EPA/600/A-95/055
Controlled Experiments for Dense Gas
Diffusion - Experimental Design and
Execution, Model Comparison
RICHARD EGAMI, JOHN BOWEN, WILLIAM COULOMBE,
DANIEL FREEMAN, AND JOHN WATSON
Desert Research Institute, University and Community College System of Nevada
DAVID SHEESLEY, BRUCE KING, JOHN NORDIN, THAYNE ROUTH
Western Research Institute, University of Wyoming Research Corporation
GARY BRIGGS' AND WILLIAM PETERSEN1
Atmospheric Sciences Modeling Division, ARL, NOAA
Research Triangle Park, North Carolina 27711
1 On assignment to the Atmospheric Research and Exposure Assessment Laboratory, U.S. Environmental
Protection Agency.
ABSTRACT
An experimental baseline C02 release experiment at the DOE Spill Test Facility
on the Nevada Test Site in Southern Nevada is described. This experiment was unique
in its use of C02 as a surrogate gas representative of a variety of specific chemicals.
Introductory discussion places the experiment in historical perspective. C02 was
selected as a surrogate gas to provide a data base suitable for evaluation of model
scenarios involving a variety of specific dense gases. Releases were conducted under
baseline conditions including a simulated "evaporating pool" release over flat
unobstructed terrain. The experiment design and setup are described, including design
rationale and quality assurance methods employed. Design conditions included
moderately low wind speed, stable atmospheric conditions. Four releases were
performed, two of which were during n' r-acutral conditions and two during slightly
stable conditions, Resulting experimental data are summarized. These include C02
cloud characteristics measured at 40 m d >wnwind from the release point. Experiment
success and effectiveness is discussed in terms of mass balance analyses. For Tests 1,
3, and 4 the measured mass accounted for at least 90% of the released mass. Measured
values for Test 2 accounted for only 60% of the released mass. Data usefulness is
examined through a preliminary comparison of experimental results with simulations
performed using the SLAB and DEGADIS dense gas models.
-------
This paper has been reviewed in accordance with the U.S. Environmental
Protection Agency's peer and administrative review policies and approved for
presentation and publication. Mention of trade names or commercial products does
not constitute endorsement or recommendation for use.
INTRODUCTION
Several large-scale experimental releases of denser-than-air gases have been
performed in recent years. These programs were designed and carried out primarily
to quantify potential hazards of specific chemicals such as liquefied natural gas (LNG),
ammonia (NH3), hydrogen fluoride (HF), and nitrogen tetroxide (N20«). Historical
dense gas dispersion field experiments are described in a review by Havens (1992).
They were performed over a wide range of release rates, initial densities, and
meteorological conditions. They provided data sufficient for a reasonable test of
models, were representative of "dense gas" effects, and were sufficiently large scale
to test the important attributes that models required for practical prediction of dense
gas dispersion. However, there are important limitations on the utility of the current
field test data base:
1. The data (and their use for model validation) are limited to description of
dispersion over level, unobstructed terrain.
2. There are important uncertainties in the specification of the source conditions,
including the release areas and rates, release momentum effects, and
specification of the properties of the aerosol and the density vs. concentration
relation.
3. The data are of questionable value for testing the submodel descriptions of
the effect of density stratification (dense gas effect) on dispersion.
4. There are very few data for releases under stable meteorological conditions;
such conditions are a principal concern in accident consequence assessment.
The Clean Air Act (CAA) Amendments of 1990 direct the Environmental
Protection Agency (EPA) to coordinate an experimental and analytical research
program at the Hazardous Chemicals Spill Test Facility (STF) operated by the
Department of Energy (DOE) at the Nevada Test Site (NTS) near Mercury, Nevada.
Although the STF is ideally suited for large-scale testing of flammability as well as
toxicity hazards, CAA provisions currently are assumed to apply primarily to toxic
dense gas hazards, which are typically associated with much lower concentrations than
those associated with flammability hazards. The STF, previously the Liquefied
2
-------
Gaseous Fuels Spill Test Facility (LGFSTF), has been the site of several test programs
involving releases of HF (Goldfish series), N204 (Eagle Series) and NH3 (Desert
Tortoise Series).
The Chemical Hazards of Atmospheric Releases Research (CHARR) Steering
Committee was formed to recommend and prioritize research tasks to be carried out
at the STF, These Research tasks should address limitations of existing field test data,
discussed above, and meet CAA requirements for research at the STF.
The CHARR Steering Committee recommended a DOE/EPA long-term research
program at STF that targets data limitations in the areas of flow over uneven and
obstructed terrain, specification of source conditions, dense gas effects, and stable
meteorological conditions. The CHARR steering committee also recommended that
data should be acquired using surrogate gases with appropriate physical characteristics,
principally density, that can be more generally applied than data acquired using very
specific gases, and that a series of baseline experiments should be conducted with a
surrogate gas and with idealized conditions of flat, level terrain without obstacles.
Baseline experiments would provide data sets useful to the world community for
evaluating models and would provide a comparative basis for evaluating the effects
on dispersion of non-ideal conditions such as uneven terrain with obstructions.
This paper describes the design and execution of a baseline experiment at STF.
The experiment differs from previous test releases in its use of carbon dioxide (C02)
as a surrogate gas with density characteristics that make it applicable to a wide variety
of dense gases. Subsequent sections of this paper describe the experiment design,
setup, execution, and some preliminary modeling results using data from C02 releases
over flat unobstructed terrain. The work was funded by EPA and conducted by the
Desert Research Institute (DRI) and Western Research Institute (WRI).
PURPOSE AND GOALS OF EXPERIMENT
The experiment reported here represents the first use of the STF for highly
controlled releases of heavy gases under stable to neutral atmospheric conditions. The
purpose of the experiment was to develop a data set for use in characterizing the source
term component of mathematical models, that is, to capture sufficient data close to the
emission point to allow for computation of a mass balance. C02 was selected as the
gas to be used because:
• C02 possesses dispersion properties similar to those of other heavy gases.
• Large quantities of C02 can be obtained for reasonable costs and can be
safely transported and handled.
3
-------
• C02 measurement methods are well established and can be efficiently
implemented.
« High concentrations of C02 present a hazard to workers in industries where
it is used.
• C02 is of relatively low toxicity, inert and chemically stable.
Specific goals of the experiment were to:
• Design and test a release, monitoring, and data management system for 0.1
to 1 ton releases of C02.
• Acquire a data base of specified accuracy, precision, and validity suitable for
data analysis and modeling of C02 releases in flat terrain under stable
atmospheric conditions.
• Establish a mass balance between the amount of C02 released and measured
downwind fluxes of C02 through the atmosphere.
EXPERIMENT DESIGN AND PREPARATION
Site Description
The STF is located on the dry lake bed in Frenchman Flat on the NTS,
approximately 100 km northwest of Las Vegas. The dry lake provides a smooth
surface with a roughness length of about 0.2 mm. The site is at an elevation of 939 m
above mean sea level (MSL) and extends 1 km to the north and west, 2 km to the
south, and 3 km to the east of the STF. Terrain slopes gradually downward toward the
dry lake bed, which is in the southeast quadrant of the valley. The surrounding
mountains are located 8 to 20 km from the dry lake at elevations ranging from 1500
to 2000 m MSL.
The STF lies near the southern edge of the Great Basin. Its climatology is similar
to that of the middle elevations of the southwestern desert area of the US.
Precipitation occurs during winter when northern storm tracks move southward and in
summer when moisture from the south causes thunderstorms (monsoons). Annual
precipitation totals about 100 mm (4 in). During summer months, southwest winds
predominate the hours from late morning to sundown with the wind direction being in
the SSW to WSW sector more than 60% of the time. Nocturnal and early morning
4
-------
winds are influenced by the surrounding terrain and become light and variable during
the hours after sunset. During winter months, southwest winds still occur from late
morning to sundown but with frequencies in the 30 to 40% range. The winter storm
pattern results in northerly winds about half the time. Terrain continues to control the
nocturnal winds during the winter when storms are not influencing the area.
The NTS is a controlled-access area operated by the U.S. Department of Energy
(DOE). As such, access by the public is restricted to a minimum distance of 25 km
from the STF. On the NTS, procedures are in place to restrict access near the STF
during experiments. The nearest major population center is Las Vegas.
The STF includes a command and operations center, fast-response data acquisition
systems, a wind tunnel, and a tank farm capable of automated high-pressure releases.
It is available for government, commercial, and academic organizations to conduct
tests on the release and mitigation of hazardous materials and has been used in the past
to understand phenomena related to large-scale spills of liquefied petroleum, chlorine,
hydrofluoric acid, and a number of other chemicals.
Experiment Design
To accomplish the objectives of the experiment, it was necessary to choose a
period during which atmospheric conditions were stable while maintaining a consistent
enough wind direction to transport the released gas towards a fixed array of sensors
and samplers. The permit issued by the state of Nevada allowing the release of
hazardous materials at the STF also required that any releases be done before sunset.
Meteorological Requirements
Data collected onsite prior to the experiment and near the site from NTS weather
stations showed that afternoon and evening winds generally follow a regular pattern
during summer. Southwesterly winds (wind directions centered around 225") typically
set up by 1200 PST, reaching maximum speeds of 4 to 8 m/s during midafternoon. At
30 to 60 minutes before sunset, the speed typically begins to decrease abruptly,
reaching speeds near 1 m/s within a 50 to 80 minute period. The southwesterly wind
direction persists as the speed decreases for as much as 1 hr later. Eventually, the
winds become light and variable. This decrease in wind speed coincides with the
change in atmospheric stability from slightly unstable or neutral to stable as the surface
heat flux reverses from positive to negative.
The regularity of the decrease in wind speed at sunset and the persistence of
southwesterly winds during this decrease established a target release window. The
deciding factor for the release time was the decrease in wind speed to 3 m/s in
conjunction with the consistent direction toward the sampling array. A release
5
-------
duration of 5 minutes was selected to maintain nearly stationary atmospheric
conditions during the release period.
Release Method for CO, and SF,
Consideration of available models of dense gas dispersion placed the following
requirements on the release system for C02: 1) the gas had to be released with low
momentum (i.e., not as a jet), 2) the release rate had to be constant in the range of I to
2 kg/s, and 3) the released material had to be in the gas phase and nearly isothermal.
These requirements were met by designing a release method with sufficient residence
time between depressurization (release from the tank) and release to the atmosphere
for the C02 to completely vaporize and reach ambient temperature. In addition to C02,
a tracer gas, sulfur hexafluoride (SFg), was released at the same time at a fixed rate as
a secondary test to evaluate the use of tracers in characterizing the dispersion of dense
gases.
COj was obtained in a refrigerated six-ton tank at 2 ®F (-17 °C) and 2103 Pa (305
psig), (Airco Gases, City of Industry). The C02 passed through a 15 kW heater and
into an insulated 1012 ft3 (28.7 m3) surge tank at a temperature of -85 °F (29 °C), A
4 in (0.1 m) diameter, 60 m long pipe connected the surge tank to a 6 in (0.2 m) control
valve placed near the release point The control valve was activated remotely to start
and end a release. A baffled discharge chamber (BDC) was connected to the outlet of
the control valve to provide an initial dampening of the flow discharge. An 18 in (0.5
m) diameter flexible hose was connected from the BDC to the bottom of a 1 m3 box
that was buried flush with the ground. The 1 m3 box had additional baffles to further
reduce the momentum of the released gas. Gaseous C02 was released from the top of
the 1 m5 box at ground level. The ground within 3 m of the discharge box was restored
to approximately the same surface roughness as the terrain of the dry lake bed.
SF6 was introduced to the C02 gas stream between the control valve and the BDC
with the BDC providing a homogeneous mixing of C02 and SFt. A Tylan Model 260
mass flow controller maintained a constant flow rate of 0.9 g/s of SF4 from a 10-lb
cylinder from Scott Environmental. The release of SF6 coincided with the release of
C02.
Thermocouples in the release line ahead of the control valve, behind the control
valve, in the box, and in the area immediately surrounding die release point quantified
temperatures at these locations during each release. Temperature and pressure were
also measured inside the surge tank. The C02 flow rate was calculated from these
quantities. The SF6 flow rate was measured and recorded from the mass flow
controller.
6
-------
Measurement Methods
C02 concentrations were measured by 1) collection of air samples in Tedlar bags
and subsequent analysis (Horiba C02 analyzer, model 355098-2) and 2) real-time
sensors (Nova Analytical Systems, Inc., micro C02 sensor, model DEO, 1% and 10%
ranges). Bag samples were used for mass balance determinations for the C02 releases.
The continuous sensors were used to detect the presence of the C02 cloud and its
instantaneous concentrations. Bag sampling was controlled remotely and was
coordinated with the release time. Air samples were collected from the time of release
to the time that all C02 had passed the sensor array. Flow rates for sample collection
were set to collect about 12 1 of air. Data from the continuous sensors were collected
every 2 seconds by the STF data acquisition system (DAS).
Bag samples were analyzed for SF6 concentrations at the same time that C02 was
measured. The SF6 measurement system consisted of a Varian gas chromatograph with
electron capture detector, a Hewlett-Packard integrator (model 3390A), and a reactor
sampling train to remove oxygen while retaining SF6. A Campbell Scientific data
logger, model CR-10, controlled sample injection and data collection.
Bag samplers and sensors were deployed downwind of the release point along a
line perpendicular to the prevailing wind direction of 225°. Preliminary modeling for
the anticipated release rate and meteorological conditions estimated that concentrations
along a line 40 m from the release point (the 40 m arc) would be within range of the
continuous sensors. Accounting for variations in wind direction and plume spread, bag
samples were collected along the center line, at 3.5° on either side of the centerline,
and at several locations 5 s to 6° apart, outside the arc. The outside samples were 20°
from the centerline. In distances from the centerline, these locations were 0 m,±2.5
m,±6 m,±10 m, and ±15 m. Preliminary modeling results indicated the best heights for
defining the cloud mass to be 0.10 m, 0.33 m, 0.67 m, 1.00 m, and 1.40 m above the
ground. A total of 43 bag samplers was deployed along the 40 m arc while 2 bag
samplers were placed upwind of the release point to collect background samples. Also
placed along the 40 m arc were 37 continuous sensors, with 30 sensors collocated with
bag samplers. The remaining 7 continuous sensors were placed along the 40 m arc at
locations outside the anticipated cloud to check for possibly wider and higher clouds
at ±30 m from the centerline, ±22 m from the centerline, and on the centerline at 2 m
above ground. The 10% sensors were located at the lower two height levels within the
inner five towers.
Wind speed and direction were measured (R.M. Young Wind Monitor-RE, model
5701) along the release centerline 5 m upwind of the 40-m arc at a height of 0.5 m
above the pound. Data from these sensors were collected by the STF DAS with other
continuous sensor data for use in subsequent mass balance calculations.
A sonic anemometer/thermometer (Applied Technology, Inc., model SAT-
211/3 K) was placed 40 m upwind of the release point to make direct measurements of
heat and momentum fluxes. Wind speed component and temperature data were
7
-------
collected at a rate of 10 Hz and averages, variances, and covariances were computed
in 1-minute blocks by a dedicated DAS.
Meteorological measurements were also made at 8 levels on a 24-m tower located
112 m south-southwest of the release point. Measurements included wind speed and
direction, temperature, relative humidity, solar radiation, net radiation, and soil
temperature. The intent of the tower has been to provide a climatology of the
atmosphere near the ground in the vicinity of the STF. Data were collected as
continuous 5-minute averages with a dedicated DAS and as instantaneous values
during the release period with the STF DAS.
Data Management
Continuous data from the real-time C02 sensors, the wind speed and direction
sensor at the array, the meteorological tower, the temperature and flow sensors for the
releases, and the control sensors, were collected by the STF DAS. Data were collected
every 2 seconds, with meteorological data updated every 10 seconds, for
approximately an hour before each release to 10 minutes after each release. Raw data
from the STF DAS, including header information, were transferred in comma
delimited format to diskettes for verification and analysis on IBM-compatible personal
computers. Following the experiment, continuous C02 data were corrected for
calibration drifts and for background readings. Continuous meteorological data were
entered into Excel spread sheets for editing. Because of intermittent data collection
problems, some meteorological data required deletion.
C02 and SF4 concentrations from bag samples were saved directly in dBase IV
data bases. In addition to concentrations, the data bases contained bag ID, sample
location, and replicate analyses. After each test, preliminary data were plotted to check
that the systems were operating. Following the experiment, calibration data provided
minor corrections to the integrated C02 and SFa concentrations.
Quality Assurance
Quality assurance (QA) procedures for this experiment were defined in a QA
Plan (DRI/WRI, 1993) to ensure that data of known and acceptable quality were
collected. Written protocols defined operating procedures for each major measurement
system. Independent field audits were performed to verify the C02 and SF6
measurement systems. Meteorological instruments were purchased new and calibrated
just prior to the experiment
Data accuracy was determined by challenging each measurement system with
blind audit standards prepared ami administered by an independent QA Officer. Data
precision was determined using replicate measurements.
To assess data accuracy, blind samples of SF6 and C02 in Tedlar bags were
introduced to each measurement system. The samples were made from standards of
8
-------
SF,, and CO- purchased t'rorn ScoU-Marrin, Inc. (Riverside, CA). traceable to the
National Institute of Standards and Technology (MIST). Standards were diluied in
each audit bag with ultra high purity nitrogen or air. using calibrated mass flow
controllers. Nominal concentrations of SF;, were verified by analysis using gas
chromatography with an electron capture detector (GC/ECD). using 60 m x 0.32 mm
i.d. DB-1 capillary column (J&W Scientific Co.). Chromatographic temperatures were
programmed; for -60°C for 3 min and then raised 6°C/min to 30°C. C02 standards
were verified using a Nickel based methanizer, coupled with a GC/flame ionization
detector (FID). A 30 m x 0.52 mm i.d. GS-Q capillary column (J&W Scientific Co)
at 40°C was used for C02 analyses.
Two SF6 standards were prepared in N: and in air at 67 pptv. Four C02 standards
were prepared in N2 at 10.12%, 0.459%, 0.304% and 0,081% concentrations.
Accuracy was calculated as a percent difference between the audit standard and
found value, according to the equation:
A = 100 . — m
where:
X = audit target value (nominal)
Y = found value (measured)
The percent differences between the audit standard in nitrogen and air and the SF6
system were -0.6 and -0.7, respectively. The percent difference between the three
audit standards and the C02 bag measurement system was -2, 0, and 1. Nine
continuous sensors were audited with the C02 standards. The maximum difference
was -19% and the minimum was 2% (mean difference of-7 ± 11).
Precision for C02, as shown in Table 1, was evaluated by measurement of sample
replicates, measurement of repeated instrument spans, and determination of the
standard deviation of four one-minute instrument averages. Reported C02
concentrations for each sample were based on the mean of four one-minute averages
(except those samples that had erratic readings or fewer averages because of small
sample volume). For each test, the average of the standard deviations gave a precision
of the measurement Measurement bias based on sample replicates and instrument
spans also is given in Table 1.
9
-------
TABLE 1
Summary QC Data for CO?
Test
Parameter
Bias
Precision
n
1
Sample Replicates
-4 %
9%
5
Four One Minute Averages
1 ppmv
43
Instrument Spans
<1%
<1%
4
2
Sample Replicates
<1%
1%
5
Four One Minute Averages
2 ppmv
42
Instrument Spans
<1%
<1%
<
3
Sample Replicates
< 1%
2
6
Four One Minute Averages
3 ppmv
42
Instrument Spans
< 1%
<1%
4
4
Sample Replicates
-3%
3%
5
Four One Minute Averages
—
1 ppmv
43
Instrument Spans
<1%
<1%
6
Note: Bias is based on the mean of the percent differences between replicates (or
target versus found span values) and precision is based on one standard
deviation of the mean. The precision of the four one-minute means is based on
mean values in ppmv.
Precision for SF6 (Table 2) was determined from sample replicates and
repeated instrument spans. For Test 1, some bags were analyzed the next day so
replicate analyses were repeated the next day to evaluate aging effects. The next
day replicate analyses suggest a negative bias (decrease in bag SF6 concentration)
and greater variability (higher standard deviation of the mean). All bags after Test
1 were analyzed immediately after the release.
10
-------
TABLE 2
Summary QC Data for SFs
Test
Parameter
Bias (%/
Precision {%)'
1
Replicate day 1
0
1
Replicate day 2
-8
15
Replicate both days
-4
11
156 pptv span (both days)
-6
4
2
Replicates
0
1
156 pptv span
-5
1
l 3
Replicates
0
2
156 pptv span
0
3
l 4
Replicates
2
6
156 pptv span
0
1
1 Bias based on the mean of the percent differences between replicates (or target versus
found span values) and precision based on one standard deviation of the mean.
RESULTS
At the time this manuscript was prepared, data from the experiment had been
validated but results are considered preliminary. This section provides an overview of
a portion of the data analysis performed to date.
11
-------
Meteorology
A summary of the meteorological conditions during each release is provided in
Table 3. Values reported are averages for periods during which the C02 plume was in
steady state. For Tests 1,3 and 4, the averaging period begins one minute after the start
of the release and ends at the time the release ends for Test 1 and one minute before the
end of the release for Tests 3 and 4. For the short release of Test 2 (70 seconds), the
averaging period is for the one-minute period beginning ten seconds after the start of the
release.
As seen from the stability classes, releases targeted to occur under stable conditions
(E-F) were performed under neutral (D) and neutral to stable (D-E) conditions, where
stability classes were estimated from graphs by Golder (1972). At the STF, chance
plays a large role in obtaining very stable conditions while maintaining the target wind
direction. As wind speeds decrease and stable conditions develop, the required wind
direction (225°) becomes more variable, as discussed previously. Therefore, it was
necessary to compromise between triggering the release at the slowest possible wind
speed (an indication of increasing stability) and limiting plume meander (so as not to
miss the downwind sensor array).
Chemical Release
The amount of C02 released in each test is the difference between the mass
calculated before and after each test The volume of the C02 before the release includes
the volume of a 30 ton surge tank (1012 ft3,28.7 m3), while the volume after the release
includes the volume of the tank plus the release line (1025 ft3,29.0 m3). The ideal gas
law was used to calculate the mass concentration using an atmospheric pressure
90.32 Pa (13.1 psia), appropriate for the elevation of the STF at 939 m above mean sea
level. A summary of the release parameters for each test is provided in Table 4.
Plume Measurements
This section gives results of the C02 and SF6 measurements. For bag samples, all
but three were successfully collected (98% recovery) during the four tests. Failure of
the three bags was related to insufficient sample collection attributed to failed solenoid
valves or loose sample container lids.
12
-------
TABLE 3
Summary Meteorology During Releases
Parameter
Test 1
Test 2
TestS
Test 4
PasquiU Stability Class
D
D-E
D-E
D
Tower Height (m)
WS WD
fm/s> (dee)
WS WD
(m/st (deg)
WS WD
(m/s) (dee)
WS WD
(m/st fdeyt
24,0 m
6.17 228
4.84 —1
5.36 —
8.32 —
16.0 m
6.18 -
4.39 -
5.06 -
7.85 —
8.0 m
6.03 -
3.78 -
4.44 —
7.03 219
4,0 m
5.70 231
3.37 237
4.06 234
6.65 218
2.0 m
5.26 224
3.00 233
3.61 229
6.08 214
1.0 m
4.75 —
2.67 236
3.19 232
5.49 215
0.5 m
4.32 226
2.35 233
2.93 ~
4.97 213
0.25 m
3.99 227
2.01 236
2.64 233
4.61 216
40 m Arc
Height=0.5 m
3.98 231
2.76 232
2.95 232
4.94 218
| Ambient Temperature at 1
I m(°Q
31.4
31.4
33-5
33.7
Ambient Pressure (mb)
903
903
903
902
Ambient Relative Humidity
<%)
10
12
9
8
'Missing Data
13
-------
TABLE 4
Chemical Release Summary
Parameter
Test 1
Test 2
Test 3
Test 4
7/22/93
7/26/93
7/27/93
7/28/93
CO: Release Start:
Stop:
18:55:01
18:58:00
19:41:01
19:42:11
19:30:00
19:34:25
19:46:00
19:50:25
SF( Release Start:
Stop:
18:55:04
18:58:04
19:41:06
19:42:16
19:30:06
19:34:28
19:46:06
19:50:30
Bag Sampler Start:
Stop:
18:55:08
19:00:01
19:41:08
19:44:39
19:30:09
19:35:41
19:46:08
19:52:01
COj/SF4 Release Duration (s)
179
70
265
265
C02 Mass Released (kg)
81.9
112.9
171.9
165.0
I
C02 Release Rate (kg/s)
0.458
1.613
0.649
0.623
SF, Mass Released (mg)
163.4
63.6
238.2
240.0
SF„ Release Rate (mg/s)
0.908
0.909
0.909
0.909
During Test 1, preliminary evaluation of the release indicated the volume was less
than the target amount and analysis of the bags would not be necessary. The bags were
analyzed nevertheless, if only to provide a shakedown of laboratory operations.
However, a portion of the bags were not analyzed until the following morning. Tests
on aging SF6 standards in bags indicated a decrease in concentration with time for some
bags. Replicate analyses performed the day after the release (Table 2) followed a
similar trend indicated by a higher negative bias and poorer precision. Bags of Tests 2,
3 and 4 were analyzed within four hours of collection. A summary of results of bag data
for C02 and SF6 is presented in Figures 1 and 2 respectively.
14
-------
4000
f 3000
s
fit
5- 2000
8 iooo
o
Test I
yl#
11-—
— -*—
1 _h- —x
-! ^
•15
¦10
0
10 15
10000
Test 2
s
Arc Location (m
5 0 5
10 15
10000
i
o
u
5000
Test 3
Arc Location (m)
-10 -5 0 5 10 15
6000
I 4000
a.
a
| 2000
Test 4
~—0.100m
0.333m
*• -0.667m
X 1.000m
X— 1,400m
-15
-10
Arc Location (m)
0 5
15
Figure 1, Summary of Integrated CQi Data for 5 Heights 40m Downwind of Release.
15
-------
Test I
Ob
Q.
w
£
CO
-15
-10
Arc Location (m)
0 5
¦10
ocation (m)
0 5
10
15
Test 2
15
U.
Crt
Test 3
-15
•10
r
Arc Location (m)
-5 0 5
10
15
3000
2500
| 2000
3 1500
| 1000
500
Test 4
•—0.100m
0.333m
* ••0.667m
X 1.000m
1.400m
Arc Location
0 5
Figure 2. Summary of Integrated SF» Data for 5 Heights 40m Downwind of Release.
16
-------
For continuous measurements of C02, the sensors generally operated successfully
during the four tests. Maximum C02 concentrations recorded for Tests 1-4 were 10,500
ppmv (1%), 54,000 ppmv (5.4 %), 21,000 ppmv (2.1%) and 16,000 ppmv (1.6%),
respectively. In regard to the adequacy of the measurement ranges of each sensor (0-
1% and 0- 10%), three sensors saturated (i.e., they were exposed to a concentration that
exceeded the maximum range) in Test 2, one in Test 3 and two in Test 4 (all 0-1%
units). One unit in Test 3 recorded an anomalous trace and the data are considered
suspect
A comparison between continuous and bag integrated C02 generally indicates good
agreement between the two measurement systems. Background C02 (approximately 366
ppmv) was measured for each test and subtracted from the bag samples and an average
was obtained from the continuous sensors for the same period the bags were open.
Linear regressions of all valid collocated bag and sensor pairs for Tests 1-4 indicate
correlation coefficients of 0.9797,0.9180,0.9883, and 0.9873, respectively.
A comparison of bag C02, averaged sensor C02 and bag SF# data, at a height of 0.1
m above ground, by arc location, is provided in Figure 3. Generally there is good
agreement between bag and averaged sensor data and the concentrations of SF# are
proportional to C02. Test 2 SF4 data levels were much lower than those in the other
tests. No anomalies were noted in the SFt sampling and analysis systems, therefore the
decrease in downwind concentrations was related to the release system.
Chemical Mass Balance
The effectiveness of the sensor and bag systems was determined by comparing the
integrated fluxes of C02 and SF6 measured at the 40-m arc to the amount of released
material. The ratio of detected to released mass, or mass balance, is an indicator of the
collective accuracy of the total system and the sufficiency of the wind speed and
concentration arrays to characterize the total mass of a release.
The mass of C02 released was calculated as described previously, using temperature
and pressure data from the tank and line collected before, after, and during each test.
The amount of released SFs was determined from the flow reading of the mass flow
controller that maintained a constant flow rate during release. Flow voltages were
collected by the STF DAS and converted to mass flow rates from calibration factors
developed for SFs. The total amount of released SF6 was calculated from mass flow rate
in mg/sec for time of the release.
17
-------
>
E
a.
a
N
O
CJ
3000
2000
1000
O Bag
O Nova
-• A---SF6
1 ^
1
est 1
--
{]—
-Q-
1
1—1
\J
>0-..
i»
2000
-30 -25 -20 -15 -10 uffUon^m) 10 15 20 25 30
10000
E
M 5000
r«
o
u
£~
A'
0
1
est 2
(>
-o-
1
1—1
h-g
li
J
i
¦¦ 1
l-,.
-B-
300
a
>1
Arc Location (m)
-30 -25 -20 -15 -10 -5 0 5 10 15 20 25 30
10000
9
E
s 5000
o
u
0
JO,
rA'
aSJ
« ~
1
est 3
-
-a-
1
1—1
K-d
A
-Q—
3000
a
>8
Arc Location (m)
-30-25 -20-15-10 -5 0 5 10 15 20 25 30
6000
| 4000
3
s 2000
u
1
est 4
!>
.~Q-
H-H
1—4
»
-Q—
3000
2000g.
Bi
lOOOg
-30 -25 -20 -15 -10 ^uyloo^n,) io 15 20 25
30
Figure 3. Comparison of COj and SF« Data at the 0.1 m height
18
-------
The amount of mass passing through the 40-m arc of the sampling array was
determined from direct measurements by the C02 sensors, integrated measurements of
C02 and SF6 in bags, and wind speed measurements from the meteorological tower.
Over the period of sampling, the combined air and released gases defined a volume with
cross-wind height and width defined by the dispersion of the released gas and along-
wind length defined by the average wind speed. The integration or summation of the
concentrations of the released gases in the volume, converted to density using ambient
pressure and temperature, gave the amount of mass for the released gases in the volume.
The integration of the mass of released gases measured at the 40-m array was
accomplished in the following steps:
1. Average concentrations for continuous C02 sensors and average wind speeds for
sensors on meteorological tower were calculated for the time that bags collected
air.
2. Background concentrations for sensors and bag samples were subtracted from all
data. For the sensors, the background was determined from measurements when
the plume was not present For the bag samples, the background was determined
from upwind samples and from samples not affected by the plume.
3. Missing gas concentrations for sensors and bags were estimated by interpolation.
The scheme started at the level nearest ground and worked in an upward
direction. Missing concentrations were estimated by assuming that the ratio of
the missing measurement to that of its nearest neighbor at the same height equals
the ratio of concentrations at the same horizontal locations at the next lower
height. For missing values equidistant between two points, the averages of the
concentrations on both sides of the missing value were used.
4. Cross-wind integrated concentrations (CIC) were calculated with a linear
interpolation of concentration across the array. Each measured concentration was
assumed to be constant over a horizontal distance defined by the midpoints
between measurements. The CIC was the sum of the products of the individual
concentrations and their horizontal spacing.
5. Average horizontal wind speeds were determined at the level of the conceniiation
measurements for the period of bag sampling. Wind speeds at 0.333,0.667, and
1.4 m above the ground were calculated using a geometric interpolation from data
collected on the meteorological tower at 0.25,0.5,1, and 2 m above the ground.
The interpolation for u„ at level z* with u, at z, and u2 at Zj was given by:
19
-------
u(zk) . Il(x,)
u(z7)
«(*,)
ta(V*iyin(V*i)
(2)
6. The average horizontal wind speed at the level nearest the ground (0,1 m) was
determined by extrapolating the logarithmic wind profile from 0,25 m using a of
0.0002 m:
«<(*). »<(0.25)
ln(x/*,j)
In (0.25 /xjj)
(3)
7. The volume flux of released gas was computed at each vertical level by multiplying
the CIC at each level by the wind speed for that level. The vertically integrated flux
was calculated with a linear interpolation of flux in the vertical direction. At each
level, the flux was assumed to be constant over a vertical distance defined by the
midpoints between measurements. To account for the gas that was above the
measurement array, the flux was assumed to decrease exponentially with increasing
height The concentrations at levels 0.667 and 1.4 m were used to determine the
rate of decrease with height The total flux of released gas was calculated as the sum
of the products of the individual fluxes and their vertical spacing.
8. The total mass measured by the sampling array was calculated from the total volume
flux using the density of C02 and SF( at the pressure and temperature conditions
during each test
Chemical mass balance results are summarized in Table 5. Results of C02 for both
continuous sensors and integrated bag samples were reasonably good. For Tests 1,3, and
4, the measured mass accounted for at least 90% of the released mass. Measured values
for Test 2 accounted for only about 60% of the mass. SF, results were not as good. For
Tests 1 and 4, the measured mass was more than the released mass. For Test 3, the
measured mass accounted for only about 50% of the released mass. Results of Test 2
indicate that there may have been a problem with the release system at that time.
20
-------
TABLE 5
Chemical Mass Balance Summary
Parameter
Test 1
Test 2
Test 3
Test 4
7/22/93
7/26/93
7/27/93
7/28/93
CO; Mass Released (kg)
81.9
112.9
171.9
165.0
CO, Mass Measured-Continuous (kg)
93.0
59.9
161.0
163.1
CO, Mass Measured-Integrated (kg)
77.4
73.2
155.3
159.6
CO, Mass Balance- Continuous (%)
114
53
94
99
CO; Mass Balance-Integrated (%)
95
65
91
97
SF, Mass Released (mg)
163.4
63.6
238.2
240.0
SF, Mass Measured (mg)
171.9
9.4
126.9
280.0
SF6 Mass Balance (%)
105
15
53
117
MODEL COMPARISON
The remainder of this paper describes an application of the dense gas models
SLAB and DEGADIS to two of the Experiment 1 releases. The SLAB (Ermak, 1990)
and DEGADIS (Spicer and Havens, 1989) models were selected because of their
popularity and wide use. They are both included as Alternative Air Quality Models
in EPA's Guideline on Air Quality Models (U.S. EPA, 1994).
Model Inputs
SLAB and DEGADIS were used to simulate Tests 3 and 4. Test 3 occurred
during Pasquill stability Class D-E (slightly stable) conditions and Test 4 occurred
during class D (neutral) conditions. Meteorological conditions during these tests are
detailed in Table 3. Actual model inputs are tabulated in Table 6. Both Tests 3 and
4 consisted of releases lasting 265 seconds. Wind speeds and wind directions are
21
-------
averages over the 3-minute (180-second) period beginning 60 seconds after C02
releases began, and ending 25 seconds before the releases ended (wind direction was
used to align model output with the sensor array, as described below). DEGADIS was
run in the isothermal steady state mode, and the last 4 items in Table 6 were used to
define the air/contaminant density profile. A linear profile was assumed, from molar
fraction equal to 0.0 to molar fraction equal to 1.0. SLAB was run in the evaporating
pool mode. It was assumed that the temperature of C02 at the moment of release was
the same as the ambient air temperature.
Model estimates and corresponding experimental data are for the nominal "40 m"
sensor array. This amy is actually 38.5 m from the center of the release pit.
DEGADIS does not allow designation of specific discrete receptors, but rather
calculates receptor locations internally. The DEGADIS- generated receptor distance
closest to the nominal 40 m array was 38.3 m downwind for Test 3 and 38.5 m for Test
4. A distance of 38.5 m was defined for both SLAB simulations as the maximum
modeled downwind distance. Also, the sensor array was centered on an axis oriented
at 225°, for prevailing southwest wind directions. The actual average wind direction
during experimental releases was off the 225° line by varying amounts. For Test 3,
the actual average wind direction was 231°, 6° off the sensor array axis and for Test
4 the array wind direction was 215°, 10° off the array axis. For presentations, model
results were thus shifted laterally to line up modeled and experimental plume center
lines. For these reasons, actual downwind distances are not exactly the same for
modeled and experimental conditions, but are within about 1 meter.
Model Results and Comparison with Experimental Data
Three different experimental data sets were used for model comparisons,
corresponding to different averaging times:
1. Integrated Bag Samples. Bag sample concentrations were adjusted to represent
the actual C02 release time rather than the total bag sampler time (see Table 4).
2. Three minute averages from the Nova sensors, for the same periods for which
wind data were averaged.
3. Maximum 30-second end-to-end average concentrations during the same 3-
minute averaging period.
22
-------
TABLE 6
Model Inputs Used for SLAB and DEGADIS Simulations
Parameter
Test 3
Test 4
Common Parameters
COj Release Rate (kg/s)
0.649
0.623
Wind Speed 9 1.0 m agl (m/s)
3.19
5.49
Wind Direction (deg)
231
215
Ambient Temperature (°K)
306
306
Ambient Pressure (rob)
903
903
Relative Humidity (%)
9
8
Surface Roughness (mm)
0.1?
0.17
Pasquill Stability Category
E
D
Averaging Time (sec)
30
30
Source Area (m!)
1.0
1.0
CO, Physical PropertiesiSlAB)
Vapor Heat Capacity (J/kg-K)
853
853
Heat of Vaporization (J/kg)
573,500
573,500 |
Liquid Specific Heat (J/kg/K)
1,276
1,276
1 Liquid Density (kg/m5)
1,564.3
1,564.3
1 CO. Physical Pro^rtUs (DEGADISt
j Mean Heat Capacity Constant (J/kg-Mol/K)
4,241
4,241
Power for Contaminant Heat Capacity Equation
1.0
1.0
Contaminant Concentration for
Molar Fraction = 0.0 (kg/to5)
0.0
0.0
Air/Contaminant Mixture Density for
Molar Fraction * 0.0 (kgto5)
1.0512
1.0512
Contaminant Concentration for
Molar Fraction = 1.0
1.5969
1.5969
3 Ait/Contaminant Mixture Density for
1 Molar Fractions 1.0
1.5969
1.5969
23
-------
Background values of 360 ppm were subtracted from bag sample concentrations.
Background values based on upwind measurements were subtracted from Nova sensor
data.
DEGADIS model output consists of ground-level (height = 0.0 m) centerline
concentration plume half-widths, parameters Sr and S, (Sigma Y and Sigma Z, not the
same as Gaussion Plume Sigmas), and a power-law wind velocity profile. These values
were used, with equations v-71 (pp. 38-39) of the DEGADIS User's Guide (Spicer and
Havens, 1989) to extrapolate cross wind profiles to heights above ground level. The
SLAB model includes height above ground as a model input variable. Separate SLAB
model runs were made for each sensor height
Trinity Consultants, Inc. software was used for DEGADIS model runs and
Bowman Environmental Engineering software was used for SLAB model runs. Trinity
and Bowman software enhancements consist primarily of menu-driven data entry
software.
The following results are qualitative. More detailed analyses, using performance
measures such as recommended in by EPA (1984) and used by Hanna et al (1991), and
a wider selection of models, should be made.
Maximum Concentrations
Figure 4 depicts maximum concentrations. SLAB and DEGADIS results are
shown, along with experimental data as described above. For the slighdy stable Test 3
case, there is a tendency toward overprediction by both models at low cloud heights
(near 0.1 m agl) and underprediction at high cloud heights (1.0 to 1.4 m agl). Both
models appear to skew the vertical distribution, i.e. underestimate cloud thickness. Near
ground level, where concentrations are highest, DEGADIS overpredicts by about a
factor of 2. For Test 4, the neutral case, the same tendencies are evident toward
overprediction at low cloud heights and underprediction at high cloud heights by
DEGADIS. The SLAB results appear to agree better with experimental results than do
the DEGADIS results.
Cloud Widths
Figures 5 and 6 show cross-wind profiles of C02 concentration at cloud heights of
0.1 m and 0.667 m for Tests 3 and 4.
24
-------
Max C02 Conmntratkm (ppm)
—— Dag ad i*
& SLAB
- ® - Max 30-Sec A*g (Nova)
- Q - 3-Mlnut* A*g (Nova)
- ~ - Integrated Avg (Bags), corrected
e» •
10000 20000 30000
Maximum 002 Conosoirafion (ppm)
40000
Test 3
1.8
1.2
0.S
0.4
0.0
10000 20000 30000
Maximum CO2 ConcenbaHbn (ppm)
40000
Test 4
Figure 4. Modeleded and Observed Vertical Profiles of Maximum CO, Concentrations at 40 m, Surface Roughness = 0.17 mm
-------
12000
-40 -20 30 40
Crossmnd Distance (m), Reference to 225 Degree Azimuth
-20 20 40
CmstwM Distance (m), Rotemnoed to 225 dogma Azimuth
Height = 0.100 m
Height = 0.667 m
Figure S. Modeled and Observed Crosswind Profiles of CO, Concentralions at 40 m Downwind for Two Heights, Test 3. Surface Roughness 0.17 nun
-------
T«rf 4 CQt i*f >m«
SlAB
HuIMmNivi
MIiNm
-40 -20 20 40
OotsrnM Dfefano* (mj, R«hnnc*d to 225 Dogma Azimuth
8000
0000
J 4000
2000
-tt -30 20
CrasmvMDislano»(mt,Re/anmo0(1 to 225dogma AlimMi
Height = 0.100 m
Height = 0.667 m
npve 4 Modeled and Observed Cms wind Profiles of CO, Caucenbatiaos at 40 m Downwind for Two Heights. Test 4, Surface Roughness 0,17 mm
-------
For the slightly stable Test 3 case, DEGADIS appears to slightly overpredict cloud
width at low cloud heights, while SLAB is more in line with experimental results. At
mid-heights, represented by the 0.667 m height, both SLAB and DEGADIS are in good
agreement with experimental results. For the neutral Test 4 case, both models are in
good agreement with experimental results.
SUMMARY AND CONCLUSION
In July, 1993 a baseline experiment involving a series of four C02 releases was
performed at the DOE Spill Test Facility in Nevada. Specific goals of the experiment
were to design and test a release, monitoring, and data management system and to
acquire a data base suitable for model testing. The data collected during the 1993
experiment began to fill data gaps in two important areas. First, releases were
performed during stable atmospheric conditions, albeit only slightly stable conditions.
Previous wind tunnel and field experiments have amassed data to evaluate dense gas
dispersion models for flow over flat or sloping terrain during neutral and unstable
atmospheric conditions. However, there is a paucity of data for model development and
evaluation for low wind speed, stable atmospheric conditions. Second, the
meteorological data collected at eight levels on the 24 m tower and from a fast response
sonic anemometer provide ample on-site boundary layer characterization during each
release. During the rest of the time meteorological data collection continues at a less
intense rate to document on-site boundary-layer climatology; this will be valuable for
optimizing future field experiments, such as experiments designed to capture data during
more stable conditions.
The design conditions for the C02 releases were moderately low wind speed, stable
atmospheric conditions. Tests I and 4 were actually performed during near-neutral
conditions with winds about 5 m/s, while Tests 2 and 3 were performed during slightly
stable conditions with winds about 3 m/s. While the Spill Test Facility provides an ideal
location for performing releases of hazardous materials because of its remote location,
variety of terrain, and relatively predictable meteorology, chance still plays a part in
attempting a release during stable conditions. Analysis of the meteorological data shows
that stable atmospheric conditions develop shortly before sunset as the wind speed
begins to drop. The regularity of the near-sunset drop in wind speed and the persistence
of southwesterly winds during the drop makes it relatively easy to target release times
for various atmospheric stabilities. The 0.0B m/s per minute drop in wind speed is not
a serious problem for release durations of 5 minutes or less. However, for much longer
releases this non-stationarity complicates data analysis.
28
-------
Dense gas dispersion models were used to design the monitoring array for a C02
release rate of about 1 kg/s. The primary sampling arc at 40 m downwind was
instrumented with integrated bag samplers and continuous C02 analyzers at 5 levels in
the vertical and out to ±30 m in the lateral. C02 gas was released from an insulated
1012 ft1 (28.7 m5) surge tank through a 4 in (0.1 m) pipe into aim3 baffled discharge
chamber, successfully releasing a steady, low momentum, nearly isothermal dense gas.
SF6 was injected into the C02 stream at 0,9 mg/s flow rate. Vertical and lateral
distributions of SF4 compare well with distributions of C02, suggesting that the SF6 was
well mixed in the COz stream. At 40 m downwind the plume clearly exhibits dense gas
behavior, with width-to-depth ratios ranging from 10 to 30. A comparison of the
continuous C02 analyzers with bag samples revealed good agreement Mass flux
estimates at the 40 m arc, excluding Test 2, averaged 90% ±1% of the released mass.
Mass flux estimates for SF6 were not in as good agreement as C02.
Preliminary comparisons of measured C02 and modeled concentrations for Tests
3 and 4 show that the SLAB and DEGADIS models tend to overpredict ground level
concentrations and underpredict the cloud depth, especially for the more stable test (Test
3). Both models are in much better agreement with observations at upper plume
elevations. For Test 4, SLAB estimates are in excellent agreement with the vertical
profile observations. Comparisons of the lateral concentration profiles and model
estimates at two heights show that both models do well at characterizing the lateral
spread of the dense cloud.
REFERENCES
Desert Research Institute and Western Research Institute, 1993: Quality Assurance
Project Plan Experiment 1: Characterization of Stable and Baseline Meteorological
Conditions Using 200 to 500 Pound Carbon Dioxide Releases. DRI Document No.
93-3305 .F2, Desert Research Institute, Reno, Nevada.
Ermak, Donald L., 1990. User's Manual for SLAB: An Atmospheric Dispersion Model
for Denser-Than-Air Releases. University of California, Lawrence Livermore
National Laboratory, Livermore, California 94550.
Golder, D., 1972, Relations Among Stability Parameters in the Surface Layer.
Boundary Layer Meteorology, 3,47-48.
Hanna, Steven R., David G. Strimaltis, and Joseph C. Chang, 1991. Evaluation of
fourteen Hazardous Gas Models with Ammonia and Hydrogen Fluoride Field Data.
Journal of Hazardous Materials, 26,127-158.
29
-------
U.S. Environmental Protection Agency, 1994. Guideline on Air Quality Models.
Supplement C. EPA-450/2-78-027R. OAQPS, U.S. EPA, Research Triangle Park,
NC 27711.
U.S. Environmental Protection Agency, 1984. Interim Procedures for Evaluating Air
Quality Models (revised). EPA-450/4-84-023, OAQPS, U.S. EPA, Research
Triangle Park, NC 27711.
30
-------
TECHNICAL REPORT DATA
1. REPORT NO.
EPA/600/A-95/055
2.
'
4. TITLE AND SUBTITLE
Controlled Experiments for Dense Gas Diffusion-
Experimental Design and Execution, Model Comparison
5.REPORT DATE
6.PERFORMING ORGANIZATION CODE
7. AUTHOR!S)
R. Egarni1, J. Bower1, W. Coulombe1, D. Freeman1, J.
Watson1, D. Sheesley2, B. King2, J. Nordin2, T.
Routh2. G. Briees3. and W. Petersen3
8.PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
'Desert Research Institute
University and Conmunity College System of Nevada
Reno, NV 69506
Western Research Institute
University of Wyoming Research Corporation
Laramie, WY 82041
'Sams as block 12
10. PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO.
12, SPONSORING AGENCY NAME AND ADDRESS
Atmospheric Research and Exposure Assessment
Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
13.TYPE OF REPORT AND PERIOD COVERED
14. SPONSORING AGENCY CODE
EPA/600/9
15, SUPPLEMENTARY NOTES
16, ABSTRACT
An Experimental baseline C02 release experiment at the DOE Spill Test Facility on the Nevada Test Site in Southern
Nevada is described. This experiment was unique in its use of C02 as a surrogate gas representative of a variety of
specific chemicals. Introductory discussion places the experiment in historical perspective, C02 was selected as a
surrogate gas to provide a data base suitable for evaluation of model scenarios involving a variety of specific
dense gases. Releases were conducted under baseline conditions including a simulated "evaporating pool" release over
flat unobstructed terrain. The experiment design and setup are described, including design rationale and quality
assurance methods employed. Design conditions included moderately low wind speed, stable atmospheric conditions.
Four releases were performed, two of which were during near-neutral conditions and two during slightly stable
conditions. Resulting experimental data are suumarized. These include C02 cloud characteristics measured at 40 ns
downwind from the release point. Experiment success and effectiveness is discussed in terms of mass balance
analyses. For Test 1, 3, and 4 the measured mass accounted for at least 901 of the released mass. Measured values
for Test 2 accounted for only 60Z of the released mass. Data usefulness is examined through a preliminary comparison
of experimental results with simulations performed using the SLAB and DEGADIS dense gas models.
17. KEY WORDS AMD DOCUMENT ANALYSIS
a. DESCRIPTORS
b.IDENTIFIERS/ OPEN ENDED TERMS
c.COSATI
18. DISTRIBUTION STATEMENT
Release to oublic
19. SECURITY CLASS (This Renortl
Unclassified
21.NO. OF PAGES
20. SECURITY CLASS (This Fane)
Unclassified
22. PRICE
------- |