3785
2322
1947
25 67
2467
2380
29 59
2856
2998
19 70
3596
s?
(deg)
2825
2641
26 20
1840
37 58
2297
19 45
2443
2436
23 73
2958
28 56
29 80
19 69
3573
N"
667
187
177
137
166
523
155
128
no
130
697
227
131
142
197
'b = Sample bias (sodar measurement-sonic measurement) Estimates accuracy
2c = comparability
3s = standard deviation of differences Estimates precision
"N - number of observations
   J. C. Kaimal and J. E. Gay nor are with the National Oceanic and Atmospheric
     Administration's (NOAA) Wave Propagation Laboratory, Boulder, CO 80303,
     the EPA author P. L. Finkelstein (also the EPA Project Officer, see below) is on
     assignment to  the Atmospheric  Sciences Research  Laboratory, Research
     Triangle Park, NC 27711 from NOAA; M. E. Graves is with Northrop Services,
     Inc.,  Research Triangle Park, NC 27709, and T. J.  Lockhart (formerly with
     Meteorological Research, Inc.) is with Meteorological Standards Institute, Fox
     Island, WA  98333.
   The complete report, entitled "An Evaluation of Wind Measurements by Four
     Doppler Sodars," (Order No PB 85-115 301, Cost $13 00, subject to change)
     will be available only from:
           National Technical Information Service
           5285 Port Royal Road
           Springfield, VA 22161
           Telephone: 703-487-4650
   The EPA Project Officer can be contacted at:
           Atmospheric Sciences Research Laboratory
           U.S. Environmental Protection Agency
           Research Triangle Park, NC 27711
                                                                                      . S. GOVERNMENT PRINTING OFFICE-1985/559-111/10780

-------
United States
Environmental Protection
Agency
Center for Environmental Research
Information
Cincinnati OH 45268
     BULK RATE
POSTAGE & FEES PAID
         EPA
   PERMIT No  G-35
Official Business
Penalty for Private Use $300

-------