United States
 Environmental Protection
Atmospheric Sciences
Research Laboratory
Research Triangle Park NC 27711
                                 II \
 Research and Development
 EPA/600/S3-86/040 Nov. 1986
 Project  Summary
 Project  CONDORS—
 Convective Diffusion
 Observed by Remote  Sensors
 J. C. Kaimal, W. L Eberhard, W. M. Moninger, J. E. Gaynor, S. W. Troxel,
 T. Uttal, G. A. Briggs, and G. E. Start
  This data report presents results from
 two diffusion experiments conducted
 at the Boulder Atmospheric  Observa-
 tory (BAO) of the National Oceanic and
 Atmospheric Administration (NOAA) in
 1982 and 1983. The objective was to
 compare diffusion in the atmospheric
 convective boundary layer with that
 observed in laboratory tank experiments
 and numerical computer models. In
 both  experiments at the BAO, two
 different tracers, oil fog and aluminized
 chaff, were released simultaneously and
 tracked by lidar and radar, respectively,
 for periods up to two hours.  In 1982,
 both tracers were released  from the
 same surface or elevated  point; in
 1983, the two were also released from
 separate levels, the oil fog from near the
 surface, the chaff from an  elevated
 point on the tower. The 1983 experi-
 ment included tracer gas releases with
 in situ samplers  measuring surface
 concentrations downwind of the tower.
 The BAO tower provided data on the
 mean and turbulent state of the atmos-
 phere, while mixing depths were moni-
 tored by balloon soundings, sodar, lidar,
 and radar. A detailed description of the
 experiment and the measurements ob-
 tained from the different sensors is
 provided. The strengths and limitations
 of the experiment are discussed in the
 context of case studies.
  This Project Summary was developed
 by EPA's Atmospheric  Sciences Re-
search Laboratory. Research Triangle
Park, NC, to announce key findings of
the research project that is fully docu-
mented in a separate report of the same
title (see Project Report ordering in-
formation at back).
  Project CONDORS (Convective Diffu-
sion Observed by Remote Sensors) was
undertaken to provide field data on dif-
fusion  in three  dimensions at short
ranges, 0.1 to 3.5 km, in convective
conditions. These conditions prevail over
land during sunny to partly cloudy days
with  light  to moderate wind  speeds,
especially when surface heating is mod-
erate to strong.  This volume is a data
report on Project  CONDORS,  later to be
followed by scientific analyses of the data
published in peer-reviewed journals.
  The need for  such an experiment
became evident after unexpected results
for vertical diffusion were obtained from
laboratory experiments  in  a convective
water tank by Willis and Deardorff (1976a,
1976b,  1978, 1981) and from computer
modeling experiments  by  Lamb (1978,
1979). The concentration (X) patterns in
downwind distance (x) and in height (z)
were substantially different from those
resulting from conventional  Gaussian
diffusion models. For releases near z=0.5
Zi, where z\ is the mixing depth, maximum
surface X was found to be about twice
that predicted by Gaussian models. (Mix-
ing depth  refers to a layer of vigorous
mixing  due to convective turbulence
produced by surface heating; this layer is
always capped by a layer of stable air,
which often is marked by a temperature
inversion just above zi.)  For elevated
releases, the centerline of  maximum
averaged X was observed to descend with
distance from the source height, nearly to
the surface; thereafter,  the plume be-
haved much like a surface release. For
surface releases, the maximum averaged
X remained near the surface up to a time

typifying the flow of air from downdrafts
to updrafts  (thermals); then  it rapidly
lifted into the upper half of the mixing
  The above-mentioned laboratory and
computer modeling experiments, as well
as turbulence measurements in the mix-
ing layer at a variety of sites, also showed
the importance of the length scale z, and
the velocity  scale w* to turbulence and
diffusion  in  convective boundary layers
(CBLs)  (w* = (H*Zi)1/3,  where  H* is the
vertical flux of  buoyant accelerations
produced by sensible heating and water
vapor flux near the surface; for surfaces
not very moist, H* = (g/T)w'T; where g is
gravitational acceleration, T is absolute
temperature, and w'T' is the  vertical
turbulent flux of temperature variations).
Average diffusion patterns in CBLs scale
best with z, and the time scale z,/w*. The
time after release, t, can be estimated as
x/u (the mean wind speed, u, is almost
constant with  z through most CBLs,
because  of  vigorous  vertical mixing).
Diffusion  results from various experi-
ments  in CBLs tend to agree well when
expressed in terms of dimensionless time
or distance,  X  = (x/u)w*/z,  = tw*/z,.
Unfortunately, very few  past diffusion
experiments in  the field included  suffi-
cient  meteorological  measurements  to
determine w* or z*; furthermore, they are
limited to measurements of X at the
surface or on towers no higher than 62 m.
Measurements of X are needed up to z —
zi, which is usually of the order of 1000m.
  To verify the new laboratory results for
convective diffusion, particularly the non-
Gaussian vertical behaviors, data from a
field study were needed. The CONDORS
field experiment was designed to go
beyond the limitations of past diffusion
experiments in  two ways. First, a large
number of  high-quality meteorological
measurements were made so that essen-
tial quantities like w*, z,, and u could be
determined  with accuracy and redun-
dancy. Many less essential  measure-
ments werealso collected; e.g., unusually
detailed  information  on wind direction
statistics. Second, remote sensors were
used to define  mean  X fields in  three
dimensions through depths up to 2000m,
easily encompassing zi. Two independent
tracer-sensor systems were used, lidar to
detect  oil fog  and Doppler  radar for
detection  of metalicized  "chaff."  Also,
limited conventional  gaseous  tracer
measurements were made at the surface;
these served primarily to test the  infer-
ences made about conservative-sourceX
on the basis of observed distributions  of
oil fog and chaff, which are not conserva-
tive (oil fog droplets tend to vaporize and
chaff tends to settle out).
Operational Plan
  The CONDORS experiment was carried
out in 1982 and 1983 at NOAA's Boulder
Atmospheric Observatory (BAO), operated
by the Wave  Propagation Laboratory
(WPL), under an interagency agreement
between  EPA and  WPL.  In 1983,  an
additional interagency agreement be-
tween EPA and NOAA's Air Resources
Laboratory Field Office (ARLFO) provided
for sampling  and analysis of gaseous
tracer releases. WPL provided the site,
most of the personnel, meteorological
measurements, radar  and lidar meas-
urements, and followup data processing.
  The 1982 experiments  consisted of
four runs of two hour's duration carried
out on separate days in September. The
four runs were intended primarily as trial
runs to test the adequacy and the limita-
tions of the two  remote sensor tech-
niques. The oil fog generator and chaff
dispenser were  located within a  few
meters of each other  to permit direct
comparisons of the observed X distribu-
tions. The runs were split evenly between
elevated releases  and  surface releases.
These experiments were successful
enough to provide six averaging  periods
with usable chaff and/or oil fog distribu-
tions plus adequate meteorological data.
  The 1983 experiments included surface
sampling of tracer gases and consisted of
eight runs  of two  hour's duration  on
separate  days in  late  August to  mid-
September. Based on conclusions follow-
ing some analyses of the 1982  runs, it
was decided to use the chaff only in the
elevated release mode in 1983. Two more
runs with collocated elevated releases
were made; during the remaining  six
runs, the oil fog was released from the
surface  so that independent  measure-
ments of  simultaneous elevated  and
surface releases could be  made. These
eight runs provided 11 averaging periods
with relatively steady meteorological
variables and good chaff and/or oil fog
  Preparation for a typical run began with
a check of daily National Weather Service
forecast maps. A forecast including clear
to partly cloudy skies, Colorado under the
influence of a surface high, and light to
moderate geostropic winds was  consid-
ered favorable for a run; all field personnel
were notified  of  the  weather outlook.
Near sunrise on promising days, a raw-
insonde was released  at the BAO and
tracked  by double  theodolite to about
3000 m AGL  The data were quickly
processed into temperature,  humidity,
wind speed and direction profiles and
transmitted to the main WPL laboratory in
Boulder. This sounding, supplemented by
the 1200Z sounding at Stapleton Airport
in Denver, was the main basis for deciding
whether to send personnel to their sta-
tions in the field, about 25 km east of
Boulder. In 1983, the sounding data were
input into a numerical model for z,  de-
velopment; this facilitated the decision
making, especially the choice of optimum
times to initiate runs (slow  z, growth
during runs was desired).
  At the CONDORS site, z, development
was monitored using tower profiles and
turbulence measurements up to 300 m,
acoustic sounder records up to about 600
m, lidar reports of haze heights, and radar
reports of heights  of natural  reflectors
(thought to be insects). When z, developed
into the desired range and the winds on
the tower entered into the desired speed
range (2 to 6 m/s) and direction sector
(NE to SE), a run initiation time (RIT) was
called. The elevated source height was
chosen on the basis of predicted z, for the
middle of the run, attempting to set this
height at about 1 /4 or 1 /2 z, to duplicate
the  laboratory and computer   modeling
runs. About 15 minutes were required for
the release height to be reached by the
carriage  on the 300-m  tower.  Tracer
releases were begun 10 to 20 minutes
prior to RITto allow plume transport out to
the distances of greatest interest (about
ZiJ/w*). The lidar  crew was advised of
preferred  lidar azimuths, which  also
depended on the expected value of the
length scale z,u/w*. Sampler, lidar, and
radar acquisition began at RIT and con-
tinued for two hours, or slightly longer if
wind velocity remained in the acceptable
range (during both  years at least one run
was aborted due to sudden wind direction

  The BAO site was chosen primarily for
the excellent  meteorological   measure-
ments available on the 300-m tower and
the convenience of a  release platform
that can quickly be elevated up to 300 m
(the carriage attached to the west side of
the tower). Another  convenient factor
was the close proximity of WPL's person-
nel and facilities, including the Doppler
radar, the lidar, acoustic sounders, data
loggers, and computers. The site has both
advantages and disadvantages for  this
type of  experiment. The terrain is gently
rolling,  neither "ideal" nor complex, and

 is somewhat typical of developed land
 surface. Convective conditions predom-
 inate during spring, summer, and fall
 days; late summer was chosen for the
 experiment because z, is likely to be in the
 desired range, and cloudy or rainy spells
 are less  likely than  in the fall. During
 convective conditions, the midday wind
 direction is almost always from the NE, E,
 or SE. This  is thought to be caused by
 upslope flow on the heated eastern slopes
 of the Front Range of the Rocky Moun-
 tains, 25 to 50 km west of the BAG. The
 prevalence of winds from one direction
 sector made it easier to effectively site the
 lidar, the radar,  and samplers during
 CONDORS. On the other hand, the easter-
 ly flow in the mixing layer was almost
 always opposed to the upper winds; as z\
 grows and entrains upper air, momentum
 from above mixes downward; the opposi-
 tion  of upper and  mixing layer  wind
 directions at this site probably causes
 more variability in wind speed and direc-
 tion than is common in CBLs farther from
 large mountain ranges.
  Siting of the sensors and samplers was
 also guided  by results of the laboratory
 convective tank  experiments and com-
 puter simulations. The distance range of
 most-needed X measurements was as-
 sociated with travel times of about 0.3 to
 1.2 z,/w*. As uzj/w* was typically 1 to 2.5
 km during CONDORS runs, the  most
 desired distances  of plume measure-
 ments ranged from 0.3 to 3 km downwind
 of the source. The  radar is  capable of
 detecting chaff at much greater distances
 than this, and the reflected signal is not
 attenuated by travel back through the
 plume. Consequently, the radar was sited
 upwind of the BAO tower, 3.5 km to the
 east, to reduce the azimuth and elevation
 angles needed to encompass the plume
 and  to reduce the  dynamic range of
 returns due to 1/r2 attenuation.  While
 the radar scans horizontally, the lidar
 scans in vertical planes, and the signal is
 attenuated appreciably by travel through
 the plume. Therefore, it was necessary to
 site the lidar as close to perpendicular to
 the plume axis as possible, at a distance
 minimalizing the elevation angle range
 needed but  without too much  signal
 attenuation. The lidar was sited 3 km
 from the BAO tower at 325° azimuth in
 1982 and was moved to 4 km away at
 346° azimuth in 1983, in anticipation of
 more frequent  southeasterly winds
 (which proved to be the case).
  Only one sampling arc could be oper-
ated  within  the  available  budget. The
sampling arc consisted of 29 samplers
 placed every 5° of azimuth from 202.5° to
 342.5°, mostly along roads. This arrange-
 ment provided adequate angular resolu-
 tion for the time-averaged plumes, which
 were 30° to 90° wide at the arc distance.
 The midpoint of the arc was 1.2 km to the
 west of the tower, a good  distance for
 intercepting nearly maximum surface X
 from the elevated release in almost every

 Tracers and Sensors
  The primary mapping of plume concen-
 tration fields was done by the two remote
 sensors, radar mapping chaff and lidar
 mapping oil fog.  These techniques were
 supplemented by conventional sampling
 and gas  chromatography  analysis of
 plumes containing conservative  tracer
 gases. Each of these three techniques
 has some areas of advantage as well as
 some substantial limitations.
  Lidar is a well-proven tool for  mapping
 atmospheric  aerosol fields and light-
 reflecting plumes. The WPL lidar uses a
 frequency-doubled Nd:YAG  laser trans-
 mitter firing at 10 pulses/s. It  is easily
 able  to map  atmospheric  haze cross
 sections several km deep; this capability
 was valuable for  making quick estimates
 of z, during CONDORS runs. Oil fog was
 used as a tracer because of the economics
 of producing enough plume particles to
 make the  plume distinguishable from
 atmospheric  haze after dilution with
 roughly 107 m3/s of air.  In the com-
 mercially built oil fogger used  in CON-
 DORS, oil is sprayed into a heated jet of
 air, causing it to vaporize. On mixing with
 ambient air, it cools and condenses  into
 drops a few microns  in diameter  which
 are very efficient reflectors of laser light.
 In 1982, a pale paraffin type oil was used,
 which was not expected to  significantly
 evaporate as it travelled downwind; how-
 ever,  the  decline of integrated return
 signal from larger x scans suggested
 otherwise.  The  maximum  x of plume
 distinguishable from  ambient haze was
 only 1.9 km,  short of the experimental
 goal. To improve this range to 2.9  km, in
 1983 a heavier oil was used; this change
 required modification of the fogger to
 preheat the oil to reduce its viscosity. In
 addition, a second oil fogger was operated
 during surface releases; however, it could
 not be accommodated  on the  tower
 carriage for elevated releases.

  The  energy  of each  pulse  and  the
digitized photomultiplier  detector signal
were recorded for later processing. Just
prior to each run, about six azimuths were
chosen  for lidar  scanning,  which was
controlled by computer. Each scan at an
azimuth  contained  about 100 pulses,
beginning at a maximum elevation angle
that fully encompassed the plume. The
scans had to be terminated a few tens of
meters above  the surface because the
laser beam was not eye-safe. The scan at
each azimuth was repeated approximately
every 210 seconds throughout the runs,
so that ensemble-averaged plume Xcould
be obtained. The spatial resolution of the
lidar was of the order of a few meters,
which was more than adequate.
  Post-experiment lildar signal  process-
ing for CONDORS achieved new levels of
quantification of plume returns in terms
of  concentration fields.  This  required
considerable computer processing time
and man-machine interaction,  as each
averaged scan  had  to be corrected for
attenuation of the signal by travel through
both background haze and the plume
itself. Because the  oil fog droplet size
distribution was unknown, only the rela-
tive concentration field at each azimuth of
scan could be determined. Consequently,
to estimate X/Qfor a conservative tracer,
where Q is the release rate, Q was
replaced by the measured downwind flux
of relative X(u times JXdydz, where y is
the  lateral dimension).  This expedient
assumes that the droplet size distribution
and the percentage evaporation is con-
stant across each scan section.
  In recent years, radar has been used to
map clear  airflow structures  using  a
tracer called "chaff," aluminum-coated
mylar threads. During CONDORS, bun-
dles of these threads were unwound from
reels, chopped into 1.6-cm lengths, and
ejected by an air jet at the rate of 38,000
filaments/s by a "chaff cutter." The
majority of  the filaments clump  together
and quickly fall to the ground, but enough
single filaments are produced to provide
very reflective targets for WPL's 3.2-cm
wavelength Doppler radar.  At typical
CONDORS  ranges, the radar could easily
detect one filament in a (50 m)3 volume of
air. Acquisition of signal was limited to
x — 3.5 km because of the experimental
focus and high data processing costs. The
chaff volume was too small to be seen by
the lidar, and oil fog droplets are extremely
poor reflectors of X-band electromagnetic
waves, so the two remote sensor/tracer
systems were quite independent.
  Chaff used as a tracer has one signifi-
cant drawback, namely, a  settling speed
of about 0.3 m/s. Convective turbulence
has vigorous upcurrents of 1 m/s and
more, so chaff is easily mixed up through
the whole mixing layer, but it does not

distribute  exactly as  a  passive, non-
buoyant tracer (oil fog droplets are small
enough to have negligible settling speed).
This is the main reason that collocated
releases of chaff and oil fog were made, to
gauge this behavior. It was decided to use
chaff only  for elevated releases in 1983
because it  is relatively easy to correct the
distribution for settling speed effect be-
fore surface contact occurs; comparisons
of plumes from  collocated  oil  fog and
chaff releases  showed the expected
magnitudes of downward displacement
of chaff z (mean height). More sophisti-
cated correcton  schemes can be applied
near the  point  of  maximum  surface
impact,  but this is  not  done  in the
CONDORS data  report. At  farther dis-
tances downwind, chaff deposition and
resuspension become additional compli-
cating factors which may frustrate at-
tempts to correct for settling  effects.
  The X-band  Doppler  radar  used  in
CONDORS has a beam width of 0.8° and
a pulse volume 90 m deep. Typically, the
radar swept horizontally through 1.2° of
azimuth during the dwell time (the time of
accumulation of sufficient return signal).
These factors plus distance determined
the typical spatial resolutions for radar
scans: 90 m  downwind; 70 m vertical;
and 160 m crosswind. The  radar swept
through enough azimuth to encompass
the plume to at lease x = 4 km  and was
raised vertically by about 0.7° increments
to encompass the depth of the plume,
repeating  the whole sequence  about
every 135  seconds. The data for selected
averaging  periods were later processed
and interpolated into  Cartesian coordi-
nates with (50 m)3  cells. Most ground
clutter (strong reflections due to surface
objects) was avoided by purposely siting
the radar so that nearby terrain blocked
the beam  below about 0.5° above hori-
zontal. This  also caused loss of signal
from plume  concentrations  near  the
ground. Remaining clutter that  was sta-
tionary was mapped by the  radar when
the chaff was absent (some clutter due to
aircraft  and surface vehicles had to be
removed at program operator discretion).
In the affected cells, the signal was kept
only if it was at least 10db (a factor of 10)
stronger than the mapped ground clutter,
or if it showed a doppler speed above a
value set for each averaging period, using
the  fact  that  uncontaminated plume
returns have speeds near u.
   Chaff could not be considered a con-
servative  tracer because of the large
dropout of chaff clumps in the  first few
hundred meters downwind  and signifi-
cant surface deposition after the point of
ground contact (in some runs, the inte-
grated plume signal, jfjfXdydz, declined as
much as 50% in the distance interval 1 to
3 km). The estimation of equivalent X/Q
for conservative tracer was done in the
same way as for oil fog, by replacing Q
with ufjfXdydz evaluated at each mean
downwind  distance.  This  estimate, of
course, does not correct for distortion of
the vertical concentration profile which is
due to deposition loss and settling.

  Tracer gas sampling and gas chroma-
tography analysis is a well established
method  for determining conservative-
tracer  X/Q accurately  (within  about
±10%). During the 1983 CONDORS runs,
SF6 was always released a few  meters
from the chaff cutter, on the BAO tower
carriage.  Freon 13B1 (CF3Br) was re-
leased from the surface, a few  meters
from the oil fog generators when they
were also on the ground. The gases were
stored in compressed gas cylinders and
piped through  linearized  mass  flow
meters to the release nozzles, with strip
charts recording release rates and digital
readouts of total  release  volume.  The
release  volume  reading  was checked
against before and after weights of the
gas cylinders.
  Each sampler box contained 12 two-
liter sample bags, each with its own pump
and  intake tubing. A  remote  switch
connected by wire to the line of samplers
initiated the sequential 10-minute sam-
pling for each 120-minute run. After each
run the sample bags were collected by a
person who had not been near the  release
points, to avoid contamination. The bags
from each  sampler were labeled  and
packed into separate 12-compartment
boxes for shipment to Idaho Falls for
analysis. The samplers were analyzed in
the laboratory by use of electron capture
gas chromatographs(GCs). Careful check-
in and handling procedures were fol-
lowed, with calibration of the  GCs with
reference gas mixtures before and after
each analysis shift. Repeatability tests
and independent audits of reference gas
mixtures suggest total errors in measured
plume  X/Q of less than 10% for  SF6.
However, uncertainty in  13B1  measure-
ments was  much greater because plume
concentrations were of the same order as
the GC threshold noise levels.
  Several months after the field work, in
spite of many precautions, 3/8 of the SF8
samples and 7/8 of  the 13B1 samples
were discovered to be severely contami-
nated. After many months of  investiga-
tions it was determined that (1) contrary
to instructions, partly filled gas cylinders
were returned to Idaho Falls in the same
truck compartment with the last three
days' samples, (2) that all gas cylinder
valves leak to some extent, and (3) that
clean  sample bags stored in  the truck
with gas cylinders do become contami-
nated at the high levels that  had been
measured (100 to 1000 times expected
plume  concentrations). Later,  it was
found that the contaminating gas does
not penetrate the bags, but resides on the
lead-in tubing attached to the bags. Four
days of  the remaining  13B1  samples
were  also contaminated.  No  proof of
cause  was discovered, but  the bags
probably were stored near some leaky gas
cylinders at the BAO just prior to use. The
remaining five days  of SF6  samples
showed  no evidence of contamination
except for occasional  "spikes" of high
concentrations  at  individual  samplers,
sometimes recurring several times among
the 12 sequential samples from a run.
This spiking occurred in about 3% of the
SFe samples and remains unexplained.

Meteorological Measurements
  As for any diffusion field experiment,
good meteorological measurements were
vital to the success of CONDORS, both
from the  operational and the scientific
analysis viewpoints. It  was particularly
helpful  to go beyond the limitations of
past field experiments in measurement of
variables important to convective turbu-
lence and diffusion, especially z, and w*.
The BAO meteorological tower provided
an excellent starting point. It is equipped
with three-component  sonic  anemom-
eters, propeller vanes, platinum wire and
quartz thermometers,  and  dew point
hydrometers at eight levels: 10, 22, 50,
100,150, 200, 250, and 300  m AGL It
has been operated more-or-less continu-
ously since 1980 with real-time logging
of 20-minute averages, including vari-
ances  and covariances of wind speed
components and temperature. During
CONDORS runs, the tower  data were
logged m fast response modes so that
many types of statistical analyses could
be made at a later date. For instance,
immediately after  the  experiments, 5-
minute  averages of wind speeds, direc-
tions, and w'T' were calculated to help
define optimum averaging periods  that
avoid large changes in these variables. To
reduce  bulk, much detailed  statistical
information for the selected 1983 averag-
ing periods was processed but not in-
cluded in the data  report, including mo-
ments of vertical velocity (w,w2, and w3),

individual and joint distributions of 10-
seconds average horizontal and vertical
wind directions (8, and 6,), and u, 0«, and
^distributions conditionally sampled dur-
ing negative 10-seconds w events.

  While the tower provided quite detailed
meteorological information for the lowest
300 m, the experiment required some
information  up to z = zi and somewhat
beyond. This was  provided mostly by
rawinsondes, which were released near
sunrise and at least twice during each
run. In 1983, a release near 1000 MST
was added to help  track z, development
and to check upper level  winds for
possible  changes  in expected  midday
winds  in the mixing layer (as zi grows,
momentum  from the entrained air  is
mixed downward). The rawinsondes were
tracked by double theodolite to about
3000 m  AGL so that wind speed and
direction profiles, as well as temperature
and humidity profiles, were obtained.
Additional meteorological  information
included solar insolation  at the surface
and acoustic sounder records.
  A summary of important meteorological
and source information for each averag-
ing period is given in Table 1.  Periods
were  numbered  separately in the 1982
and 1983 experiments. Start times were
Mountain Standard (MST), which was,
fortuitously, extremely close to true solar
time at the BAO. Averaging period dura-
tions  were  usually 30 to 50  minutes.
limited by rapid changes in wind or Zior, in
a few cases, by breaks in chaff or oil fog
releases. For convecti ve experi ments, the
most critical measurement is the mixing
depth, Zi.  For  1982 periods,  this was
determined entirely from the chaff /Xdy
profiles at x > 2 km by a "zero projection"
method. Oil fog detection did not extend
far enough in 1982, but this method was
applied to 1983 oil fog jXdy profiles. A
second method using JXdy vertical pro-
files of oil fog and chaff was also used in
1983; it set zi as the height of  dropoff to
40% of a  peak or  plateau value in the
upper half of the profiles. Rawinsonde
measurements of virtual potential tem-
perature, dew  point, and wind velocity
profiles and lidar measurements of haze
dropoff heights provided additional indi-
cators of zi. Using a consensus of these
estimates, the accuracy of the 1983 z\
estimates is thought to be ±20 m to ±50
m.  The  convective scale  velocity (w*)
depends only on the 1 /3 power of Zj and
w'T'. The latter quantity was taken as the
average of the 10- and 22-m level tower
measurements; at higher levels it tended
to be erratic. The mean wind speed and
direction (u and 0a) were taken as the
average of the upper four levels of sonic
anemometer measurements.  Based  on
past meteorological experiments, this is
believed  to  represent well the whole
mixing layer from 0.1  to 0.9 zi (wind
velocity shear was very slight or negligible
above 100 m during CONDORS runs).
  An underlined release height for oil fog
or chaff  in Table 1  indicates that the
remote sensor data  were processed for
that period. SF« heights are listed only for
uncontaminated periods;  13B1 X/Q  is
available only for period 1-83. The last
column shows the ratio of source height
(z,) to zi for the elevated release, if there
was one. The post-experiment estimates
of zi show that the experimental targets of
z,/zi — 1 /4 or 1 /2 were usually missed.
However, the groupings of periods near
z,/zi = 0.17,0.33, and 0.43 make  possible
the combination of  periods  to achieve
better ensemble averaging for elevated
sources.  For surface sources, data from
all periods can be combined  in terms  of
the dimensionless  convective scaling
coordinates z/zi, y/z(, and X =  (x/u)w*/zi.
Weighting of the periods can be done
using absolute  duration or durations
normalized by zi/u, which typifies the
passage  time  of individual  convective

Data Reported
  Although the final  report gives detailed
explanations of  the  CONDORS  experi-
ment planning, siting, instrumentation,
operating procedures and  processing
methodologies,  it is  primarily  a data
report. Consequently, the bulk of the final
report consists of tables  and figures—
approximately 230tables and 240 figures.
This quantity of information can hardly be
summarized, but a reasonably complete
Table 1.    Meteorological Measurements and Source Summaries for CONDORS A veraging Periods
Source Height (m)

description of it can be given. This de-
scription will be done  in the following
order, representing an increasing amount
of informational detail reported: meteoro-
logical  parameters,  gaseous tracers,
chaff, and oil fog.
  The basic meteorological parameters
reported for each averaging period have
been repeated in this project summary in
Table 1, exceptf or the derived parameters
u/w* (shows degree of convectiveness),
ZjG/w* (length scale for  transport dis-
tance), and period duration times u/z,
(dimensionless duration in terms of "eddy
passage time"). Detailed listings of 1983
zi estimates are given in the final report to
show the consistency of differing meth-
ods; the nine types of estimates never all
agree, but, in each period, at least several
methods give  approximately  the  same
value. Finally, some  examples of the
statistical information available for 1983
periods are given, namely, actual distri-
butions  of 10-s average  wind azimuth
and elevation angles, by 5° bins, for z =
250 m. Most of the azimuth distributions
are approximately Gaussian, but  most
elevation angle distributions are skewed
strongly toward negative values. Much
more statistical  information  was  pro-
cessed but was too bulky to be included;
the same is true of the 5-minute averaged
tower measurements and the rawinsonde
  Essentially all of the processed results
of the 1983 gaseous tracer  sampling,
except for the highly contaminated runs,
are included in the final data report. Basic
information includes the range and azi-
muth of each sampler from the center of
the BAO tower and the average release
rate(s) of the two  gases for each usable
run. The rest of the information concerns
measured X/Q values, the  ratio of con-
centration  to release  rate. In the final
report tables are  given  for  every 10-
minute sample of X/Q at each sampler for
each entire two-hour  run that was not
contaminated (one run for 13B1, five runs
for SF8). However, a small fraction of such
samples are missing  due to sampler
inoperation. Two tables of averaged X/Q
for the  chosen averaging  periods are
presented: one with all the available 10-
minute averages and one with "spikes"
of anomalously  large  X/Q values re-
moved. These spikes are identified in two
tables, one listing multiple spikes at a
single sampler (exceeding both  back-
ground and the averaged two neighboring
samplers by at least a factor of 5) and one
listing especially large single spikes (ex-
ceeding both background and the aver-
aged two  neighboring samplers  by at
least a factor of 10). A figure is shown for
the frequency of occurrence of X/Q
values during each run; this was used to
set "background," or noise level, values.
There are also figures for each usable
1983 averaging period showing average
X/Q values, both with and without spikes,
versus azimuth position of the samplers.
  The basic information for the radar/
chaff results includes the Cartesian grid
range and increment chosen in process-
ing each period, the direction chosen for
the x-axis, and the velocity chosen for
thresholding out return signals contami-
nated with ground clutter. Then  chaff
plume statistics are presented in tables
for each processed  averaging period as
functions of x, mean distance downwind,
incremented by 250 m or 300 m. Statistics
include  y and z, a> and  az (standard
deviations), the y  of maximum /Xdz and
the z of maximum /Xdy,  the value of
maximum/Xdy, and JJXdz. An appendix to
the final report shows plots of each of
these quantities (except y) versus x in 50-
m increments. It also shows normalized
/Xdy or  /Xdz versus z or versus y and
versus x, in semi-tabular form, for each
processed averaging period; x is usually
in 250-m increments.  Finally,  a time
history of the contour /Xdz  = 100 fila-
ments/(50 m)2 versus x and y is shown
for selected periods from 1983.
  The  basic  information presented for
the lidar/oil fog results, besides specifics
on the lidar and laser beam, consists of
logs of release and scan periods. Then oil
fog statistics for each lidar azimuth are
presented for each processed averaging
period. Statistics  include the mean dis-
tances from the  source and from the
tower base (when  used as a  surface
source, oil fog  was released 134 m west
and 44 m north of the tower), the corre-
sponding mean azimuth angles, 
Lidar and Oil Fog.  Both the lidar and the
oil foggers performed well, with only a
few very short lapses. A large release rate
was required to produce a plume  that
could be distinguished from background
haze at 2 to 3 km downwind in convective
conditions. This goal was  marginally
achieved in 1983 by switching to a  very
heavy oil and adding a second fogger for
the surface releases. The paraffin type oil
used in 1982 apparently evaporated to a
significant degree. The lidar provided very
good resolution of plume scans, but with
limited coverage of azimuths.

Samplers and Gases.  The single sam-
pling  arc with 5° azimuth spacing  pro-
vided adequate "ground truth" in  this
experiment to test the X/Q assumptions
applied to the observed  distributions of
chaff and oil fog, which are not conserva-
tive tracers. The loss of more than half the
data due to contamination was unfortu-
nate and avoidable, in retrospect. Sample
bags  should not have been stored  near
source gas cylinders, especially in an
enclosed  space, because the common
type of cylinder valves do leak  slightly.
The contaminated samples could have
been saved had the discovery been made
in time that the contaminating gases
resided only on the intake tubing, not
inside the bags. At any rate, the 1381 gas
released from the surface was inadequate
because the concentrations at the sam-
pling arc were of the same magnitude as
the QC threshold noise levels.  The SF6
gas  released from the tower was  ade-
quate and provided five  runs of useful

Tracer Comparisons.  As already men-
tioned, with collocated releases the chaff
plumes generally tended to sink lower
than the oil fog plumes, to the expected
degree,  due to gravitational  settling.
However, the horizontal patterns of/Xdz
tended to agree very well. Several litera-
ture references also compared SF6  with
the lowest layer oil fog X/Q at the azimuth
nearest the sampling arc for period 9-83,
finding disagreement on the location of
plume boundaries and peak by only about
1° of azimuth, or 20 m.  The/(X/Q)dy of
the oil fog was only 16% larger than that
of SF6; this is very satisfactory considering
the difference in methods and the inexact
coincidence in space.
  Preliminary and  partial analysis of
selected periods from both  1982   and
1983  lead to the  following tentative
conclusions about diffusion in convective
conditions (WD stands for the convective
tank experiments of Willis and Deardorff):
Plume Width and Depth.  The final data
report shows that ffi/z, vs. X = (x/u)w*/zi
for period 9-83 (an  example) is in very
good agreement with WD for both oil fog
and chaff measurements. On the other
     J. C. Kaimal,  W. L Eberhard, W. M. Moninger, J. E. Gaynor, S. W. Troxel.
       and T. Uttal are with the NOAA/ERL Wave Propagation Laboratory, Boulder,
       CO 80303;  G. E. Start is with NOAA/ERL Air Resources Laboratory, Idaho
       Falls, ID 84301; and Gary A. Briggs (also the EPA Project Officer, see below),
       is with Atmospheric Sciences Research Laboratory, Research Triangle Park,
       NC 27711.
     The complete report, entitled "Project CONDORS—Connective Diffusion Observed
       by Remote Sensors," (Order No. PB 86-222 221 /AS; Cost: $28.95, 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
United States
Environmental Protection
Center for Environmental Research
Cincinnati OH 45268
Official Business
Penalty for Private Use $300
       0000329    PS