EPA 55O-B99-O17
TECHNICAL BACKGROUND DOCUMENT FOR OFFSITE
CONSEQUENCE ANALYSIS FOR
ANHYDROUS AMMONIA, AQUEOUS AMMONIA, CHLORINE, AND
SULFUR DIOXIDE
Chemical Emergency Preparedness and Prevention Office
U.S. Environmental Protection Agency
-------
Prepared by:
Geoffrey D. Kaiser, Joseph D. Price, and Jos6 Urdaneta
Science Applications International Corporation
11251 Roger Bacon Drive
Reston, Virginia 20190
-------
Background Document for Offsite Consequence Analysis, April 1999
TABLE OF CONTENTS
CHAPTER 1: INTRODUCTION 1-1
1.1 BACKGROUND 1-1
1.2 PURPOSE • 1-1
1.3 GENERAL APPROACH • , 1-1
1.4 ASSUMPTIONS • - 1-2
CHAPTER 2: UNCERTAINTIES AND MODELS USED 2-1
2.1 ANHYDROUS AMMONIA 2-1
2.7.7 SAIC Proprietary Model 2-1
2.1.2 USEPA RMP Offsite Consequence Analysis Guidance (OCAG) 2-3
2.1.3 TFI 2-3
2.1.4 DNV-UDM-Technica 2-5
2.1.5 AWWARF Approach 2-6
2.1.6 Additional ALOHA Run 2-6
2.1.7 Comparison with Available Data 2-6
2.1.7.1 Data from Accidents 2-6
2.1.7.2 Experimental Data-Desert Tortoise 2-9
2.1.7.3 Data from Modeling 2-10
2.1.7.4 Interpretation of Figures 2-2 and 2-3 ; ......2-10
2.1.7.5 Choice of a Single Curve for AR and WWTP Guidance 2-11
2.1.7.6 10-Minute vs. 60-Minute Releases 2-11
2.7.5 Anhydrous Ammonia-Urban Site, Worst-Case 2-75
2.7.P Anhydrous Ammonia—Alternative Scenarios -2-73
2.2 AQUEOUS AMMONIA • 2-17
2.3 CHLORINE 2-17
2.3.7 Worst-Case Scenarios 2-17
2.3.2 Alternative Scenario 2-23
2.4 SULFUR DIOXIDE 2-23
2.5 BACKGROUND DESCRIPTION OF SENSITIVITY STUDIES 2-24
2.5.7 Dry Deposition - • • 2-24
2.5.2 Puff Releases 2-25
2.5.3 Qualitative Uncertainties and Conservatisms...:. 2-25
2.5.3.1 Duration of Worst-Case Weather Conditions 2-25
2.5.3.2 Pooling 2'25
2.5.3.3 Time Varying Toxic Endpoints 2-25
2.5.4 Conclusion-Sensitivity Studies 2-26
CHAPTERS: GASES LIQUEFIED UNDER PRESSURE .— 3-1
CHAPTER 4: ADJUSTMENT OF MEAN CONCENTRATION FOR AVERAGING TIME 4-1
CHAPTER 5: AMMONIA/MOIST-AIR THERMODYNAMICS .'. 5-1
5.1 CALCULATION OF THERMODYNAMIC PROPERTIES OF MIXTURES OF
AMMONIA AND AIR 5'1
5.7.7 Methods. S~J
5.1.1.1 Physical Property Data.. 5"2
5.1.1.2 Algorithm for Determination of Final Cloud Conditions 5-5
5.7.2 RESULTS -. 5~10
5.2 EFFECT ON PREDICTION OF DISTANCES TO TOXIC ENDPOINT 5-12
5.3 POTENTIAL FOR LIFT-OFF 5-13
11
-------
Background Document for Offsite Consequence Analysis, April 1999
CHAPTER 6: EFFECT OF AMMONIA RELEASES ON STRUCTURES 6-1
6.1 PROLONGED RELEASES 6-1
6.L1 Building Structural Response. 6-2
6.1.2 Building Attenuation of Release 6-2
6.2 SUMMARY OF CONCLUSIONS 6-3
REFERENCES...
ill
-------
Background Document for Offsite Consequence Analysis, April 1999
LIST OF FIGURES
Page
Figure 2-1. Sensitivity Studies for Worst-Case Anhydrous Ammonia Scenarios - Predicted
Distances to Toxic Endpoint, Rural Site, Atmospheric Stability F, Windspeed
1.5 m/s 2-4
Figure 2-2. Ammonia Dispersion Data Accidents and Desert Tortoise Experiments 2-8
Figure 2-3. Ammonia Dispersion Data (Enlarged) 2-12
Figure 2-4. Sensitivity Studies For Worst-Case Anhydrous Ammonia Scenarios Predicted
Distances To Toxic Endpoint, Urban Site, Atmospheric Stability F,
Windspeed 1.5 m/s 2-14
Figure 2-5. Sensitivity Studies for Alternative Anhydrous Ammonia Scenarios Predicted
Distances to Toxic Endpoint, Rural Site, Atmospheric Stability Category D,
Windspeed 3 m/s 2-15
Figure 2-6. Sensitivity Studies for Alternative Anhydrous Ammonia Scenarios Predicted
Distances To Toxic Endpoint, Urban Site, Atmospheric Stability Category D,
Windspeed 3 m/s 2-16
Figure 2-7. Sensitivity Studies for Worst-Case Aqueous Ammonia Scenarios Predicted
Distances to Toxic Endpoint, Rural Site, Atmospheric Stability F,
Windspeed 1.5 m/s 2-18
Figure 2-8. Sensitivity Studies For Worst-Case Aqueous Ammonia Scenarios Predicted
Distances to Toxic Endpoint, Urban Site, Atmospheric Stability F,
Windspeed 1.5 m/s 2-19
Figure 2-9. Sensitivity Studies for Alternative Aqueous Ammonia Scenarios Predicted
Distances to Toxic Endpoint, Rural Site, Atmospheric Stability D, Windspeed
3 m/s • 2-20
Figure 2-10. Sensitivity Studies for Alternative Aqueous Ammonia Scenarios - Predicted
Distances to Toxic Endpoint, Urban Site, Atmospheric Stability Category D,
Windspeed 3 m/s : 2-21
Figure 2-11. Sensitivity Study Predicted Distances To Toxic Endpoint For Chlorine.... 2-22
Figure 3-1. Fraction of Liquid Chlorine Falling to the Ground as a Function of Superheat 3-2
Figure 4-1. Illustration of Meandering 4-3
. Figure 5-1. Algorithm for Determination of Final Cloud Conditions for Mixing of
Ammonia and Moist Air Clouds 5-6
Figure 5-2. Algorithm for Calculation of Dew Point Pressure 5-10
\ •
IV
-------
Background Document for Offsite Consequence Analysis, April 1999
LIST OF TABLES
Table 2-1. Maximum Center Line Concentrations Measured in the Desert Tortoise
Experiments (ppm) 2-9
Table 2-2. Comparison of Worst-Case Hazard Assessments for Anhydrous Ammonia 2-10
Table 2-3. Examples of AWWARF and SACRUNCH/AR Predictions Worst Case,
Urban Site 2-13
Table 2-4. Ratios of AWWARF Alternative Case Predictions 2-13
Table 2-5. Distances to Toxic Endpoint (ft)--Sensitivity Studies for Chlorine 2-23
Table 4-1. Example of Effect of Meandering of Anhydrous Ammonia Releases, Worst-
Case, Rural Conditions 4-4
Table 4-2. Example of Effect of Meandering of Anhydrous Ammonia Releases, Worst-
Case Urban Conditions 4-5
Table 5-1. Comparison of Measured and Predicted Vapor Pressure of Water 5-4
Table 5-2. Comparison of Measured and Predicted Vapor Pressure of Ammonia 5-4
Table 5-3. Comparison of Measured and Predicted Vapor Pressures Above
Ammonia/Water Solutions 5-4
Table 5-4. Heat Capacity and Heat of Vaporization Data 5-5
Table 5-5. Final Cloud Conditions for a Worst-Case Scenario, Two-Phase Release with
Moist Air* 5-11
Table 5-6. Final Cloud Conditions for a Worst-Case Scenario, Two-Phase Release with
Dry Ak* 5-11
Table 5-7. Final Cloud Conditions for All Vapor Release with Dry Air*.. 5-12
Table 5-8. Example of the Effect of New Thermodynamic Model and Meandering of
Anhydrous Releases, Worst-Case, Rural Conditions, 75% RH 5-12
Table 5-9. Illustration of the Potential for Lift-Off. 5-13
Table 6-1. Ten-Minute Building Release Attenuation Factors for Continuous Releases
of Ammonia 6-4
Table 6-2. Ten-Minute Building Release Attenuation Factors for Prolonged Releases of
Chlorine and Sulfur Dioxide 6-5
-------
Background Document for Offsile Consequence Xnafyses, May {999
CHAPTER 1: INTR6DUCTION
1.1 BACKGROUND
Following the publication of the US Environmental Protection Agency's (EPA's) Risk
Management Program (RMP) regulations, 40 CFR Part 68, EPA developed generic guidance for
the offsite consequence analyses required by the regulation. This document, RMP Offsite
Consequence Analysis Guidance (OCAG), is intended to provide simple methods and reference
tables for determining distances to toxic and flammable endpoints for worst-case and alternative
release scenarios. The generic approach is based on parameters required by the rule and on
conservative assumptions about other conditions and may not reflect site-specific conditions.
Use of the guidance is not required; facilities may conduct their own air dispersion modeling,
provided that they use the parameters specified in the rule and a model appropriate for the
substance.
EPA also developed industry-specific guidance for ammonia refrigeration (AR) and
wastewater treatment plants (WWTPs). In developing these documents, EPA conducted
chemical-specific modeling for anhydrous ammonia, aqueous ammonia, chlorine, and sulfur
dioxide, including consideration of liquid droplet formation (except in the case of aqueous
ammonia). This chemical-specific modeling was incorporated into the OCAG. The modeling for
these four toxic substances is different from, and less conservative than, the generic modeling that
applies to other regulated substances covered in the OCAG.
1.2 PURPOSE
The purpose of this document is to provide the technical background of the methodology and
assumptions used to develop the chemical-specific tables.
1.3 GENERAL APPROACH
Modeling the consequences of large-scale accidental releases of toxic vapors involves many
uncertainties. These uncertainties may arise from the capability of different models to describe
the physical phenomena, the selection of input parameters, and the lack of data to validate the
models. When the same inputs are used, different models may produce widely varying results;
the same model may also produce widely varying results if the input parameters are varied across
then* range of uncertainty. The range of predicted distances can be as much as a factor of 10.
The modeling conducted to develop the chemical-specific tables differs from the modeling
for the generic tables found in the OCAG in the following ways:
(1) Models developed by SAIC (referred to as SACRUNCH and SADENZ) were used rather
than SLAB. Sensitivity analyses were conducted using various models, experimental data, and
accident data to evaluate the reasonableness of the results: Chapter 2 provides the results of these
analyses, which illustrate the range of outcomes possible when performing analyses of the type
required by EPA. Because SACRUNCH, SADENZ, and SAPLUME (1994) are proprietary
1-1 .
-------
Background Document for Offsite Consequence Analyses, May 1999
dispersion models, and thus not readily available for review, some information about these
models is provided in Appendix A.
(2) Liquid anhydrous ammonia, chlorine, and sulfur dioxide are frequently stored as gases
liquefied under pressure. In the OCAG, for the worst-case release, gases liquefied under
pressure are assumed to behave similarly to gases. Based on relevant studies, absent obstacles,
liquid anhydrous ammonia, chlorine, and sulfur dioxide released at typical ambient temperatures
are assumed to become and remain airborne as a mixture of vapor and fine: liquid droplets and,
for the purposes of RMP, can be modeled as a gas. Chapter 3 discusses this issue.
(3) The effect of averaging tune on plume spread was considered and a method for adjusting
the predicted mean concentration for averaging time developed. Chapter 4 discusses this issue.
(4) The thermodynamics of mixtures of moist air and anhydrous ammonia were analyzed
using the techniques reported by Wheatley (1987). (See Chapter 5, which also discusses whether
the ammonia/moist air rnixing will generate enough heat to cause the plume to become buoyant.)
(5) For scenarios in which the release from a vessel is indoors, the effect of hold-up of vapors
within a building has been incorporated into the industry-specific models, but not the OCAG,
which uses a simpler approach. Chapter 6 discusses this issue.
1.4 ASSUMPTIONS
As previously mentioned, the RMP rule requires that certain parameters be used in the offsite
consequence analysis modeling (40 CFR 68.22). The analyses presented in this document use
these required assumptions.
• For anhydrous ammonia, chlorine, and sulfur dioxide (i.e., gases liquefied under
pressure), the worst-case scenario consists of the sudden release of the whole contents of
the largest vessel or pipeline. For the purposes of the modeling, it is assumed that the
release is spread over 10 minutes, whether the release is outside or inside a building.
• The worst-case weather conditions consist of Atmospheric Stability Category F, with a
windspeed of 1.5 m/s, unless it can be shown that such conditions have not occurred at
the site during the past three years.
• The toxic endpoints are 200 ppm for ammonia, 3 ppm for chlorine., and 3 ppm for sulfur
dioxide, irrespective of the duration of exposure. EPA is currently developing Acute
Exposure Guideline Levels (AEGLs), which will consist of different values of toxic
endpoint for a number of exposure times. However, until the AEGLs have been
published and the rule has been changed, toxic endpoints are fixed,,
1-2
-------
Background Document for Offsite Consequence Analyses, May 1999
CHAPTER 2: UNCERTAINTIES AND MODELS USED
Developing offsite consequence analysis guidance that is simple and easy to use, yet
scientifically defensible, is difficult because a large range of uncertainty exists for predictions of
distances to the toxic endpoint. For example, 40 CFR Part 68 requires that worst-case modeling
be carried out assuming atmospheric stability category F and a windspeed of 1.5 m/s. However,
very few experimental data exist for these weather conditions with which to validate models. To
develop an understanding of the plausible uncertainty range, a comparative study was conducted
in which the same input parameters were used in the different available models, release rate
varied, and the outputs compared. The following is a description of analyses performed to
support the reference tables and provides an explanation of how reference tables, plots, and
formulas were selected from within the range of possibilities. Anhydrous ammonia is discussed
first because, for this particular chemical, there are many examples and calculations available
from which to develop an understanding of the range of uncertainties.
2.1 ANHYDROUS AMMONIA
In many parts of a typical refrigeration system, ammonia is liquefied under pressure. If the
pressure and temperature are sufficiently high, and if there is a sudden release of ammonia, it
will become and remain airborne as a mixture of vapor and very fine liquid droplets that do not
fall to the ground. The droplets evaporate quickly cooling the air so that a cold mixture of air
and ammonia vapor is formed. The mixture is initially denser than air.
The comparative study was conducted for a worst-case scenario release of anhydrous
ammonia at a rural site. The toxic endpoint for ammonia as specified in the RMP Rule is 0.14
mg/L (200 ppm). For the purposes of the RMP, this is a fixed value no matter the duration of
release. The worst-case weather conditions consist of Atmospheric Stability Category F, with a
windspeed of 1.5 m/s. The worst-case scenario consists of an outdoor, sudden release of the
whole contents of the largest vessel or pipeline. For the purposes of this comparative study, the
worst-case release was varied from 1,000 to 400,000 Ibs. It is assumed that the release is spread
over 10 minutes, therefore, the rate of release varied from 100 to 40,000 lbs./min.
, Figure 2-1 displays several different answers to the question: "For worst-case scenarios at a
rural site, what is the predicted distance to the toxic endpoint as a function of the rate of release
of anhydrous ammonia?" The various models used to prepare Figure 2-1 are described below.
[On the tables and plots in this chapter, it was sometimes necessary to extrapolate data presented
' by other authors. This was done by assuming a linear relationship between distance and release
rate on a log-log plot.]
2.1.1 SAIC Proprietary Model
The May 1996 draft guidance for ammonia refrigeration (USEPA, 1996b) made use of two
SAIC proprietary computer models—SACRUNCH, which is suitable for the modeling of
ground-level, horizontal releases of denser-than-air vapors, and a companion model, SADENZ,
for denser-than-air puffs. These models are described in SAIC (1994), and a summary is
2-1
-------
Background Document for Offsite Consequence Analyses, May 1999
provided in Appendix A. These models were used because they allow the easy use of sensitivity
studies, including the phenomenon of dry deposition, a highly effective mechanism for depleting
clouds of reactive gases such as ammonia, and because they allowed the easy insertion of an
ammonia/moist air thermodynamics module.
1. The models are 2-D "box" models with gravitational slumping, edge entrainment, and top
entrainment of air given by simple but reasonably well-established, formulas in the initial,
heavier-than-air phase. The model was essentially "tuned" by comparison with the
Thorney Island experiments (McQuaid, 1986).
2. The model finally evolves from being denser-than-air to being neutrally buoyant when
Ap/pa = 0.001, where pa is the density of ah* and Ap is the difference between the density
of the vapor cloud (averaged across a continuous-plume cross section or throughout a
puff) and the density of the surrounding air1.
3. In the neutrally buoyant phase, the models evolve to the "Green Book" horizontal and
vertical standard deviations of cry and crz (i-e., the rural or urban parameterizations
proposed by Briggs (1973a) and reproduced in the "Green Book." The "Green Book" is
EPA's Technical Guidance for Hazards Analysis (USEPA, 1987), which contains a
Gaussian model with vertical and horizontal standard deviations from Briggs (1973a).
4. For ammonia, chlorine, and sulfur dioxide released from the liquid space of vessels in
which they are liquefied under pressure at 25 °C or at the highest daily temperature, it is
assumed that a portion of the released liquid immediately flashes to vapor (e.g., 20
percent). The user calculates the percentage outside the model from thermodynamic
principles. The remaining liquid atomizes and remains airborne. See Chapter 3 for
justification of this assumption. See Chapter 6 for a discussion of how buildings could
mitigate this effect.
5. SACRUNCH makes a simplifying assumption: the turbulence generated by the flash
atomization process is such that, almost immediately, the mass mixing ratio is 10 (i.e., the
ratio of entrained air to airborne ammonia, chlorine, or sulfur dioxide mass is 10). For
anhydrous ammonia, the density and temperature of this mixture are calculated using the
ammonia/moist air thermodynamic model described in Chapter 5. For chlorine and
sulfur dioxide, the mixture is assumed to be ah" and Ck vapor or air and SC>2 vapor,
respectively, at their atmospheric boiling points. The initial horizontal momentum of the
escaping liquid jet and the entrained air is conserved to define the initial conditions for
SACRUNCH. A similar assumption is made for an instantaneous puff release in
SADENZ. The predictions of the model at distances at which the toxic endpoints of Ck
or SO2 are encountered (3ppm) are not sensitive to this assumption, although it does
mean that predictions near the source may not be accurate.
1 Some reviewers criticized this assumption because it is a simpler transition criterion than is found in other models.
However, as is'shown in Appendix A, the models do a reasonable job of fitting the large-scale experimental data-
bases. They evolve in the far field into a well-established Gaussian model with well-known standard deviations
provided by Briggs (1973a). In addition, sensitivity studies (not shown here) indicate that the results do not change
significantly when the Ap/p, criterion is varied between 0.01 and 0.001.
_ __
-------
Background Document for Offsite Consequence Analyses, May 199?
6. In SACRUNCH, the toxic endpoint is compared to the peak centerline concentration,
which, for worst-case scenarios, is assumed to jump up to that value when the puff
arrives and to remain constant for exactly 10 minutes, independent of location. In
SADENZ, the model calculates the average centerline concentration over the duration of
cloud passage. This duration is a function of distance downwind.
7. SACRUNCH and SADENZ have simple dry deposition modeling algorithms
(see Section 2.5.1). None of the other computer programs discussed herein have these
capabilities, which is one of the reasons the authors consider it useful to use
SACRUNCH and SADENZ.
8. There is an issue concerning the use of models such as SACRUNCH with high surface
roughness length (this issue is discussed below in some detail in the context of the use of
DEGADIS). The concern is that, at a truly urban site, a heavy vapor will flow in among
the obstructions on the surface (e.g., large buildings) and will not be exposed to the
turbulence in the atmosphere above those obstructions. For the present work, it has been
assumed that, while in the denser-than-air phase, the surface roughness length is 10 cm at
both urban and rural sites. When Ap/pa < 0.001, the model is a Gaussian one in which
ay and az are different for urban and rural sites. This approach should be somewhat
conservative for the urban site.
Three sensitivity studies are shown on Figure 2-1:
A. A conservative case, in which SACRUNCH defaults into the "Green Book" rural
dispersion model in the far field, when the initial denser-than-air behavior has been
"forgotten".
B. A case in which a dry deposition velocity of 1 cm/sec has been used. See Section 2.5.1
for further discussion of dry deposition. The authors also looked at a case in which the
dry deposition velocity was 0.3 cm/sec, but that case is not reproduced on Figure 2-1.
C. A case in which the puff model SADENZ has been used.
2.1.2 USEPA RMP Offsite Consequence Analysis Guidance (OCAG)
The OCAG (USEPA, 1996a) was developed using the SLAB model (Ermak, 1989). It is
intentionally conservative. The distances are obtained simply by reading from tables provided in
the OCAG. The nearest entry in the OCAG table that is conservative is the one that is chosen.
In addition, a second OCAG curve has been provided - one that has been interpolated between
the discrete values of release rate and toxic endpoint that are given hi the OCAG lookup tables.
This gives somewhat less conservative predictions.
2.1.3 TFI
The Fertilizer Institute (TFI) has produced its own guidance on large-scale releases of
anhydrous ammonia. TFI used the DEGADIS model (USEPA, 1989), with its transient option.
In this option, the initial ten-minute "slug" of ammonia gradually evolves into a puff as it travels
2-3
-------
Background Document for Offsite Consequence Analyses, May 1999
100.00
10.00
1,00
0,10
WORST-CASE
RURAL
ANHYDROUS AMMONIA
SACRUNCH, CASE B
PUFF.SADENZ.CASE C
OCA-geneHo
-^-AWWA
-*-TFI
DNV-UDM
ALOHA
-SACRUNCH CONSERV, CASE
A
1000 10000
Rate Of Release (Ibs/min)
100000
Figure 2-1. Sensitivity Studies for Worst-Case Anhydrous Ammonia Scenarios - Predicted
Distances to Toxic Endpoint, Rural Site, Atmospheric Stability F, Windspeed 1.5 m/s
downwind because there is along-wind lengthening of the slug due to the action of atmospheric
turbulence. This lengthening effect is most dramatic in Atmospheric Stability Category F
conditions with a low windspeed. The transient release model that TFI has used is conceptually
realistic.
Note that TFI uses roughness lengths of 3 cm and 1 m to characterize rural and urban areas,
respectively. There are two potential concerns about this:
1. The authors of DEGADIS have previously expressed the opinion that DEGADIS should
not be used with surface roughness lengths in excess of 10 cm. This issue was
extensively discussed in 1990/91 during the South Coast Air Quality Management
Districts rulemaking on hydrogen fluoride storage and use (SCAQMD, 1991a,b).
SCAQMD states:
"The slumping and stably stratified flow characteristic of dense gas releases
produces dense gas plumes that have height scales significantly less than the
height of the atmospheric boundary layer. The surface roughness parameter is
used by the models to characterize the dense gas vertical dispersion. The
mathematical concept of the surface roughness parameter dictates the use of a
value that is much less than the height scale of the dispersing cloud. This is not a
problem when simulating dense gas dispersion in a desert environment, but it
becomes more complicated when applying the models in urban areas.
2-4
-------
Background Document for Offsite Consequence Analyses, May 1999
"Since dense gas models have been developed to simulate test releases conducted
in non-urban (desert) settings, most models, including DEGADIS, are designed to
simulate rural dispersion -when the cloud enters the passive phase. A surface
roughness value characteristic of urban scale roughness elements is
inappropriate unless the dense gas cloud height is approximately 30 times the
height of the surface roughness elements. Until the issue of extrapolating the use
of the models from non-urban settings to urban settings is better understood, a
surface roughness value of 0.1 meter shall be used as the input to DEGADIS for
the entire transport and dispersion calculation. "
This limitation on the use of DEGADIS was supported by one of the original authors of
DEGADIS, Jerry Havens, in testimony to SCAQMD.
2. When moving from a rural to an urban area, increased intensity of atmospheric
turbulence arises from two sources, mechanical (due to the presence of buildings) and
convective (due to the presence of large heat sources). It is questionable whether a model
in which changed surface roughness alone is responsible for the enhanced intensity of
turbulence at urban sites correctly characterizes the physics of the situation (this comment
also applies to SLAB).
2.1.4 DNV-UDM-Technica
The work presented by Woodward (1998) is of considerable interest because the model used
is based on experimental data obtained at very low windspeeds in stable weather conditions.
There are few such data available for any hazardous vapor; DNV made use of a database of
large-scale propane releases (Heinrich et al., 1988/1989). The releases ranged in size from a few
hundred kilograms to several thousand kilograms, and the duration of release varied from
40 seconds to 600 seconds. It is pertinent to try to understand why the DNV predictions on
Figure 2-1 are relatively low.
1. The model that DNV "tuned" based on the TUV experiments is known as UDM
(Unified Dispersion Model). It has considerable merit because, as noted above, it was
actually based on experiments at low windspeeds in stable weather conditions. However,
the appreciably lower predictions of the UDM model in Figure 2-1 are, in part, due to an
assumption about averaging. Basically, the author appears to have divided the model's
predicted concentrations by a factor of six to take account of the 10-minute duration of
release, whereas the ammonia toxic endpoint of 200 ppm is valid for an exposure time of
60 minutes. This amounts to assuming that Haber's law is valid for ammonia. As noted
above, such exposure tune-dependent relationships for toxic endpoints are not permitted
under the current rule.
2. The principal aim of the original TUV papers from the Journal of Hazardous Materials
(Heinrich et al., 1988; 1989) was to examine the lower flammable distance (LFD)
(i.e., the distance to the lower flammable limit, which is 2.1 v% [~ 20,000 ppm] for
propane). Experimental measurements were taken down to concentration levels of a few
thousand ppm. To make predictions for ammonia at 200 ppm, extrapolations of more
than an order of magnitude are required. Therefore, it should be noted that the
2-5
-------
Background Document for Off site Consequence Analyses, May 1999
experimental results upon which the "tuning" of the UDM models is based are strictly
near-field results and do not provide information about concentrations at or near the toxic
endpoint of ammonia. (UDM is no different from any of the other models in this respect;
however, it is also true to say that it is no better than the others, either.)
3. Two types of instruments were used to record propane concentration - "catalytic-type"
instruments (details not given) that were regularly distributed across the field, and
infrared (IR) spectrometers that used the 3.7 m propane absorption band for detection. It
turned out that, in the original publication (Heinrich et al., 1988), the IR measurements
were incorrectly interpreted because the results were distorted by the presence of ice
crystals, which led to considerable overestimates of the LFDs. These overestimates were
corrected in 1989 (Heinrich et al., 1989).
Woodward points out inconsistencies between the readings of the catalytic sensors and
the IR sensors hi experiments in which the rate of release and other conditions were
nearly identical and, on this basis, states that the IR results are preferable. The IR results
appear to be generally lower than the catalytic sensor results, presumably biasing the
tuning of the UDM model towards lower predicted distances. It would seem that caution
is advisable in ignoring one set of results just because the experimental fluctuations
appear to be large, while accepting another set of results that required major post facto
corrections.
2.1.5 AWWARF Approach
The American Water Works Association Research Foundation (AWWARF, 1998) approach
is based on the ALOHA model (NOAA and USEPA, 1995) and is provided here as a
representative application of that computer model. It is likely that many facilities that do their
own modeling will use ALOHA.
2.1.6 Additional ALOHA Run
When preparing the RMP Guidance for Ammonia Refrigeration (USEPA, 1998), EPA
engaged hi continuous dialog with the International Institute of Ammonia Refrigeration (IIAR).
Early in 1996, one of IIAR's consultants provided an ALOHA output2, which is also shown on
Figure 2-1.
2.1.7 Comparison with Available Data
2.1.7.1 Data from Accidents
No data set (or sets) was identified that unequivocally distinguishes among all of the models
on Figure 2-1. However, there are sufficient data available to make some judgments about
where to place reasonably conservative guidance.
Markham (1986) provides an instructive review of the consequences of quite a large number
of accidents that have resulted in the release of anhydrous ammonia. The results of Markham's
'• IIAR, Private Communication, March 1996
__ — —
-------
Background Document for Offsite Consequence Analyses^ May 1999
work are summarized on Figure 2-2, which is a little complicated, but which is worth further
study. The letters A, B, MM identify the estimated concentrations from a list of 12 accidents
that Markham analyzed. The lengths of the bars represent the uncertainty in recovering data
from accident descriptions. Note that these bars do not represent actual measured
concentrations. They represent post-facto reconstructions from observations of effects on plants
and birds.
Markham defines the releases as follows:
• A, B, C and D-45-ton release of cold product (presumably refrigerated) in approximately
30 minutes, Bainesville, MN (6/10/81)
• E-puncture in 82-ton railcar, ambient temperature, Belle, WV (1/21/70)
• F and G-160 tons of cold product over 2D hours, Blair, NE (11/16/70)
* H, J and K-truck train collision, 18 tons of anhydrous ammonia at ambient temperature
released in a few minutes, Boitte, LA (12/15/70)
• L, M and N-pipeline rupture, 230 tons at ambient temperature released in under eight
hours, Conway, KS (12/6/73)
• O-160 tons released instantaneously, presumably ambient temperature, Crestview, FL
(4/8/79) <
• P-70 tons from train wreck, ambient temperature, Crete, NE (2/18/69)
• Q and R-pipeline rupture, 400 tons over four hours, ambient temperature, Enid, OK
(5/7/76)
• U, V and W-train wreck, 50 tons rapidly released, ambient temperature, Pensacola, FL
(11/9/77)
• X, Y, Z, AA, BB and CC-bullet tank failure, instantaneous release of 30 tons, ambient
temperature, Potchefstroom, South Africa (7/13/73)
• DD and EE-railcar failure, 75-ton release, ambient temperature, Verdigis, OK (6/10/79)
• FF, GG, HH, II, JJ, KK, LL and MM-19 tons instantaneous release from a tank truck,
ambient temperature, Houston, TX (5/11/76)
Thus, most of the data from accidents on Figure 2-2 are from spills of anhydrous ammonia at
ambient temperature, with two releases of refrigerated ammonia. Markham does not specify the
weather conditions associated with each specific release.
2-7
-------
Background Document for Offsite Consequence Analyses, May 1999
1000000
100000
10000
209,00
1000
10
100 1000 100010
Distance From Spill Source (rreters)
100000
Figure 2-2. Ammonia Dispersion Data Accidents and Desert Tortoise Experiments
2-8
-------
Background Document for Offsite Consequence Analyses, May 1999
2.1.7.2 Experimental Data—Desert Tortoise
The solid curves on Figure 2-2 are data from the so-called "Desert Tortoise" (DT) large-scale
experimental releases of flashing liquid ammonia, which were carried out at the Nevada Test Site
(Goldwire et al., 1985). The release rates in the experiments were about 100 kg/sec (220 Ib/sec,
13,200 Ib/min), and the durations of release were a few minutes. The masses released in the four
DT experiments were ~ 24,500 Ib, 66,000 Ib, 50,000 Ib, and 90,000 Ib in atmospheric stability
classes D, D, D, and E, respectively, with windspeeds 7.42 m/s, 5.76 m/s, 7.38 m/s, and 4.5 m/s,
respectively. The surface roughness length was 0.003 m. The actual data points are reproduced
on Table 2-1. No estimates of experimental error were provided by Goldwire et al.
Some pertinent observations are as follows:
1. The solid curves connecting the Desert Tortoise data points on Figure 2-1 are drawn by
eye to connect the points. They are not intended to be model fits to the data.
2. The Desert Tortoise data points themselves are peak concentrations taken from plots in
the Desert Tortoise Series Data Report (Goldwire et al., 1985). Comparisons with some
other publications show the folio whig: (1) in a comparison with the numerical computer
model, FEM3, Chan et al. (1987), use almost the same concentrations as in Table 2-1 for
DT4; (2) Spicer et al. (1987) use 75,000, 21,000, and 5,000 ppm for the three DT4
measurements at 100, 800, and 2,800 m, respectively, in DT4, also close to the values
given in Table 2-1. Therefore, there is precedent for the interpretation of the Desert
Tortoise data in the way they are presented in Table 2-1.
Table 2-1. Maximum Center Line Concentrations Measured in the
Desert Tortoise Experiments (ppm)
Distance
Downwind (m)
100
800
1,400
2,800
3,500
5,600
DTla
50,000
10,000
—
—
650
150
DT2b
80,000
15,000
5,000
—
—
—
DT3C
80,000
12,500
—
600
—
—
DT4d
65,000
17,500
—
5,000
—
—
"24,500 Ib over 2 minutes (12,250
b66,000 Ib over 4 minutes (16,500
C50,000 Ib over 3 minutes (16,700
d90,000 Ib over 6 minutes (15,000
Ib/min), stability category D, windspeed 7.42 m/s
Ib/min), stability category D, windspeed 5.76 m/s
Ib/min), stability category D, windspeed 7.38 m/s
Ib/min), stability category E, windspeed 4.51 m/s
3. Beyond 800 m, the ammonia concentration was measured by portable sensor stations.
These data should be regarded as less reliable than those taken at 800 m and 100 m
(with a full range of stationary instruments), but, nonetheless, do provide information that
is helpful when making judgments.
2-9
-------
Background Document for Of/site Consequence Analyses, May 1999
2.1.7.3 Data from Modeling
Refer to Table 2-2 for a brief summary of some of the modeling data plotted on Figure 2-1.
The predicted distance to the toxic endpoint is given for five discreet total masses, 10,000,
50,000,100,000,150,000 and 200,000 Ib, respectively. Note that, for OCAG, two columns are
presented. One, "without interpolation," consists of reading the predicted distances from the
OCAG tables using the nearest conservative value that is directly tabulated therein. The other,
"with interpolation," involves interpolating between values in the OCAG tables, assuming linear
relationships on log-log plots. The AR data is the SACRUNCH, Case B from Figure 2-1 and is
close to the OCAG guidance published in 1996.
The data from Table 2-2 are plotted on Figure 2-2. They should all lie: along the 200-ppm
line, but have been broken apart for greater clarity. For each model, the points labeled 1,2, 3,4,
and 5 correspond to total mass released of 10,000, 50,000,100,000,150,000, and 200,000 Ib,
respectively. A portion of Figure 2-2 has been enlarged on Figure 2-3.
Table 2-2. Comparison of Worst-Case Hazard Assessments for Anhydrous Ammonia
Total Mass
Released (ib)
10,000
50,000
100,000
150,000
200,000
PREDICTED DISTANCE TO TOXIC ENDPOINT (m)
QCAG"
4,800f
9,200
12,000
14,300
16,000
5,800s
11,000
15,000
17,700
19,300
ARb
2,900
6,500
9,400
11,600
13,500
AWWARF*
2,700
5,300
7,200
8,500
9,700
TFId
1,200
2,400
3,400
4,200
4,700
DNVe
720
1,500
2,300
3,100
3,600
•OCAG
"AR
CAWWARF -
dTFl
CDNV
fOCAG with interpolation
*OCAG without interpolation
Offsite Consequence Analysis Guidance
Risk Management Program Guidance for Ammonia Refrigeration
American Water Works Association Research Foundation
Fertilizer Institute
Det Norske Veritas-Technica
2.1.7.4 Interpretation of Figures 2-2 and 2-3
Recognizing that there is great uncertainty in the data on Figures 2-2 and 2-3, it is
nevertheless pertinent to try to come to some tentative conclusions.
• Bars F and G represent the farthest observed distance to which accidental releases of
ammonia have been seen to generate vapor clouds hi the 1,000 to 10,000 ppm range,
namely about 2,000 m. The data from Markham represent a prolonged release of
refrigerated ammoniaDl60 tons over 22 hours, or about 2,000 Ib/min. This is the release
rate that would be expected from an RMP worst-case release of 20,000 Ib, although the
comparison is not quite apt because bars F and G come from a steady-state release,
whereas the worst-case release is transient. Nonetheless, one would expect worst-case
releases with larger release rates than 2,000 Ib/min (e.g., points 2,3,4, and 5 with release
rates of 5,000,10,000,15,000, and 20,000 Ib/min, respectively) to be farther to the right
2-10
-------
Background Document for Offsite Consequence Analyses^ May 1999
beyond bars F and G along the 200 ppm level. Therefore, at the 200-ppm level, the
largest releases (e.g., 100,000 Ib, representative of a railcar-sized release) ought to give
predicted distances considerably in excess of this. For DNV and TFI, the predicted
distances for a 100,000-lb release are 2,300 and 3,400 m, respectively. From this
perspective, they seem a little low.
• The Desert Tortoise experiments have provided some data, albeit uncertain, in the 100 to
1,000 ppm range. These data were taken in Stability Categories D and E, with
windspeeds considerably in excess of 1.5 m/s. If experiments had been performed in
Atmospheric Stability Category F, with a windspeed of 1.5 m/s, the distances would be
expected to increase. The surface roughness length at the Desert Tortoise was 0.003 cm,
characteristic of a very smooth rural site (e.g., TFI gives rural terrain a surface roughness
length of- 0.03 m, while DNV gives a surface roughness length of 0.003 m). Therefore,
it seems reasonable that model predictions for releases of about the size of Desert
Tortoise releases should propagate somewhat farther than do the Desert Tortoise curves
on Figures 1 and 2. Both the DNV and TFI models give predictions that seem a little low
from this perspective (e.g., DTI shows a 150-ppm result at 5,600 m).
• In the 10 to 100 ppm range, the worst-case data from accident scenarios propagates out to
about 12,500 m. Assuming that the worst-case accident data come close to the worst
theoretically possible case, the predictions at the 200-ppm level should not propagate as
far as (or, at least not much beyond) this distance. From the perspective, the OCAG
predictions are perhaps too high.
2.1.7.5 Choice of a Single Curve for AR and WWTP Guidance
In conclusion, based on an analysis of what are admittedly highly uncertain data, it appears
that the AWWARF and AR models fit well with what is observed. The OCAG model is more
conservative (as intended), and the TFI and DNV models seem perhaps a little optimistic.
Therefore, given the paucity of currently available data in the few hundred ppm range, it would
seem reasonable to choose something in the region of the AWWARF/AR predictions. In the AR
and WWTP guidance, the SACRUNCH, Case B, curve has been chosen.
2.1.7.6 10-Minute vs. 60-Minute Releases
In the OCAG, a distinction is drawn between releases that last for 10 minutes and releases
that last for 60 minutes, and separate lookup tables are provided for each. However, hi the
guidance for WWTPs and ARs, no distinction is made. The main reason for this is that
differences between the two are expected to be small relative to the uncertainties that have been
identified in this section.
2-11
-------
Background Document for Offsite Consequence Analyses, May 1999
\
\
M
1000
a^^
n
Concent
.A
luu > —
MM
•"•jl
10
\
\l >
N
C
I 1
\
500
s
\
\
\
"4
\
\
^
T
V
i\
R N
F
r/
-1
V
\
\
\
\
\
if \
|
'c
1
\
\
\
»••
<
/
\
4
y
V
I
f
\
Q
D
)T1
&
'
r4
i
\
•
/
«-H
r
^-•
3
W
R
\
Maximum Concentration
in the Desert Tortoise
Experiments
D DT1
0 DT2
0 DT3
....!
T
II
^
.
•
5
1
*
DT4
.
— (
/
2
fr,
D
J
,
>
Data Points Denoted
A, B, C, ect from
Markham (1986)
Curves are Maximum
Concentration Measured
in The Desert Tortoise
Experiments
^ DNV
• TFT
i n
W AWWARF
If AR
* OCA (Interpolated)
/9k SV*A /K^A*. T«*-«t-«rtl-il-rt^\
-s
5
-------
Background Document for Offsite Consequence Analyse^ May 1999
2.1.8 Anhydrous Ammonia—Urban Site, Worst-Case.
The discussion so far has been for a rural site. Figure 2-4 is similar to Figure 2-1, except that
it is for worst-case anhydrous ammonia on an urban site. (There is no modeling available for
DNV at an urban site.) The pattern is similar to that for the rural site, except that the range of
uncertainties is not so great. Again, the SACRUNCH, Case B, has been chosen for the generic
worst-case, urban-site guidance for AR and WWTPs.
2.1.9 Anhydrous Ammonia—Alternative Scenarios
There is alternative scenario guidance for flashing liquid releases of anhydrous ammonia in
AR and WWTPs. This guidance is displayed on Figures 2-5 and 2-6 in comparison with data
from TFI, AWWARF, and the OCAG. These curves do not lie as far apart as do the curves for
the worst case and much less attention has been devoted to justifying the choice of the
SACRUNCH curve than was done for the worst-case scenarios. However, it is pertinent to make
the following observations.
Examples of AWWARF and SACRUNCH/AR predictions from Figure 2-6 are given in
Table 2-3. The distances generated by ALOHA as used by AWWARF are about a factor of 4
higher. It is instructive to look at the additional examples of data taken from Figures 2-5 and 2-6
and shown on Table 2-4. These show that, within the context of the large uncertainties that exist
in the modeling, there is essentially no difference between the AWWARF predictions at rural
and urban sites. For SACRUNCH, the corresponding ratios lie between 2 and 3 (i.e.,
SACRUNCH does show that there is a difference between an urban and a rural site in
alternative-case weather conditions).
Table 2-3. Examples of AWWARF and SACRUNCH/AR Predictions
Worst Case, Urban Site
Release Rate
100 Ib/min
1,000 Ib/min
3,000 Ib/min
AWWARF
0.32 mi
1.0 mi
1.8
SACRUNCH
0.08 mi
0.24 mi
0.4
AWWARF/SACRUN
CH Ratio
4.0
4.2
4.5
Table 2-4. Ratios of AWWARF Alternative Case Predictions
Release Rate
100 Ib/min
1,000 Ib/min
3,000 Ib/min
AWWARF
(Rural)
0.4 mi
1.2 mi
2.0 mi
AWWARF
(Urban)
0.32 mi
1.0 mi
1.8 mi
Ratio
Rural/Urban
1.25
1.2
1.1
2-13 '
-------
10.00
to
C
'5
a
•a
UJ
u
0)
o
I
1.00
0.10
WORST-CASE
URBAN
ANHYDROUS AMMONIA
.SACRUNCH,
CASEB
• PUFF, SADENZ,
CASEC
• OCA-generic
-HX-AWWA
-JK-TFI
-ALOHA
SACRUNCH
CONSERV,
CASE A
1000 10000
Rate Of Release (Ibs/min)
100000
Figure 2-4. Sensitivity Studies For Worst-Case Anhydrous Ammonia Scenarios Predicted Distances To Toxic Endpoint,
Urban Site, Atmospheric Stability F, Windspeed 1.5 m/s
-------
to
•5
a
1
.a
g
ALTERNATIVE-CASE
RURAL
ANHYDROUS AMMONIA
WWTFS/SACRUNCH CASE B
AWWA
0.01
100
Release Rate (Ibs/min)
1000
10000
Figure 2-5. Sensitivity Studies for Alternative Anhydrous Ammonia Scenarios Predicted Distances to Toxic Endpoint,
Rural Site, Atmospheric Stability Category D, Windspeed 3 m/s
-------
(0
o>
•s
o
a
1
0.001
URBAN
ANHYDROUS AMMONIA
0.01
-•-WWTFs/SACRUNCH CASE B
-•-AWWA
-nfc-TFI
-^-OCA-generic
100
Release Rate (Ibs/min)
1000
10000
Figure 2-6. Sensitivity Studies for Alternative Anhydrous Ammonia Scenarios Predicted Distances To Toxic
Endpoint, Urban Site, Atmospheric Stability Category D, Windspeed 3 m/s
*!
I
-------
Background Document for Offsite Consequence Analysis^ May 199P
2.2 AQUEOUS AMMONIA
Aqueous ammonia is sometimes found at WWTPs, but not at ammonia refrigeration
facilities. In the WWTP guidance, it is assumed that a solution of 30 percent ammonia spills
onto the ground: this is conservative for the range of concentrations found at such sites. Both
diked and undiked areas are considered. The WWTP guidance contains methods for predicting
the rate of evaporation. These methods are taken from the OCAG and are not discussed further
here. The discussion that follows concerns how to predict the distance to the toxic endpoint,
assuming that the rate of evaporation is known.
The principal difference between aqueous and anhydrous ammonia, in the context of
atmospheric dispersion modeling, is that the former evaporates relatively slowly from a pool,
entirely as vapor, whereas the latter consists of a mixture of vapor and liquid droplets that is
initially denser than air. By contrast, the vapor from a pool of aqueous ammonia is neutrally
buoyant, or even marginally lighter than air. Therefore, it is appropriate to use the passive
Gaussian dispersion model for a neutrally buoyant plume (which will be somewhat conservative
if the plume is buoyant).
Figure 2-7 shows the worst-case SACRUNCH Case B for aqueous ammonia at a rural site,
and Figure 2-8 for that at an urban site. Here, the SACRUNCH Case B is the "Green Book"
Gaussian model, modified by assuming a dry deposition velocity of 1 cm/s. The other models
shown are those proposed by TFI, AWWARF, and OCAG.
In Figure 2-9, the alternative scenario SACRUNCH case for aqueous ammonia is displayed,
together with the AWWARF, TFI, and draft OCAG suggestions. Figure 2-10 is a similar plot for
an urban site.
2.3 CHLORINE
2.3.1 Worst-Case Scenarios
The results of various sensitivity studies are shown on Figure 2-11 and Table 2-5, taking
chlorine with a 150-lb cylinder, a one-ton cylinder, a 17-ton cylinder, and a 90-ton railcar as
examples. These are container sizes that are most common at WWTPs.
None of the sensitivity studies on Table 2-5 is the "right" sensitivity study to choose for a
"point estimate." The approach adopted here has been to exclude the SACRUNCH
conservative case and the OCAG as being too conservative, and then to choose values that are
approximately hi the middle of the range defined in the various sensitivity studies. This leads
to the choice of the SACRUNCH case with 1 cm/s dry deposition velocity as the representative
choice for the guidance tables, the same as was the case for anhydrous ammonia
(SACRUNCH Case B).
2-17
-------
10.0 -r
to
>—»
00
¥
WORST-CASE
RURAL
AQUEOUS AMMONIA
10
1000
10000:
Release Rate (Ibs/min)
•SACRUNCH
CONSERVATVE,
CASEA
•VWVIPS
AVWARF
•OCA-generic
Figure 2-7. Sensitivity Studies for Worst-Case Aqueous Ammonia Scenarios Predicted Distances to Toxic Endpoint,
Rural Site, Atmospheric Stability F, Windspeed 1.5 m/s
-------
c
'5
a
•o
UJ
u
I
a>
u
J
(0
WORST-CASE
URBAN
AQUEOUS AMMONIA
0.01
100 1000
Release Rate (Ibs/min)
10000
•SACRUNCH
CONSERVATIVE
CASE A
-WWTFs
-AWWA
-*-TFI
•OCA-generic
Figure 2-8. Sensitivity Studies For Worst-Case Aqueous Ammonia Scenarios Predicted Distances to Toxic Endpoint,
Urban Site, Atmospheric Stability F, Windspeed 1.5 m/s
-------
10.00
to
to
o
ALTERNATIVE-CASE
RURAL
AQUEOUS AMMONIA
•WWTFs
•AWWA
: TFI
—*-OCA-generic
1000
10000
Release Rate (Ibs/min)
Figure 2-9. Sensitivity Studies for Alternative Aqueous Ammonia Scenarios Predicted Distances to Toxic Endpoint,
Rural Site, Atmospheric Stability D, Windspeed 3 m/s
-------
10.00
Is)
I
I. 1.00
o
CL
•o
o
I
,2
8
0.10
0.01
ALTERNAIIVt-UAbh
URBAN
AQUEOUS AMMONIA
1000
-4-WWTFs
Hfr-AWWA
-A.-TFI
—^—OCA-generic
10000
Release Rate (Ibs/min)
Figure 2-10. Sensitivity Studies for Alternative Aqueous Ammonia Scenarios - Predicted Distances to Toxic Endpoint,
Urban Site, Atmospheric Stability Category D, Windspeed 3 m/s
-------
to
*• mnnn .
tance To Toxic Endpoint
-x
3
§ :
U)
Q
100
4
1
•
A
4
•
i
•<
y
4
R 4
•i
' "L i \. A /
\ / \ / \ /\'»
" •
R
^ U
A
5 L o
.. - .,_ _..... -i „
' A B
A
U u N
R R
U 4 B
R A
A u N
L R
B
A ,
N " -
R
..
L R
B
A
i . - .-
1
150 Ib
Cylinder
1 Ton 17 Ton 90 Ton
Cylinder Railcar Railcar
• Base Case
• 0.3 cm/s Vg
1.0 cm/s Vg
Puff
-EPA
-AWWARF
Figure 2-11. Sensitivity Study Predicted Distances To Toxic Endpoint For Chlorine
-------
Background Document for Qffsite Consequence Analysis, May 1999
Table 2-5. Distances to Toxic Endpoint (ft)-Sensitivity Studies for Chlorine
- EPAOCAG1
•• AWWARF^
*• SACRUNCH
Conservative Case3
o 0.3 cm/s vg4
A 1.0 cm/s vg5
+ Puff
150-lb Cylinder
Rural
10400
6900
8200
6200
4500
2500
Urban
7800
6200
2100
2000
1790
1400
1-ton Cylinder
Rural
36000
16000
78000
36000
16000
6600
Urban
26000
13600
7700
7400
6800
3500
17-ton Tank Car
Rural
**
>6mi
**
**
**
20400
Urban
**
> 6 mi
**
**
29000
8400
90-ton Cylinder
Rural
**
>6 mi
**
**
**
46700
Urban
**
>6mi
**
**
**
• 15700
** Not evaluated or beyond limits of model
1. From EPA's 1996 OCAG Guidance
2. From AWWARF Guidance for the Water Industry
3. Conservative Case Run with SACRUNCH Model
4. SACRUNCH with a dry deposition velocity of 0.3 cm/s
5. SACRUNCH with a dry deposition velocity of 1 cm/s
6. Puff Case
Pertinent conclusions and observations are as follows:
• The results of worst-case scenario modeling should not be quoted without a caveat that
states the range of uncertainty. As can be seen from Figure 2-11 and Table 2-5, the range
of uncertainty is not necessarily the same for each prediction, but a reasonable statement
for the predictions made using the methods presented in the WWTP guidance is that the
result is uncertain by up to a factor of 2-3 below and a factor of 2-3 above. However,
based upon the analysis of uncertainties provided above, it is reasonable to choose a
single, point estimate that is towards the middle or lower end rather than the higher end
of the range.
• The 17-ton tank truck and 90-ton railcar case illustrates a difficulty with essentially all
models that are available for modeling worst-case scenarios at low toxic endpoints like the
3 ppm for chlorine, namely that the predicted distances become increasingly uncertain.
2.3.2 Alternative Scenario
As for ammonia and aqueous ammonia, the WWTP guidance for alternative scenarios for
chlorine is simply based upon the SACRUNCH with vg = 1 cm/s case for alternative weather
conditions, atmospheric stability category D and windspeed 3 m/s. See Section 2.5.1. for a
definition of dry deposition velocity.
2.4 SULFUR DIOXIDE
Once the characteristics of the source term have been determined, vapor clouds formed from
flashing liquid chlorine or sulfur dioxide releases should disperse in much the same way
(the molecular weights are similar, and the toxic endpoints are the same (3 ppm)). Therefore, the
2-23
-------
Background Document for Offsiie Consequence Analysis, May 1999
sulfur dioxide guidance for WWTPs (for both the worst-case and alternative scenarios) has been
calculated using the SACRUNCH case with a dry deposition velocity vg = 1 cm/s.
2.5 BACKGROUND DESCRIPTION OF SENSITIVITY STUDIES
As discussed above, the biggest single difficulty encountered when attempting to provide
guidance on how to calculate the distance to the toxic endpoint is that there are large
uncertainties in the predictions of atmospheric dispersion models. This section contains further
background on uncertainties.
2.5.1 Dry Deposition
The toxic gases that are under discussion here — ammonia, chlorine, and sulfur dioxide —
are highly reactive. They will interact with vegetation, moisture, and surfaces as they travel
downwind. This mechanism depletes the vapor cloud and can effectively reduce predicted
downwind distances. This phenomenon is known as dry deposition.
Deposition is often expressed in terms of an empirical deposition velocity (Hanna and
Hosker, 1980). The dry deposition velocity is used as follows:
CD = vgCA (2-1)
where CD is the rate of deposition of the material onto unit area of the ground, and CA is the
airborne concentration immediately above the ground.
Erisman and Draaijers (1995) have published a book titled Atmospheric Deposition in
Relation to Acidification and Eutrophication. They reviewed more than 30 experiments on the
dry deposition of sulfur dioxide and concluded that "In the literature, average values for the
deposition velocity range from 0.1 to over 2 cm/s with daytime values usually between 0.8 and
1.2 cm/s. Large values (> 2 cm/s) are observed above water surfaces and forests and relatively
small values (< 0.13 cm/s) are measured above snow and bare soil." Therefore, a choice of
0.3 cm/s or 1 cm/s is within the observed range.
Sehmel (1984) has written a chapter on deposition and resuspension in Atmospheric Sciences
and Power Production. He reports that measured dry deposition velocities for all gases range
from 0.002 up to 26 cm/s. He quotes one result for chlorine in the range 1.2 - 2.1 cm/s. Erisman
and Draaijers, in the reference cited above, quote a dry deposition velocity of 0.8 cm/s for
ammonia, but with a large range of uncertainty that includes 1 cm/s.
In SACRUNCH, the dry deposition model does not start until the plume has evolved out of
the heavy vapor phase because very little work has been done on models for dry deposition in the
denser-than-air phase. This approach should be conservative when predicting the distance to the
toxic endpoint.
2-24
-------
Background Document for Offsite Consequence Analysis, May 1999
2.5.2 Puff Releases
The worst-case release is assumed to occur over a period of 10 minutes. Close to the source,
a continuous release model is a good approximation. However, as the vapor cloud travels further
downwind, it begins to look more and more as it would if it had been released as a puff. As a
puff travels downwind, the action of atmospheric turbulence lengthens it along the wind, as well
as causing the width and height to grow. This causes the average concentrations seen by an
individual far downwind to be lower than they would be if modeled as a quasi-continuous "slug"
that goes by in 10 minutes. Some atmospheric dispersion computer programs, such as
DEGADIS, model this transition explicitly. In the present work, a further sensitivity study has
been undertaken in which the worst-case contents of a vessel are released as a puff. It shows that
the case chosen for use as guidance has an element of conservatism to it.
2.5.3 Qualitative Uncertainties and Conservatisms
There are a number of other uncertainties that have not been explicitly modeled in this
chapter, but which add strength to the proposition that many atmospheric dispersion models have
considerable conservatisms built into them.
2.5.3.1 Duration of Worst-Case Weather Conditions
For the very largest releases (e.g., chlorine from a 90-ton railcar), almost all available models
predict very large worst-case distances, usually 25 miles or more. However, traveling at 1.5 m/s,
a plume would take ~ 7 hours to travel 25 miles. It is unlikely that atmospheric stability category
F weather conditions with a windspeed of 1.5 m/s will persist for this long. Before the vapor
cloud has traveled anything like 25 miles, the weather is likely to change to a condition that will
cause more rapid dilution.
2.5.5.2 Pooling
In very low windspeeds, heavy vapor clouds often "pool" on the ground (this is not the same
as a liquid pool). This was explicitly demonstrated in the early heavy vapor experiments at
Porton Down (Picknett, 1978), consisting of puff releases of freon-12, which, at low windspeeds,
slumped until they were only a few inches deep and then remained on site, barely moving. This
might well happen to some or all of the vapor clouds in worst-case conditions.
2.5.3.3 Time Varying Toxic Endpoints
The toxic endpoint established by the rule is valid for an exposure time of one hour, but used
even if the duration of exposure is much less than one hour, as it would be for a worst-case gas
release that takes place hi 10 minutes. As a general rule, for a given health effect, an individual
can withstand higher concentrations at smaller exposure times. Consequently, using a 60-minute
endpoint adds to the conservatism of the predictions.
EPA has begun the process of developing concentrations that will have different toxic
endpoints for various exposure times. These alternatives are known as Acute Exposure
Guideline Levels (AEGLs) for Hazardous Substances. Proposed AEGLs for 12 chemicals have
been published in the Federal Register (62 FR 58839-58851, October 30,1997) (notice
__ :
-------
Background Document for Offsite Consequence Analysis, May 1999
published by the National Advisory Committee for Acute Exposure Guideline Levels for
Hazardous Substances). The 12 chemicals are 1,1-dimethylhydrazine; methylhydrazine; aniline;
ethylene oxide; hydrazine; 1,2-dichloroethene; 1,2-dimethylhydrazine; nitric acid; fluorine,
chlorine, arsine; and phosphine. The only one of these relevant to the present work is chlorine.
For chlorine, the proposed AEGL-1 is 1 ppm for a one-hour exposure (the same as ERPG-1);
the proposed AEGL-2 is 2 ppm (just below the toxic endpoint [ERPG-2] of 3 ppm); and the
proposed AEGL-3 is 20 ppm (the same as the ERPG-3). Thus, AEGLs and ERPGs are roughly
equivalent. To incorporate exposure-time dependence, the National Advisory Committee for
AEGLs states that
C2t = k (2-2)
where k is a constant that has different values for AEGL-1, AEGL-2, and AEGL-3, C is the
average airborne concentration and t is the exposure time. For AEGL-1, C2t = 60 ppm2-min; for
AEGL-2, C2t = 240 ppm2-min; and for AEGL-3, C2t = 24,000 ppm2-min. Focusing on the
AEGL-2 as being closest to the EPA's toxic endpoint, C for chlorine is 2 ppm for t = 1 hour (as
noted above), 2.8 ppm for t = 30 minutes and 4.9 ppm for t = 10 minutes.
It is also pertinent to address the question of whether Haber's law applies to substances such
as chlorine, sulfur dioxide, and ammonia. Gephardt and Moses (1989) looked at published
literature and focused on the effects of airborne concentrations of 3-20 ppm of chlorine over a
duration of exposure of 1 hour (i.e., concentrations in the ERPG-2 to ERPG-3 range). They
concluded that Haber's law is valid as an extrapolation of the 3 ppm/1 hour exposure (i.e., the
ERPG-2) with Ct = 180 ppm-min.
Gephardt and Moses expressed the caveat that Ct = k is not expected to apply for C > 100
ppm, where different types of more severe health effects begin to occur. For ERPG-2, C = 100
ppm corresponds to an exposure time of less than 2 minutes. For t = 10 min, C = 18 ppm and for
t - 30 min, C = 6 ppm.
Gephardt and Moses also consider ammonia, for which the Haber's law constant k for the
ERPG-2 is Ct = (200)(60) = 12,000 ppm-min, provided that C < 5,000 ppm (equivalent to t < 2.4
min.)
2.5.4 Conclusion-Sensitivity Studies
The qualitative sensitivities discussed above would reduce the predicted distances to the toxic
endpoint, if they were analyzed quantitatively. This gives added confidence that the choice of
guidance for AR and WWTPs still contains some elements of conservatism.
2-26
-------
Background Document for Offsite Consequence Analysis* May 1999
CHAPTERS: GASES LIQUEFIED UNDER PRESSURE
The purpose of this chapter is to discuss how the phenomenon of aerosolization from liquid
chlorine, sulfur dioxide, or anhydrous ammonia releases is handled in the AR and WWTP
guidance documents.
Chlorine, sulfur dioxide, and anhydrous ammonia in WWTPs and anhydrous ammonia in
such vessels as the high-pressure receiver in ammonia refrigeration facilities are kept liquefied
under pressure. If the pressure and temperature are sufficiently high, if there is a sudden liquid
release of one of these materials, and if there are no obstructions, it will all become and remain
airborne as a mixture of vapor and very fine liquid droplets that do not fall to the ground.
Experimental results clearly show that this is a real physical phenomenon (Goldwire et al., 1985;
Kaiser, 1989). The airborne droplets evaporate quickly as air is entrained. The evaporation
process cools the air so that a cold mixture of air and vapor is formed. The mixture is denser
than air, even in the case of ammonia, and a heavy vapor dispersion model is required to
adequately predict airborne concentrations downwind of the point of release.
Figure 3-1 shows the results of some experiments that were carried out on liquid chlorine and
reported by Johnson (1991). Similar experiments were not performed for ammonia, but
ammonia results should look similar because chlorine and ammonia have similar density ratios
of liquid to vapor and have similar atmospheric boiling points. It is also a reasonable assumption
that sulfur dioxide will exhibit the same type of behavior.
Figure 3-1 shows the percentage of liquid chlorine that falls to the ground as a function of
superheat, which is the difference between the temperature of the chlorine initially in the vessel
and its atmospheric pressure. Figure 3-1 also shows for comparison the results of the Dow
Model (Dow, 1993), which predicts that the fraction of airborne liquid droplets is five times the
vapor flash fraction (the fraction of chlorine that immediately vaporizes as it is released to the
atmosphere). As can be seen, the Dow Model appears to be non-conservative (i.e., it predicts
that too much chlorine falls back to the ground).
Figure 3-1 also shows the results of a model (lanello 1989), known as the "RELEASE"
model, that was used by Johnson (1991) to try to reproduce the experimental results. As can be
seen, agreement is poor3. Other models that take into account this evaporation lead to better
agreement with experiments (e.g., Woodward and Papadourakis, 1991; Woodward et al., 1995).
The principal conclusion is that, even at superheats of only 10 °C (which would be a
temperature of only about -23 °C for ammonia and chlorine and about 0 ° C for sulfur dioxide),
only a small fraction of released liquid would fall to the ground. Therefore, at most, a small
degree of conservatism is introduced if it is assumed that, for superheats exceeding 10 °C, all of
the released chlorine, sulfur dioxide, or ammonia remains airborne as a mixture of vapor and fine
liquid droplets.
3 CCPS has been funding further development of the RELEASE model. A private communication from Johnson,
D.J., Quest Consultants, Norman Oklahoma (October 1997) indicates that RELEASE has been modified so that
agreement with experiment is much improved. However, at the time of writing, RELEASE was not available to the
authors.
, . , _
-------
Background Document for Offsite Consequence Analysis, May 1999
100—I
80
i.
«
O
•a
3 40
cr
"Release" Model
Dow Model
20 —
X - Experimental Results from Johnson (1991)
X
10
50
60
20 30 40
Liquid Superheat
Figure 3-1. Fraction of Liquid Chlorine Falling to the Ground as a Function of Superheat
3-2
-------
Background Document for Offsite Consequence Analysis, May 1999
If a jet of liquid droplets and vapor impinges upon a surface close to the point of release,
there can be efficient recovery of droplets, which will form a relatively slowly evaporating pool
on the ground. Experiments with ammonia have shown that up to 75 percent of the airborne
droplets can be removed in this way (Resplandy, 1969: Kaiser, 1989). For worst-case scenario
modeling, the use of this or similar reduction factors is probably not justified if it is possible that
the release would not encounter obstructions. However, when considering mitigation, this
phenomenon can be taken into account. For example, if vessels are indoors, there will almost
certainly be a surface upon which the jet will impinge and the jet will likely change direction
(e.g., impingement of the floor and subsequent upward movement of the vapor cloud). This
arrangement is similar to the design of separators in chemical processes and would be expected
to be very efficient at removing liquid droplets from the vapor stream. This issue is further
discussed in Chapter 6. For the purposes of discussing the effects of obstacles, it is assumed that
obstacles cause the recovery of 75 percent of any airborne liquid droplets. Thus, if the initial
airborne release consists of 20 percent vapor and 80 percent liquid droplets, an obstacle-impeded
release would consist of the original 20 percent vapor plus 20 percent liquid droplets (i.e., a
mixture split equally between vapor and liquid and containing only 40 percent of the mass in an
unobstructed release).
The above amounts only to a rule of thumb. In fact, the percent capture of the liquid depends
on several factors, among which is the path length of the jet before it encounters an obstruction. A
model that shows this effect explicitly has been developed by Muralidhar et al. (1995). This model
was specifically developed for escaping jets of a mixture of hydrogen fluoride and a proprietary
additive. The mixture is being used as a catalyst in the alkylation unit at Mobil's Torrance, CA,
refinery. However, the general picture is applicable to all jet releases that consist of a mixture of
fine droplets and vapor. The droplets present a very large surface area for evaporation; as long as
they remain airborne, they evaporate rapidly. On encountering an obstacle, they run down to form
a pool on the ground, which has a much smaller surface area to volume ratio, so the rate of
evaporation is much decreased. Some typical results show that the airborne reduction factor
(essentially the percentage of hydrogen fluoride that ends up on the ground) is 50-70 percent for a
40' travel distance, 75-85 percent for a 5' travel distance, and ~ 90 percent for a 3" travel distance.
These figures cannot be directly applied to ammonia, chlorine, or sulfur dioxide, but illustrate
the general idea. They show that the percentage collecting on the.floor (or the percentage of the
original release remaining airborne) is going to be highly configuration dependent. However,
compressor rooms in ammonia refrigeration facilities and chlorine or sulfur dioxide rooms in
WWTPs are often highly crowded.
The rule-of-thumb presented above, that 60 percent of the initial flashing liquid release of
' ammonia, chlorine, or sulfur dioxide ends up on the floor as a slowly evaporating pool, is,
therefore, offered as a simple means of taking some advantage of the presence of obstructions in
buildings. This number is highly uncertain, but it is not possible to produce configuration-
specific guidance that is also simple to use.
3-3
-------
Background Document for Offsile Consequence Analysis, May 1999
CHAPTER 4: ADJUSTMENT OF MEAN CONCENTRATION FOR AVERAGING
TIME
The work in this section is based on a monograph by D.J. Wilson (1995), Concentration
Fluctuations and Averaging Time in Vapor Clouds, that contains the most comprehensive
existing summary of the theories and experimental data relating to the subject of the effect of
averaging time on mean concentrations.
This issue is easily illustrated in the context of the Gaussian model, in which the cloud
centerline concentration C is inversely proportional to the crosswind standard deviation ay:
C a 1/cjy
There is a widely used relationship for the dependence of ay on averaging time ta:
ay a (ta)p
(4-1)
(4-2)
There are few direct measurements of the exponent p. Wilson documents experiments by
Yersel, Gobel, and Morrill (1983) in which the value of p close to the source varied from 0.0 to
0.18. Studying releases from a high stack, Mueller and Reisinger (1986) found far downwind
(25 km) an average p = 0.25, with at least a factor of two variability in p from one test to another.
Wilson has proposed a more realistic estimate of averaging time effects on plume spread
using a power law model that accounts for initial source size and plume travel time on averaging
time effects. Basically, a wide plume (with large ay) meanders less than does a narrow plume
because meandering is caused by turbulent eddies that are larger than the plume, and there are a
wider range of turbulent atmospheric eddies to push the smaller plume around. Wilson's
recommended working equation is, for an atmospheric sampling time tj equal to the mean
concentration averaging time ta,
y _
CT y.ref
f
1 ^ cr°
1.5
1 O v * =00
V *»l* y
f
1 ^ a°
la
I + *• +r «•
^ ^ ll
/ TLV TLV
I_L tt _L ts,ref
"r "r il
TLV TLV ,
(4-3)
where ay>ref is the plume spread for a short sampling time, typically - t^f = 180s. As an
approximation in the source size cr0 term, use ay,ts=~ ~ 2 ay>ref to estimate the long-sampling-time
plume spread. Equation 4-3 was derived for stationary processes and should be limited to
sampling times less than about 3 hours in the atmosphere. The parameter ri is given by
4-1
-------
Background Document for Offsite Consequence Analysis, May 1999
rr
+ 0.6-^- +0.3-ts
(4-4)
TLV TLV
The power law in Equation 4-3, with atmospheric sampling time ts equal to the averaging
time ta is more physically realistic than the purely empirical one-parameter Equation 4-2. The
major difficulty applying this power law is estimating the Lagrangian turbulent time scale TLV.
Here, a value of TLV — 10,000 seconds is recommended, based on plume dispersion data for long
travel times (Barr and Gifford (1987)).
Surprisingly, considering the importance of averaging time effects, there is a lack of
available data with statistically independent measurements of plume spread cry needed to validate
(4-3). The only data used by Wilson for validation are from Briggs (1993) for plume spread
measurements in unstable atmospheric conditions. Equation 4-3 gave a plausible fit to this data,
but there was so much scatter that other functional forms would have done equally well.
It has been conventional wisdom in models for dense gas dispersion that the strong density
gradients across the top of a dense plume reduce its entrainment and suppress both meandering
and vertical mixing inhomogeneities that produce concentration fluctuations. To refute this
claim, Hanna, Chang, and Strimaitis (1993), in a review of hazard assessment models, showed
that large one-second averaged concentration fluctuations were present in the repeat realizations
of various dense vapor field test experiments. In fact, it is apparent that fluctuations caused by
inhomogeneous mixing, excluding meandering, are as strong in dense plumes as they are in
neutrally buoyant releases.
In addition, dense plumes should also experience crosswind meandering, because they
entrain atmospheric air with crosswind velocity fluctuations. These crosswind fluctuations
should not be much affected by the density gradient that suppresses the vertical turbulence
fluctuations and effectively reduces the rate of entrainment of air across the top surface of the
vapor cloud (see Britter (1989)). Once the dense plume has been diluted by about a factor of
five, it is mostly ambient air and should meander in much the same way as a passive plume.
One important obstacle to applying meandering plume models to dense releases is the
substantial difference in shape of crosswind velocity profiles between a dense plume and a
passive plume. Gravity-driven spreading produces a wider, more uniform concentration across
the center of a dense plume, as shown in Figure 4-1.
Crosswind meandering simply flips this relatively uniform core back and forth across the
centerline, causing considerably smaller fluctuations in the center and considerably more near
the edges than would be caused by a Gaussian profile plume, shown superimposed on the dense
plume developed. More work is needed in this area before an operational model can be
recommended.
4-2
-------
dense plume
U)
Source : Wilson (1995)
Meandering of the broad, flat crosswind concentration profile in a dense
phiine should cause lower concentration fluctuations on the centerline and
higher fluctuations near the edges than observed in a passive plume.
passive
plume
Figure 4-1. Illustration of Meandering
-------
Background Document for Offsite Consequence Analysis, May 1999
Tables 4-1 and 4-2 show how the implementation of Wilson's meandering model affects
predicted distances to the toxic endpoint for worst-case anhydrous ammonia releases. An
exposure duration of 10 minutes has been assumed. In addition, to take account the lack of a
good meandering model for dense plumes, the meandering effect is not allowed to start until the
plume has essentially become neutrally buoyant, at which point CTO is determined from the half-
width L of the plume at the transition point:
cj0 = L/2.14
That is, at the transition to neutral buoyancy, the plume becomes Gaussian with the 10
percent edge at a crosswind distance L from the centerline.
(4-5)
It can be seen that the distances are not much affected (the influence of meandering would be
considerably larger for durations of exposure of an hour or more). The SACRUNCH Case B that
was chosen as the guidance nevertheless contains the Wilson meandering correction.
The predicted effect of meandering is smaller for chlorine and sulfur dioxide than it is for
ammonia, given the same rate of release. This is because the toxic endpoints of chlorine and
sulfur dioxide are much smaller than for ammonia (3 ppm vs 200 ppm). Therefore, a chlorine or
sulfur dioxide plume has to travel much further to the toxic endpoint than does an ammonia
plume. Equation 4-3 shows that the ratio cjy/(Ty;ref tends to unity for very large travel times.
Table 4-1. Example of Effect of Meandering of Anhydrous
Ammonia Releases, Worst-Case, Rural Conditions
Rate of Release
(Ibs/min)
10
20
50
100
200
500
1,000
2,000
5,000
10,000
20,000
50,000
Distance to Toxic Endpoint
(miles)
"Old"
Model
0.19
0.27
0.41
0.57
0.80
1.28
1.85
2.72
4.59
6.74
9.59
14.27
With Wilson
Meandering
0.19
0.26
0.40
0.56
0.78
1.26
1.82
2.68
4.54
6.69
9.55
14.23
4-4
-------
Background Document for Offsite Consequence Analysis, May 1999
Table 4-2. Example of Effect of Meandering of Anhydrous
Ammonia Releases, Worst-Case Urban Conditions
Rate of Release
(Ibs/min)
10
20
50
100
200
500
1,000
2,000
5,000
10,000
20,000
Distance to Toxic Endpoint
(miles)
"Old"
Model
0.10
0.14
0.21
0.28
0.39
0.61
0.84
1.18
1.86
2.64
3.89
With Wilson
Meandering
0.10
0.13
0.20
0.28
0.39
0.60
0.83
1.17
1.84
2.62
3.83
4-5
-------
Background Document for Offsite Consequence Analysis, May 1999
CHAPTER 5: AMMONIA/MOIST-AIR THERMODYNAMICS
A rigorous consideration of ammonia/moist air thermodynamics will have some effect on
predicted distances to the toxic endpoint. In this chapter, the basic thermodynamic model is
described in Section 5.1, and the effect on predicted distances to the toxic endpoint is discussed
in Section 5.2.
A related issue is that of potential lift-off of the plume. A flashing liquid release of ammonia
is initially denser than air, but could potentially become buoyant as it travels downwind and
entrains air. This issue is addressed in Section 5.3.
5.1 CALCULATION OF THERMODYNAMIC PROPERTIES OF MIXTURES OF
AMMONIA AND AIR
A quantity of material released into the atmosphere entrains air both during the release
process and the atmospheric transport process. In the case of ammonia, the release may involve
pure vapor or a mixture of vapor and liquid, and considerable turbulence is generated in the
release process as ammonia storage pressures are usually well above atmospheric pressure.
Estimation of cloud concentration, position, velocity, and dimensions during atmospheric
transport involves consideration of a variety of physical processes, including rate of entrainment
of air, gravitational slumping, transfer of heat from the atmosphere and the ground, and fall-out
of entrained aerosols. Performance of these calculations is facilitated by independent analysis of
the changes in thermodynamic properties that occur when ammonia is mixed with air containing
water vapor. This chapter discusses calculation of the temperature, density, quantities, and
compositions of vapor and liquid phases comprising a cloud formed by mixing known quantities
of ammonia and moist air. In all cases, the total pressure is one atmosphere, and both the initial
clouds and the final cloud are at thermodynamic equilibrium.
5.1.1 Methods
The equilibrium state of a mixture of known quantities of ammonia, air, and water is
completely determined by specification of two independent variables. If the number of
independent Intensive variables4 determined by Gibb's phase rule5 is greater than two,
specification of temperature and pressure determines the state of the system, including
composition and quantities of vapor and liquid phases. If the number of Intensive variables
determined by Gibb's phase rule is less than two, specification of pressure and relative quantities
• of vapor and liquid phases determines the state of the system for the cases considered in this
analysis. Given specification of the properties of the initial ammonia cloud and the initial moist
air cloud, the conditions of the final cloud may be calculated using vapor/liquid equilibrium
relationships and mass and energy balances. The following sections describe physical property
4An intensive variable is a property of a system that does not depend on the quantity of material comprising the
system. Temperature and pressure are examples of intensive variables.
5Gibbs's Phase Rule determines the minimum number of independent variables that must be specified to uniquely
establish the intensive state of a system at equilibrium comprising given numbers of components and phases.
_
-------
Background Document for Offsite Consequence Analysis, May 1999
data and the algorithm combining mass and energy balances with vapor/liquid equilibrium
relationships to determine the final cloud conditions.
S.J.LI Physical Property Data
Physical property data required to calculate the state of a system includes
pressure/volume/temperature (PVT) relationships, vapor/liquid equilibrium data, and enthalpy
data. Because the mixing process occurs at atmospheric pressure, the ideal gas law was used to
represent PVT behavior for vapor phases. For liquid phases, densities were taken to be constant
at a value of 690 kg/m3 for ammonia and 1,000 kg/m for water.
The vapor pressure exerted by ammonia/water systems depends on temperature and the
composition of the liquid phase. Correlations proposed to represent this dependence (Wheatley,
1987) are:
Pi(X2,T) = (1-X2) PS>1(T) EXP [-Ai/T + Bi] (5-1)
and
P2(X2,T) = X2 PS,2(T) EXP (-Az/T + B2) (5-2)
where:
Pi(X2,T) = partial pressure of water (component 1) as a function of
temperature and liquid phase composition, Pa
X2 = mole fraction of ammonia (component 2) in the liquid phase
T = temperature, K
Ps,i(T) — partial pressure of pure water as a function of temperature, Pa
P2(X2,T) = partial pressure of ammonia as a function of temperature and liquid
phase composition, Pa
= partial pressure of pure ammonia as a function of temperature, Pa
and AI, BI, A2, and B2 are estimated as:
Ai = (l+3Ra/2-RaX2)X22Wa (5-3)
A2 = (1 + Ra - RaX2) (1-X2)2 Wa (5-4)
Bi = (1 + 3Rb/2 - RbX2) X22 Wb (5-5)
B2 = (l+Rb-RbX2)(l-X2)2Wb (5-6)
Ra, Rb, Wa, and Wb are empirically determined constants with recommended values of -14, -
14, -174 and'-0.74 (Wheatley, 1987).
5-2
-------
Background Document for Offsite Consequence Analysis, May 1999
Pure component vapor pressures were estimated using Antoine's equation (Smith and
Van Ness, 1975):
PSji(T) = exp [ Aa,, - Ba/(T + C^) ] • (5-7)
where:
PSji(T) = equilibrium vapor pressure of pure component i at temperature T, Pa
T = temperature, K
and AS,}, Ba,; and Ca,i are empirically determined coefficients with recommended values of
16.4981,2132.5 and -32.98 for ammonia and 18.3036, 3816.44 and -46.13 for water,
respectively. The utility of these relations is supported by the comparison with pure component
and mixture vapor/liquid equilibrium data presented in Tables 5-1, 5-2 and 5-3. The pure
component predictions of Equation 5-7, which are summarized in Tables 5-1 and 5-2, are
incorporated into the Wheatley model as indicated in Equations 5-1 and 5-2. The accuracy of the
comparison of the mixture data (Table 5-3) is limited by interpolation of the reported measured
data.
The dependence of enthalpy on temperature for vapor phase components is represented as:
hVji = CpV;i(T-Tr) + AHV;i (5-8)
where:
hv,i = enthalpy per unit quantity of component i in the vapor phase, J/g-mole
CpV,i = heat capacity at constant pressure of component i in the vapor phase
J/g-mole/K
T = temperature of vapor phase, K
Tr = reference temperature, K
AHv,j = latent heat of vaporization of component i, J/g-mole.
The enthalpy of pure liquid components was calculated as:
hu = Cpu(T-Tr) (5-9)
where:
by = enthalpy per unit quantity of component i in the liquid phase, J/g-mole
Cpi,i = heat capacity at constant pressure of component i in the liquid phase,
J/g-mole/K
T = temperature of the liquid phase, K
Tr = reference temperature, K
5-3
-------
Background Document for Offsite Consequence Analysis, May 1999
Table 5-1. Comparison of Measured and Predicted Vapor Pressure of Water
Temperature
(K)
273
283
293
303
313
Saturation Pressure (Pa)
Measured*
611.3
1,164.5
2,290.4
4,165.5
7,253.4
Predicted
593.1
1,204.5
2,313.3
4,219.9
7,358.8
* data from Table 1 of Keenan and Keyes, 1969
Table 5-2. Comparison of Measured and Predicted Vapor Pressure of Ammonia
Temperature
(K)
210
230
250
270
290
310
Saturation Pressure (MPa)
Measured*
0.01775
0.06044
0.16496
0.38100
0.77413
1.4235
Prediicted
0.01793
0.06091
0.16515
0.37841
0.76211
1.3873
*data from Table 15 of ASHRAE, 1981
Table 5-3. Comparison of Measured and Predicted Vapor Pressures Above
Ammonia/Water Solutions
T
(K)
273
253
233
213
Liquid Phase
NH3 Mole
Fraction
0.06
0.25
0.29
0.37
0.50
0.63
0.28
0.53
0.74
0.42
0.63
0.87
Measured*
Vapor Phase
NH3 Mole
Fraction
0.74
0.98
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
Total
Pressure
(Pa)
2.0xl03
2.0xl04
9.8xl03
2.0xl04
4.9xl04
9.8xl04
2.0xl03
2.0xl04
4.9xl04
2.0xl03
9.8xl03
2.0xl04
Predicted
Vapor Phase
NH3 Mole
Fraction
0.73
, 0.99
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
Total
Pressure
(Pa)
2.0x10'
2.6xl04
1.2xl04
2.4xl04
5.6xl04
9.8xl04
2.6xlOJ
2.2xl04
4.9xl04
2.5xlOJ
l.OxlO4
1.9xl04
*adapted by interpolation of Figure 3-17 of Perry and Chilton, 1973
5-4
-------
Background Document for Offslte Consequence Analysis, May 1999
Values of heat capacities and heats of vaporization are presented in Table 5-4.
Table 5-4. Heat Capacity and Heat of Vaporization Data
Component
Ammonia
Water
Air
Vapor Phase Heat
Capacity
(J/g-moIe/K)
23.0
32.5
29.0
Liquid Phase
Heat Capacity
(J/g-moIe/K)
77.1
75.4
—
Heat of Vaporization
(J/g-mole)
21459.9
45009.2
—
The enthalpy of ammonia/water solutions was calculated as:
hst,ln = (1-X2) hi,H2O + X2 h^NHS + AHmjx
(5-10)
where:
hsoin = enthalpy per unit quantity of ammonia/water solution, J/g-mole
X2 = mole fraction of ammonia in the final solution, dimensionless
hi,H2o = enthalpy per unit quantity of pure liquid water, J/g-mole
hi,NH3 = enthalpy per unit quantity of pure liquid ammonia, J/g-mole
AHmiX = enthalpy of mixing of ammonia and water per unit quantity of final solution,
J/g-mole
x is estimated as (Wheatley, 1987):
AHmix = - (1 + Ra - RaX2/2) X2 (1-X2) R Wa (5-
and
where R is the universal gas constants and all other terms are as defined above.
5.1.1.2 Algorithm for Determination of Final Cloud Conditions
Mixing of initial clouds of ammonia and moist air produces a final cloud whose temperature,
composition, and physical state are initially unknown. In addition, the nature of the equilibrium
relationships and mass and energy balances describing the mixing process is such that direct
calculation of the final conditions is not possible. Thus, the iterative procedure represented in
Figure 5-1 was adopted to solve this problem. The procedure involves three primary elements:
calculation of final cloud conditions assuming that the final cloud contains solely vapor;
calculation of dew point pressure for the all- vapor final cloud; and calculation of conditions for a
final cloud containing both vapor and liquid phases. Calculation of the dew point pressure of the
assumed all-vapor final cloud is used to identify the physical state of the final cloud. If the sum
of the vapor pressures of ammonia and water in the all-vapor final cloud is less than their
pressures at the dew point, the final cloud is all vapor. If the sum of the pressures of ammonia
5-5
-------
Yes
t
Set Conditions of Initial
Ammonia and Moist Air
Clouds
V
Assume Final Cloud
All Vapor
1
Calculate Final
Cloud Temperature, Tf
V
Calculate Dew
Point Pressure
/ Is \w
v Liquid y
\ Present? /
No
Figure 5-1. Algorithm for Determination of Final Cloud Conditions for Mixing of Ammonia and Moist Air Clouds
-------
Ul
(continued)
Guess Tf
Guess liquid phase
ammonia mole fraction, X2
Solve mass and
equilibrium relations
Yes
Calculate Fin
Cloud Density
Figure 5-1. Algorithm for Determination of Final Cloud Conditions for Mixing of Ammonia and Moist Air Clouds
(Continued)
-------
Background Document for Offsite Consequence Analysis, May 1999
and water in the all-vapor final cloud is greater than their dew point pressure, the final cloud
contains both vapor and liquid phases. The following paragraphs describe execution of the three
elements comprising the solution algorithm.
For given initial conditions, the temperature and composition of a final cloud containing
solely vapor may be determined using mass and energy balances. The mass balance for each
component was expressed as:
(5-12)
where:
Nv,i = number of moles of component i in the vapor phase, g-mole
NI,J = number of moles of component i hi the liquid phase, g-mole
and the vertical brackets indicate evaluation at initial and final conditions. For the case of no
liquid present in the final cloud, Nl,i final is equal to zero in each of Equations 5-12, and Nv,i
-Una] can be calculated directly for each component. The energy balance for the mixing process
was expressed as:
£ (Nv,ihv,i + Ni.ihv.0 -final + £ (Ni,D AHmix -fmai = I. (Nv>ihv>i + Ni,ihu)-initiai + S (Nu) AHmix -initial (5-13)
where all quantities are as defined above, the summations are taken over components, and the
vapor and liquid phase enthalpies are evaluated according to Equations 5-8 and 5-9. For the
assumed conditions of an all-vapor final cloud and the final cloud component masses determined
by Equation 5-12, Equation 5-13 can be directly solved for the final cloud temperature.
Calculation of the dew point pressure is based upon specification of the temperature and
vapor phase mole fractions of the cloud and identification of the liquid phase composition in
equilibrium with the specified vapor phase composition. By definition, the liquid phase mole
fractions sum to unity:
SXi=l (5-14)
where X{ are liquid phase mole fractions. The vapor/liquid equilibrium constraints may be
expressed as:
(5-15)
where Yj and Xj are vapor and liquid phase mole fractions, respectively, and the KI are
vapor/liquid equilibrium constants calculated as:
i = (l/Xi)Pi/Pt (5-16)
5-8
-------
Background Document for Offsite Consequence Analysis, May 1999
where the Pj are calculated using Equations 5-1 and 5-2, Pt is total pressure, and air is taken to be
non-condensible. The dew point constraint, Equation 5-14, becomes:
E(Yi/Ki) = l (5-17)
where the Yj are the specified vapor cloud mole fractions, and the Kj are functions of liquid
phase mole fraction and the specified temperature. Because air is effectively non-condensible
under conditions occurring in ammonia releases, the vapor phase mole fractions may be
expressed on an air-free basis, and use of Equations 5-14, 5-15 and 5-16 was reduced to the
condition of locating the liquid phase ammonia composition (X2) in equilibrium with the
specified vapor phase at the specified temperature. The algorithm for this procedure is presented
in Figure 5-2. If the calculated dew point pressure is less than the sum of the specified partial
pressures of water and ammonia, the final cloud is solely vapor, and the density was calculated
using the ideal gas law. If the sum of the partial pressures of water and ammonia is greater than
the dew point pressure, a liquid phase will form, and the procedure described in the following
paragraphs was used to determine final cloud conditions.
For a final cloud containing both vapor and liquid phases, the algorithm presented in the
portion of Figure 5-1 below connection point B was used to determine final cloud conditions.
The mass balances of Equation 5-12 were combined with the vapor/liquid equilibrium
relationships of Equations 5-15 to derive the constraint:
0 = 2 { [Zi (1-KO]/[1-(1-KOP] } = 0 (5-18)
where:
-------
Background Document for Offslte Consequence Analysis, May 1999
Specify Vapor Phase
Temperature (T) and
Composition OQ
Shift Vapor Phase
Composition to
Air-Free Basis (V^{)
Set Yf>2 — 1^2
Guess
Calculate Psl, P& and
YC>2 = PS2/(PS1+PS2)
No
Dew Point Pressure = P.J + P
Figure 5-2. Algorithm for Calculation of Dew Point Pressure
5.1.2 RESULTS
Representative results were developed for three cases bounding potential conditions. The
first case, representative of worst-case scenario conditions, involves release of vapor and liquid
ammonia at the atmospheric pressure boiling point (240 K), with 80 percent of the ammonia in
the liquid phase and mixing with warm, moist air (relative humidity = 50 percent). The results
presented hi Table 5-5 indicate that the final cloud initially experiences evaporative cooling and
remains denser than ambient air even after entrainment of enough air to fully vaporize the
ammonia. Liquid droplets, comprised primarily of condensed water, persist at high air-to-
ammonia mixing ratios at moderate relative humidities. The second example, also a worst-case
-------
Background Document for Offsite Consequence Analysis, May 1999
scenario, involves release of a two-phase cloud of ammonia and mixing with warm, dry air. The
results presented in Table 5-6 show the evaporative cooling step, but indicate that the ammonia
liquid phase is completely evaporated after mixing with ll kilograms of air per kilogram of
ammonia. Comparison of the results of Tables 5-5 and 5-6 indicates that heat of mixing offsets
the evaporative cooling effect and decreases the density of the final cloud. The third case
involves release of vapor ammonia into dry air of variable temperature. The results presented in
Table 5-7 indicate that the final cloud remains less dense than air even at relatively low ambient
air temperatures.
Table 5-5. Final Cloud Conditions for a Worst-Case Scenario,
Two-Phase Release with Moist Air*
Mass of Air
Mass of Ammonia
1
3
6
9
11
20
100
Final Cloud Conditions
T(K)
221.5
214.7
214.0
226.7
238.8
266.8
293.3
Density(k
g/m3)
2.053
1.711
1.597
1.492
1.424
1.294
1.190
Fraction of NH3in
Liquid Phase (%)
67.1
48.7
23.7
9.5
6.8
3.3
0.6
Mole Fraction of
NH3 in Liquid
Phase
0.986
0.945
0.808
0.530
0.397
0.176
0.021
* Initial Conditions
• 80% of ammonia in Liquid Phase, T = 240
• air temperature = 298 K
• relative humidity = 50%
• air density =1.186 kg/m3
Table 5-6. Final Cloud Conditions for a Worst-Case Scenario,
Two-Phase Release with Dry Air*
Mass of Air
Mass of Ammonia
1
5
11
20
100
Final Cloud Conditions
T(K)
221
209.3
205.9
243.1
286.4
Density
(kg/m3)
2.100
1.689
1.626
1.407
1.225
Fraction of NH3 in
Liquid Phase
69.3
41.6
2.4
0.0
0.0
*Initial Conditions
• 80% of ammonia in Liquid Phase, T = 240
• air temperature = 298 K
• relative humidity = 0%
• air density =1.186 kg/m3
TTT
-------
Background Document for Offsite Consequence Analysis, May 1999
Table 5-7. Final Cloud Conditions for All Vapor Release with Dry Air*
Ambient Air Conditions
T(K)
263
273
283
293
Density
(kg/m3)
1.344
1.295
1.249
1.219
Final Cloud Conditions
T(K)
249
253
258
263
Density
(kg/m3)
1.048
1.031
1.014
0.997
"Initial Conditions
• ammonia release all vapor, T = 240 K
• air-to-ammonia mass mixing ratio equal to unity
• 0% relative humidity
5.2 EFFECT ON PREDICTION OF DISTANCES TO TOXIC ENDPOINT
Table 5-8 shows an example of the effect on SACRUNCH runs of including the new
thermodynamics, Wilson meandering, and both. As can be seen, the effect is small, but does
lead to a small reduction in predicted distances, which has been taken into account in the AR and
WWTPs guidance.
Table 5-8. Example of the Effect of New Thermodynamic Model and Meandering of
Anhydrous Releases, Worst-Case, Rural Conditions, 75% RH
Rate of Release
(Ibs/min)
10
20
50
100
200
500
1,000
2,000
5,000
10,000
20,000
50,000
Distance to Toxic Endpoint (miles)
Old
Model
0.19
0.27
0.41
0.57
0.80
1.28
1.85
2.72
4.59
6.74
9.59
14.27
With Wilson
Meandering
0.19
0.26
0.40
0.56
0.78
1.26
1.82
2.68
4.54
6.69
9.55
14.23
With New
Thermodynamics
0.20
0.28
0.42
0.58
0.81
1.27
1.79
2.54
4.11
5.94
8.60
13.8
With Wilson
Meandering &
New
Thermodynamics*
0.20
0.27
0.42
0.58
0.80
1.25
1.77
2.52
4.07
5.90
8.54
13.74
* These results do not exactly match those in the AR. Minor details of input have been changed.
-------
Background Document for Offsite Consequence Analysis, May 1999
5.3 POTENTIAL FOR LIFT-OFF
There are certain types of initially heavy plumes that can potentially become buoyant and
may lift off the ground. These include hydrogen fluoride (HF) releases in moist air, where the
heat liberated by condensation of HF/water droplets can cause the plume's buoyancy to become
positive as air is entrained. Another possibility is the uranium hexafluoride (UF^/moist air
system because the UFg-water reaction is highly exothermic. Finally, it is conceivable that
initially denser-than-air ammonia plumes could become buoyant as they dilute.
Briggs (1973b) has developed a simple approach that requires the calculation of a lift-off
parameter:
Lp = ghAp/(pau2*)
(5-19)
where lift-off will occur if Lp > 30 (Meroney, 1984). Here, g is the acceleration due to gravity
(m/s2), h is the height of the cloud (m), Ap is the difference hi density between the air and the
plume (kg/m3), pa is the density of air (kg/m3), and u* is the friction velocity (m/s).
For UFg, HF, and ammonia, the SACRUNCH computer model calculates Lp at distances
downwind of the transition point at which the plume ceases to be denser than air. If Lp exceeds
30, in any weather condition, this information will be printed out, and the user must judge
whether there will be more than trivial plume rise. Table 5-9 contains an illustration of the
potential for lift-off for a 1,000 Ib/min release of flashing liquid ammonia. As can be seen, there
appears to be no potential for lift-off until the plume has diluted below the toxic endpoint
Table 5-9. Illustration of the Potential for Lift-Off
Distance
Downwind (m)
1,950
2,400
3,000
4,500
LP
6.4
20.5
31.9
48.0
Average Concentration
(kg/m3)
2x10-"
S.lxlO'5
4.4x1 0'5
1.9xlO'5
Peak
Concentration
(kg/m3)
5.9X1Q-4
2.4X1Q-4
1.3X1Q-4
3.8X10'5
Assumptions:
Anhydrous Ammonia, flashing liquid release
Release Rate 1,000 Ib/min
Worst-Case Weather Conditions, Rural Site
RH = 75%
Results:
Cloud evolves out of denser-than-air phase ~ 1800 m downwind with mean
concentration ~ 4x 10"4 kg/m3
Toxic endpoint: 1.4x10^kg/m3 (200 ppm) at ~ 2,900 m based on peak
concentration
Conclusion:
No lift-off until after toxic endpoint
No counter-examples found in sensitivity studies
5-13
-------
Background Document for Offsite Consequence Analysis, May 1999
CHAPTER 6: EFFECT OF AMMONIA RELEASES ON STRUCTURES
The purpose of this chapter is to discuss the consequences of releases Inside rooms, such as
the compressor room in an ammonia refrigeration plant or a room containing chlorine or sulfur
dioxide vessels at WWTPs. The work presented here has been used as the basis for building
mitigation models in the AR and WWTP guidance, but not in OCAG.
This chapter provides methods for prediction of pressures inside buildings and the
subsequent rate of release to the atmosphere. Important parameters include the quantity of
ammonia, chlorine, or sulfur dioxide available for release, the time over which the release takes
place, the volume of the room, the presence of airborne liquid droplets, leakpaths in the structure,
and the characteristics of the ventilation system.
As discussed in Chapter 3, experimental evidence indicates that hi the case of unobstructed
releases, ammonia, chlorine, or sulfur dioxide droplets formed in the relea.se remain suspended in
the vapor phase and evaporate. Obstructions can be effective in facilitating the rain-out of
droplets with removal efficiencies on the order of 75 percent of the liquid reported. The
remaining 25 percent of the liquid release is assumed to be entrained in the vapor release.
Chapter 3 presents an approximate rule-of-thumb: if released inside a building as a flashing
liquid release, 60 percent collects on the floor as a relatively slowly evaporating pool, and 40
percent remains airborne as a 50:50 mixture of vapor and liquid droplets.
Evaluation of building structural integrity and effectiveness of mitigation for intact structures
was based on consideration of single-phase releases of vapor into buildings containing leakpaths
and engineered vents capable of relieving pressure. The vapor release for the analysis included
the contribution of vapor formed during depressurization of liquefied gas stored under pressure
(flashing). Building release attenuation factors estimated in this chapter are applicable to both
the entrained liquid and gaseous portions of the release.
This document considers releases over 10 minutes. EPA also analyzed sudden releases. Its
analysis indicated that for pressure-tight buildings, a sudden release could damage the building
sufficiently to eliminate its ability to mitigate the release unless the size of the building is large in
relation to the mass of the chemical. In the more likely case of a building with leakpaths and
ventilation, a sudden release will generally not damage the building.
6.1 PROLONGED RELEASES
Gradual releases of stored material as vapor may not be capable of producing the pressure
differentials predicted for instantaneous releases. Over periods of 10 minutes, ammonia,
chlorine, or sulfur dioxide may escape through leakpath and ventilation system flowpaths at rates
large enough to relieve the initial pressure disturbance. The potential magnitude of this behavior
was investigated for a leakpath flow that would produce a room change-over rate of one-half
volume per hour at undisturbed flow conditions. This assumption does not preclude the
possibility that there may be significantly different change-over rates to be investigated on a
case-by-case basis. The approach applied was to estimate leakpath resistance factors for
-------
Background Document for Offsite Consequence Analysis, May 1999
representative conditions and use these resistance factor estimates to evaluate the building
pressure response to a specified rate of either ammonia, chlorine, or sulfur dioxide release from a
vessel inside the building.
Leakpath flow areas were estimated based on the assumption that one-half of a volume of the
room is lost per hour (ASTM, 1986) and that the driving force for this loss is developed by air
flow around the building. For a given windspeed, the pressure differential may be estimated
using correlations based on experimental data (Blevins, 1984). The pressure differential may
then be used in conjunction with the assumed normal condition leakage rate to estimate
resistances for the in- and out-leakage paths. This leakage resistance calibration procedure also
assumed that the cross-sectional area of each leakpath was proportional to the length of the
building and that the building length was twice the building width. Engineered vent areas were
estimated based on the assumption that linear air velocities through the vent were 8 m/s (1,600
ft/mm), a value consistent with standard practice (ACGIH, 1986). Release rate of stored material
into the room (i.e., Rj) was represented as a piecewise continuous function, allowing simulation
of transient release with variable dependence on time. Room density and pressure conditions
and vent flow rates were estimated using FIRAC (Gregory and Nichols, 1986), a computer code
capable of simulating ventilation system response to accident conditions. FIRAC is a
node/branch network model in which nodes represent rooms and branches represent ducts,
blowers, and filters. In this case, leakpaths were modeled as ducts of small size, and the
ammonia release was represented as mass injection with associated evaporative energy loss.
6.1.1 Building Structural Response
Potential conditions that could be established were investigated for a 10-minute release of
liquid ammonia stored at 310 K (98 °F) and 1.4 MPa (206 psia). The simulation estimated the
ammonia injection rate required to produce an overpressure large enough to threaten structural
integrity. An overpressure value of 6,895 Pa (1 psia) was adopted for this criterion. Release
modeling predicted immediate evaporation of approximately 20 percent of the ammonia flow
with subsequent evaporation of the remaining mass. High-accident condition leakpath flows
were predicted for room volumes from 500 m3 to 10,000 m3 with 6,710 kg/10 min (14,790 lb/10
min) required to produce the 6,895 Pa (1 psia) overpressure for the 500 m room. Very large
release rates were required to approach the overpressure criterion for rooms in the 1,000 m to
10,000 m3 range. This general conclusion was verified using the single-room mass balance
model. Because ammonia has lower density than either chlorine or sulfur dioxide, the results of
this ammonia analysis also indicate that prolonged releases of chlorine or sulfur dioxide will not
threaten building integrity.
6.1.2 Building Attenuation of Release
Continuous release of either ammonia, chlorine, or sulfur dioxide into a ventilated room will
produce an increase in concentration of gas which approaches a constant level determined by the
gas injection rate, ventilation rate, and room volume. Once steady state conditions are
established the removal rate equals the injection rate and the fractional release factor equals
unity. For transient conditions occurring before establishment of a steady state, such as may
occur for ten-minute release periods, the fractional release factor (FRio) may be less than unity.
A single-room mass balance model was used to estimate FRio for set of values of 9 (ratio of
6-2
-------
Background Document for Offsite Consequence Analysis, May 1999
room volume to the amount of ammonia, chlorine, or sulfur dioxide vapor release) and Nv
(ventilation rate as number of room changes per hour) at constant rate of gas addition. The
results of these calculations are presented in Table 6-1 for ammonia and Table 6-2 for chlorine
and sulfur dioxide. Results for chlorine and sulfur dioxide are similar due to similarity in
molecular weight of the gases. As expected, for high ventilation rates and small rooms, steady
state is quickly established and fractional release factors approach unity. For larger rooms and
lower ventilation rates, building attenuation of a release can be appreciable. For cases in which
overpressurization effects are small, an analytical solution to Equation 6-1 may be used to
corroborate the results presented in Tables 6-1 and 6-2.
Pi^FsatOu) (6-1)
where:
PJti = equilibrium partial pressure of the i'th component, either ammonia, chlorine, or
sulfur dioxide at temperature Ty, Pa
Fsat = equilibrium relation between pressure and temperature for the i'th component,
either ammonia, chlorine, or sulfur dioxide
TI,J = temperature of ammonia, chlorine, or sulfur dioxide at the end of the first step K
The vapor/liquid equilibrium relationship is available in tabular form for ammonia
(ASHRAE, 1981), chlorine and sulfur dioxide (Perry and Chilton, 1973).
For negligible overpressurization, constant gas addition and constant ventilation rate,
fractional release rate is given by:
FR = 1 - (l/Nvt){ 1 - exp (-Nvt)} (6-2)
where all variables are as defined above. Evaluation of Equation 6-2 for time of 0.1667 hours
(10 minutes) and values of Nv of 1, 5,10,20, 30, and 40 hr'1 yields values of FRio of 0.08, 0.32,
0.51,0.71,0.80, and 0.85, respectively. These values are in agreement with the values of Tables
6-1 and 6-2 for the high room volume to vapor mass release case (i.e., 6 = 10) which would
produce the smallest overpressures.
6.2 SUMMARY OF CONCLUSIONS
For prolonged releases, including those occurring over ten-minute periods, failure of
buildings of industrial size would not be expected. The presence of the building serves to
" attenuate transient releases and attenuation is effective for relatively small buildings ventilated at
low rates. Attenuation is small for buildings ventilated at high rates as would occur if emergency
ventilation systems were used.
"6-3
-------
Background Document for Offsite Consequence Analysis, May 1999
Table 6-1. Ten-Minute Building Release Attenuation Factors for Continuous
Releases of Ammonia
e*
(m3/kg)
10.0
5.0
2.0
1.0
Nv
(hr-1)
0
.1
5
10
20
30
40
0
1
5
10
20
30
40
0
1
5
10
20
30
40
0
1
5
10
20
30
40
FR,0
(dim)
0.07
0.08
0.32
0.51
0.71
0.80
0.85
0.13
0.13
0.32
0.51
0.71
0.80
0.85
0.29
0.29
0.32
0.51
0.71
0.80
0.85
0.47
0.47
0.47
0.51
0.71
0.80
0.85
e*
(m3/kg)
0.5
0.25
0.05
0.02
Nv
(hr1)
0
1
5
10
20
30
40
0
1
5
10
20
30
40
0
1
5
10
20
' 30
40
0
1
• 5
10
20
30
40
FR10
(dim)
0.67
0.67
0.67
0.67
0.71
0.80
0.85
0.83
0.83
0.83
0.83
0.83
0.83
0.85
0.97
0.97
0.97
0.97
0.97
0.97
0.97
0.99
0.99
0.99
0.99
0.99
0.99
0.99
* Values of 9 in m3/kg can be converted to values of 6 in tf/lb by multiplying by 16.
6-4
-------
Background Document for Offsite Consequence Analysis, May 1999
Table 6-2. Ten-Minute Building Release Attenuation Factors for Prolonged Releases
of Chlorine and Sulfur Dioxide
0*
(m3/kg)
10.0
5.0
2.0
1.0
Nv
(hr1)
0
1
5
10
20
30
40
0
1
5
10
20
30
40
0
1
5
10
20
30
40
0
1
5
10
20
30
40
FR10
(dim)
0.02
0.08
0.32
0.51
0.71
0.80
0.85
0.03
0.08
0.32
0.51
0.71
0.80
0.85
0.08
0.08
0.32
0.51
0.71
0.80
0.85
0.15
0.15
0.32
0.51
0.71
0.80
0.85
e*
(m3/kg)
0.5
0.25
0.05
0.02
Nv
(hr-1)
0
1
5
10
20
30
40
0
1
5
10
20
30
40
0
1
5
10
20
30
40
0
1
5
10
20
30
40
FR10
(dim)
0.28
0.28
0.32
0.51
0.71
0.80
0.85
0.46
0.46
0.46
0.51
0.71
0.80
0.85
0.85
0.86
0.86
0.86
0.86
0.86
0.86
0.94
0.94
0.94
0.94
0.94
0.94
0.94
; Values of 6 in m3/kg can be converted to values of 9 in fiMb by multiplying by 16.
-------
Background Document for Offsite Consequence Analysis, May 1999
REFERENCES
American Conference of Governmental Industrial Hygienists (ACGIH), Industrial Ventilation,
ACGIH, Lansing, MI, 1986.
American Institute Hygiene Association (AIHA), "Emergency Response Planning Guidelines,"
Akron, OH, 1988-1992.
American Society of Heating, Refrigerating, and Air-Conditioning Engineers, Inc., (ASHRAE),
"ASHRAE Handbook 1981 Fundamentals," ASHRAE, Atlanta, GA, 1981.
American Society of Testing and Materials (ASTM), "Measured Air Leakage of Buildings,"
ASTM, Philadelphia, PA, 1986.
American Waterworks Research Association Research Foundation (AWWARF), "Model RM
Plan for the Water Industry" February 1998 (Draft).
Barr, S. and Gifford, F.A., "The Random Force Theory Applied to Regional Scale Tropospheric
Diffusion," Atmospheric Environment 26A pp. 1053-1062, 1987.
Blevins, R.D., "Applied Fluid Dynamics Handbook" Van Nostrand Rheinhold, New York, NY,
1984.
Blewitt, D.N., J.F. Yohn and D.L. Ermak, "An Evaluation of SLAB and DEGADIS Heavy Gas
Dispersion Models Using the HF Spill Test Data," in CCPS, 1987.
Briggs, G.A., "Diffusion Estimates for Small Emissions," ATDL Contribution File No. 79,
National Oceanic and Atmospheric Administration, Atmospheric Turbulence and Diffusion
Laboratory, Oak Ridge, TN, 1973a.
Briggs, G.A., "Lift Off of Buoyant Gas Initially on the Ground," Environmental Research
Laboratories, Air Resources, Atmospheric Turbulence and Diffusion Laboratory, ATDL
Contribution File No. 87 (draft), National Oceanic and Atmospheric Administration, Oak Ridge,
TN, 1973b.
Briggs, G.A., "Plume Dispersion in the Convective Boundary Layer. Part III: Analysis of
CONDORS Field Experiment Data," Journal of Applied Meteorology, 32 pp. 1388-1425,1993.
Brighton, P.W.M., "Continuous Chlorine Releases Inside Buildings: Concentrations on Emission
to Atmosphere," SRD R 468, United Kingdom Atomic Energy Authority Health and Safety
Directorate, Culcheth, UK, 1989.
Britter, R.E.^ "Atmospheric Dispersion of Dense Gases," Annual Review of Fluid Mechanics, 21.
pp. 317-344, 1989.
R-l
-------
Background Document for Offsite. Consequence Analysis, May 1999
Britter, R.E. and S.T. Cole, "The Evaluation of Technical Models Used for Major Accident
Hazard Installations," Report # EUR 14774 EN, Directorate-General, Science, Research and
Development, Commission of the European Communities, Brussels, 1994.
Carpenter, R.J., R.P. Cleaver, P.J. Waite and M.A. English, "The Calibration of a Simple Model
for Dense Gas Dispersion Using the Thorney Island Phase I Trials Data," United Kingdom
Health and Safely Executive, Second Symposium on Heavy Gas Dispersion Trials at Thorney
Island, Sheffield, UK, 1986.
Center for Chemical Process Safety (CCPS), "Guidelines for Use of Vapor Cloud and Source
Dispersion Models, -with Worked Examples", American Institute of Chemical Engineers, New
York, NY, 1986.
Center for Chemical Process Safety (CCPS), International Conference on Vapor Cloud
Modeling, Boston, MA, November 2-4, 1987, American Institute of Chemical Engineers, New
York, NY, 1987.
Center for Chemical Process Safety (CCPS), "Guidelines for Chemical Process Quantitative
Risk Analysis" American Institute of Chemical Engineers, 1989.
Center for Chemical Process Safety (CCPS), International Conference and Workshop on
Modeling and Mitigating the Consequences of Accidental Releases of Hazardous Materials, New
Orleans, LA, May 20-24, 1991, American Institute of Chemical Engineers, New York, NY,
1991.
Center for Chemical Process Safety (CCPS), International Conference and Workshop on
Modeling and Mitigating the Consequences of Accidental Releases of Hazardous Materials, New
Orleans, LA, September 26-29,1995, American Institute of Chemical Engineers, New York, NY
1995.
Chamberlain, A.C. and R.C. Chadwick, "Deposition of Airborne Radio-iodine Vapour,"
Nucleonics 8, 22-25: Chamberlain, A.C. (1966), "Transport of Gases To and From Grass and
Grass-Like Surfaces," Proceedings of the Royal Society A290. 236-265,1953.
Chan, S.T., H.C. Rodean and D.N. Blewitt, "FEM3 Modeling of Ammonia and Hydrofluoric
Acid Dispersion," in CCPS (1987) pp. 116-154,1987.
1 , ;', r , ,in
Clough, P.N., D.R. Grist and CJ. Wheatley, "The Mixing of Anhydrous Hydrogen Fluoride with
' Moist Air," International Conference on Vapor Cloud Modeling, American Institute of Chemical
Engineers, New York, pp. 39-55,1987.
Cox, R.A. and R.J. Carpenter, "Cloud Dispersion Model for Hazard Analysis," Proc. 1st Heavy
Gas and Risk Assessment Symposium, Battelle Institute, Frankfurt, West Germany, 1979.
Erisman, J.W. and G.P.J. Draaijers, "Atmospheric Deposition in Relation to Acidification and
Eutrophication," ISBN 0-444-82247-X, Elsevier, Amsterdam, The Netherlands, 1996.
R-2
-------
Background Document for Offsite Consequence Analysis, May 1999
Ermak, D.L., "User's Manual for the SLAB Model, An Atmospheric Dispersion Model for
Denser-than-Air Releases," Lawrence Livermore Laboratory, Livermore, CA, 1989.
Fryer, L.S. and G.D. Kaiser, "DENZ - A Computer Program for the Calculation of the
Dispersion of Dense Toxic or Explosive Gases in the Atmosphere," United Kingdom Atomic
Energy Authority and Reliability Directorate Report SRD R152, Culcheth, Cheshire, UK, 1979.
Gephart, L. and S. Moses, "An Approach to Evaluate the Acute Imports from Simulated
Accidental Releases of Chlorine and Ammonia," Plant/Operation Progress, Vol. 8, no. 1, pp. 8-
11,1989.
Gregory, W.S. and B.D. Nichols, "FIRAC User's Manual: A Computer Code to Simulate Fire
Accidents in Nuclear Facilities," NUREG/CR-4561, Los Alamos National Laboratory, Los
Alamos, NM, 1986.
Goldwire, Jr., H.C., T.G. McRae, G.W. Johnson, D.L.Hipple, R.P. Koopman, J.W. McLure, L.K.
Morris, and R.T. Cedarwall, "Desert Tortoise Series Data Report — 1983 Pressurized Ammonia
Spills," Lawrence Livermore National Laboratories Report UCID-20562, Livermore, CA, 1985.
Hanna, S.R. and R.P. Hosker, "Atmospheric Removal Processes for Toxic Chemicals," ATDL
Contribution File No. 80/25, Air Resources, Atmospheric Turbulence and Diffusion Laboratory,
National Oceanographic and Atmospheric Administration, Oak Ridge, TN, 1980.
Hanna, S.R., T.C. Chang and D.G. Strimaitis, "Hazardous Gas Model Evaluation with Field
Observations," Atmospheric Environment, 27A pp. 2265-2285, 1993.
Havens, J.A. and T.O. Spicer, "/4 Dispersion Model for Elevated Dense Gas Chemical
Releases," Volume I & II, EPA 450/4-88-006a/b, U.S. Environmental Protection Agency,
Research Triangle Park, NC, 1988.
Heinrich, M.E., E. Gerold and P. Wietfeld, "Large Scale Propane Release Experiments Over
Land at Different Atmospheric Stability Classes," Journal of Hazardous Materials 20, pp 287-
301, 1988: Corrigendum, Journal of Hazardous Materials 22, pp 407-413, 1989. For more
details, see the report by the same authors, APraxisgerechte Bestimmung der Zundentfernungen
bei der Freisetzung Schwere Gases," BMFT 326-7591-01-RG-8402, TUV (Technischer
Uberwachungs-VereinNorddeutscheland e.V.), Hamburg, 1988.
Hoot, T.G., R.N. Meroney and J.A. Peterka, "Wind Tunnel Tests of Negatively Buoyant
Plumes," Report CER73-74TGH-RNM-JAP-13, Fluid Dynamics and Diffusion Laboratory,
Colorado State University, 1973.
Hosker, Jr., R.P., "Estimates of Dry Deposition and Plume Depletion over Forests and
Grassland," in Physical Behavior of Radioactive Contaminants in the Atmosphere, IAEA
STI/PUB/354, pp. 291-, 1974.
R-3
-------
Background Document for Offsite Consequence Analysis, May 1999
lanello, V., P.M. Rothe, G.B. Wallis, R. Diener and S. Schreiber, "Aerosol Research Program:
Improved Source Term Definition for Modeling the Ambient Impact of Accidental Releases of
Hazardous Liquids,"6th International Symposium on Loss Prevention arid Safety Promotion in
the Process Industries, June, 1989, Oslo, Norway, 1989.
Iman, R.L. and M.J. Shortencarier, "A Fortran 77 Program and User's Guide for the Generation
of Latin Hypercube and Random Samples for Use with Computer Models," NUREG/CR-3624
(SAND83-2365), prepared for the U.S. Nuclear Regulatory Commission, Washington, D.C.,
1984.
International Institute of Ammonia Refrigeration (IIAR), "Guidelines for: Ammonia Machinery
Room Ventilation," Bulletin No. Ill, IIAR, Washington, DC, 1991.
Jagger, S.F., "Development of CRUNCH: A Dispersion Model for Continuous Releases of
Denser-than-Air Vapor into the Atmosphere," United Kingdom Atomic Energy Authority Safety
and Reliability Directorate Report SRD R229, 1983.
Johnson, D. J., "Prediction of Aerosol Formation from the Release of Pressurized, Superheated
Liquids to the Atmosphere" in CCPS, 1991.
Kaiser, G.D., "Examples of the Successful Application of a Simple Model for the Atmospheric
Dispersion of Dense, Cold Vapors to the Accidental Release of Anhydrous Ammonia from
Pressurized Containers," United Kingdom Atomic Energy Authority Safety and Reliability
Directorate Report SRD Rl 50, Culcheth, UK, 1979.
Kaiser, G.D., "A Review of Models for Predicting the Dispersion of Ammonia in the
Atmosphere," Plant/Operations Progress, Vol. 8 No. 1, pp 58-64,1989.
Keenan, J.H. and F.G. Keyes, Steam Tables, John Wiley and Sons, Inc., New York, NY, 1969.
Lonsdale, H., "Ammonia Tank Failure - South Africa," Ammonia Plant Safety 17 (1975), 126-
131,1975.
Lees, F.P., "Loss Prevention in the Process Industries" Butterworths, London, UK, 1980.
Markham, R.S., "A Review of Damages from Ammonia Spills," presented at the 1986 Ammonia
Symposium on Safety in Ammonia Plants and Related Facilities, American Institute of Chemical
Engineers, Boston, MA, 1986.
" McMullen, G., "A Review of the llth May Ammonia Truck Accident," City of Houston Health
Department Report (unnumbered), Houston, TX, 1976.
McQuaid, J. (Ed.), "Heavy Gas Dispersion at Thorney Island," Journal of Hazardous Materials
H (Special Issue), 1986
Meroney, R N., "Lift-Off of Buoyant Gas Initially on the Ground," J. Industrial Mathematics,
Vol. 5,1979.
R-4
-------
Background Document for Offsite Consequence Analysis, May 1999
Mueller, S.F. and L.M. Reisinger, "Measured Plume Width versus Sampling Time: A Look
Beyond 10 Kilometers," Atmospheric Environment 20 pp> 895-900,1986.
Muralidhar, R., G.R. Jersey, FJ. Krambeck and S. Sundaresan, "A Two-Phase Model for
Subcooled and Superheated Liquid Jets," in CCPS, 1995.
Nair, S.K., D.B. Chambers, S.H. Park, Z.R. Radonjic, P.T. Courts, C.J. Lewis, J.S. Hammonds,
and P.O. Hoffman, "Review of Models for Determining Consequences of UFg Release,"
NUREG CR/6481, U.S. Nuclear Regulatory Commission, Washington, D.C., 1997.
National Oceanic and Atmospheric Administration (NOAA) and U.S. Environmental Protection
Agency (USEPA), "User's Manual for the ALOHA Model, ALOHA 5.0, Areal Locations of
Hazardous Atmospheres," Washington, D.C., 1995.
National Transportation Safety Board (NTSB), "Railroad Accident Report; Chicago, Burlington
and Quincy Railroad Company Train 64 and Train 824 Derailment and Collision with Tank Car
Explosion, Crete, Nebraska, February 18, 1969," Report Number NTSB-RAR-71-2, Washington,
D.C., 1971.
National Transportation Safety Board (NTSB), "Railroad Accident Report; Louisville and
Nashville Railroad Company Freight Train Derailment and Puncture of Anhydrous Ammonia
Truck Cars at Pensacola, Florida, November 9, 1977," Report Number NTSB-RAR-78-4,
Washington, D.C., 1978.
Ooms, G., "A New Method for the Calculation of the Plume Path of Gases Emitted by a Stack,"
Atmospheric Environment 6, pp. 899-909, 1972.
Perry, R.H. and C.H. Chilton (ed.), Chemical Engineer's Handbook, McGraw-Hill Book Co.,
New York, NY, 1973.
Picknett, R.G., "Field Experiments on the Behavior of Dense Clouds," Chemical Defense
Establishment Report PTN, IL 1154-78-1, Porton Down, UK, 1978.
Resplandy, A., Chimie et Industrie Genie Chimique, 102. pp. 691-702 (1969).
Scarborough, J.B., Numerical Mathematical Analysis, Johns Hopkins Press, Baltimore, MD,
1966.
Schotte, W., "Fog Formation of Hydrogen Fluoride hi Air," Ind. Chem. Eng. Chem. Res. 26, pp
300-306,1987.
Science Applications International Corporation (SAIC), "SAIC's Computer Programs for
Modeling the Atmospheric Dispersion of Hazardous Vapors-Model Description and User's
Guide," prepared by Science Applications International Corporation, McLean, VA, 1994.
R-5
-------
Background Document for Offsite Consequence Analysis, May 1999
Smith, J.M. and H.C. Van Ness, Introduction to Chemical Engineering Thermodynamics,
McGraw-Hill Book Co., New York, NY, 1975.
, • " ' ' ! ' !
South Coast Air Quality Management District (SCAQMD), "Supporting Document for:
Proposed Rule 1410: Hydrogen Fluoride Storage and Use," El Monte, CA, 1991a.
South Coast Air Quality Management District (SCAQMD), "Guideline to Comply with Proposed
Rule 1410: Hydrogen Fluoride Storage and Use," El Monte, CA, 1991b.
ii.i ' "' , '
Spicer, T.O., J.A. Havens and L.E. Kay, "Extension of DEGADIS for Modeling Aerosol
Releases," in CCPS (1987) pp. 416-438,1987.
The Fertilizer Institute (TFI), "Guidance Document for Fertilizer. Retailers/Wholesalers/
Distributors, SIC 5191, EPA's Risk Management Program, Clean Air Act Section 112(r),
Anhydrous Ammonia, Aqua Ammonia," Washington, D.C., 1998.
United States Environmental Protection Agency (USEPA), "Technical Guidance for Hazards
Analysis," in collaboration with the Federal Emergency Management Agency and the U.S.
Department of Transportation, Washington D.C., 1987.
United States Environmental Protection Agency (USEPA), "User's Guide for the DEGADIS 2.1
Dense Gas Dispersion Model," EPA-450/4-89-019 (NTIS PB 90-213893), Research Triangle
Park, NC, 1989.
United States Environmental Protection Agency (USEPA), "Evaluation of Dense Gas Simulation
Models," EPA-450/4-90-018, prepared by TRC Environmental Corporation under EPA Contract
68-02-4399,1991.
Wheatley, C.J., P.W.M. Brighton and A.J. Prince, "Comparison Between Data from the Heavy
Gas Dispersion at Thorney Island and Predictions of Simple Models," United Kingdom Atomic
Energy Authority Safety and Reliability Directorate Report, Warrington, UK, 1986.
Wheatley, C.J., "Discharge of Liquid Ammonia to Moist Atmospheres - Survey of Experimental
Data and Model for Estimating Initial Conditions for Dispersion Calculations", SRD/HSE/R 410,
United Kingdom Atomic Energy Authority, Culcheth, UK, April, 1987.
Wilson, D.J., Concentration Fluctuations and Averaging Time in Vapor Clouds, Center for
Chemical Process Safety of the American Institute of Mechanical Engineers, New York, NY,
1995.
Woodward, J.L. and A. Papadourakis, "Modeling of Droplet Entrainment and Evaporation in a
Dispersing Jet" in CCPS (1991), pp. 147-167,1991.
Woodward, J.L., J. Cook and A. Papadourakis, "Modeling and Validation of a Dispersing
Aerosol Jet,'* Journal of Hazardous Materials 44, pp 195-207,1995.
R-6
-------
Background Document for Offsite Consequence Analysis, May 1999
Woodward, J.L., "Improving the Effect of Atmospheric Stability Class for Dispersion
Modeling," Process Safety Progress 17, pp 1-8, 1998.
•\ _•,_
Yersel, M., R. Goble and J. Morrill, "Short Range Dispersion Experiments in an Urban Area,"
Atmospheric Environment, 1/7 pp. 275-282, 1993.
R-7
-------
APPENDIX A
SAIC'S COMPUTER PROGRAMS FOR
MODELING THE ATMOSPHERIC DISPERSION
OF LARGE SCALE ACCIDENTAL
RELEASES OF HAZARDOUS VAPORS IN
INDUSTRIAL ENVIRONMENTS
-------
Appendix A: May 1999
TABLE OF CONTENTS
Page
A.1 INTRODUCTION A-l
A.2 SACRUNCH A-l
A.3 SADENZ... A-l
A.4 SAPLUME A-2
A.5 COMMON FEATURES OF THE COMPUTER PROGRAMS A-3
A.5.1 VAPORS WITH COMPLEX THERMODYNAMIC PROPERTIES A-3
A.5.2 BEHAVIOR IN THE PASSIVE REGIME A-3
A.5.3 PROBABILISTIC SHELL A-4
A.5.4 DRY DEPOSITION A-4
A.5.5 CHEMICAL AGENTS A-4
A.5.6 MODELING OF MIXTURES A-4
A.5.7 MODELING VAPOR CLOUD EXPLOSIONS A-4
A.6 DEFINITION OF "SOURCE TERMS" FOR THE CODES A-4
A.7 COMPARISONS WITH EXPERIMENTS AND OTHER MODELS A-5
LIST OF FIGURES
Page
Figure A-l. Predictions for Concentration as a Function of Non-Dimensionalized Time
from the SRD and SADENZ Models Compared with Area-Averaged
Concentrations Determined from the Data for Trials 13 and 17 A-10
Figure A-2. A Comparison of the BG/C&W and SADENZ Models with Data Derived
from the Photographic Records of Trial 13 by theUKAEA SRD A-ll
Figure A-3a. Temperature Data at 100m, Height of 1.0m A-12
Figure A3-b. Temperature Data at 100m, Height of 2.5m A-13
Figure A-3c.DTI Temperature Data at 800m, Height of 1.0 m A-14
Figure A-4. Maximum Observed Concentration and Maximum Predicted Concentrations
Using DEGADIS and Pasquill-Hanna Gaussian Plume Model for DT4 A-l5
Figure A-5. Comparison Among DEGADIS, SACRUNCH, and Data "Goldfish" HF
Experiments A-16
Figure A-6. FEM3, SACRUNCH, and AMOCO Series Test 1: (a) 300 m, (b) 1000 m -
Height: 1 m A-17
Figure A-7. Maximum HF Concentration vs. Time at an Elevation of 1m for Two Downwind
'Locations for AMOCO HF Spill Test 3: (a) x = 300 m, (b) x = 1000 m A-18
-------
Appendix A: May 1999
LIST OF TABLES
Page
Table A-l. Comparison Between SAPLUME and Experiment A-7
Table A-2. Comparison Between Schotte's Results and Theory Mixing HF Vapor and Air
with 80% RH A-8
Table A-3. Comparison Between Experimental Results and Theory A-9
n
-------
Appendix A: May 1999
A.1 INTRODUCTION
SAIC makes use of three atmospheric dispersion computer programs to simulate the range of
possible large-scale accidental releases of hazardous vapors from units in the chemical,
petrochemical or refining industries. They are known as SADENZ, SACRUNCH, and
SAPLUME. The purpose of this appendix is to describe the models that are included in the
computer codes, to give the reader a representative selection of comparisons with data, and to
assess how the computer programs compare with other state-of-the art models.
A.2 SACRUNCH
SACRUNCH is a computer program that models the atmospheric dispersion of continuous
releases of heavy vapors at ground level. SACRUNCH has its origins in an earlier code,
CRUNCH (Jagger, 1983)1, that was developed in the UK with .funding provided by the UK
Health and Safety Executive. SACRUNCH has also undergone extensive modifications during
the 1980s. It has been validated against:
• Continuous releases of heavy vapors that were done at Thorney Island;
• The "Desert Tortoise" series of large scale releases of anhydrous ammonia at the Nevada
Test Site; and
• The "Goldfish" series of large-scale releases of anhydrous HF.
The important features of SACRUNCH are as follows:
• Accepts input describing the dimensions of the source (which is approximated as a
rectangle through which the vapor is flowing), the initial mass flux of heavy vapor, the
initial mass flux of air (if any), the initial momentum flux, and the initial temperature.
• Takes account of:
Gravitational slumping
Entrainment of air at the edge of the cloud during the slumping
- Entrainment of air at the top of the cloud that is strongly influenced by the presence of
a density gradient
- Conservation of momentum
- Heating of the cloud by the ground
- Eventual transition to behavior as a passive plume when the difference in density
between the cloud and the surrounding air becomes small
A.3 SADENZ
SADENZ models the atmospheric dispersion of puff releases of heavy vapors. SADENZ is
based on a code called DENZ (Fryer and Kaiser, 1979) which was developed during the late '70s
1 See References in the main report for a list of references.
__
-------
Appendix A: May 1999
and early '80s at the Safety and Reliability Directorate in the United Kingdom. Funding for the
development of DENZ was provided by the United Kingdom Health and Safety Executive,
which is the body that regulates the chemical industry in the UK. DENZ has been used in the
UK's CIMAH (Control of Industrial Major Accident Hazards) program. CIMAH is a risk
management program, which has elements in common with California's Risk Management and
Prevention Program and other state-mandated risk management programs in the United States.
SADENZ has been updated from tune to time during the 1980s to take account of
experimental and theoretical developments. In particular, considerable modifications were made
on the basis of the Thorney Island Experiments, which are the best available large-scale
experimental results on the atmospheric dispersion of puffs of heavy vapor. SADENZ compares
well with the Thomey Island results.
The important features of SADENZ are as follows:
• Accepts input on the initial configuration, density, temperature, and composition of a
heavy vapor puff
• Takes into account:
- Gravitational slumping
- Entrainment of air at the edge of the cloud during the slumping
- Entrainment of air at the top of the cloud that is strongly influenced by the presence of
a density gradient
- Conservation of momentum
- Heating of the cloud by the ground
- Eventual transition to behavior as a passive plume when the difference in density
between the cloud and the surrounding air becomes small
A.4 SAPLUME
The SAPLUME computer program simulates the atmospheric dispersion of heavy or buoyant
vapors emitted from stacks, relief valves, or ruptured pipework. It is based on the well-known
Ooms (1972) model that was originally developed by workers at Shell's research laboratories in
the Netherlands.
The important features of SAPLUME are as follows:
• Accepts data describing the initial dimensions of the plume or jet, its orientation, its
momentum flux, the initial mass flux of heavy gas, the initial mass flux of air (if any),
and the temperature and density.
* Takes into account entrainment of air by the following mechanisms:
- Entrainment into a high speed jet
- Entrainment into a rising or falling buoyant or dense plume due to the action of its
internally generated turbulence
-------
Appendix A: May 1999
- Action of the ambient turbulence in the surrounding atmosphere
• Also takes into account:
Temperature gradient in the atmosphere
Velocity gradient in the atmosphere
Conservation of momentum in both the horizontal and vertical directions
- The influence of gravitational forces
The rate of change of temperature of the plume due to the entrainment of air
Transition to behavior as a passive plume when either the density difference becomes
small or plume's velocity falls to close to that of the surrounding air
For buoyant plumes, termination of plume rise by a number of alternative
mechanisms that take into account both mechanical and convective turbulence in the
atmosphere and the presence of inversion lids.
S APLUME has been successfully compared to wind tunnel experiences on the release of
heavy plumes from stacks. There are no large-scale data against which a comparison can be
made.
A.5 COMMON FEATURES OF THE COMPUTER PROGRAMS
A.5.1 Vapors with Complex Thermodynamic Properties
For vapors with complex thermodynamic properties, such as HF and ammonia, the user can
input an array of vapor cloud properties: air/gas mass mixing ratio, cloud temperature and cloud
density. The computer programs then interpolate among the values in this array each time a
temperature or a density is needed. The array must be independently calculated.
A.5.2 Behavior in the Passive Regime
In all three computer programs, once the difference hi density between the plume and the
surroundings becomes small, the model reverts to the traditional Gaussian formulation. There
are two user-selected alternatives for the vertical and horizontal standard deviations. The first is
due to Briggs (1973), who allows a further choice between a rural and an urban case. Briggs'
model is virtually identical to that in ALOHA, the Gaussian model that was developed by the
National Oceanic and Atmospheric Administration (NOAA, 1988) and has been widely
disseminated among agencies and industry. Test runs have shown that the SAIC codes and
ALOHA give the same results for the dispersion of passive plumes. The Briggs model has also
been included in Technical Guidance issued by the EPA (USEPA, 1987).
The second option that a user has is to call on standard deviations in a formulation given by
Hosker (1974) wherein the vertical standard deviations are functions of meteorological roughness
length, thus allowing the user to account for sites where the terrain differs widely. However,
Hosker's model does not take into account the additional dilution caused by thermal effects in
urban areas, so that the Hosker approach tends to be conservative for city sites.
_
-------
Appendix A: May 1999
A.5.3 Probabilistic Shell
All three of SAIC's computer programs are embedded within a probabilistic "shell" that
allows the user to calculate various measures of risk. These measures include "f-n" lines, which
are plots of the frequency f with which a given consequence equal exceeds a magnitude n. The
quantity n is the number of people exposed above various levels chosen by the user, such as the
LCJO (the concentration that would prove fatal to 50 percent of those exposed to it), the IDLH
(Immediately Dangerous to Life or Health) or the ERPG (Emergency Response Planning
Guideline), "n" can also be the areas affected by concentrations exceeding the levels given
above, or the distances downwind to which those areas extend. The codes can also be used to
calculate quantities such as mean societal risk and individual risk.
The computer programs generate the information on risk by carrying out the calculations for
a variety of different weather conditions (the codes accept the meteorological data in the STAR
format) and wind directions, as well as different release scenarios. The codes keep track of the
frequencies associated with each combination of scenario, atmospheric stability category,
windspeed, and wind direction and contain subroutines that store the results and manipulate them
into the various measures of risk.
A.5.4 Dry Deposition
AH of the computer programs contain the option of activating a dry deposition mechanism
that depletes the plume by deposition of reactive gases or particulates onto the ground, using the
well-established concept of dry deposition velocity.
A.5.5 Chemical Agents
All of the computer programs contain the option of modeling chemical agent releases. The
models contain chemical agent-specific vapor pressure relationships and probit equations for
consequence assessment.
A.5.6 Modeling of Mixtures
All of the computer programs can handle the releases of mixtures of ideal vapors.
• A.5.7 Modeling Vapor Cloud Explosions
All of the computer programs can predict the potential for vapor cloud explosions. They can
predict if a cloud will make contact with a given ignition source. If such contact is predicted,
then the models also calculate the inventory of materials in the cloud, the cloud geometry and the
time elapsed since the release.
A.6 DEFINITION OF "SOURCE TERMS" FOR THE CODES
SAIC believes that, among them, the three computer programs described above are capable
of simulating the atmospheric dispersion of most of the accidental releases of toxic or flammable
A-4
-------
Appendix A: May 1999
vapors that might occur at industrial facilities. Howeter, the use of codes is not trivial, because
the codes are deliberately not set up as "black boxes" that purport to simulate every conceivable
case without much thoughtful effort by the user. The range of possible releases from industrial
facilities is so large, and the variety of subtle physical and chemical phenomena to be considered
is so wide, that no dispersion model can be used without careful consideration of whether it is
appropriate for the type of release being considered and what approximations are needed to
formulate the input in a way that the code can handle without sacrificing physical reality.
SAIC uses the full array of models that are available for calculating release rates, such as the
standard formulas for evaporating pools. The techniques are reviewed in the "Guidelines for Use
of Vapor Cloud Dispersion Models," published by the Center for Chemical Process Safely
(Hanna and Drivas, 1987) or the CCPS' Quantitative Risk Assessment Guidance Document
(CCPS, 1989). The user makes use of these techniques to generate input for the atmospheric
dispersion computer programs.
A.7 COMPARISONS WITH EXPERIMENTS AND OTHER MODELS
SAIC's computer programs, including their theoretical basis and how they compare with
experiment, are documented in SAIC's internal report, "SAIC's Computer Programs for the
Atmospheric Dispersion of Hazardous Vapors." For the purposes of this summary, a selection of
comparisons is shown in Figures A-l to A-7. Much of the modeling basis for SADENZ and
SACRUNCH depends on the series of large-scale releases of Freon-12 (refrigerant-12) that were
carried out at Thorney Island (McQuaid, 1985). Figure A-l shows how SADENZ compares with
the concentration results of two of the trials from Thorney Island and with the DENZ model that
has been independently updated in the UK by the Safety and Reliability Directorate (Wheatley et
al., 1986). Agreement between experiment and both models is good. Figure A-2 shows
experimental results for the position of the puff centerline, the puff height, and the area of the
base of the puff as a function of time. In addition, Figure A-2 shows the predictions of a model
developed in the UK by British Gas Corporation and Cremer and Warner (Carpenter et al.,
1986). Again, agreement between both models and the data is good.
Figure A-3 shows some results from the "Desert Tortoise" series of large-scale anhydrous
ammonia releases (Goldwire et al., 1985). The figure contains experimentally measured
temperatures as a function of tune, compared with the predictions of SACRUNCH. The
predicted temperature of a cloud is a very sensitive function of, for example, the predicted rate of
entrainment of air. Small differences in the entrainment can easily lead to temperature
•• differences of tens of degrees. Therefore, it is gratifying that SACRUNCH does so well with
modeling that was fixed by comparison with the Thorney Island Experiments, hi which
temperature effects were unimportant. Figure A-4 shows how SACRUNCH and DEGADIS
(Spicer et al., (1987) predict the peak centerline concentrations in the fourth of the Desert
Tortoise experiments, DT4. SACRUNCH reproduces the data with a goodness of fit at least
comparable to that of DEGADIS. Incidentally, the differences between the predictions of the
two models and between each of the models and the data lie within the combined error band that
is to be expected from the models and the data.
Figure A-5 shows how SACRUNCH and DEGADIS reproduce the predicted peak centerline
concentrations that were observed hi the "Goldfish" series of HF experiments (Blewitt et al.,
A-5
-------
Appendix A: May 1999
1987). Again, the way the two codes fit the data is comparable. Figures A-6 and A-7 show how
SACRUNCH and FEM3 (Chen et al., 1987) fit some of the "Goldfish" data. These figures are
instructive because, while SACRUNCH (and DEGADIS) are examples of relatively simple
"box" models that, hi the main, deal with averaged quantities, FEM3 is a sophisticated three
dimensional code that solves the underlying conservation equations and turbulence closure
assumptions numerically. The goodness of fit achieved by SACRUNCH, both to the
concentration data and the time of arrival and time of departure of the plume, is at least
comparable to that achieved by FEM3. This confirms an opinion that is widely accepted in the
community of those who develop and use heavy vapor dispersion models, namely that, for the
purposes of risk assessment, the simple box models do at least as good a job as the more
sophisticated three dimensional codes at a fraction of the cost.
As discussed above, the code SAPLUME is based on the well-known Ooms model for stack
releases. The comparable module hi DEGADIS is based on exactly the same model.
Unfortunately, there are relatively few data against which to compare the releases of heavy
vapors from stacks or as jets from relief valves. One good data set of heavy vapor releases from
stacks has been prepared by Hoot et al. using a wind tunnel at Colorado State University (Hoot et
al., 1973). Table A-l shows a selection of data from those experiments and how they compare
with the predictions of SAPLUME. As can be seen, agreement is quite good. The authors of
DEGADIS used the same data set to validate the Ooms module. SAIC has reviewed the
DEGADIS manual (Havens and Spicer, 1988), where the comparisons are presented, and has
concluded that SAPLUME reproduces the data with a goodness of fit that is at least as good as
that achieved by DEGADIS. Finally, Tables A-2 and A-3 show a comparison between the
predictions of the HF model in SACRUNCH and data on the mixing of HF and air or nitrogen
collected by workers at DuPont (Schotte, 1987). Agreement is good.
In conclusion, SAIC's toxic vapor dispersion models compare well with data from a variety
of sources and with other state-of-the-art models.
A-6
-------
Table A-l. Comparison Between SAPLUME and Experiment
Exit
Velocity
3.712
15.022
11.137
7.424
7.511
7.597
3.799
6.216
Velocity
Ratio
5.1
20.72
15.3
10.2
10.32
10.1
5.05
8.80
Specific
Gravity
1.077
1.572
2.158
3.278
4.16
1.608
2.217
3.044
Height of Rise
(IN)
EXPT.
3.7
10.8
7.78
5.24
5.12
3.22
1.38
2.88
SAPLUME
3.6
13
9.45
5.88
6.05
3.77
1.64
3.2
Touchdown Dist
(IN)
EXPT.
SAPLUME
Touchdown Cone.
(kg/m3)
EXPT. SAPLUME
beyond experimental range
56
28
13
9.7
31.5
21.23
14.05
50
24
10.5
7.8
31.2
14
13
0.006
0.015
0.038
0.054
0.0052
0.013
0.017
0.004-
0.015
0.055
0.08&
0.0037
0.01
0.017
Spot comparisons of SAPLUME predictions with experiments of Hoot, Meroney and Peterka. "Wind Tunnel Tests of
Negatively Buoyant Plumes" PB-231 590 (1973)
I
-------
00
Table A-2. Comparison Between Schotte's Results and Theory
...... Mixing HF Vapor and' Air with 80% RH
Total Mol% Mixing
HF
0.373
0.696
1.13
1.52
1.90
2.86
5.55
8.85
12.07
15.68
19.63
1 Air/HF
Ratio1
387
207
127
94.7
75.8
50.3
25.9
16.3
11.9
9.18
7.34
Air
298.9
299.1
299
298.9
298.8
299
298.7
298.7
298.5
25.6
25.7
Temperature
HF-N2
299.6
299.7
299.7
299.8
299.7
298.9
299.2
298.8
298.1
25.0
25.2
FogExp
300.1
303.8
305.5
306.8
307.2
308.9
308.6
305.0
296.2
12.0
7.0
Fog (SAIC)
300.8
302.8
304.2
305.1
305.7
306.9
307
303.8
299.2
15.1
8.8
y 2
AHF
expt
0.261
0.306
0.346
0.370
0.389
0.423
0.490
0.553
0.620
0.687
0.737
(SAIC)
0.263
0.302
0.340
0.351
0.374
0.421
0.476
0.524
0.597
0.677
0.726
% HF Condensed
expt
26.7
29.5
29.8
29.8
29.7
29.3
28.7
28.8
29.5
29.9
29.2
(SAIC)
26.29
28.33
29.11
28.04
28.35
29.21
27.92
26.83
28.69
28.4
27.4
mass mixing ratio
2 Mole percent of HF in
Source:
Schotte(1987)
HF/water droplets
-------
Table A-3. Comparison Between Experimental Results and Theory
Mixing a 50/50 Mixture of Nitrogen and HF with Air at 80% RH
Total Mol% Mixing
HF
0.197
0.482
0.951
1.41
1.87
2.85
5.26
7.29
10.18
12.76
14.95
1 Air/HF
Ratio1
732
298.
151
101
75.8
49.2
26.0
18.4
12.8
9.9
8.2
mass mixing ratio
Air
26.4
26.4
26.3
26.3
26.2
25.8
25.9
25.4
25.4
25.5
25.7
Temperature
HF-N2
26.7
26.6
26.6
26.4
26.4
25.7
25.6
26.7
26.2
25.8
25.8
Fog Exp
28.5
29.7
33.5
35.0
35.9
39.7
41.3
42.7
43.4
42.1
40.8
Fog (SAIC)
27.4
30.1
33.7
34.7 ,
35.8
38.7
40.4
42.7
41.9
41.1
39.9
X 2
AHF
expt
0.215
0.282
0.33
0.36
0.38
0.41
0.46
0.49
0.52
0.55
0.57
(SAIC)
0.213
0.280
0.331
0.361
0.382
0.415
0.465
0.493
0.526
0.554
0.577
% HF Condensed
expt
16.0
21.5
21.8
21.0
20.1
18.6
15.8
13.1
12.2
11.7
11.1
(SAIC)
10.3
17.3
18.7
18:1
18.1
17.5
15;9
14.6
14.0
13.7
13.3
2 Mole percent of HF in HF/water droplets
Source:
Schotte(1987)
-------
Appendix A: May 1999
x Experiment
_ 3RD Model
• SADENZ
i i i i i i i i
i i i i i i i i
10
Dimensionless Time
x Experiment
SRD Model
SADENZ
i i—| | | I III . I 1—I I I I I II., 1 1—I I II III., 1 1—I I I I II
Dimensionless Time
Source: Wheatley et al. (1985)
Figure A-1. Predictions for Concentration as a Function of Non-Dimensionalized
Time from the SRD and SADENZ Models Compared with Area-Averaged
Concentrations Determined from the Data for Trials 13 and 17.
A-10
-------
Appendix A: May 1999
80-,
60-
40-
0-5
•DO.
go
•*; —
o
-------
Degree Centigrade
10
I
•>
H
n
B
v
n
»s
I
d
o
o
3
\*
W
o
!•*»
I
-------
Degree Centigrade
eg
s
fe
i-
I
-------
Degree Centigrade
t
3
t
n
H
I
I
I
o
00
O
o
B
n>
s,
O
5
I
!k.
t
-------
Appendix A: May 1999
SACRUNCH.
p Experimental Data
10°.
icentration (%)
.A
<^
i
Maximum Cor
_&
o
IsJ
I
io-3
n
V
Gau
X
;sia
\
\
x
»
\
\
iP
L
\
\
\
\
"2
S
\
Ul
.
\
v,
"
S
V
n<
\
V
\
L
s
*
I
h
L\
s
y
. N
N
V
Do
\\
\ \\
k \ N
*\ \
N\
iel
-------
Appendix A: May 1999
10s-,
£: 10* -
g 10s
O
u.
10*
105^
I"
1 10"
c
1
o
O
u.
10Z
102
103
Distance (m)
10"
102
103
Distance (m)
104
Comparison of Observed and DEGADIS Predicted Plume Comparison of Observed and DEGADIS Predicted Plume
Contorllno Concentrations forTest 1 (1 Meter Height) Centerline Concentrations forTest 2 (1 Meter Height)
10=
!
¥
103
102
Legend:
p = Data
^ = DEGADIS
o =SACRUNCH
10Z
10*
Distance (m)
Comparison of Observed and DEGADIS Predicted Plume
Centorllno Concentrations forTest 3 (1 Meter Height)
Figure A-5. Comparison Among DEGADIS, SACRUNCH, and Data "Goldfish"
HF Experiments.
A-16
-------
HF Concentration (%Vol)
p
at
O £
o ft
O M
«§•
2« J^
era' b
B
cc
»
s-
ce
8
s°
O
T* ••-
eg
3
u
c
o
p
CO
1—
•n
m
8
u
at
(D
=1
8-
8
HF Concentration (%Vol)
l_
o
I
I
-------
Appendix A: May 1999
1 1 1 I 1 ' 1
(a) FEM3 (Phase Change)
40 60 80 100
Time (s)
"50 TOO 150 200 250 300 350
Time (s)
Figure A-7. Maximum HF Concentration vs. Time at an Elevation of 1m for Two
Downwind Locations for AMOCO HF Spill Test 3:
(a) x = 300 m, (b) x = 1000 m
A-18
------- |