Work Study 20/12/2005 #10072

WORK STUDY 20/12/2005
N° 10072
AMMONIA
Large-scale atmospheric dispersion tests
Réf. : INERIS - DRA - DRAG – 2005 - 10072 – RBo – Ammonia

Page 1 sur 130
This document forms a non-dissociable entity. It may only be used in its complete form.
AMMONIA
Large-scale atmospheric dispersion tests
INERIS – Accident Risks Division
Translation of the French report:
AMMONIAC : Essais de dispersion d’ammoniac à grande échelle –

INERIS-DRA-RBo-1999-20410. R. BOUET

Page 1 sur 130
FOREWORD
This report was drawn up on the basis of the information provided to INERIS and
the objective (scientific or technical) data available as well as applicable
regulations.
INERIS cannot be held liable if the information received was incomplete or

erroneous
Any findings, recommendations, suggestions or equivalent that are recorded by

INERIS as part of the services it is contracted to perform may assist with decision
making. Given the tasks entrusted to INERIS on the basis of the decree founding
the organisation, INERIS cannot be involved in the decision making process itself.
INERIS cannot therefore take responsibility in lieu and place of the decision-
maker.
The addressee shall use the results comprised in this report in whole or at least in
an objective manner. Using this information in the form of excerpts or summary
memos may take place only under the full and complete responsibility of the
addressee. The same applies to any modification made to this report.
INERIS declines any liability for any use of this report outside of the scope of the
service provided.
Author Verificateur Approbateur

NAME Rémy BOUET Olivier SALVI Bruno FAUCHER
Programme manager Scientific manager Director
Qualification Test activity
Accident Risks Division Accident Risks Division Accident Risks Division
Visa Signed Signed Signed

Page 2 sur 130
TABLE DES MATIÈRES
1. INTRODUCTION ..............................................................................................7
2. PROPERTIES OF AMMONIA...........................................................................9
2.1 Physical properties ........................................................................................9
2.1.1 General points ........................................................................................... 9
2.1.2 Thermodynamic data ................................................................................. 9
2.1.3 Solubility .................................................................................................. 10
2.1.4 Density..................................................................................................... 10
2.2 Explosibility and flammability.......................................................................11
2.2.1 Explosibility limits..................................................................................... 11
2.2.2 Self-ignition temperature.......................................................................... 11
2.2.3 Minimum ignition energy.......................................................................... 11
2.2.4 Extinguishing agents................................................................................ 11
2.3 Reactions with contaminants.......................................................................12
2.3.1 Halogens and inter-halogens................................................................... 12
2.3.2 Heavy metals ........................................................................................... 12
2.3.3 Oxidants and peroxides ........................................................................... 12
2.3.4 Acids........................................................................................................ 13
2.3.5 Other aspects .......................................................................................... 13
2.3.6 Stability .................................................................................................... 14
2.4 Toxicity ........................................................................................................14
2.4.1 General.................................................................................................... 14
2.4.2 Acute toxicology....................................................................................... 15
2.4.3 Toxicology at the workplace .................................................................... 16

Page 3 sur 130
2.4.4 Summary of results on human toxicology ................................................ 17
2.4.5 Toxicology of fauna and flora................................................................... 20
2.4.5.1 Animals.............................................................................................. 20
2.4.5.2 Aquatic life ......................................................................................... 20
2.4.5.3 Plant life............................................................................................. 21
3. ATMOSPHERIC DISPERSION...................................................................... 23
3.1 Context ....................................................................................................... 23
3.2 Atmospheric dispersion tests...................................................................... 24
3.2.1 Characteristics of accident ammonia releases......................................... 24
3.2.2 Tests by A. Resplandy ............................................................................. 26
3.2.2.1 Context .............................................................................................. 26
3.2.2.2 Test conditions .................................................................................. 27
3.2.2.3 Results .............................................................................................. 29
3.2.3 Tortoise Desert tests................................................................................ 30
3.2.3.1 Context .............................................................................................. 30
3.2.3.2 Test conditions .................................................................................. 31
3.2.3.3 Results .............................................................................................. 32
3.2.4 Unie van Kunstmest Fabrieken bv tests, Netherlands, 1972 ................... 33
3.2.5 Imperial Chemical Industries tests, United Kingdom, 1974...................... 34
3.2.6 Unie van Kunstmest Fabrieken bv tests, Netherlands, 1980 ................... 34
3.2.7 .Landskrona tests, Sweden, 1982............................................................ 35
3.2.8 FLADIS programme tests ........................................................................ 35
3.2.9 Tests conducted at Ecole des Mines, Alès .............................................. 36
3.2.10 Other atmospheric dispersion tests ................................................... 36
3.2.10.1 Summary ........................................................................................... 39
3.3 Modelling of atmospheric ammonia dispersion........................................... 39

Page 4 sur 130
3.3.1 Gaussian models ..................................................................................... 40
3.3.2 Three-dimensional models ...................................................................... 40
3.3.3 Integral models ........................................................................................ 40
3.3.4 Residual problems and improvements to be made.................................. 40
3.3.4.1 Two-phase flows ............................................................................... 41
3.3.4.2 Aerosol formation and progression.................................................... 41
3.3.4.3 Pool formation and evaporation......................................................... 41
3.3.5 Summary ................................................................................................. 42
3.4 Issues examined in this programme............................................................43
3.4.1 Impacting jets .......................................................................................... 43
3.4.2 Influence of orifice geometry.................................................................... 44
4. DESCRIPTION OF LARGE-SCALE TESTS ..................................................45
4.1 Presentation of tests conducted ..................................................................45
4.2 Description of test resources.......................................................................47
4.2.1 Description of release point ..................................................................... 47
4.2.2 Weather measurement equipment .......................................................... 52
4.2.3 Ammonia measurement chain composition and tests.............................. 53
4.2.4 Sensor instrumentation on test site ......................................................... 57
4.2.5 Acquisition equipment.............................................................................. 60
4.2.6 Video equipment...................................................................................... 62
4.2.7 Safety ...................................................................................................... 62
5. RELEASE CONDITION MEASUREMENTS...................................................63
5.1 Flow conditions in pipe ................................................................................63
5.1.1 Experimental release conditions.............................................................. 63
5.1.2 Release flow condition analyses.............................................................. 67
5.2 Weather conditions during the tests ............................................................73

Page 5 sur 130
6. MEASUREMENTS RECORDED LEEWARD FROM RELEASES................. 81
6.1 Ammonia concentration sensors ................................................................ 81
6.2 Analyses..................................................................................................... 88
6.2.1 Influence of release orifice ....................................................................... 89
6.2.2 Influence of atmospheric stability ............................................................. 90
6.2.3 Influence of obstacle placed in near field ................................................. 91
6.2.4 Influence of retention dike........................................................................ 95
6.2.5 Influence of water curtain produced using peacock tail hoses ................. 97
6.2.6 Experiment / modelling comparisons ....................................................... 98
7. CONCLUSION AND PROSPECTS ............................................................. 103
8. BIBLIOGRAPHY.......................................................................................... 105
9. LISTE DES ANNEXES ................................................................................ 111

Page 6 sur 130
1. INTRODUCTION
Within the scope of its activities relating to accidental risk, INERIS particularly
needs to determine safety perimeters around industrial facilities. Following a loss
of confinement on a facility, the risks involved may be instantaneous, such as the
explosion of flammable substances, or delayed, such as toxic substance releases.
This research programme has concentrated more specifically on the progression
of a toxic cloud formed after an accidental release of pressurised liquid ammonia
in storage.
The atmospheric dispersion study on ammonia is of major interest for two reasons.
Firstly, ammonia is a widely used substance with a large number of applications,
due to its chemical or physical properties. Secondly, ammonia is a highly toxic,
corrosive, flammable and explosive substance under certain conditions. Note that
the loss of confinement of a 22 tonne tank of ammonia on 24 March 1992 in Dakar
caused a large number of deaths (129 deaths and over 1100 injured subjects),
some of which occurred several weeks after the accident due to the toxicity of
ammonia.
This research programme was essentially financed by the French Ministry for
Territorial Development and the Environment. Six European industrial firms also
contributed: Norsk Hydro (Norway), Grande Paroisse (France), Veba Oel
(Germany), SKW Piesteritz (Germany), CEA-CESTA (France) and Rhône-Poulenc
(France). At INERIS, the programme was organised and coordinated in the
Accident Risks Division (DRA). It was started in 1995 and completed in 1999.
The main aims of this programme were:
to analyse the risks represented by facilities using quantities of ammonia of

up to a few dozen tonnes ;
to complete knowledge on the atmospheric dispersion of ammonia in open

and obstructed environments ;
and to compare the test results with atmospheric dispersion models.
To this end, a large-scale test programme was conducted using a pressurised

liquid ammonia tank. The tests took place on the CEA-CESTA site over a period
from December 1996 to April 1997. The purpose of this test campaign was to
measure the ammonia concentrations leeward from the release in order to obtain
a clearer idea on the dispersion of ammonia in the atmosphere in the case of
realistic releases, particularly in open or semi-obstructed environments.
The best-known pressurised liquefied ammonia dispersion tests are the Tortoise
Desert tests conducted by Goldwire et al. (1985) and the FLADIS tests conducted
by RISØ (1993/1994). The flow rates used during the Tortoise Desert tests and the
FLADIS tests were of the order of 100 kg/s and 0.5 kg/s, respectively. The INERIS
tests were carried out with intermediary flow rates of the order of 2 to 4 kg/s.
Page 7 sur 130
The report is organised as follows. After a description of main properties of
ammonia in chapter 2, a summary of atmospheric dispersion in general and
ammonia dispersion in particular is given in chapter 3. Then, the large-scale tests
conducted within the scope of this programme are described in chapter 4. Finally,
a summary of the analysis of the release conditions measurements and the
measurements recorded leeward from the releases is given in chapter 5 and 6,
respectively.

Page 8 sur 130
2. PROPERTIES OF AMMONIA
2.1 PHYSICAL PROPERTIES
This section is particularly based on the Air Liquide (1980), ULMANN (1985) and
SNIE (1991) documents
2.1.1 General points
Ammonia is identified as follows (see table 1):
name ammonia
CAS number 7664-41-7
EEC number 007-001-00-5
RTMD hazard code 268
UN number 1005
chemical formula NH3
molar mass 17.03 g
table 1
At atmospheric pressure and 20°C, ammonia is a colourless gas with a pungent,

irritant characteristic odour.
2.1.2 Thermodynamic data
The main thermodynamic data for ammonia is as follows:
Melting point: -77.7°C ;
Boiling point: -33.4°C to 1.013 bar abs ;
Variable vapour tension as a function of temperature (see table 2):
Temperature (°C) -77.71 -33.4 -18.7 0 4.7 20 25.7 30 50.1 78.9
Absolute 0.06077 1.013 2 4.29 5 8.56 10 11.66 20 40

pressure (bar)
table 2: Vapour tension of ammonia as a function of temperature

Page 9 sur 130
Critical temperature : 405.55 K ;
Critical pressure : 114.80 bar ;
Melting heat at 1.013 bar : 332.3 kJ.kg-1 ;
Vaporisation heat at -15°C: 1210 kJ.kg-1 (289.5 kcal.kg-1) ;
Vaporisation heat at -33.4°C : 1370 kJ.kg-1 (328 kcal.kg-1) ;
Dynamic viscosity of liquid at -33.5°C : 10.225 mPa.s.
One litre of liquid releases 947 litres of gas (expanded at 15°C, at a pressure of
1 bar).
2.1.3 Solubility
The solubility of ammonia in water is high. Table 3 gives the solubility of ammonia
as a function of temperature (WHO, 1986).
Temperature Solubility
(°C) (g/l)
0 895
20 529
40 316
60 168
table 3: Solubility of ammonia as a function of temperature
In addition, the dissolution of ammonia in water is highly exothermic: 2000 kJ per

kilogram of ammonia dissolved in water (i.e. 478.5 kcal.kg-1). As an indication, the
dissolution of one kilogram of ammonia releases enough energy to evaporate
almost one and a half kilograms.
2.1.4 Density
Gas 0.772 kg.m-3 at 0°C

0.610 kg.m-3 at 20°C
i.e. a density of 0.597 with reference to air.
Liquid variable as a function of temperature as indicated in table 4:
Temperature -40 -33.4 -20 -10 0 10 15 20 30 50 100 132.4
(°C)
Density of 690 679 659 647 634 621 617 607 592 558 452 235
liquid ammonia

Page 10 sur 130
-3
(kg.m )
table 4: Density of ammonia as a function of temperature
2.2 EXPLOSIBILITY AND FLAMMABILITY
The information given below is essentially based on the study conducted by

INERIS (Abiven, 1991) at the request of the French Ministry for the Environment.
2.2.1 Explosibility limits
In the literature, different values exist in relation to the Lower and Upper
Explosibility Limits (LEL and UEL). The French Ministry for Territorial Development
and the Environment's Industrial Environment Department data sheet referenced
TOX 003-06-1998 gives the following values: LEL 16% and UEL 25%. These
values are also given by INRS (1989) and SAX (1996). Other authors give slightly
different values: LEL 15% and UEL 28% according to NFPA (1994) and Medart
(1979), LEL 15.5% and UEL 27% according to Weiss (1985).
2.2.2 Self-ignition temperature
The self-ignition temperature of a mixture of gas or vapour is the minimum

temperature from which the mixture is the site of a sufficiently rapid chemical
reaction for a flame to appear spontaneously and propagate throughout the
mixture. The self-ignition temperature of ammonia recorded in the literature is
650°C (Chaineaux, 1991).
2.2.3 Minimum ignition energy
The ignition of a flammable mixture requires the local introduction of a certain

quantity of heat. The evaluation of this quantity of energy required to ignite the
mixture is relatively difficult.
Buckley and Husa (1962) determined the minimum ignition energy of ammonia
using a capacitive system. In this way, they obtained a minimum ignition energy of
680 millijoules for ammonia.
Kramer (1985) obtained a minimum energy of 14 millijoules for an ammonia-air

mixture. This value, very different to that obtained by Buckley and Husa,
demonstrates the difficulty quantifying the release energy of the experimental
systems used.
However, in spite of this lower value for the minimum ignition energy determined
by Kramer, an air-ammonia mixture has a higher minimum ignition energy (1 to 2
orders of magnitude) compared to most air-hydrocarbon mixtures.
2.2.4 Extinguishing agents

Page 11 sur 130
The extinguishing agents that must be used consist only of CO2 or powders
whenever there is a possibility of liquid ammonia being present. In fact, if water
comes into contact with liquid ammonia, heat is transmitted to the latter, thus
favouring its vaporisation.
2.3 REACTIONS WITH CONTAMINANTS
Ammonia is used in the manufacture of a large number of nitrous substances.

However, hazardous reactions exist with some compounds or families of
compounds.
The mixture of ammonia with a contaminant may lead to the formation of explosive
substances.
In its documentary note records (note 1024-84-76), INRS (1976) mentions

substances inducing hazardous chemical reactions with ammonia, such as nitric
acid, oxygen, boron or oxidants
2.3.1 Halogens and inter-halogens
Halogens (fluorine, chlorine, bromine, iodine) react strongly with ammonia and its
solutions. Explosive reactions may also occur with the following substances:
acetaldehyde, hypochlorous acid, potassium ferrocyanides. Explosive compounds
such as nitrogen trihalides are produced.
For example:
with chlorine, Cl2, mixtures are explosive if they are heated or if chlorine is
in excess, due to nitrogen trichloride formation ;
with bromine pentafluoride, BrF5, explosions are likely.
2.3.2 Heavy metals
Ammonia is capable of reacting with some heavy metals (silver, gold, mercury,
etc.), to produce materials liable to explode violently when they are dried:
with gold chloride (III), AuCl3, under a wide variety of conditions, the
presence of ammonia results in explosive or fulminant compounds which
explode when heated ;
with silver oxides, AgO, Ag2O, explosive substances are formed ;
with mercury, Hg, the reaction gives substances which are highly explosive
and detonate easily. No instrument containing mercury should be used if it
is liable to come into contact with ammonia (B.I.T., 1993).
2.3.3 Oxidants and peroxides

Page 12 sur 130
Ammonia reacts with a large number of oxides and peroxides: cold chlorine
peroxide, hot iodic anhydride, perchlorates, which induce a violent reaction at
around 250°C. The mixture of an oxidant compound and liquefied ammonia is
liable to explode under the effect of a shock.
For example:
with hydrogen peroxide, H2O2, ammonia dissolved in 99.6% of peroxide

gives an unstable solution which explodes violently ;
with nitryl chloride, ClNO2, the interaction is very violent, even at -75°C ;
with trioxygen difluoride, F2O3, the reaction is liable to cause inflammations

and explosions, even at -183°C. With solid ammonia, it reacts to produce
inflammations or explosions;
with oxygen, O2, if they are placed in contact in a cooling device, an

explosion is liable to occur. In addition, in the presence of ammonia, oxygen
can accelerate or induce corrosion.
2.3.4 Acids
With some acids, violent reactions are observed, such as:
with pure hypochlorous acid, HClO, ammonia, in gaseous form, explodes

on contact and releases chlorine;
with nitric acid, HNO3, an ammonia jet burns in a nitric acid atmosphere
2.3.5 Other aspects
Ammonia can also cause incandescent reactions, for example:
with boron, B, heated in a dry ammonia atmosphere;
with chromic anhydride, CrO3, ammonia gas breaks down dry trioxide with
incandescence at normal temperatures.
Ammonia can also form self-igniting mixtures:
with boron, B, heated in a dry ammonia atmosphere;
with chromic anhydride, CrO3, ammonia gas breaks down dry trioxide with
incandescence at normal temperatures.
Ammonia can also form self-igniting mixtures:
with nitric acid, HNO3, (see section 2.3.4.);

Page 13 sur 130
with chromyl dichloride, CrO2Cl2, ammonia can be ignited with this
substance.
2.3.6 Stability
At normal temperatures, ammonia gas is a stable compound; its dissociation into

hydrogen and nitrogen only starts at around 450 - 500°C. In the presence of some
metals such as iron, nickel, osmium, zinc, uranium, this decomposition starts from
a temperature of 300°C and is almost complete at around 500 to 600°C.
2.4 TOXICITY
2.4.1 General
With respect to ammonia toxicity, different published values are available for the
same given effect.
There are three fundamental reasons for this:
the lack of human testing for high concentrations ;
the disparity of the subjects making up a human sample, compared to a

sample of selected laboratory animals for which the response varies very
little from one animal to another ;
and the difficulty extrapolating results obtained from animal testing to

humans.
The toxicity of ammonia gas is associated with its very high solubility and the
alkalinity of the resulting solutions, rendering it an aggressive agent for mucosa
and lungs.
While reliable values on daphnids or rats are available, toxicity on humans is more
difficult to determine. In particular, the influence of the concentration or exposure
time is difficult to assess. With this respect, a study on "acute ammonia toxicity"
was conducted by INERIS (Auburtin, 1999).
Exposure to an ammonia-charge atmosphere may induce various types of

physical effects detailed below:
Ocular effects: may be induced by the effect of vapours, but also by liquid
projections. They are conveyed by tear production, conjunctivitis liable to be
accompanied by more or less deep corneal effects ;
Skin effects: in the form of contact dermitis ;

Page 14 sur 130
Respiratory effects: the inhalation of ammonia vapours induces irritation of
the upper respiratory tract with sneezing, dyspnoea and coughing, the most
severe stage being acute pulmonary oedema (APO). APO is an accident
which occurs following the inhalation of vesicant gases (Cl2, NH3, SO2) by
breaking down the walls of the pulmonary alveoli which are then inundated
with blood plasma. Fortunately, the olfactory detection threshold of
ammonia is well below the concentrations considered to be hazardous ;
Digestive burns: the ingestion of ammonia is followed by very intense

painful phenomena with gastric intolerance, a state of shock sometimes
accompanied by erythema or purpura. The most dangerous complication is
swelling of the glottis.
In direct contact with the skin, liquid ammonia freezes the tissues and causes
burns. Ammonia solutions are strongly alkaline and as such highly irritating for the
mucosa, skin and eyes.
Recurrent long-term exposure to ammonia induces a higher tolerance. Odours and

irritants are perceived with more difficulty.
The olfactory perception thresholds are very variable according to the subject,
ranging from a few ppm to several dozen ppm.
2.4.2 Acute toxicology
On this point, the results are based on observations made in the case of accidents
and not rigorous experiments; however, the irritation threshold has been measured
on groups of volunteers.
The median lethal concentration (LC501) values, 50 %, are essentially used to

evaluate the fatal risks for humans, as it is impossible to conduct such
experiments on humans.
Data relating to toxicity for humans can only be used if it is associated with:
a concentration ;
an exposure time ;
and a likelihood of effects occurring.
1 LC50: value calculated on the basis of the substance concentration assumed to cause the death
of 50% of the test population for a specified exposure time.

Page 15 sur 130
INERIS report (Tissot, 2003) provides the following values:
Time 1 min 3 min 10 min 20 min 30 min 60 min
Lethal Effect Threshold2 17 710 10 290 5 740 4 083 3 337 2 380
3
(mg/m )
Irreversible Effect Threshold3 1 050 700 606 428 350 248
3
(mg/m )
Odour 3.62 – 36.2 (i.e. 5 – 50 ppm)

3
(mg/m )
-3 -3
Note: for ammonia, 1 ppm = 0.70 mg.m and 1 mg.m = 1.43 ppm
table 5: Ammonia toxicity values from INERIS data sheet (Tissot, 2003)
The detection of ammonia by human smell depends on the subject's sensitivity.

This olfactory limit is generally from a concentration between 5 and 25 ppm.
This value is well below the threshold corresponding to irreversible damage which
is in turn markedly lower than the threshold for lethal effects. It is important to note
that this aspect is not verified for other gases, such as carbon monoxide, for
example, which is odourless.
2.4.3 Toxicology at the workplace
The authorised concentration for ammonia in working atmospheres is limited and

monitored according to regulations in most industrialised countries. The values
currently permitted, which may vary between countries, are generally between 25
and 50 ppm. Higher values are defined for short-term exposure or in emergency
situations. The values selected in France are given in table 6 below.
2 Lethal effect threshold: Maximum concentration of pollutant in air at a given exposure time below
which, in most subjects, no risk of death is observed.
The term "most subjects" excludes "hypersensitive" subjects such as those suffering from
respiratory failure.
3 Irreversible effect threshold: Maximum concentration of pollutant in air at a given exposure time
below which, in most subjects, no irreversible effect is observed.
An irreversible effect corresponds to the persistence over time of a lesional or functional effect
directly following exposure in an accidental situation (single and short-term exposure) resulting
in incapacitating effects.

Page 16 sur 130
Name Definition Value Observations
(ppm)
TLVC Threshold Limit Value Ceiling at the 50 Labour Ministry Circular

workplace for 15 minutes dated 19 July 1982 France
MLV Mean Limit Value at the workplace for an 8- 25 Labour Ministry Circular
hour shift dated 19 July 1982 France
Note: for ammonia, 1 ppm = 0.70 mg.m-3 and 1 mg.m-3 = 1.43 ppm
table 6: Limit values for workplaces in France
2.4.4 Summary of results on human toxicology
The two tables below show different effects observed according to the
concentration and exposure time.
Main results of experimental studies on humans
Concentration Time Number of Effects Ref.

subjects
(ppm) (minutes)
Odour
Perception (sensitive subjects) IPCS, 1986
Perception IPCS, 1986
30 8 Nuisance Verbeck, 1977
Unspecified irritation
50 10 6 Moderate irritation Mac Ewan, 1972
30-50 5 10 Nasal dryness Report, 1973
Ocular irritation and tear production
130 5 10 Ocular irritation (5 subjects/10) Report, 1973
110-140 30 8 Nuisance Verbeck, 1977
150-200 360 6 Tear production, transitory discomfort Ferguson, 1977
150-200 7 Onset of ocular effects (not specified) Wallace, 1978
400 <1 7 Ocular irritation Wallace, 1978
600 30 sec 7 Tear production Wallace, 1978
700 few sec. 7 Tear production Wallace, 1978
1000 immediate 7 Tear production, impaired vision Wallace, 1978

Page 17 sur 130
Respiratory tract irritation and respiratory function disorders
130 5 10 Throat irritation (8 subjects out of 10) Report, 1973
150 8-11 16 Respiratory function signs (exercise) Cole, 1977
110-140 30 8 Throat irritation: nuisance Verbeck, 1977
500 30 7 Respiratory tract irritation Silverman, 1949
Functional signs
700 few sec. 7 Still breathable Wallace, 1978
1000 1-3 7 Breathing intolerable Wallace, 1978
Tolerable / intolerable concentration
140 30-75 8 Departure from exposure chamber Verbeck, 1977
Intolerable score for non-experts
1500 instantaneous 7 Departure from exposure chamber Wallace, 1978
table 7: Effects of ammonia on humans
Table 7 is taken from the INERIS study (Auburtin, 1999) on the acute toxicity of
ammonia.
As an indication, the NIOSH (1987) proposes an IDLH value (Immediately

Dangerous to Life or Health) of 500 ppm for ammonia. Note that this IDLH value
corresponds to a maximum concentration in air up to which a subject exposed for
not more than 30 minutes may escape without risking irreversible effects.

Page 18 sur 130
Concentration in Effects Exposure time
-3
mg.m
3.5 Perceptible odour for some subjects.
18 Perceptible odour for most subjects. Mean occupational limit value in France
and in a number of countries
35 - 70 Perceptible irritation for most subjects, in Tolerable for up to 2 hours for subjects
the eyes. not accustomed to exposure;
accustomed subjects may tolerate
higher concentrations over the same
period
87 - 100 Irritation of the eyes, nasal tract and Exposure time of over 1 hour.
mucosa.
Irritation of the throat and respiratory

tract.
140 Nausea and headaches.
280 - 490 Immediate irritation of the eyes, nose, Exposure for ½ hr to 1 hr does not cause
throat and upper respiratory tract. serious damage although irritation of the
upper respiratory tract may persist for 24
hrs following a 30-min exposure period.
Aggravation of pre-existing respiratory
problems may occur.
700 - 1400 Severe coughing. Severe irritation of the Damage to the eyes and respiratory
eyes, nose and throat, bronchitic system may occur if they are not treated
spasms. quickly.
A 30-min exposure period may induce

very serious effects on subjects with a
predisposition for respiratory problems.
2100 - 2800 Severe coughing. Severe irritation of the May be fatal after 30 min.
eyes, nose and throat.
3500 - 8400 Respiratory spasm. Rapid asphyxia, Fatal in a few minutes.

severe oedema, strangulation.
table 8: Effects on ammonia on humans according to concentration (EFMA-IFA, 1990)
Table 8 is taken from the "guide on risks represented by ammonia" issued by

SNIE (1991).

Page 19 sur 130
2.4.5 Toxicology of fauna and flora
2.4.5.1 Animals
Table 9 shows some effects observed on animals
Species Concentration in Effects

-3
mg.m
Pigs 75 - 100 Coughing, nasal, oral and lachrymal secretions.
140 Severe symptoms of irritation and convulsion after 36 hrs

of exposure with recovery 7 hrs after end of exposure.
Rabbits 280 A 10-min exposure period induces an irreversible

cessation of ciliary activity.
350 Same effects as above for a 5-min exposure period.
3500 LC 50 – 1 hr.
Cats 3500 LC 50 – 1 hr.
table 9: Toxicological effects of ammonia on some animals

according to concentration (EFMA-IFA, 1990)
2.4.5.2 Aquatic life
Free (non-ionised) ammonia in surface water is toxic for fish. However, ammonium
ions are not. In this way, in the event of water contamination by ammonia,
ammonia salts representing no toxic risk may be formed. The pH value of the
water is important as free ammonia is formed for pH values greater than 7.5-8.0.
For trout, disorders occur from 0.3 mg.l-1. Percidae (or castotomidae) and
salmonidae are the most sensitive varieties for freshwater species. For seawater,
prawns appear to be the most sensitive invertebrate species, Depending on the
type of fish, a risk of mortality arises between 1.2 and 5 mg.l-1 (these values refer
to non-ionised ammonia).
For ionised ammonia (ammonia in solution), lethal concentrations (LC50) for some
species are given in the table below:
Species Threshold Exposure time Concentration

-1
daphnids LC50 24 hrs 27 mg.l
-1
fish LC50 24 hrs 182 mg.l

Page 20 sur 130
-1
algae LC50 5 days 185 mg.l
table 10: lethal concentrations (LC50) for some species
2.4.5.3 Plant life
Ammonia is considered as biodegradable and not accumulable, However, it

requires monitoring in the form of an annual evaluation of water, air and soil
releases for annual uses of over 10 tonnes.
Ammonia can be toxic for some plants as it cannot be excreted. It is assimilated by

combining with carbonate chains, and, in this way, the excess ammonia can pass
into the sugar metabolism.
Other plants have functions enabling the assimilation of ammonia thus allowing
them to tolerate ammonia or use it preferentially.

Page 21 sur 130
3. ATMOSPHERIC DISPERSION
3.1 CONTEXT
Knowledge of the atmospheric dispersion mechanisms of ammonia is applied in

the following three cases:
The predictive nature of an atmospheric dispersion study makes it possible

to envisage the potential risks of an industrial facility, particularly within the
scope of hazard studies or impact studies ;
In the event of an accidental situation, modelling the atmospheric dispersion

makes it possible to evaluate the measures to be taken in real time ;
Following an accident, the analysis of the dispersion conditions in the atmosphere

can give a clearer understanding of the causes of these accidents, and in some
cases, validate atmospheric dispersion models used for the above mentioned
studies.
There are different approaches to the study of atmospheric dispersion:
full-scale tests ;
model (hydraulic or aeraulic) simulation ;
use of mathematical calculation codes.
The last way to approach gas transportation and diffusion phenomena in the
atmosphere offers advantages over the techniques mentioned above such as the
fact that complex experimental procedures are not used, the rapid nature of the
study, and the possibility of envisaging a large number of cases.
However, in order to be able to access simulation, it is necessary to have

extensive knowledge of the phenomena governing atmospheric dispersion. With
this respect, full-scale tests are required.
Studying the atmospheric dispersion of ammonia is of major interest for two

reasons. Firstly, ammonia is a very toxic, corrosive, flammable and explosive
substance under certain conditions. Secondly, ammonia is a very widely used
substance with numerous applications, due to its chemical or physical properties.
With respect to its chemical properties, the main application of ammonia is the
manufacture of fertilisers. Ammonia represents the main concentrated source of
nitrogen for agriculture, which consume approximately 85 to 90% of production.
Industry uses ammonia as a raw material for the manufacture of explosives, fibres
and plastics. It is also used in the manufacture of paper, rubber, in refineries, the
leather industry and the pharmaceutical industry.

Page 23 sur 130
With respect to applications associated with physical properties, ammonia is used
as a coolant in compression and absorption systems. Some characteristics of
ammonia, such as its high latent heat, its low vapour density and its chemical
stability, favour its use in major industrial facilities.
These two specificities of ammonia, its harmfulness and its widespread use
require, in terms of personal safety, details on its behaviour in the event of loss of
confinement, particularly in the case of accidental releases into the atmosphere.
Therefore, to this end, this chapter proposes to:
provide a status report on atmospheric dispersion tests conducted with

ammonia ;
present the state of the art and gaps in the knowledge of the modelling of
the atmospheric dispersion of ammonia, as the modelling of the
atmospheric dispersion represents a key stage in the study of toxic
substance release ;
and justify the study areas developed within the scope of this research
programme.
3.2 ATMOSPHERIC DISPERSION TESTS
After a description of the main characteristics of ammonia releases, this section

first presents atmospheric ammonia dispersion tests and then briefly presents
dispersion tests conducted with heavy gases.
3.2.1 Characteristics of accident ammonia releases
The atmospheric dispersion of a gas is influenced by the release conditions, the

weather conditions and also by the type of gas released.
There are different possible types of accidental ammonia releases:
ammonia gas jet from a pressurised vessel (release from gaseous phase) ;
two-phase ammonia jet from a pressurised vessel (release from liquid

phase) ;
evaporation of a pool of liquid ammonia in which the temperature is less

than or equal to its boiling point4 ;
leakage of liquid ammonia from a cryogenic tank (liquid ammonia at a

temperature below the boiling point and at atmospheric pressure).
4 Note that the boiling point of ammonia at normal pressure is -33.4°C.

Page 24 sur 130
Ammonia has specific dispersion characteristics and depending on the type of
release, the dispersion of the ammonia released will be different. The type of
release, which was more specifically studied within the scope of this programme,
is the two-phase ammonia release from the liquid phase of a pressurised tank.
All the studies conducted on two-phase releases break down the jet into three
main zones as shown in figure 1 below:
figure 1: two-phase jet
These three main zones are respectively:
the expansion zone: In this short zone, between 0.5 and 4 times the
diameter of the opening (IANELLO, 1989), the fluid is expanded from the
pressure at the opening to atmospheric pressure. Due to this sudden
depressurisation, the liquid phase of the release is in a superheated state
and a fraction of this liquid phase is vaporised almost instantaneously. This
phenomenon is known as the "thermodynamic flash". The gaseous phase
created by this flash phenomenon will have, due to its lower density, a
higher velocity than that of the liquid. This difference in velocity between the
two phases induces the entrainment of the liquid phase and its
fragmentation into fine droplets. These droplets are entrained by the
gaseous phase at a high velocity and form what is referred to as the
"aerosol".
In this way, at the end of this zone, the jet consists of a gaseous phase and
a liquid phase in aerosol form and it is commonly admitted that the entire jet
(liquid and gaseous phases) is at the boiling point of the released
substance.

Page 25 sur 130
the entrainment zone: In this zone, the turbulent jet induces the
entrainment of the ambient air in the jet. The energy supplied by this air
which is warmer than the jet is initially used to vaporise the liquid droplets
present in the jet. This vaporisation cools the jet which then behaves like a
jet of heavy gas in the atmosphere.
Once the liquid phase has been entirely vaporised, this energy is used to
heat the jet which has become entirely gaseous.
the passive dispersion zone: With the entrainment of air, the velocity of
the jet progressively decreases until it reaches the wind velocity. From this
point, the jet is said to be dispersed passively in the atmosphere.
The presence of obstacles is also a factor in the atmospheric dispersion of heavy

gases, particularly liquefied ammonia.
If the obstacle is sufficiently close to the release, it will modify the initial dispersion
phase and subsequently, it will influence the dispersion in the atmosphere. For
example, a two-phase release in jet form, which encounters an obstacle such as a
wall or floor, may result in pool formation. In this case, the dispersion of the
ammonia from the pool is different to that which would have followed the free jet.
On the other hand, if the obstacle is at a distance from the release point, and if the
size of the cloud formed by the ammonia is comparable to the obstacle, the
obstacle only modifies the gaseous mass flow and will not modify the physical
state of the release. In this case, the effect on dispersion will be less than in the
previous case, decreasing according to the dilution and size of the cloud when it
encounters the obstacle.
3.2.2 Tests by A. Resplandy
The information contained in this section is based on the A. Resplandy's

summaries (1967, 1968, 1969).
3.2.2.1 Context
In order to determine the measures to be recommended for the safety of the

vicinity of liquefied ammonia storage tanks, A. Resplandy, Official Inspector of
Classified Facilities and Chairman of the Consultative Committee for Classified
Facilities, conducted two series of liquid ammonia dispersion experiments. They
took place, firstly at the Boissise-la-Bertrand army camp in June 1967 and,
secondly, at the Mourmelon army camp in March 1968.
The purpose of the experiments was to:
specify the conditions under which ammonia is diffused into the atmosphere
from a liquid phase spillage in order to determine the radius of the isolation
zones to be stipulated around pressurised storage tanks ;
examine the extent to which refilling of liquid ammonia can be envisaged ;

Page 26 sur 130
test the effectiveness of the intervention resources provided to Civil
Protection teams called upon to intervene in the event of an accidental
liquid ammonia leakage.
3.2.2.2 Test conditions
A – Tests at Boissise-la-Bertrand, June 1967
The liquid ammonia tanks consisted of:
a semi-trailer tanker containing 15 tonnes of ammonia at a pressure of 7

bar, equipped with a 2 inch pipe with a flow rate of approximately 20,000
kg/hr (5.5 kg/s) in liquid phase and a 1 inch conduit with a flow rate of
100 kg of liquid ammonia in approximately 70 seconds (1.4 kg/s) ;
a tractor-drawn replenishment vehicle comprising two ammonia tanks at 7

bar with a unit capacity equal to 2000 kg, each tank being equipped to
release in liquid phase via a 1¼ inch pipe or in gaseous phase via a ¾ inch
pipe.
The tests were conducted on 28 June 1967 between 6 am and 11 am. The ground
temperature rose gradually from 13°C to 17°C. The relative humidity was
maintained in the vicinity of 85%. The wind velocity varied from 0 to 3 m/s.
Two types of measuring device were used:
17 Dräger manual detectors:

The Dräger units essentially consisted of a bellows pump used to pass a
given volume of gas through a calibrated tube filled with coloured reagent.
Three types of calibrated tubes were used:
0.5/a type: sensitivity from 5 to 70 ppm (non-interpretable
measurements) ;
25/a type: sensitivity from 50 to 700 ppm ;
0.5%/a type: sensitivity from 0.5 to 10% (non-interpretable
measurements).
Note that these measuring devices only gave mean values calculated over
variable periods (since the number of pump strokes is not constant
according to Resplandy) and only allowed a single measurement per test.
Hartmann and Braun Caldos 2 continuous thermal conductibility analyser:

This unit equipped with a 4-6 mm diameter and 60 m long probe-conduit
was operated coupled with a Philips potentiometric recorder over the
following three sensitivity ranges: 0 to 2%, 0 to 4%, 0 to 10%.
Three types of releases were used in order to study:
the atmospheric diffusion of ammonia from a liquid phase leakage:

Page 27 sur 130
without curtain interposition, using a 2 inch conduit ending with a 90°
bend to release vertically upwards, and producing approximately 9
releases per minutes, i.e. a released quantity of approximately 330 kg of
liquid ammonia ;
with curtain interposition, using a 1 inch conduit ending with a 90° bend
fitted to a cylindrical vessel in order to be able to guide the jet upwards,
as above, or towards the base of the vessel, located 15 cm from the end
of the conduit, and performing 2 tests of approximately one minute (one
with an upward jet, the other with a downward jet) ;
retrievable liquid ammonia refilling possibilities after a storage tank
accident,
intervention resources in the event of a liquid phase (aerosol) ammonia

leakage via the effect of water and heat.
B – Tests at Mourmelon, March 1968
The ammonia was supplied by a 15000 kg semi-trailer tank identical to that for the
previous tests, able to release in liquid phase via a 50 mm diameter pipe, and in
gaseous phase via a 30 mm diameter pipe.
The tests were conducted on 28 March 1968 between 8 am and 12:30 pm. The
ground temperature rose gradually from 8°C to 18°C. The relative humidity varied
from 95% to 80%. The wind velocity was between 0 and 3 m/s.
Three recording analysers were used:
Hartmann and Braun Caldos 2 continuous thermal conductibility analyser

identical to that used above ;
Kuhlmann conductimetric analyser equipped with a 50 m probe operating

over the 0 to 0.5% range ;
"ONERA SO" type IR spectrometry analyser equipped with a 50 m probe operating

over the 0 to 0.1% range.
Three types of releases were studied:
the atmospheric diffusion of ammonia from a liquid phase leakage:

in the case of a release into a stable atmosphere with a vertical ejection ;
by varying the orientation of the ejection conduit ;
by monitoring the progression of the aerosol in terms of stabilised flow
rate ;
by observing the contamination of the atmosphere of a residential
building (caravan) flushed with the ammonia cloud.
gaseous phase leakage simulation in a pressurised storage tank:

Page 28 sur 130
by varying the direction of ejection ;
by inserting a curtain during vertical ejection.
handling of liquid ammonia in the open air:
by studying the effectiveness of the clay and cement retention dikes ;
by observing the effect of water on liquid ammonia.
3.2.2.3 Results
In these publications, A. Resplandy (1967, 1968, 1969) gives the following main
results.
Liquid phase leakages induce the formation of an aerosol, which, before

dissipating, may travel several hundred metres at ground level propelled by the
wind. In fact, ammonia's particularly high latent heat of vaporisation only allows
very slow evaporation of the micro-drops.
The ammonia aerosol is carried by the wind, even if the wind velocity is extremely
low, covering a relatively narrow area, without any diffusion upstream from the
release point. Propagating against the ground, the ammonia droplets may be
deviated from the trajectory given to them by the wind by large obstacles.
An ammonia aerosol is dissipated completely a few minutes after the release

causing it has stopped and there is no residual hazardous concentration for
humans on the zone covered.
For the same mass of liquid ammonia ejected into the atmosphere from a
pressurised storage tank, aerosol formation decreases as the temperature of the
liquid approaches its boiling point at atmospheric pressure.
In the open air, it is not possible to block the progression of ammonia droplets
effectively with a water spray, even if it is considerable in size.
The relatively rapid diffusion of gaseous ammonia in the atmosphere makes it

possible to envisage, if required, stabilisation by decompressing the contents of
non-cooled tanks, followed by handling of the stabilised liquid ammonia in the
open air.
Finally, it was demonstrated how dangerous it could be to intervene with water on

a sheet of liquid ammonia. In fact, this action results in a significant release of gas
and aerosol.
In a liquid phase release from a pressurised ammonia storage tank, by adjusting

the ejection mode (inserting a curtain, direction control), there are limited options
available to reduce aerosol formation and decrease the distance covered by the
aerosol front.
Pollution at ground level can be limited by discharging gaseous ammonia vertically

into the atmosphere from a pressurised tank.
Page 29 sur 130
A well-designed system makes it possible to equilibrate the liquid from a
pressurised tank in the open air by eliminating the vapour phase in the
atmosphere.
Liquid phase ammonia equilibrated at atmospheric pressure can be effectively

collected and stored in the open air in a retention dike; however, the temperature
rise on the casing walls is accompanied by gas and aerosol emission.
The atmospheric dispersion tests conducted by A. Resplandy were among the

first. They resulted in qualitative observations rather than genuinely interpretable
quantitative data. In fact, not many parameters were measured and recorded. In
particular, there was no precise characterisation of the atmosphere, or precise
measurements of the ammonia concentration leeward from the release.
These tests particularly made it possible to give ideas on the behaviour of

ammonia in the event of an accidental release and outline the perspectives to be
studied. In this way, A. Resplandy proposes three areas of study to complete the
study conducted:
specify the influence of the leakage flow rate on the characteristics of the
aerosol formed ;
observe the influence of weather conditions on aerosol characteristics in

more detail ;
determine the optimal conditions to return pressurised tanks to atmospheric

pressure.
3.2.3 Tortoise Desert tests
The releases produced for the Tortoise Desert tests (Goldwire, 1996) were very
different to those used by Resplandy.
For example, the quantity of ammonia released on average during the tests
conducted by Resplandy in 1967 and 1968 was of the order of 300 to 500 kg with
a maximum of 1000 kg for times between 1 and 6 minutes, while for the Tortoise
Desert tests, the quantities released were between 10,000 and 41,000 kg.
3.2.3.1 Context
A series of 4 ammonia dispersion tests was conducted during the summer of 1983
by the Lawrence Livermore National Laboratory in the United States. These tests
are frequently referred to as the Tortoise Desert Tests in the literature. They were
used to study the dispersion of pressurised liquid ammonia releases in the
atmosphere.

Page 30 sur 130
3.2.3.2 Test conditions
Two tankers with a capacity of 41.5 m3 were used for the tests. Nitrogen tube
trailers were connected to the two tankers in order to maintain the pressure inside
the tankers.
The first test was conducted with a 3.19 inch (81 mm) release orifice and the three
others were conducted with a 3.72 inch (94.5 mm) diameter orifice. The release
was horizontal, in the direction of the wind and at a height of 0.79 metres above
ground level.
Table 11 below shows the main characteristics of these releases:
Test Date Release Release Release Flow Wind Wind Stability

volume mass in time in s rate in velocity direction class
3
in m kg kg/s in m/s
1 24/08/83 14.9 10,200 126 81 7.4 224° D
2 29/08/83 43.8 29,900 255 117 5.7 226° D
3 01/09/83 32.4 22,100 166 133 7.4 219° D
4 06/09/83 60.3 41,100 381 109 4.5 220° E
table 11: Characteristics of releases produced during Tortoise Desert tests
Two main devices were used to measure the meteorological data and the
concentrations for the dispersion study.
With respect to the dispersion study, the vapour concentrations were measured
by:
20 MSA NDIR gas sensors positioned 100 m from the source and such that
the gas plus the aerosol passed through a heating device intended to
vaporise the aerosol so as to determine the total quantity of NH3 ;
31 LLNL IR gas sensors positioned 100 m and 800 m from the source and
initially devised for liquid natural gas (LNG) dispersion tests and not
optimised for ammonia detection, but they made it possible to detect
concentrations above 1% ;
24 IST sensors located 0, 800, 1450, 2800 and/or 5500 m from the source.
The data acquisition and storage system used UHF radio waves for data
transmission. All the data acquisition stations and all the stations were portable,
battery-operated and gas-tight.

Page 31 sur 130
3.2.3.3 Results
The first two tests were conducted on a water-saturated site due to heavy rain.
The site was almost completely dry for the third test and completely dry for the
fourth, such that the overall weather conditions were more variable than planned.
Test No. 4, the most significant in terms of released quantity, was conducted
under the most stable atmospheric conditions. It demonstrated that the jet was
visible up to a distance greater than 100 m from the source. The dimensions of
this cloud, which was heavier than air, are approximately 70 m wide and less than
6 m high. The maximum gas concentration observed at 100 m was 6.5%
according to Goldwire (1986) and 10% according to Koopman et al (1986). The
cloud had spread over a width of 400 m when it reached 800 m, its height was still
6 m. The heavy gas and jet effects dominated the dispersion.
Table 12 below defines some orders of magnitude based on experimental results.

It shows maximum concentrations recorded according to the leeward distance
from the release site.
Distance 100 m 800 m 1450 m 2800 m 5500 m
Maximum 6 to 10% 1.4 to 1.6% > 0.5% > 0.5% 100 to

concentration 200 ppm
recorded
table 12: Maximum concentrations observed as a function of leeward distance from the release site
according to Koopman et al. (1986) and Goldwire (1985, 1986).
Goldwire et al. (1986) demonstrated the need for "enhanced" simple models to
predict the effects of accidental releases. In addition, Hanna et al. (1993) also use
Goldwire's results (1985) to compare the results of the models and the
experimental results. The table below gives the concentrations observed on the
plume axis.
Tortoise Desert tests 1 2 3 4
Flow rate (kg/s) 81 117 133 109
Concentration at 100 m (ppm) 49,943 83,203 76,881 57,300
Concentration at 800 m (ppm) 8,843 10,804 7,087 16,678
table 13: Concentrations observed on plume axis during Tortoise Desert tests according to Hanna
et al. (1993).
These results should be compared to the information contained in table 12.

Page 32 sur 130
In relation to these results, several remarks should be noted:
The highest concentration at 100 m is obtained for test 2 despite a higher

flow rate for test 3. The wind velocity for test 2 was lower (5.7 m/s) than for
test 3 (7.4 m/s) and the outdoor temperature was 3°C lower for test 2
compared to test 3 ;
At 800 m, the highest concentration is obtained for test 4, with the lowest
flow rate among the releases produced with the 3.72 inch orifice (tests 2, 3,
4). On the other hand, the concentration at 100 m was the lowest. This
result is explained by most stable weather conditions, stability class E
instead of class D for tests 2 and 3 ;
Among all the tests, the most unstable weather conditions were recorded
during test 3 (Goldwire, 1985). Under these conditions, in spite of a higher
release flow rate, the concentration observed on the plume axis at 800 m
proves to be the lowest.
These remarks demonstrate that the weather conditions are a very important
factor in atmospheric dispersion at longer distances, i.e. beyond 100 metres. It is
not sufficient to characterise the source term in detail to model the atmospheric
dispersion of ammonia, the weather conditions are also required.
Note: The 4 sections below (§ 3.2.4 to 3.2.7) are essentially based on the
document produced by Wheatley (April 1987).
3.2.4 Unie van Kunstmest Fabrieken bv tests, Netherlands, 1972
The Unie van Kunstmest Fabrieken bv tests, Netherlands, 1972 were conducted
by J. W. Frenken and were reported by Blanken (1980).
The test conditions were as follows:
vessel containing ammonia at 3.5 bar / 0°C ;
8 mm diameter release nozzle ;
ambient temperature of -10°C ;
and wind velocity of the order of 2 to 3 m/s.
The purpose of these tests was to measure the quantity of ammonia collected on
the ground in the case of two one-minute releases each releasing 38.4 kg (0.64
kg/s) in the following configurations:
a horizontal release perpendicular to the wind, located 1800 mm from

ground level. This release made it possible to collect 5.5 kg of liquid
ammonia on the ground, i.e. 14.3% of the ammonia released ;

Page 33 sur 130
a vertical release directed towards the ground, 2000 mm from ground level,
making it possible to collect 27.3 kg of liquid ammonia on the ground, i.e.
71.1% of the ammonia released.
The measurement method for the quantity of ammonia collected is not specified.
No measurements of the concentration in the air were made
3.2.5 Imperial Chemical Industries tests, United Kingdom, 1974
The Imperial Chemical Industries tests, United Kingdom, 1974, were reported by
Reed (1974).
The purpose of the tests was the characterisation of the effectiveness of

pressurised storage tank retention dikes.
Two series of tests were conducted:
one series intended to simulate the disaster-related rupture of a container

by suddenly removing a lid. The results of these tests are not described by
Wheatley (April 1987);
another series of releases via a 1 mm diameter orifice linked by a tube to a

tank containing ammonia at 6.5 bar/16°C, located 1 m from ground level at
various angles. When the orifice was horizontal or pointing slightly upwards,
an aerosol cloud was formed and no liquid formation was observed on the
ground. When it was pointing slightly downwards, a small fraction of liquid
ammonia was collected. The quantity is not specified in Wheatley's report
(April 1987).
No other observations were made, and no air concentration measurements were

made.
3.2.6 Unie van Kunstmest Fabrieken bv tests, Netherlands, 1980
The Unie van Kunstmest Fabrieken bv tests, Netherlands, 1980, reported by

Blanken (1980), used ammonia stored at 13.4 bar at 38°C and released via a 100
mm long and 2 mm diameter capillary tube into moist air.
The purpose of the tests was to study the size of the droplets in the jet. Only
qualitative results were obtained.
In an ammonia atmosphere, the jet was transparent and the liquid separated from
the jet and formed a pool on the ground.
In the moist air, the jet was opaque and the liquid ammonia was not collected. It
was assumed that the ammonia aerosol initially formed was rapidly vaporised
when it was diluted with air. In addition, it was concluded that the opacity was
caused by water condensation, which forms an aqueous ammonia aerosol.

Page 34 sur 130
No concentration measurements were made.
3.2.7 .Landskrona tests, Sweden, 1982
The Landskrona tests, Sweden, 1982 were conducted by the Swedish National
Defence Research Institute and were reported by Nyrén et al (1983).
The test conditions were as follows:
a commercial vessel containing 1400 kg of ammonia at 6 bar and 9°C ;
an ambient temperature of 8°C and 42% relative humidity ;
and a wind velocity of 13 m/s.
The purpose of these tests was to measure the flow rate of a two-phase ammonia
release and compare it to a theoretical model.
The releases were produced from a tube with an internal diameter varying
between 32 and 40 mm. Six tests were conducted with a 2 m long tube and five
with a 3.5 m tube. In each case, the tube was located 2 m above ground level.
Each release lasted 60 to 90 s.
For all the tests, except for four, the tank pressure varied considerably. For the
remaining four, the outlet pressure was 2.2 bar and the mass flow rate 2.2 kg/s.
The jet touched the ground at a point located between 6 and 10 m from the
release point. No pool was observed.
No concentration measurements were made.
3.2.8 FLADIS programme tests
Ammonia releases were produced within the scope of a European test programme
called FLADIS (Research on the dispersion of two-phase flashing releases). This
programme was conducted under the aegis of DG XII in Brussels (Nielsen & Ott,
1996).
Three measurement campaigns were conducted in April and August 1993 and in
August 1994. In total, 27 horizontal releases of pressurised liquefied ammonia
were produced.
The release flow rates varied from 0.25 to 0.55 kg/s via 4.0 and 6.3 mm diameter
orifices. The longest test lasted 40 min.
One of the aims of the FLADIS programme was to study in detail the aerosol jet in
the near field, the heavy gas dispersion phase and the transition from passive
dispersion. For this, sensors were positioned in an arc shape centred on the
release point at 20 m, 70 m and 240 m.

Page 35 sur 130
Aerosol composition measurements in the two-phase jet were made during the
tests. On the basis of these measurements, it was observed that the aerosol,
which consisted of almost pure ammonia close to the release point, consisted,
within a few meters, of almost pure water.
In addition, no rise in the ammonia cloud was ever observed during the releases.
3.2.9 Tests conducted at Ecole des Mines, Alès
Ecole de Mines, Alès, has conducted several ammonia dispersion test campaigns
since 1996. The tests are conducted using one or two 44 kg ammonia cylinders
placed upside down so as to produce liquid phase releases (Bara, Dussere, 1996).
The purpose of these tests is to study the effectiveness of the peacock tail hoses
used by the fire services.
Downstream from the peacock tail hoses located approximately 1 metre from the
cylinders, a reduction in the ammonia concentrations was observed on the release
axis, by a factor greater than 10 at a 13 m distance from the cylinders and a factor
of at least 3 at a 20 metre distance. On the other hand, no significant reduction
was observed beyond 50 metres.
3.2.10 Other atmospheric dispersion tests
Within the scope of this programme, substance dispersion tests in the atmosphere
with gases other than ammonia were taken into consideration.
As mentioned above, ammonia released into the atmosphere generally results in

the formation of a cold cloud, heavier than air. Therefore, dispersion is carried out
as for that of a heavy gas until a sufficient level of dilution is reached so that the
ammonia dispersion is governed solely by the characteristics of the atmospheric
flow (passive dispersion).
For this reason, atmospheric dispersion tests conducted on heavy gases are of
interest for the study of the atmospheric dispersion of ammonia. This is particularly
true for all aspects relating to the initial release phase.
Table 14 overleaf summarises the main dispersion tests relating to heavy gases.
Note: The Tortoise Desert tests are included in this table for comparison purposes

Page 36 sur 130
Thorney Thorney
Tortoise Maplin Prairie Island Island
Burro Coyote Goldfish Handford
Desert Sands Grass (instantaneou (continuou
s) s)
Hydrogen LNG and Sulphur Freon and Freon and

Substance LNG LNG Ammonia Krypton 85
fluoride LPG dioxide nitrogen nitrogen
Release Boiling Boiling Two-phase Two-phase Boiling

Gaseous Gaseous jet Gaseous Gaseous
type liquid liquid releases releases liquid
Number of
8 3 4 3 5 4 and 8 44 9 2
tests
LNG: 2000
10700 6500 10000 3500 11 to 6600 23 3150
Quantities
to to to to to to to 4800
in kg
17300 12700 36800 3800 24 Curies LPG: 1000 63 8700
to 3800
Time in s 79 to 190 65 to 98 126 to 381 125 to 360 598 to 1191 60 to 360 600 instantaneous 460
Release on Release on Release on

Surface small small Dry lake Desert with shallow
Moist sand Grass Grass Grass
type stretches of stretches of bed bushes stretches of
water water water
Stability
C-E C-D D-E D C-E D A-F D-F E-F
classes
Maximum
observatio 140 - 800 300 - 400 800 3000 800 400 – 650 800 500 - 580 472
n distance
table 14: Summary table of main dispersion tests relating to heavy gases
Note: Thorney Island, 19 tests during phase I; 9 tests during phase II and 3 tests for continuous releases

Page 37 sur 130
The study of the tests mentioned in table 14 suggests the following remarks:
Only the Tortoise Desert and Goldfish tests involved two-phase releases ;
The liquid release tests on water (Burro, Coyote and Maplin Sands) do not
offer any conclusions for ammonia which has a very different behaviour with
water ;
Of the tests, only those at Thorney Island used obstacles.
The information obtained from these tests, in terms of dispersion, relates to

general aspects. Within the scope of these tests, it is not possible to account for
the specificity of ammonia, such as jet formation, aerosol generation, reactions
with air moisture, etc.
3.2.10.1 Summary
The atmospheric dispersion tests already completed have demonstrated:
the influence of weather conditions ;
the behaviour of ammonia which is dispersed like a heavy gas when it is

released from a pressurised tank (no upward dispersion) ;
a release impacting the ground made it possible to collect 70% of the

ammonia released (Unie van Kunstmest Fabrieken bv tests, Netherlands,
1972).
However, data is particularly lacking on:
the characterisation of the source term of the release and the formation of
the two phase jet in the case of pressurised releases and particularly
aerosol generation and progression;
release tests impacting a surface with higher flow rates. The tests such as
the Unie van Kunstmest Fabrieken bv tests, Netherlands, 1972, did not
study ammonia dispersion. These tests just studied the phenomenon in a
qualitative manner.
3.3 MODELLING OF ATMOSPHERIC AMMONIA DISPERSION
As seen in the previous section, modelling particularly needs to account for:
gravitational effects due to the cloud temperature ;
the presence of aerosol ;
reactions with air moisture.

Page 39 sur 130
Different types of dispersion models exist: Gaussian models, integral models and
three-dimensional models. These models are described extensively in the
literature, e.g.: CCPS (1999) or RIOU (1989).
3.3.1 Gaussian models
Purely Gaussian models are, in the majority of cases, unsuitable for the detailed
modelling of pressurised liquid ammonia releases. In fact, ammonia does not
behave like a passive gas in the initial release phases.
3.3.2 Three-dimensional models
Three-dimensional models make it possible to solve the physical equation system

governing atmospheric dispersion directly. In this way, a large number of situations
of varying complexity can be simulated. However, this requires precise
incorporation of all phenomena such as, for example, atmospheric turbulence, site
topography or obstacles.
These models are more complex to implement and require qualified personnel for
their use. The limitations of these models include the choice of turbulence models,
mesh refinement or convergence criteria. For this reason, more simple digital
models are used more frequently.
3.3.3 Integral models
Half-way between Gaussian models and three-dimensional models, integral

models give interesting results with shorter implementation times than three-
dimensional models. They make it possible to account for the specificities of the
gas and release in the initial phase, until the dispersion becomes passive.
However, such programs do not account for obstacles such as buildings,

topography and vegetation. Only a roughness parameter can be used to modulate
the results according to the average site configuration.
3.3.4 Residual problems and improvements to be made
Lantzy (1992) states that it should be borne in mind that the purpose of
atmospheric release modelling is to predict, as accurately as possible, the effects
of a release of hazardous substance. For this purpose, the characterisation of the
source term is decisive, because the best dispersion models will always give
incorrect results if the source term is incorrect.
Some points presented in this article may be repeated for the study of ammonia
dispersion, more specifically for the study of the source term for ammonia
releases. These points relate to two-phase flows, aerosol formation and
progression and pool formation and evaporation.

Page 40 sur 130
3.3.4.1 Two-phase flows
With respect to two-phase flows, according to Lantzy (1992), future research

should focus on:
improved understanding of the effects of non-equilibrium for releases via

small tubes (less than 100 mm long). This is the scope of interpolation
between flows in genuine non-equilibrium and homogeneous flows ;
a more accurate determination of the discharge coefficient. This value is

generally between 0.6 and 1.0 and, for this reason, has a significant
influence on the calculated flow rate ;
the creation of an experimental database for ruptures of pipes with a high

L/D ratio. To date, the research carried out in this area has been limited to
computer simulations. Data is required to verify and extend the existing
theories ;
the development of entrainment and mixing models for disaster-related

ruptures. For example, there is no entirely satisfactory theory to estimate
the drop size and release velocity correctly. Large-scale data is required to
complete the small-scale data currently available.
3.3.4.2 Aerosol formation and progression
To model releases from pressurised storage tanks correctly, the models must
account for the aerosol and the progression of the droplets along the jet. However,
for this, it is necessary to be able to estimate the initial size of drops and their
distribution accurately. Therefore, most of the research to be developed relates to:
droplet evaporation. It is necessary to predict the quantity of the substance

that will be transported correctly ;
droplet coalescence ;
the effects of viscosity on shearing. No data is available to date ;
release orifice non-uniformity.
3.3.4.3 Pool formation and evaporation
Pool formation and evaporation is a subject which should be examined in order to

understand and model the phenomena liable to be encountered during accidents.
In particular, the following points must be studied:
freezing of ground moisture. This phenomenon affects thermal conduction

towards the pool. It is not very clearly understood at the present time ;

Page 41 sur 130
infiltration of substances into the ground. For a cryogenic substance (such
as ammonia), this phenomenon increases the contact surface area which
favours heat transfer and, for this reason, the evaporation rate. For a non-
cryogenic substance, this phenomenon results in a pool with a smaller
surface area and less evaporation. As for the above point, this phenomenon
is not well known ;
the effect of the substrate on pool size. There are no good theories linking
the roughness of the substrate and the pool size directly for a smooth and
irregular surface (i.e. in the majority of actual cases) ;
evaporation with light winds. Data would be required to develop suitable

models for these conditions.
In addition, with respect to the evaluation of source term models, Lantzy (1992)
points out that, generally, the evaluation is not conducted on the source term
model itself, but on the combination of the source term model and dispersion
model.
Lantzy also specifies that most of the data available relates to distances of over
300 m. Well-designed and well-instrumented tests are required in the vicinity of the
release point so as to compile quality data for the source term model evaluation.
In conclusion, Lantzy expresses the need for satisfactory experimental data for the
evaluation and development of all the aspects of source term modelling.
3.3.5 Summary
In sum, a number of technical problems still need to be dealt with in detail in

various areas relating to the study of atmospheric pollutant dispersion.
With respect to source term comprehension and modelling, it is necessary to

examine in detail:
the two-phase release ;
pre-mixing in the jet phase ;
aerosol formation and behaviour in dispersion.
With respect to the specificity of ammonia:
density ;
reactions with air moisture.
With respect to the extrinsic conditions for the release:
definition of weather conditions (wind profile, temperature gradient,

hygrometry, etc.) ;

Page 42 sur 130
topography and orography ;
influence of roughness.
With respect to problems directly associated with modelling:
limit conditions with the choice of closing equations for three-dimensional

models ;
influence and importance of the different empirical settings for simplified

models and the suitability of these settings with respect to reality ;
influence of the software user on modelling results.
3.4 ISSUES EXAMINED IN THIS PROGRAMME
This research programme particularly consists of completing knowledge on

atmospheric ammonia dispersion in open environments and in obstructed
environments and making comparisons between results obtained by means of
digital models and measurements made during tests. To this end, a large-scale
test programme was initiated.
It aims to provide solutions for the problems raised in the previous chapter, more
specifically on the two points presented below.
3.4.1 Impacting jets
Within the scope of an ammonia release study, it is difficult to characterise the

source term correctly, particularly the quantity flowing and forming a liquid pool,
referred to as the "rain-out" in the literature.
To try to understand large-scale phenomena, INERIS used ammonia releases in

configurations differing only by the presence of obstacles in the very near field:
two-phase jet in open field ;
two-phase jet impacting a wall at a distance of less than 3 metres ;
two-phase jet impacting the ground.
The difference between these types of release should make it possible to estimate
the pool formation rate. This potential pool formation in the case of an impacting
jet will enable a more realistic approach within the scope of safety studies.

Page 43 sur 130
Tests conducted on gaseous jets impacting a surface have demonstrated that,
under test conditions, the volume for which the concentration is greater than a
given concentration was much higher than in the case of a free jet, up to 10 times
greater during tests with methane (Chaineaux, 1995). In fact, when the jet impacts
an obstacle, the air entrainment is, in principle, reduced as the cloud tends to be
"diluted" with gas-charged air.
If this result is confirmed, it will be necessary to take this aspect into account for
safety in the near field.
However, an impacting jet may result in the formation of a pool on the ground and,
therefore, reduce the gas flow rate passing in the atmosphere. The distances
corresponding to a given concentration may then be less than those that would be
obtained in the case of an open field release.
Facility safety studies characterised by a configuration similar to these tests will be

more realistic if these phenomena are taken into consideration.
3.4.2 Influence of orifice geometry
Within the scope of the safety studies, releases are generally considered to be
circular (due to guillotine pipe rupture).
In this programme, two-phase releases via different orifices, but with the same
open cross-section surface area, were used, in order to compare the results
obtained in terms of concentration leeward from the mission.
The reference orifice is a circular orifice (nozzle). Then, an orifice formed by a

flange in which the seal was removed beforehand and replaced by blocks of the
same thickness was produced as shown in figure 2 below.
figure 2: Diagram of device used to simulate a leakage on a flange.

Page 44 sur 130
4. DESCRIPTION OF LARGE-SCALE TESTS
In this research programme, fifteen large-scale ammonia dispersion tests were
conducted. In order to obtain configurations similar to industrial situations, the
tests were defined with the 6 industrial firms taking part in the programme.
4.1 PRESENTATION OF TESTS CONDUCTED

The fifteen release configurations produced during this test campaign are described briefly
below next to each of the block diagrams. These tests are numbered from 1 to 8 b in the
chronological order in which they were carried out. For each of these configurations, the
pipe diameter is 2 inch (50.8 mm). Unless specified otherwise (case of tests 2, 3, 4p and
2b), the circular orifice also has a diameter of 2 inch.
Horizontal release at a height of 1 metre in

1 2″ gaseous phase. Only test conducted in gas
phase. The subsequent tests all relate to
ammonia releases initially in liquid phase.
20 mm
2 2″ Horizontal release at a height of 1 metre in

liquid phase via a 20 mm orifice.
Release via a 2 inch flange at a height of 1

3 2″ metre. The distance between the two parts of
the flange is 1.6 mm such that the release
surface is equal to 18 mm.
30 mm
4p 2″ Horizontal release at a height of 1 metre via a
30 mm orifice.
4 2″
Horizontal release at a height of 1 metre.
Horizontal release at a height of 1 metre

5 2″ impacting a wall with a surface area of 3 m x 3
m. This wall is located 3 m from the end of the
release system.

Page 45 sur 130
Horizontal release at a height of 1 metre
impacting a wall with a surface area of 3 m x 3
6 2″ m. This wall is located 1 m from the end of the
release system.
7 2″ Vertical release impacting the ground from a

height of 3 m.
8 2″
Vertical release impacting the ground from a
height of 1 m.

9 2″ height of 1 m in a retention dike with a surface
area of 2 m x 2 m and 1 mm high (4 m3).
Release identical to release No. 4. A water

curtain located 25 m from the release point was
10 2″ started up during the test.
Release identical to release No. 4. A water

11 2″ curtain located 60 m from the release point was
started up during the test.
Release identical to No. 4. The temperature of

12 2″
the liquid ammonia inside the tank was lowered
to -12°C before the test.
18 mm
Release identical to No. 2 with a 18 mm
2b 2″ diameter orifice.

height of 1 m in a retention dike with a surface
8b 2″ area of 10 m x 10 m and 5 cm high (5 m3).

Page 46 sur 130
4.2 DESCRIPTION OF TEST RESOURCES
To conduct these large-scale ammonia dispersion tests, the CEA-CESTA (Centre
d'Etudes Scientifiques et Techniques d' Aquitaine, Centre of Scientific and
Technical Studies of Aquitaine) Outdoor Testing Site (TEE) was selected. It is
located approximately sixty kilometres south of Bordeaux. The site covers a
surface area of 950 hectares. It is completely enclosed and access to it is
continuously monitored. The safety study conducted demonstrated that the tests
envisaged could be carried out under satisfactory safety conditions particularly due
to the space available.
The TEE test site is completely flat and free from any obstacles. At the centre of
this site, which has a radius of approximately two kilometres, there is a square
concrete slab with 10 metre sides and approximately 15 cm thick. The releases
were produced from this slab.
4.2.1 Description of release point
In the immediate vicinity of the release slab, three ammonia tanks, of 12 m3 each,
were set up. The liquefied ammonia was stored in the tanks at its saturating
vapour pressure. The dimensions of the tanks were as follows:
total length: 4810 mm
shell length: 3700 mm
internal diameter: 1874 mm
maximum liquid height: 1500 mm
During the tests, only one of these tanks was connected to the release system
located at the centre of the slab, which was equipped with 5 thermocouples and a
pressure sensor, in order to monitor the progression of the pressure and
temperature conditions inside the tank during each test.
The 5 thermocouples were installed on a vertical line from the base of the tank at
the following heights:
50 mm
400 mm
750 mm
1100 mm
1450 mm
The pressure sensor was located at the top part of the tank, in the gaseous phase.
This sensor is referenced P3 in figure 3 overleaf.

Page 47 sur 130
The tank was also connected to a nitrogen bundle used either to pressurise it or to
keep the pressure constant during a test, or to purge, after a test, the pipe
connecting it to the release system.
# ( )
! " # $
% !
' %$ &# !
+$ ,
$ *
figure 3: release point diagram
To produce the fifteen releases defined within the scope of this programme, the
two 2 inch diameter manual valves installed on the tank were used: one was
connected to the gas phase and the other to the liquid phase. This two identical
valves are not full bore and their bore diameter is 3/2 inch.
After the 2 inch valve connected to the liquid phase, a 300 mm long and 41 mm
internal diameter flow-meter was installed. Intended to operate in liquid phase, it
did not give any interpretable results during the tests as the ammonia was two-
phase from the valve located just upstream for almost all the tests.
The connection between the flow-meter and the release system was provided by a
50 mm diameter and 10.4 m long pipe.
Finally, the release system consisted of a 2 inch pipe instrumented with valves,
thermocouples and pressure sensors (see figure 4 overleaf).

Page 48 sur 130
release pipe
flange (to receive
2" diameter different outlet
10.4 m long types)
445
th8 P2 th7
130
110
280
475
1015/ground
600
715
815
1050
1370
2" manual valve (3/2” internal diameter)
2" full-bore pneumatic valve

th7 temperature measurement: 1 mm thermocouple at centre of pipe
P2
pressure measurement: 15 cm offset sensor
Note: all the distances are given in mm.

figure 4: description of release system
The axis of the release system was located 1015 mm above ground level.
The end of the release system was equipped with different types of outlets (see
figure 5) according to the desired test configuration.

Page 49 sur 130
! $ $ /
/00 000 0 0 01 / $ 2$
! !
#
-
.
-
1 -/ $ 3
!
!
#
#
.
.
! / " $ $ ! ! 2
#
.
! !
figure 5: different outlet types for each configuration

Page 50 sur 130
For the tests in which the jet was pointing towards the ground, the concrete slab
surface was instrumented with thermocouples. For the jets against the wall, the
slab was also instrumented with thermocouples. The instrumentation installed and
the references of the thermocouples are described in appendix 1
photo 1: horizontal release impacting a wall
In the case of tests 2, 4p and 2b, the end of the release system was equipped with
a 50 mm pierced cap of a diameter corresponding to the test, i.e. 20 mm, 30 mm
and 18 mm for these three tests, respectively.
For test 3 with the flange, the seal was removed. The thickness between the 2
parts of the flange was 1.6 mm. This is equivalent to a leakage surface area of 255
mm2. Therefore, the equivalent surface diameter is 18 mm.
For tests 10 and 11, water curtains were produced with "peacock tail" hoses. The
characteristics of these peacock tails were as follows:
feed diameter: 70 mm ;
feed pressure: 8 bar ;
water flow rate: 1200 l/min ;
x span: 32 m ;
y span: 10 m.
figure 6: spans of peacock tails used for tests 10 and 11.

Page 51 sur 130
For these two tests, two peacock tails were installed as described in figure 7
below.
0
0 0
0
4 4 4
" "3 $
figure 7
Test 10: Layout of both peacock tails 25 m Test 11: Layout of both peacock tails 60
from release point. m from release point.
The measurement mast position is specified in detail further in this report.
4.2.2 Weather measurement equipment
To measure the atmospheric conditions, a 10-metre high weather mast was set up
350 metres from the release point in the direction noted as 3 (reference M in figure
11). It consisted of three vane anemometers, located at a height of 1.5, 4 and 7
metres, a weather-cock at 7 metres and an ultrasonic anemometer at 10 metres.
A weather station was also set up next to the control station where the
temperature, relative humidity and solar flux measurements were made. These
measurements were made at a height of 1.5 metres. The acquisition frequency of
these parameters was 1 Hz, except for the ultrasonic anemometer, which had a
frequency of 10 Hz.

Page 52 sur 130
The type of equipment used is described below:
1 PULSONIC 429 weather-cock ;
3 PULSONIC 438 vane anemometers ;
1 PULSONIC CM6B pyranometer (solar flux measurement) ;
1 PULSONIC temperature probe ;
1 PULSONIC hygrometer ;
1 METEK – USA-1 ultrasonic anemometer.
An installation diagram is given in figure 8 below.
10m
ultrasonic anemometer
, ! " "3
7m
vane anemometer
4m
1.5
m
figure 8: Weather mast
The ultrasonic anemometer was installed with the x axis facing north and the y
axis facing east.
The weather-cock was facing the opposite direction to the ultrasonic anemometer.
4.2.3 Ammonia measurement chain composition and tests
To conduct the large-scale tests, INERIS chose to use a standalone measurement

chain both to minimise wiring, and also to be easy to move according to weather
conditions (particularly wind direction variations). For this reason, the sensor

Page 53 sur 130
electrical power supply was provided by batteries and cells. For each series of
measurements, the data was stored on "EEPROM5".
In this way, the ammonia measurement chain consisted of an electronic unit on

which it was possible to connect up to 4 sensors with different technologies. In
total, 70 identical electronic units were produced for this programme.
An electronic unit comprises three parts arranged on an assembly plate inside an airtight
housing:
An electronic signal acquisition card, used to amplify the signal, perform the
digital processing, display various parameters, program the configuration
and save data. The main component of the card is a microcontroller
specially adapted for the instrumentation.
An EEPROM saving device designed for the primary acquisition of the data
from the sensors in digital signal form (8192 data items).
An electrical power supply (cells) to power the electronic circuits
The electronic card is equipped with an internal clock which starts up and shuts
down the system at set times. During this period, the EEPROM saves the sensor
measurements at a predetermined rate. For the tests, a 0.25 Hz frequency was
selected, i.e. one measurement every 4 seconds, which made it possible, in view
of EEPROM capacities, to save for 8 hours. At the end of each day of tests, the
data collected and stored in the EEPROM was copied to a PC hard disk via a
specific interface.
To select the sensors to be installed on the 70 electronic units, the releases

envisaged in section 4.1 were modelled using atmospheric dispersion software.
Some general information was obtained from these simulations:
In principle, it is not necessary to fit sensors capable of measuring

concentrations above 1% gaseous ammonia beyond 300 metres from the
release point on the plume axis ;
however, up to 200 metres from the release point, the sensors located on
the plume axis should all be subject to concentrations above 1% ;
finally, beyond 750 metres, the calculated concentrations are below 0.1%,
i.e. 1000 ppm.
These results were considered as orders of magnitude and were adjusted

according to the progress of the programme.
In this way, the sensors were selected according to the concentration to be

measured, which is associated with the distance to the release point leeward from
5 EEPROM: Electrically Erasable Programmable Read Only Memory

Page 54 sur 130
the emission and the distance to the plume axis.
Several ammonia concentration measurement techniques could be envisaged,

such as:
electrochemical cells ;
semiconductors ;
catalytic sensors including pellistors ;
indirect ammonia measurement by measuring oxygen in the air, based on

the paramagnetic properties of oxygen ;
measurement in infrared (IR) ;
chromatography.
Each of these techniques applies to a specific concentration range as shown in

figure 9 below. The concentration ranges are orders of magnitude which can be
extended according to the characteristics of the sensors.
Logarithmic concentration scale in ppm and %
1 10 100 1000 1% 10% 100 %
Electrochemical cell
Semiconductor
Catalytic sensor
Indirect O2
measurement
Infrared and
photoacoustic
spectrometry

Page 55 sur 130
Chromatography
figure 9: Representation of specific concentration ranges for sensors
Appendix 2 describes the operating principle of the first three techniques

mentioned above, which are the most widely used for combustible gas
measurements.
Finally, the sensors, which were connected to the electronic units during the test
campaign, are:
pellistor type catalytic sensors, specific for ammonia (EEV VQ 41) used to
measure ammonia concentrations between 0.1% and 15% (between 1000
ppm and 150,000 ppm). These sensors, 60 in number, were placed in the
near field (up to 200 metres from the release point) ;
electrochemical cells (SENSORIC NH3 3E 1000) measuring ammonia

concentrations between 10 and 1000 ppm. These sensors, 50 in number,
were arranged in the far field (200 metres to 1700 metres from the release
point) ;
60 thermocouples (K type, 1 mm diameter, class 1, stainless steel sheath, 5

m long), used to measure temperature fluctuations. These thermocouples
were installed next to each catalytic sensor up to a distance of 100 metres
from the release point, particularly in order to assess temperature variations
in the jet.
The EEV VQ41 pellistors and the SENSORIC NH3 3E 1000 electrochemical cells
first of all underwent a series of laboratory tests at INERIS in order to verify their
characteristics. The behaviour of these sensors was then tested at INERIS in a 50
metre long gallery with a 10 m2 cross-section. Releases via 0.5 mm to 8 mm
diameter orifices were carried out for these tests and made it possible to validate
the measurement chain (Bouet, 1996).
To facilitate data processing, the microcontroller was programmed to supply the

data directly in the unit (ppm vol.) and in the scale corresponding to the sensor.
The data was stored in EEPROM as follows:
electrochemical cell: in 10 ppm intervals up to 2000 ppm ;
pellistor: in 250 ppm intervals from EEPROM value 0 to 130, and then in
560 ppm intervals from EEPROM value 131 to 255 ;
thermocouple: in 0.3125°C intervals, from -40°C to +40°C.

Page 56 sur 130
Finally, a programme used to read and backup the data saved by the EEPROM
was developed based on Labview software. A processing tool for this data was
also developed based on this software.
4.2.4 Sensor instrumentation on test site
To enable easier installation of the electronic units before a release, the test site
was also equipped beforehand with 150 x 3 m high masts. These masts offered
the possibility of attaching the sensors at 4 different heights, i.e. 0.1, 1, 2 and 3
metres. They were arranged on 7 arcs 180° in amplitude centred on the release
slab, with radii of 20, 50, 100, 200, 500, 800 and 1700 metres. On the first two
arcs located at 20 and 50 metres, 41 masts were installed every 4.5°, and on all
the arcs, 21 masts were installed every 9°.
To enable easier subsequent processing of the tests, the masts installed on the
test site were identified by a letter and a number.
photo 2: Masts installed on the tests site
The letter corresponds to the radius of the arc, i.e. the distance from the release
point (see table 15).

Page 57 sur 130
distance from
release point 20 50 100 200 500 800 1700 4 350 650
(m)
letter A B C D E F G S M T
table 15: Correspondence between distance from release point and associated letter.
The last three distances of 4, 350 and 650 metres identified in the above table
correspond to the sensors installed on the release slab, the weather mast and a
19 metre high mast, respectively. In the case of the latter, from test 8, two
electrochemical cells are installed at heights of 6 and 10 metres.
The number corresponds to a direction from the release point. We defined 80

directions, i.e. an angle of 4.5° between each. In the selected configuration, the
north corresponds to the direction 50 (see figure 10). The layout of the masts was
carried out in the directions between 1 and 41.
The numbering of the angles from the release slab is given in figure 10 below.
'6 3
'6 3
'6 3
$ *
5
figure 10: Identification of angles on release slab

Page 58 sur 130
The numbering of the angles on the test site and the position of the masts are
given in figure 11 below.
Life zone
41 (1 620 m)
39
37
35
Release system
Control room
A
B 33
C
Meteorological mast
D
E
M 31
T 29 (1 680 m)
G 27 (1 580 m)
(1580 m) 1
25 (1 530 m)
3 23 (1 515 m)
21 (1 540 m)
5
7 19 (1 605 m)
9
11 17
13 15
figure 11
Due to the favourable wind direction identified during the initial weather data
analysis and the test site configuration, the measurement field defined in principle
was located between the directions numbered 1 to 41.

Page 59 sur 130
The arc at 1700 metres is limited by the dimensions of the test site for the
directions 1, 19, 21, 23, 25, 27, 29 and 31. For these directions, the distance at
which the masts were positioned is given in brackets.
In each test, the sensors were installed on measurement masts in their final
position leeward from the release approximately 1 hour before the start of the test.
In this way, the sensors were all distributed along each side of the release axis on
an angular sector ± 35° on either side of the axis.
For the impacting releases and the release via the flange, the sensor distribution
angle was greater for the arcs located at 20, 50 and 100 metres (up to 80° on
either side of the release axis). In these configurations, sensors were also added
to the concrete slab over a 4-metre radius.
4.2.5 Acquisition equipment
The data collected during these tests was saved or processed by three separate
types of acquisition equipment for the following points:
the electronic units (measurement chain described in section 4.2.3) used for
the acquisition, firstly, of ammonia concentration values using catalytic
sensors and electrochemical cells connected to them and, secondly,
temperature values measured via thermocouples which were also
connected to them ;
the LabView acquisition system for the weather sensors and for the
pressure sensors located on the release point ;
acquisition system for the thermocouples located on the release point, i.e.
the thermocouples arranged in the tank, on the release system and on the
ground or concrete blocks (acquisition by scanning).
The acquisition system for the data from the weather sensors and the pressure
sensors was installed in the control station, in order to monitor the parameters
during the test, at the same time as the acquisition, in particular:
the pressure in the tank ;
the pressure at the centre of the release system ;
the pressure at the release orifice ;
the wind velocity ;
and the wind direction.

Page 60 sur 130
The location and acquisition frequency for each of these sensors are given in table
16 below.
distance from acquisition

z
sensors release point (°) – dir No. frequency
(m)
(m) (Hz)
vane anemometer 350 211.5 - 3 7 1
vane anemometer 350 211.5 - 3 4 1
vane anemometer 350 211.5 - 3 1.5 1
ultrasonic anemometer 350 211.5 - 3 10 10
weather-cock 350 211.5 - 3 7 1
tank T (50 mm) release system 0.05 0.2
pipe T (th8) release system 1 0.2
pipe T (th7) release system 1 0.2
orifice T (th6) release system 1 0.2
thermocouples 11 to 30 Ground and obstacle 0.2
solar flux 450 270 - 70 1.5 1
relative humidity 450 270 - 70 1.5 1
temperature probe 450 270 - 70 1.5 1
ammonia sensors Leeward from releases 0.25
tank pressure (P3) release system 1.9 1
pipe pressure (P2) release system 1 1
orifice pressure (P1) release system 1 1
table 16: sensor position and acquisition frequency

Page 61 sur 130
4.2.6 Video equipment
To monitor the tests from the control station, five CEA-CESTA cameras were
placed on the test site:
one surveillance camera at a height of 30 metres located between the

release point and the control station ;
one camera located 10 metres behind the release point, at a height of 10

metres ;
one camera located 40 metres behind the release point, at a height of 26

metres ;
one camera located 1000 metres behind the release point in direction 33, at
a height of 1 metre ;
one mobile camera at a height of 1.5 metre, positioned perpendicular to the

release direction.
4.2.7 Safety
To conduct these tests, the site was divided into four zones:
the "forefront zone" where the releases were carried out ;
the "support zone", which was located in a 200 metre radius around the
release point ;
the "control room", located 450 metres windward from the release point ;
and the "living zone", located at the entrance to the site approximately 1400
metres from the release point.
Each person present on the site was equipped with a filtering mask. In the "control
room", the personnel present wore "tyvec" clothing. Six people equipped with
special clothing and an individual breathing apparatus were prepared to intervene
in the "support zone". Two of these people were also equipped with a pressure
suit to intervene in the "forefront zone", in particular, if required, in the cold
ammonia cloud. This never occurred.
No personnel were present to the ammonia release outside the "control room",
apart from the six people located in the "support zone".
The different teams working on the site communicated with radio equipment.
A shower system was set up at the entrance to the "forefront zone".
A medical service was available at any times next to the "control room".

Page 62 sur 130
5. RELEASE CONDITION MEASUREMENTS
In order to understand the atmospheric dispersion of a pressurised liquefied
ammonia release, it is important, firstly, to determine the ammonia flow conditions
in the pipe, and, secondly, to evaluate the weather conditions at the time of
release correctly.
5.1 FLOW CONDITIONS IN PIPE
Initially, we will present the experimental conditions in which the releases carried
out within the scope of this programme were conducted. Then, these release
conditions are analysed.
5.1.1 Experimental release conditions
The experimental conditions under which the tests were conducted are frequently
referred to as the "source term" of the release. The source term particularly
specifies the flow rate, temperature, pressure, velocity and quality6 values at the
opening. In this programme, the measured source term values are essentially
temperature and pressure values, along with the quantity of ammonia released.
During each test, the pressure and temperature conditions were measured and
recorded continuously every second, both in the tank and at different points of the
pipe (see chapter 4).
As an illustration, the two graphs in figure 12 below show:
the typical progression of the pressure conditions in the tank and in the pipe
during a release ;
the progression of the temperatures located at different heights in the tank

and located in the pipe.
6 release quality: a single-constituent two-phase flow is essentially characterised by its quality, i.e.
the mass fraction of its vapour phase

Page 63 sur 130
These two graphs show the conditions measured and recorded during test 4.
!"
(essai 4) #
7 2" #
7 3 #
6
4
"
0
0 100 200 300 400 500 600 700 800
# "
(s)
!"
(essai 4)
20
10
0
$"
-10
-20
7 # 3( )
7 # 3( )
-30
#
7 # 3( )
#$ # (! )
7 # 3( )
-40 7 # 3( )
7 # 1 2 #$ $ (! )
# (! )
-50
0 100 200 300 400 500 600 700 800
Time(s)
figure 12: Progression of pressure and temperature conditions in tank and in release pipe

Page 64 sur 130
For test 1, the gaseous release was produced from the vapour pressure inside the
tank. The tank was then depressurised and the test was stopped when the
pressure had reached approximately 1.8 bar absolute.
With respect to test 2, a 1 bar pressure surge was produced beforehand with
nitrogen inside the tank. The nitrogen regulation system then kept this pressure
constant throughout the test.
During test 3, via the flange, no nitrogen pressure surge was produced
beforehand. However, the initial existing pressure in the tank was maintained
throughout the test by means of nitrogen injection.
For all the subsequent tests, no nitrogen pressure surge was produced before the
test and the pressure was not kept constant throughout the test. In this way, for
each test, a varying pressure drop was observed in the tank according to the
release conditions.
For each test, the behaviour of the pressure and temperature sensors was
substantially the same. As a general rule, with respect to the values measured
during the tests, it is possible to note that:
inside the instrumented tank, the value of the temperature in the gaseous
ceiling was identical to the temperature of the liquid ammonia from the
second part of the night. On the other hand, after sunrise, temperature
differences between the two phases could be observed. In this way,
depending on the outside temperature, the sunlight and tank filling rate, the
temperature of the gas phase sometimes increased more rapidly than that
of the liquid phase. Maximum differences of 10°C between the two phases
were measured within the scope of this programme ;
before a release, the value of the pressure in the tank rarely corresponded
to the saturating vapour pressure of the liquid ammonia temperature. It was
generally between the saturating vapour pressure corresponding to the
temperature of the gaseous ceiling of the tank and the saturating vapour
pressure corresponding to the temperature of the liquid ammonia
depending on the time of day, the ambient temperature, the sunlight
conditions and the tank filling rate. In some configurations, the difference
between the pressure in the tank and the saturating vapour pressure
corresponding to the temperature of the liquid ammonia was of the order of
1 bar ;
in the case of a release in which the temperature of the gaseous phase was
initially greater than the temperature of the liquid phase, a decrease in the
temperature of the gas phase to the temperature of the liquid phase was
conventionally observed. Depending on the initial and ambient conditions,
the temperature was liable to fall rapidly after the start of the test, and then
became equal to the temperature of the liquid phase ammonia during the
test ;

Page 65 sur 130
the pressure inside the tank, progressing under similar conditions to the gas
phase saturating vapour curve, could, therefore, in the same way as the
temperature values, change rapidly during the first minute depending on the
time of day and sunlight conditions ;
once the first minute of the release had elapsed, the pressure and
temperature conditions inside the tank progressed in a constant and slower
manner for the remainder of the release: drop of 200 to 400 mbar for a
pressure of the order of 6 bar absolute, and 0.5 to 2°C temperature
decrease ;
while the ammonia was flowing in the pipe, a drop in pressure and,
therefore, temperature was observed between the tank and the orifice.
Typically, for a temperature value of approximately 10°C in the tank, the
temperature at the orifice was approximately -15°C ;
unlike the conditions in the tank during a test, the pressure and temperature
conditions at the opening changed more significantly: on average, drop in
pressure between 0.5 and 1 bar for an initial orifice pressure below 3 bar
absolute, and decrease in orifice temperature between 5 and 10°C ;
the pressure and temperature measurements made at the same point of the
release pipe demonstrated that the pressure was similar to the saturating
vapour pressure corresponding to the temperature measured at this point ;
after the end of the release, since the valves were closed, the temperature
and pressure values progressed more or less rapidly to their pre-test
conditions, except for the pressure and temperature at the orifice. In fact,
since the orifice is at atmospheric pressure, the liquid ammonia remaining
at the orifice continued to be vaporised by lowering its temperature further.
During the tests, no satisfactory continuous flow rate measurement was made.
Only the mean release flow rate value was estimated on the basis of the quantity
of ammonia released and the release time.

Page 66 sur 130
The general characteristics of all the releases carried out are described in the
table below.
Release characteristics
Test No. Time Quantity released Mean flow rate
(sec) (kg) (kg/s)
1 460 300 0.65
2 440 1980 4.5
3 540 1400 2.6
4 prel. 660 2380 3.6
4 600 2520 4.2
5 600 2520 4.2
6 600 2520 4.2
7 670 2560 3.8
8 875 3420 3.9
9 605 2370 3.9
10 690 2150 3.1
11 600 1800 3
12 660 2000 3
2b 720 1400 1.95
8b 780 2120 2.7

table 17 General characteristics of release conditions
5.1.2 Release flow condition analyses
When ammonia is stored in the liquid state at its saturating vapour pressure, its
use generally results in a pressurised liquid flow in a pipe. In the event of a rupture
on this pipe, the flow conditions are modified, with the ammonia changing from
operating pressure and temperature conditions to conditions corresponding to
atmospheric pressure and a temperature below that prior to the rupture.

Page 67 sur 130
These new conditions at the opening site must be determined in order to estimate
the characteristics of the release, particularly the mass flow rate, the release rate,
the release quality, pressure and temperature. These values form what is referred
to as the "source term" of the release. The difficulty establishing the mass flow rate
of the release at the opening particularly lies in the evaluation of the flow state in
the pipe, If the liquid ammonia encounters a low pressure wave, it is vaporised.
Ammonia is said to undergo a thermodynamic flash generating a mixture of liquid
and vapour which affects the flow.
A distinction is essentially made between three different flow states, i.e.:
bubble state: the liquid phase is continuous and contains gas bubbles ; in
this case, liquid is in the majority ;
annular or semi-annular state: the gas velocity is greater than that of the
liquid against the walls of the pipe ;
dispersed flow: the gas flow rate is high, the liquid is dispersed in the gas
phase; in this case, the liquid is the minority phase in the mixture.
In order to determine the source term at the release point, it is necessary to

understand and determine the characteristics of the flow in the pipe.
There are a large number of models used to describe the release in the case of
ejection via a conduit. These models, mentioned for example by Langard (1995),
Bigot (1996), U.I.C. (1987), can be classified into three categories in order of
increasing complexity:
homogenous equilibrium models (HEM) ;
models with differences in velocity (between gaseous phase and liquid

phase) ;
non-equilibrium models.
In the homogeneous equilibrium models, the gas and liquid phases are assumed
to:
have the same velocity;
be distributed homogeneously;
be at the saturation temperature.
These assumptions imply that:
the two-phase flow is considered as a homogeneous fluid with average

properties between those of the gas and liquid ;
the flow is at thermodynamic equilibrium (no heat or mass exchange takes

place between the phases).
Page 68 sur 130
In addition, the flow is assumed to be stationary and one-dimensional.
There are two main categories of equilibrium models:
isentropic expansion models: in this case, the entropy of the fluid subject to
the expansion remains constant while its enthalpy decreases. In this case,
the loss of enthalpy is assumed to be balanced by the gain in kinetic energy
in the flow.
isenthalpic expansion models: in this case, the enthalpy of the fluid remains
constant while its entropy decreases. This loss is balanced by the friction
energy applied by the flow.
Van den Akker et al (1983) demonstrated that these two assumptions are equally
unrealistic but that the error generated in the "constant enthalpy" assumption is
less significant.
The purpose of these models is to determine the flow rate at the opening. Within
the scope of this programme, the tests conducted make it possible to determine
the temperature and pressure at the opening and the mean flow rate throughout
the test. A HEM model makes it possible to determine, in addition to these values,
the quality of the flow at the opening. This parameter was not measured directly
during the tests. A complete model, proposed by Wheatley (April 1987) and
belonging to the HEM model category, is presented below and its results are
compared to the experimental results obtained during the tests.
Modelling of flow rate at opening: Wheatley approach (April 1987)
The purpose of this modelling is to determine on the basis of the tank conditions
(Pint, Tint):
the pressure-temperature conditions at the opening ;
the mass flow rate of the release ;
the quality at the opening.
The basic hypotheses for this model are that:
the fluid expansion is considered to be isentropic and stationary
the two-phase fluid is considered to be homogenous and in liquid-vapour

equilibrium
In this way, the model proposed is based on the following equations:
98 (71 )
entropy equation: Cpliq.Ln(Tint) = Cpliq.Ln(Tb) + (1)
71

Page 69 sur 130
:1
Euler's equation: =α 9 + 9! (2)
1
internal pressure equation as a function of temperature and liquid level:

Pint = Psat(T) + liq.g.hint = 4 − + liq.g.hint (3)
7
Specific volume equation: v = (1 – x).vliq + x.vgas (4)
Clausius-Clapeyron's equation: Lv(T) T(vgas - vliq) (5)

7
After integration, Euler's equation is expressed as follows:
:1 − 7
=α 1
+; $ & (7 − 71 ) − ; $ & 971 98 + 9! (6)
ρ$ & 71
This expression is used to determine the velocity of the fluid at the opening as a
function of the difference in pressure between the tank and the outlet. Since the
fluid is assumed to be at saturation, the temperatures are known once the
pressures are known.
The coefficient p is an empirical corrective factor used to account for the effects of
friction; as a general rule, it is taken to be equal to 0.6 in the case of a wall
opening and equal to 0.8 or 0.9 in the case of a pipe rupture.
The problem then consists of determining the pressure of the fluid at the opening.
The existence of a pressure gradient between the tank and the opening induces a
significant acceleration of the fluid, i.e. potential energy is converted into kinetic
energy. The velocity of the fluid at the outlet will be limited by the propagation rate
of the pressure waves in the fluid, i.e. the speed of sound.
It is accepted that, if the pressure gradient is sufficiently high, the fluid will reach
the speed of sound at the outlet (May, 1996). If this is the case, the flow is said to
be critical.
A critical (or sonic or blocked) single-phase flow is characterised in that a

downstream decrease in pressure, below a certain threshold, has no influence on
the flow. This pressure threshold is referred to as the "critical pressure". It is
achieved in a well-defined section of the conduit, generally at the outlet. In this
section, the flow rate is equal to the speed of sound in the fluid, so much so that
disturbances generated downstream do not affect the upstream flow.
Experience shows a similar phenomenon for two-phase flows. On the other hand,
it is more difficult to determine the value of the pressure wave propagation rate in
a two-phase fluid given that these waves do not move at the same speed in each
of the two phases. However, the experiments conducted for the two-phase flow
study demonstrated that the speed of sound of a two-phase flow is markedly lower

Page 70 sur 130
than the speed of sound of a single-phase flow (Delhaye, 1981). As a result, the
critical flow is achieved more easily in the case of a two-phase flow (May, 1996).
The modelling of the flow rate at the opening using the method proposed by
Wheatley (April 1997) assumes that the two-phase flow is critical which is
equivalent to considering that the velocity of the fluid at the outlet, Ub, is equal to
the speed of sound in the two-phase fluid of quality xb.
Due to the flow isentropy hypothesis, the speed of sound can be obtained using
the equation:
∂
: = (7)
∂ρ
#
<= "
In the case of a two-phase fluid, integration at constant entropy is relatively

complex. Finally, it was demonstrated that the speed of sound is expressed as
follows:
19
: # = (8)
7
ρ 1 971 ; $ & 971 + + 98
71 71
The density of the fluid at the opening, b, is obtained using the flowing equation:
7
8 ; $ & 971
71
= + (9)
ρ1 ρ$ & 9 1
In this case, the pressure Pb such that Ub = Usound (equations 6 and 8) is sought
using an iterative process. Once this pressure is obtained, the release temperature
is known (saturation) and the quality at the opening is determined using
equation 4:
+
ρ1 ρ$ &
1 = (10)
+
ρ ρ$ &
The mass flow rate, Gb, is then determined using the conventional equations:
Gb = b.Ab.Ub (11)
where Ab is the opening surface area
The release pressure value found by testing the equality between the release
velocity and the speed of sound will determine the type of release. In fact, if this
value is above atmospheric pressure, the flow will be critical and the jet will
accelerate further after the opening due to its expansion to atmospheric pressure.

Page 71 sur 130
On the other hand, if the pressure obtained is below atmospheric pressure, it has
no physical significance. In this case, the flow will not be critical (the fluid at the
opening does not reach the speed of sound) and the pressure at the opening is
considered to be equal to the atmospheric pressure.
Non-equilibrium models
Models closer to reality exist, referred to as "non-equilibrium models". They make

it possible to account for thermodynamic differences existing with liquid-vapour
equilibrium.
In addition, they include the fact that the two phases are separate, and are
considered separately, i.e. that they can have different velocities and that phase
changes may be limited by heat transfers or accelerated by the presence of
irregularities.
An example includes the "frozen" model which makes the simple assumption that
there is no change of phase. The flow is then modelled using Bernoulli's equation.
The actual separate phase model evaluates the material, energy and quantity of
movement on each of the two phases and applies a vaporisation law in view of the
flow state. This model is based on empirical correlations qualifying the flow states.
More comprehensive models such as Yan's model (1991), which models

nucleation, also exist. It particularly makes it possible to estimate the frequency of
occurrence of bubbles and the progression of their size.
In order to compare the experimental values to theoretical values, two models

(Fauske and Wheatley) were used on the basis of the test conditions for the tests
conducted by INERIS and described in chapter 4 above. As a general rule, the
results obtained using these models differ substantially from the values recorded
during the tests, where the values obtained using the models were generally
greater than the experimental values.
These results confirm the fact that the modelling of the flow rate in the case of a
two-phase flow is highly complex and that simplified models are unable to account
for all the phenomena arising during flow. In addition, within the scope of these
tests, which were not intended to study the flow rate at the opening in an in-depth
way, but simply to determine it, the release system used consisted of significant
pressure drops on a pipe over 12 metres in length.
We verified experimentally that the fluid at the orifice was indeed under saturation
conditions, i.e. that the (Tb,Pb) pairs measured experimentally corresponded to the
(Tb,Psat(Tb)) pairs, the basic hypothesis for HEM models.

Page 72 sur 130
In conclusion, these models are suitable for tank leakages or releases after a short
pipe. They do not account for pressure drops and, therefore, ammonia
vaporisation during its transport outside. If an opening is considered to be flush
with the tank wall, ammonia is vaporised instantaneously into the atmosphere
without passing via the three flow states (bubble state, annular state and
dispersed flow).
In the tests conducted, the pipe is sufficiently large for the flow to pass via these
three flow states. In the literature, it is admitted that dispersed flow is established
when the condition L>2D is met (L pipe length, D pipe diameter), which is the
case. In this way, the dynamic behaviour of ammonia is influenced by these three
states and the phenomena are complex to model.
5.2 WEATHER CONDITIONS DURING THE TESTS
The weather conditions are described using a number of parameters, the main
one being associated with the wind velocity, on one hand, and atmospheric
turbulence, on the other.
For each test, the atmospheric conditions were recorded using the equipment
described in chapter 4 above. During the recording period, some values measured
were subject to little and slow variation. This particularly applied to the ambient
temperature, relative humidity and solar flux. Note that these values were
measured on the weather station next to the control station.
However, other values, such as the wind velocity and direction measured on the
weather mast referenced M in figure 11, could be very variable during the
recording period.
In order to give an order of magnitude of the weather conditions during the tests,
the mean values associated with each of the releases are given in the table below.
Mean atmospheric conditions
Test No. Wind velocity Wind

Sheltered Relative Solar flux
2 at height of direction at
temperature humidity (kW/m )
7m height of 7 m
1 14°C 76% 0.2 5 m/s 260°
2 8.5°C 84% 0.1 3 m/s 305°
3 5°C 90% 0.04 2.5 m/s 5°
4p 10°C 61% 0.5 3 m/s 10°
4 12.5°C 82% 0.25 3 m/s 290°
5 20°C 50% 0.65 3.5 m/s 310°
6 12°C 75% 0.5 5 m/s 285°

Page 73 sur 130
7 13°C 60% 0.7 6 m/s 310°
8 20°C 37% 0.7 5 m/s 275°
9 14°C 50% 1.0 3 m/s 355°
10 24°C 20% 1.0 3 m/s 10°
11 24°C 24% 0.7 5 m/s 310°
12 15°C 33% 0.6 6 m/s 310°
2b 18°C 31% 0.7 4 m/s 315°
8b 18°C 31% 0.6 5 m/s 310°
table 18: mean atmospheric conditions during each release
In addition, to evaluate the dispersion of ammonia in the atmosphere, it is

important to characterise the atmospheric turbulence well.
The atmospheric turbulence is due both to turbulences of mechanical origin,

generated by ground roughness, and turbulences of thermal origin associated with
the vertical stratification of the temperature of the atmosphere. It can be evaluated
by recording wind velocity and direction fluctuations at one point.
A distinction can be made between two types of turbulences of mechanical origin:

small-scale turbulence and large-scale turbulence.
Small-scale turbulence, generated by the friction of wind on the ground, is

influenced by roughness. It is applied within a layer of some hundred metres in
height. This type of turbulence is a predominant factor in the dispersion
phenomenon.
Large-scale turbulence is associated with distant weather phenomena. It

influences the transport or diffusion of particles, if the study site covers several
dozen kilometres, which is not the case in this programme.
In the case of turbulences of thermal origin, an important characteristic is the

atmospheric stability. It is associated with the vertical air density gradient, in turn
associated with the vertical temperature gradient. The adiabatic vertical
temperature variation is between -0.6°C/100 m and -1°C/100 m according to the
air humidity level. To simplify, when the vertical temperature gradient is below that
of an adiabatic atmosphere, the atmosphere is said to be unstable. If the gradient
corresponds to the adiabatic, the atmosphere is said to be neutral. Finally, if the
thermal gradient is positive (therefore, greater than that of the adiabatic), the
atmosphere is said to be stable, in other words, unfavourable to effective vertical
diffusion of a pollutant.

Page 74 sur 130
With this respect, Pasquill proposes a classification comprising 6 stability classes,
referenced A to F, covering cases of very unstable to unstable atmospheres
(classes A and B) to cases of stable to very stable atmospheres (classes E and F),
including neutral cases (classes C and D).
If an atmosphere is unstable (Pasquill's classes A and B), atmospheric dispersion

is favoured, resulting in the formation of a very open plume (see figure 13,
Pasquill, 1974).
figure 13: Dispersion in an unstable atmosphere (taken from Pasquill, 1974
On the other hand, if an atmosphere is stable, its ability to diffuse a pollutant

vertically is low, resulting in the formation of a very closed plume (see figure 14,
Pasquill, 1974).
figure 14: Dispersion in a stable atmosphere (taken from Pasquill, 1974
From an experimental point of view, there are different methods available to

determine the stability of the atmosphere, such as, for example, the determination:
of the standard deviation of the wind direction ( ) ;
of the vertical temperature gradient ( T) ;
of the Richardson number (Ri) ;
of the Monin-Obukhov length ;

Page 75 sur 130
or, methods based on criteria such as the mean wind velocity at a fixed
height, the altitude of the sun, cloud coverage, etc.
On the basis of the different values measured during the test campaign, we
applied two of these methods: the determination of the standard deviation of the
wind direction ( ) and the method based on the determination of the Monin-
Obukhov length. After a brief description of these methods, we will present the
results obtained for all the tests.
Standard deviation of the wind direction ( )
The standard deviations of the wind direction essentially represent the horizontal
stability of the atmosphere (the vertical turbulence of the atmosphere being
estimated on the basis of the vertical temperature gradient).
For each test, the standard deviation of the wind direction ( ) was calculated over
a period comprising approximately 5 minutes before the test, the duration of the
test and approximately 15 minutes after the test, i.e. a period of approximately one
half-hour.
Determination of Monin-Obhuckov length
For each of the tests, we also determined the Monin-Obukhov length which is a
characteristic length of atmospheric turbulence. This length, referenced L, is
defined as follows:
L = - Cpu*k(g/T0)H
where is the density of the fluid, Cp is its heat capacity, u* is the friction
velocity, k in the von Karman constant, g is the gravity constant, T0 is the
temperature at ground level and H is the vertical heat flux.
The values of u* and H were determined using the values measured by the
ultrasonic anemometer with an acquisition frequency of 10 Hz.
Golder (1972) defined a relationship between Pasquill's stability classes, the mean
site roughness z0, and the length L.

Page 76 sur 130
figure 15: Relationship between the Monin-Obukhov L and the roughness z0
for different Pasquill stability classes (Golder, 1972).

Page 77 sur 130
On the basis of the values recorded during the tests, the values of and L were
calculated. Then, for each test, the Pasquill stability classes were estimated using
each of these two methods. The results obtained are given in the table below. In
addition, for each of the tests, an estimation of the visible plume length is given.
Estimated atmospheric stability classes

estimation of visible
Test No. by determining the
by determining the Monin- plume length (m)
standard deviations of the
Obukhov length (L)
wind direction ( )
1 D - 5m
2 D C/D 500 m
3 C D 350 m
4 prel. C - 400 m
4 C D 500 m
5 A B 30 m
6 D D 60 m
7 C C/D 60 m
8 D C 25 m
120 m then
9 A B/C
30 m
10 A A/B 70 m
11 C C 70 m
12 D C/D 70 m
2b A A/B 60 m
8b C B/C 15 m
-: not determined
table 19: Atmospheric stability classes estimated during tests
The use of these methods, for given atmospheric conditions, may result in different
stability classes. However, during the tests, no significant differences were
observed.

Page 78 sur 130
In addition, in relation to the general atmospheric conditions during the tests, the
following points were noted:
From test No. 4, there was no rain on the test site. In this way, relatively low
daytime relative humidity levels were observed (up to 20% for test No. 10) ;
The visible part of the plume varies considerably. As mentioned above, this
is essentially due to the relative humidity level of the atmosphere. In fact,
the ammonia cloud is relatively cold (approximately -40°C on the release
axis 20 metres from the release point) due to the presence of ammonia
droplets which take up the heat, particularly in the ambient air, to vaporise.
The low temperatures reached mean that the water vapour present on the
plume path is condensed and forms a visible cloud until the plume is heated
by diluting with ambient air.
In this way, for almost identical release conditions (tests 4 and 11), the
visible part of the plume changed from approximately 500 metres to
approximately 70 metres for a relative humidity level which changed from
82 to 24% ;
The solar flux varied considerably between the tests, with very low values at
the start of the campaign (0.04 kW/m2) and high values in April (1 kW/m2) ;
For all the tests, the mean wind velocity oscillated between 2 and 6 m/s.
The mean wind direction for each test was very variable. These mean wind
directions for each test are given in figure 16 overleaf.
photo 3: Typical ammonia plume

Page 79 sur 130
Life zone
41
39
37
35
Release system
Control room Trial 1
33
E 31
29
Trial 8
G 27
Trial 6
1
25 Trial 4
3 23
21 trials 2,7,11
5
trial 5
7 19
9 trial 12
11 17
13 15
trial 3 trial 2bis
trial 9
trial 4p
trial 10 trial 8bis
figure 16: mean wind direction for each test

Page 80 sur 130
6. MEASUREMENTS RECORDED LEEWARD FROM RELEASES
The values recorded by the ammonia concentration sensors leeward from the
releases are given in the sub-section below.
These values are then analysed in sub-section 6.2.
6.1 AMMONIA CONCENTRATION SENSORS
Note that the ammonia concentration measurement equipment used in the test
campaign is described in chapter 4. These concentrations were essentially
measured leeward from the releases on 7 arcs.
The responses from the concentration sensors were very different according to
their positions with reference to the plume axis and with reference to the distance
from the release point.
For example, the two figures below give two signals representing the ammonia
concentrations (in ppm) recorded during test No. 2 leeward from the release, by a
pellistor located 100 metres from the release point (figure 17) and by an
electrochemical cells located at a distance of 800 metres (figure 18). Note that the
release time was 440 seconds.
figure 17: Progression over time of ammonia concentration (in ppm) 100 m from release point
observed during test 2.

Page 81 sur 130
figure 18: Progression over time of ammonia concentration (in ppm) 800 m
from release point observed during test 2.
Depending on the phenomena under study, there are different options for the
analysis of the fluctuations in concentration recorded during the tests. In this
document, the concentration values given are mean values of the concentrations
measured on the plume axis.
To obtain these values, we recorded, at each time interval (4 s), the maximum
concentration value on each of the 6 arcs located 20, 50, 100, 200, 500 and 800
metres from the release point. Then, on each arc, we calculated the mean of all
the maximum values recorded over time, excluding the first and last minute of
exposure of each sensor. The start of sensor exposure was determined by
accounting for a cloud advection time. This time was taken to be equal to the ratio
of the distance between the arc in question and the release point over the mean
wind velocity during the release.

Page 82 sur 130
The table below gives the mean concentration values determined in this way on
the plume axis for each of the tests conducted.
mean concentrations in ppm
measured at height of 1 m on release axis
Test No. 20m 50m 100m 200m 500m 800m
1 15000 4500 1500 400 40 10

2 55000 31000 20000 9000 900 300
2b 33000 18000 6600 1600 60 20
3 33000 20000 800 2700 250 100
4 65000 27000 16000 10000 1200 500
5 28000 14000 6500 1500 140 30
6 28000 13000 7300 2500 450 150
7 25000 11000 5300 1000 160 60
8 26000 12000 - - - -
8b 20000 8500 4000 800 70 30
9 20000 5000 2000 450 30 5
10 65000 29000 18000 4000 80 20
11 65000 27000 15000 3500 300 80
12 60000 20000 7000 1500 120 -
-: concentration value on plume axis not measured
table 20: Mean ammonia concentrations on plume axis at height of 1 m
Test 4p described in chapter 4 was an intermediate test to validate the procedure

for the subsequent tests. The concentration values measured during this test did
not make it possible to determine the mean values leeward from the release.
Therefore, this test is not mentioned in the above table.
The ammonia concentration values given in table 20 above were rounded off in
view of the precision of the sensors. The progression of the ammonia
concentrations on the plume axis as a function of the distance to the release point
was plotted for each test. These progressions are given in figure 19 overleaf.

Page 83 sur 130
Page 84 sur 130
1
1
-
"
"
figure 19: Progression of ammonia concentrations on "plume axis as a function of distance to release point

Page 85 sur 130
As a general rule, after the tests, several remarks can be noted:
Liquefied ammonia releases behave like heavy gas releases as illustrated

in figure 20 below. In this figure, we have indicated the response of 4
sensors located on the same mast 50 metres from the release point and on
the plume path axis during test 10.
9 =
91
9 = 91
91
9 = 91
"
9 =
9 =
9 =
9 =
9 =
9 =
9 =
9 =
-
"
Legend ' 10pe-01.b11’: The first number ‘10’ corresponds to the trial number, the two letters ‘pe'(or
‘tc'or ' ec'
) corresponds respectively to the sensor type (pellistor, thermocouple or electrochemical
cell), the number ‘01’ corresponds to the height level, and the extension corresponds to the mast
number.
figure 20: Progression of ammonia concentration values

for sensors located on the same mast 50 metres from the release point.
On this mast, the sensor heights, referenced 01 to 04 in the legend associated

with these curves, are 0.1 m, 1 m, 2 m and 3 m above ground level, respectively.
The vertical ammonia concentration gradient is high 50 metres from the release
point. In fact, at ground level, the concentration is greater than 40,000 ppm (4%)
and is more than 10 times lower at a height of 3 m for this test.
In addition, since the weather conditions were unstable on the day of this test,
considerable fluctuations in the ammonia concentration values are observed at
times.

Page 86 sur 130
The vertical ammonia concentration gradient is still pronounced 100 metres from
the release point, as illustrated in figure 21 below for two sensors at heights of 1
and 3 metres.
9"
9"

for sensors located on the same mast 100 metres from the release point
As a general rule, a vertical concentration gradient within the cloud was still
observed 200 metres from the release point.
That being said, from 500 metres, the concentration values measured at heights of
1 and 3 metres above ground level were practically identical as illustrated in figure
22 below.
9 =
9 =
" 9
9 = " 9
9 =
"
9 =
9 =
9 =
9 =
9 =
9 =
9 =
-
"

for sensors located on the same mast 500 metres from the release point

Page 87 sur 130
For the gaseous phase ammonia release, heavy gas type behaviour was
also observed on the sensors located 20 metres from the release point (see
figure 23). However, 50 metres from the release point, this effect is
practically not noticeable.
2.50E+04
1pe-01.a38
1pe-02.a38
1pe-03.a38
2.00E+04 1pe-04.a38
")
1.50E+04
1.00E+04
5.00E+03
0.00E+00
600 650 700 750 800 850 900 950 1000 1050 1100
"

20 metres from the release point
In addition, throughout the tests, no rise in the ammonia cloud formed was
observed.
6.2 ANALYSES
In this section, we will study the influence of some parameters on the atmospheric
dispersion of ammonia. These parameters are as follows:
the type of orifice ;
the atmospheric stability ;
the presence of obstacles ;
the presence of a retention dike ;
and the use of peacock tails.
Finally, in the final section, the experimental results measured during this test
campaign are compared to calculation results obtained using an integral
atmospheric dispersion software program.

Page 88 sur 130
6.2.1 Influence of release orifice
During the tests, 4 types of orifice were used on the release system, i.e.:
A) a 18 mm diameter orifice Test 2b
B) a 20 mm diameter orifice Test 2
C) a 2 inch flange (50.8 mm) Test 3
D) and a 50.8 mm diameter orifice Tests 1, 4, 10, 11, 12 (5 open-field releases)
Tests 5, 6, 7, 8, 8b, 9 (6 impacting releases)
These 4 types of orifice will be referenced A to D hereafter.
In order to evaluate the influence of the type of orifice on atmospheric dispersion,

an ammonia plume width was estimated for the 8 open-field releases at the arc
located 20 metres from the release point. To estimate this width, for each test, we
determined the zone in which the ammonia concentration was greater than or
equal to half the maximum concentration recorded at 20 metres. The results
obtained are given in the table below.
plume width associated

orifice type test
where C > Cmax/2 angular sector
A 2b (diam. 18 mm) 9m 25°
B 2 (diam. 20 mm) 9m 25°
C 3 (flange) 49 m 140°
1 4m 11°
4 5m 15°
D 10 6m 18°
11 5m 16°
12 7m 20°
table 21: plume width for open-field releases
Several remarks can be made on the basis of this table:
The release via the flange (orifice C) produces, in the near field, the widest
cloud of all the open-field tests. This is essentially due to the fact that the
nozzle orients the ammonia projection along directions perpendicular to the
release axis. The resulting plume is extended in width, progressing axially
with the wind. In this way, despite a 33% greater flow rate than test 2b, the
ammonia concentration values measured during test 3 with the flange are of
the same order of magnitude as for test 2b over the first 50 metres;
The tests in which the end if free (50.8 mm diameter orifice D) result in jets
referred to as "linear jets". Under these conditions, the width of the plume is
narrower than under other conditions;
Page 89 sur 130
Finally, the releases in which a diameter restriction was added at the end
(orifices A and B) result in greater plume widths. This may be due to the fact
that, for these tests, restricting the diameter at the orifice results in a pipe
pressure at the opening (> 5 bar) which is greater than the pressure at the
opening for tests with no diameter restriction (between 1.5 and 3 bar). As a
result, the thermodynamic flash that takes place at the outlet into the
atmosphere is more significant and, therefore, the resulting volumetric
expansion results in a larger plume size.
In sum, the orifices affect the physiognomy of the cloud (width, height) and, as a
result, the mean concentrations. This is clearly visible and measurable during the
first 50 metres. At greater distances, this influence wears off in favour of that of the
atmospheric stability, as demonstrated below in this report.
6.2.2 Influence of atmospheric stability
The influence of the release conditions in the near field may be considerable over
a few dozen metres after the release point. At further distances, the atmospheric
stability becomes predominant as the passive dispersion phase is then involved.
In this way, significant differences in concentrations may be observed in the far

field for releases with identical flow rates but with different atmospheric stability
conditions. For example, tests 5 and 6 were conducted with similar release
conditions and with different atmospheric conditions, type (A/B; 3.5 m/s) for test 5
and type (D; 5 m/s) for test 6, respectively.
In addition, tests 10 and 11 were conducted on the same day, the first in the early
afternoon and the second in the late afternoon, with identical release conditions
and different atmospheric conditions, type (A/B; 3 m/s) for test 10 and type (C; 5
m/s) for test 11, respectively.
The progressions of the concentrations in tests 5 and 6, on one hand, and tests 10
and 11, on the other, are described in the graph in figure 24 below
( .4 0) ( .4 )
(> ) (; )
""
""
" "
figure 24: Progression of concentrations on release axis as a function of distance

for different weather conditions.

Page 90 sur 130
In these graphs, a difference in concentration of a factor of 4 to 5 can be observed
800 metres from the release point while the concentrations are practically equal in
the first 100 metres.
In addition, tests with the same type of atmospheric conditions were conducted
during the test campaign. When these tests are compared to each other, a similar
plume dilution is observed for all the tests. For example, in the figure below, we
have reproduced two graphs showing the dispersion conditions of tests 2 and 4, in
which the release conditions were similar (similar flow rate, different cross-
sections) and tests 8b and 11, in which the release conditions were different.
(;.> ) 1 (4.; )
(;.> ) (; )
"
"
" "
figure 25: Progression of concentrations on release axis as a function of distance

for similar weather conditions.
The progression of the ammonia concentrations as a function of the distance to

the release point is almost identical for the tests. It is important to note that, in
reality, the atmospheric flow is difficult to reproduce, and, as such, it is unlikely that
two tests with the same source term would give identical results in terms of
concentrations leeward from a release.
6.2.3 Influence of obstacle placed in near field
During the test campaign, six releases impacting a surface in the near field were
used, i.e. tests 5, 6, 7, 8, 8b and 9. Note that tests 5 and 6 were horizontal
releases against a wall positioned 3 metres and 1 metre from the release point,
respectively. The 4 other tests related to vertical releases falling and impacting the
ground. For tests 7 and 8, there was no retention dike on the ground. For tests 8b
and 9, a retention dike was installed. The last two tests will be studied in the
section below entitled "Influence of retention dike".
Note that, during tests 5 to 9, the atmospheric stability conditions determined

experimentally were practically all different, ranging from class A/B to class D.
In the case of the two horizontal releases impacting a wall, the release conditions
are similar to the conditions for test 4 (open-field release). Therefore, a
comparison of these three tests was conducted. The progression of the ammonia
concentration values leeward from the release is given in the figure below for
these three tests.

Page 91 sur 130
100000
4 (C/D3)
5 (A/B3,5)
10000 6 (D5)
concentration (ppm)
1000
100
10
1
0 100 200 300 400 500 600 700 800 900
distance (m)
figure 26: Progression of concentrations as a function of distance

for tests 4, 5 and 6
It is noted that the ammonia concentration values recorded for tests 5 and 6 after
the wall 20 and 50 metres from the release point are very similar, whether the wall
is located 1 or 3 metres from the release point.
However, when the release is produced in an open field, the ammonia

concentration values recorded by the sensors are greater by a factor of 2 up to 50
metres from the release point. After that, it is more difficult to compare the values,
since the influence of the weather conditions is predominant on the dispersion of
the plume.
Two phenomena can explain the decrease in the concentration in the near field
(less than 50 metres from the release site in this case) in the changeover from test
4 to tests 5 and 6.
Firstly, there was an increase in the level of mechanical turbulence, because the
plume is obliged to deviate the obstacle formed by the wall. In this case, the air
flows via the sides and the top of the wall.

Page 92 sur 130
Secondly, the impact of the ammonia jet consisting of fine droplets on the front of
the wall results in the formation of a very cold liquid ammonia film which favours
the trickling of the droplets and the formation of a pool at the base of the wall. With
this respect, the thermocouples installed on the release slab for tests 5 and 6 give:
temperature values between -40°C and -60°C for the thermocouples

installed on the wall ;
and temperature values of the order of -60°C throughout the release for the
majority of those installed at ground level.
In this way, one of the effects of the presence of a wall in the near field of a
liquefied gas release is to favour the recondensation of part of the ammonia
released and, through the formation of a liquid ammonia film, induce the trickling of
the liquid phase. This immediately results in a decrease in the ammonia
concentrations downstream from the wall. It is important to note that, for these two
tests, there was no retention dike. In this way, part of the liquid ammonia was
dispersed over a surface area, the size of which is difficult to estimate. This liquid
ammonia was then vaporised. From this point of view, no calculations were made.
In section 6.1, we saw that the ammonia cloud behaved like a heavy gas up to a
distance of more 200 metres from the release point when it was released in liquid
phase under our test conditions, particularly in the form of an open-field jet. On
analysing the results, it has emerged that this heavy gas type behaviour is less
pronounced for releases impacting a wall in the near field.
Typically, when the release was carried out in an open field, two sensors installed
at two different heights, on the same measurement mast located 50 m from the
release point, give concentration values at a height of 3 m that are 5 to 10 times
lower than those measured at a height of 1 m.
After an impact against a wall, the vertical concentration gradient is such that the
concentration values measured at a height of 3 m are only approximately two
times lower than those measured at a height of 1 m.
In the case of the two vertical releases impacting the ground with no
retention dike (tests 7 and 8), their release conditions were similar to those
for tests 4, 5 and 6. The graph below shows the progression of the
ammonia concentrations leeward from the releases for tests 4, 5, 6, 7
and 8.

Page 93 sur 130
100000
4 (C/D3)
5 (A/B3,5)
10000 6 (D5)
7 (C/D6)
concentration (ppm)
8 (C/D5)
1000
100
10
1
0 100 200 300 400 500 600 700 800 900
distance (m)
figure 27: Progression of concentrations as a function of distance

for tests 4, 5, 6, 7 and 8.
In the above figure, test 8 is not shown after 50 m due to a change in the wind
direction during the test such that the measurements made could not indicate the
mean concentration on the shifted cloud axis. However, the ammonia
concentration level appears to be of the same order of magnitude up to 50 m from
the release point for the four tests in which an obstacle was encountered.
The release conditions for tests 7 and 8 are such that the mass flow rate for these
tests (3.9 kg/s) is slightly lower with reference to tests 4, 5 and 6 (4.2 kg/s). As for
tests 5 and 6, roughly the same difference in concentration is found between tests
7 and 8 and test 4, i.e. a decrease of approximately 50% up to 50 m from the
release point. Beyond this point, we do not have sufficient information to be able to
compare with test 8.
The release conditions for tests 7 and 8 are such that the mass flow rate for these
tests (3.9 kg/s) is slightly lower with reference to tests 4, 5 and 6 (4.2 kg/s). As for
tests 5 and 6, roughly the same difference in concentration is found between tests
7 and 8 and test 4, i.e. a decrease of approximately 50% up to 50 m from the
release point. Beyond this point, we do not have sufficient information to be able to
compare with test 8.
In the case of tests 4, 6 and 7, the stability of the atmosphere was still considered
as neutral (Pasquill classes C or D). However, the wind velocity is different for
these three tests. The mean velocity during these three tests was calculated
between 3 and 6 m/s, with velocity ranges between 1 and 10 m/s. The difference
in the progression of the ammonia concentrations leeward from releases 6 and 7
can be explained in two ways:

Page 94 sur 130
by improved plume dilution due to more favourable atmospheric conditions;
by a larger quantity of liquid ammonia trapped around the release point.
In addition, temperatures of the order of -60°C were recorded on the ground for
each test. It was not possible to obtain an accurate estimation of the size of this
pool and even less so of its thickness (thin in principle). During the release, the
pool was evaporated. Once the release had been completed, some sensors still
detected temperatures below -50°C 4 minutes after the end of the release, which
underlines the size of the pool on the ground, without having the possibility to
quantify it in the absence of a retention dike.
In sum, orienting a release against a wall or against the ground appears to induce
a significant reduction in ammonia concentrations leeward from the release
(approximately 50% in our release configurations) in both the near field and the far
field.
6.2.4 Influence of retention dike
Of the fifteen tests conducted during this campaign, a retention dike was set up on
the release slab for tests 8b and 9. The release configuration is identical for these
two tests, i.e. the vertical releases impacting the ground from a height of 1 m. Only
the size of the retention dike is different in these two tests. For test 8b, the surface
area of the retention dike was 100 m2 and the height 5 cm, i.e. a retention volume
of 5 m3. For test 9, the surface area of the retention dike was reduced to 4 m2 and
the height was 1 m, i.e. a retention volume of 4 m3.
In the previous section relating to releases impacting an obstacle (tests 5, 6, 7 and

8), we saw that, as a general rule, at the end of each impacting release, the
concrete slab was extensively wetted with liquid ammonia. In particular, for
releases directly impacting the ground, the entire slab was wetted. However, the
quantity of liquid ammonia on the ground was not estimated. However, tests 8b
and 9 can be used to estimate the quantity of liquid ammonia trapped on the
ground.
In test 8b, 2120 kg of ammonia was released throughout the test. At the end of this
test, a height of approximately 1.5 to 2 cm of liquid ammonia could be observed in
the retention dike, i.e. a volume of approximately 1.5 to 2 m3. The temperature of
the liquid ammonia at the end of the release was approximately -60°C as for the
previous tests. In this way, it is possible to estimate that the quantity of liquid
ammonia present on the release slab was between 1050 and 1400 kg, which
represents between 50% and 65% of the total mass quantity released.
In test 9, 2370 kg of ammonia was released throughout the test. At the end of test
9, a height of approximately 55 cm of liquid could be observed in the retention
dike, i.e. a volume of approximately 2.2 m3. The temperature of the liquid ammonia
at the end of the release was approximately -42°C. It is important to note that,
during this test, minor leakages were observed at the bottom of the retention dike.
In addition, part of the liquid ammonia was projected outside the retention dike
under the effect of the velocity of the jet on the ammonia sheet. Liquid ammonia
projections outside the retention tank were not accurately quantified.
Page 95 sur 130
3
photo 4: Vertical release in a 4m retention dike
In this way, it is possible to estimate that, at the end of the release, the quantity of
liquid ammonia present in the retention tank was greater than 1520 kg, which
represents approximately 65% of the total mass quantity released.
In addition, the liquid ammonia in a retention dike is not evaporated quickly. In fact,
ammonia needs a significant quantity of heat to vaporise. It obtains this energy
from external sources (temperature of ambient air, wind velocity, solar radiation,
rain, etc.), and by lowering its temperature. In this way, during test 9, the
temperature of the liquid ammonia fell from -40°C at the end of the release to -
60°C one hour afterwards (temperature reached much more quickly for the other
tests).
It is difficult to compare the progression of the ammonia concentrations leeward

from releases impacting into a retention dike with impacting releases with no
retention dike, because:
The measurements made during test 8 (change of wind direction) are

insufficient to compare with test 8b;
in test 9, a decline in the ammonia concentration values over time is

observed;
and the weather conditions are different for the different tests.
The figure below represents the progression of the ammonia concentration

measured in the near field 20 m from the release point over time for test 9.
%
35000
30000
25000 In this figure, a mean of the ammonia

calculation values was calculated on the
Concentration (ppm)
20000
15000
basis of the measurements made every
200 seconds.
10000
5000
0
100 300 500 700 900 1100
temps (s)
figure 28

Page 96 sur 130
The decline in the ammonia concentrations over time for test 9 can be interpreted
as follows. At the start of the release, the liquid ammonia impacting the retention
dike is vaporised very quickly due to the energy supplied by the concrete
contained in the retention tank. Over time, the temperature falls progressively,
enabling the retention of liquid ammonia and the decrease in the quantity of
ammonia vaporised over time.
In this way, a retention tank may represent an effective means to reduce the
concentration level in the environment following a release, provided it is suitably
positioned and designed. In particular, it is important to note that a retention dike
installed under a leakage point may prove to be ineffective if the jet does not
impact it directly or indirectly following an impact on an obstacle.
In addition, the retention dike must be well-maintained. In particular, the presence

of water must be prevented as the water would provide additional energy to
vaporise a larger quantity of liquid ammonia.
6.2.5 Influence of water curtain produced using peacock tail hoses
During this test campaign, two releases used peacock tail hoses installed on the
release path. For these two tests, referenced 10 and 11, two 70 mm diameter
peacock tail hoses were located 25 m and 60 m from the release point,
respectively, as described in chapter 4. The purpose of these two tests was to try
to evaluate the influence of the presence of a water curtain on atmospheric
dispersion.
To study this influence for each of the two tests, we produced an open-field
release for approximately 10 minutes. In order to be able to compare the ammonia
concentration value results under atmospheric dispersion conditions which were
as similar as possible before and after the start-up of the peacock tail hoses, we
started up the peacock tail hose only 5 minutes after the start of the release.
On analysing the results, it has emerged that significant fluctuations in the

atmospheric conditions were recorded during the tests, particularly the wind
direction, which did not remain constant. In this way, during the tests, the ammonia
plume sometimes passed by the water curtain.
In addition, under our release conditions during test 10, the ammonia jet passed
through the water curtain located at 20 cm, rendering this type of curtain practically
ineffective. At times, the jet was pointing towards the peacock tail in which the
water was output at its highest velocity. Under these conditions, mechanical
mixing of the plume could be observed as the jet was deviated from its path.
However, during this test, the cloud generally passed between the two peacock
tails, meaning that the concentration at the centre of the plume was not affected
very much.
As an illustration of the previous pages, the graph in the figure below shows the
progression of the mean ammonia concentrations leeward from release 11 before
and after the start-up of the water curtain.

Page 97 sur 130
10000
1 2
2
concentration (ppm) 1000
1000
100
10
1
0 100 200 300 400 500 600 700 800 900
distance (m)
figure 29: Progression of ammonia concentrations leeward from release 11

before and after starting up the peacock tails
In the above figure, it is noted that there are no major difference before and after
the peacock tails were started up. However, it is important to note that these
concentration values are a factor in the direction adopted by the cloud.
In addition, it is important to note that, during test 11, the wind was relatively
strong. The mean wind velocity was 5 m/s, with maximum velocities of 9 m/s
recorded. During this test, the water curtain itself was flattened under the effect of
the wind.
In this way, in this test campaign, the presence of a water curtain produced with
peacock tail hoses did not significantly decrease the concentrations before and
after the water curtain. This may be explained by the fact that an ammonia jet can
pass through a water curtain if the momentum is sufficient, and by changing
weather conditions during the tests. A release of the same type as those used in
these test campaigns, but with a much lower momentum when passing through
the water curtain may result in lower concentration values in the environment. This
point is worth studying in more detail.
6.2.6 Experiment / modelling comparisons
In this section, the experimental results obtained during the test campaign were
compared to calculation results obtained using a software program used to
evaluate the effects of the accidental atmospheric release of toxic or flammable
substances. This type of software program particularly makes it possible to
calculate the dispersion of these substances in the atmosphere.

Page 98 sur 130
Of the software available, PHAST software version 4.2 was used. This model is of
the integral type, i.e. it can be used to solve conservation equations based on
simplifying hypotheses. With this respect, the limitations inherent to this software
are particularly as follows:
the site must be flat, of uniform roughness and not obstructed with
obstacles ;
the wind velocity must be at least equal to 1 m/s to obtain plausible results ;
the weather conditions must be invariable throughout the emission and

dispersions time ;
the release source term must be constant over time.
To make the dispersion calculations for a substance in the atmosphere, it is

necessary to define the source term of the release and the weather conditions in
which it is dispersed in the model beforehand.
In chapter 5, we saw that the determination of the source term of a release with a
model could give different results to that measured in reality. In order to do away
with the source term problem (the subject of this section being the study of
dispersion in the atmosphere), the comparisons were made as follows. Firstly, the
parameters measured in the tank for a given release were entered in the software
and, secondly, the other parameters such as the regular and irregular pressure
drops were adjusted so as to calculate a mass flow rate identical to the flow rate
measured experimentally.
Then, the meteorological data measured during the test were entered on the basis
of the experimental data (ambient temperature, relative humidity, etc.). The mean
wind velocity during the test was selected for the calculations. With respect to the
atmospheric stability, we selected those which were calculated using the two
methods described in chapter 5. If the two methods gave different stability classes,
both classes were used for the calculations.
With these conditions entered in the software, firstly, we calculated the dispersion
of ammonia in the atmosphere for open-field releases, i.e. tests 1, 2, 2b, 4, 10, 11
and 12.

Page 99 sur 130
The 4 graphs given in the figure below show a comparison between the
experimental results (blue curve) and the results of dispersion calculations in the
atmosphere.
100000 100000
4 (C/D3) exp 10 (A/B3) exp
4 (C3) 10 (A3)
10000 4 (D3) 10000 10 (B3)
concentration (ppm)
concentration (ppm)
1000 1000
100 100
10 10
1 1
0 100 200 300 400 500 600 700 800 900 0 100 200 300 400 500 600 700 800 900
distance (m) distance (m)
100000 100000
2 (C/D3) exp 2b (A/B4) exp
2 (C3) 2b (A4)
10000 2 (D3) 10000 2b (B4)
concentration (ppm)
concentration (ppm)
1000 1000
100 100
10 10
1 1
0 100 200 300 400 500 600 700 800 900 0 100 200 300 400 500 600 700 800 900
figure 30: Comparisons between calculation results

and experimental results for tests 4, 10, 2 and 2b
In these graphs, it can be seen that the software calculates ammonia

concentration values of the same order of magnitude as the values measured
experimentally. Only test 2b, among the open-field releases, resulted in the
calculation of two concentration values 100 m and 200 m from the release point
which were significantly lower than the experimental values.
However, as a general rule, the modelling of the dispersion in the atmosphere of a

liquefied gas release such as ammonia gives comparable orders of magnitude to
that observed for large-scale releases, provided that:
the source term is determined correctly in the software ;
the weather conditions are defined correctly, given that they may vary
rapidly over time (both the wind velocity and the atmospheric stability ;
the release is carried out horizontally, in the direction of the wind, on a flat
site and with no obstacles.
Secondly, we made calculations with impacting releases. The figure below shows
the progression of the experimental concentrations (blue curve) and the
concentrations calculated with the model for releases against a wall (tests 5 and 6)
and on the ground (tests 7 and 8b).

Page 100 sur 130
100000 100000
5 (A/B3,5) exp 6 (D5) exp
5 (A3,5)
6 (D5)
10000 5 (B3,5) 10000
concentration (ppm)
concentration (ppm)
1000 1000
100 100
10 10
1 1
0 100 200 300 400 500 600 700 800 900 0 100 200 300 400 500 600 700 800 900

100000 100000
7 (C/D6) exp
8b (B/C5) exp
7 (C6)
7 (D6) 8b (B5)
10000 10000 8b (C5)
concentration (ppm)
concentration (ppm)
1000 1000
100 100
10 10
1 1
0 100 200 300 400 500 600 700 800 900 0 100 200 300 400 500 600 700 800 900
figure 31: Comparisons between calculation results

and experimental results for tests 5, 6, 7 and 8b
In this figure, it can be seen that, 20 m from the release point, the model
systematically calculates a concentration approximately two times greater than
that measured experimentally. Then, the progression of the concentrations
depends on the atmospheric conditions. For test 6, in which these conditions were
probably (D; 5 m/s) type, both curves (experimental and calculated) always show
the same difference during the progression of the cloud leeward from the release.
The mixing of the cloud following the impact of the release, on one hand, and the
retention of ammonia at the release point, on the other, may explain this
difference.
At the present time, integral type models do not account for releases impacting
solid surfaces. Differences of the order of 50% may be observed between large-
scale test results and calculations made with this type of model. Even greater
differences may be observed if the configuration of the impact is complex or if a
retention dike is suitably positioned with respect to the release.

Page 101 sur 130
7. CONCLUSION AND PROSPECTS
Within the scope of this research programme, INERIS essentially conducted a test
campaign in order to improve knowledge on the atmospheric dispersion of
ammonia, particularly in open or obstructed environments.
Of all the results obtained from this test campaign, the following points are worth
noting:
15 tests with different configurations were conducted over the period

between December 1996 and April 1997 ;
the quantity of ammonia released from the liquid phase was, depending on
the tests, between 1400 kg and 3500 kg per test over a time of between 7
and 14 minutes, i.e. flow rates between 2 and 4.5 kg/s ;
in the case of the thermodynamic progression of ammonia in the tanks,

temperature and pressure measurements were made during the releases.
These measurements and the interpretations are given in chapter 5 ;
in the case of the thermodynamic progression of ammonia in the pipes, the

measurements made at the same point demonstrated that the value of the
pressure is similar to the saturating vapour pressure corresponding to the
temperature measured at this point ;
approximately 200 sensors, to measure the ammonia concentration and the

temperature, were set up for each test ;
the ammonia cloud formed behaves like a heavy gas, and no rise in the
cloud is observed ;
the temperature in the ammonia jet can typically fall to -70°C ;
solid obstacles (wall or ground) placed in a two-phase ammonia jet at a

distance of a few metres from the release point have a major influence on
the concentration values measured after the obstacle. In this test campaign,
approximately two times lower concentrations were measured after the
obstacle with respect to an open-field release ;
If the release is pointed towards the retention dike, the dike can retrieve a
large quantity of ammonia in liquid form. For two tests, quantities greater
than 50% of the total mass released were collected in liquid form at a
temperature of approximately -60°C ;
The liquid ammonia sheet formed in this way is not evaporated rapidly ;
water curtains produced with peacock tail hoses and located on the plume
path had little effect on ammonia dispersion under the test conditions ;

Page 103 sur 130
in addition, the results obtained demonstrate that the integral type
dispersion models currently available can give correct orders of magnitude
of concentrations leeward from a free release. In the case of an impacting
release, the results obtained are only correct if the models receive
information beforehand on the quantity released into the atmosphere (in
vapour or aerosol form) after the impact.
The results obtained during these tests led us to develop a new programme
intended to study the "near-field dispersion of a liquefied gas in the presence of
obstacles". The purpose of this programme, initiated in 1999 at INERIS, is
particularly to create a mathematical model used to account for the interaction of a
liquefied gas release against an obstacle. In the long-term, this model could be
incorporated in integral type software.

Page 104 sur 130
8. BIBLIOGRAPHY
Air Liquide, Encyclopédie des gaz, 1980.
AUBURTIN, PICHARD, SCHORSCH, Toxicité aiguë de l'ammoniac, rapport

INERIS, May 1999.
BARA, DUSSERRE, The use of water curtains to protect firemen in case of heavy
gas dispersion, Journal of Loss prevention in the process industries, vol. 10, No. 3,
pp 179-180, 1997
B.I.T. (Bureau International du Travail), La Maîtrise des Risques d'Accident

Majeur, guide pratique, Geneva, 1993.
BLANKEN, Behaviour of ammonia in the event of a spillage, Ammonia safety

plant, 22, 1980, quoted by Wheatley (April 1987).
BIGOT, Modélisation de la perte de confinement d'un gaz liquefiés, Ecole des

Mines de Saint-Etienne, Euroforum, June 1996.
BOUET, Essais de dispersion atmosphérique, INERIS, Rapport annuel, 1996.
BOUET, Essais de dispersion atmosphérique, INERIS, Rapport annuel, 1997.
BUCKLEY and HUSA, Combustion Properties of Ammonia, Chemical

Engineering Progress vol. 58, No. 2, p. 81-84, 1962.
BRETHERIK, Handbook of Reactive Chemical Hazards, ed Butterworths, 4th ed,

1990.
CCPS, Guidelines for consequence analysis of chemical releases. AIChE, CCPS,

0003028, p-741, 1999.
ABIVEN, CHAINEAUX, Caractéristiques d'inflammabilité et d'explosibilité de

l'ammoniac, Rapport INERIS, EXP-FAB/DG F42 e/349 91.78.1120, 1991.
CHAINEAUX, EMERGE project (Extended Modelling and Experimental into Gas

Explosion), 2nd progress report, INERIS, January 1995.
DELHAYE, Les écoulements diphasiques gaz, liquide en conduite, Entropie, n°

99, pp 3-25, 1981, quoted by May (1996).
EMFA (European Fertilizer Manufacturer's Association, Zurich), IFA (International

Fertilizer Industry Association Limited, Paris), Hazardous properties of ammonia,
1990, quoted by SNIE (1991).
FAUSKE, Two phase and one component critical flow. Symp. on two phase flow,
Ed. P.M.C. Lacey, U. of Exeter, vol 1, pp101-114, 1965.

Page 105 sur 130
FAUSKE, Flashing flows or some practical guidelines for emergency releases,
Plant operation progress, vol. 4, No. 3, p. 132, 1985.
FLETCHER, Flashing flow through orifices and pipes, CEP, March 1984.
FULLERINGER, Evaluation de la sûreté des installations utilisant de l'ammoniac,

rapport IPSN, DES n° 252, 1995.
GIFFORD, Diffusion in the Diabatic Surface Layer, J. Geophy. Res., vol. 67, no. 8,
pp. 3207-3212, 1962.
GIFFORD, An outline of theories of diffusion in the lower layers of the atmosphere.

Meteorology and Atomic Energy, US Atomic Commission, D.H. Slade Edition,
1968.
GOLDER, Relations among stability parameters in the surface layer, Boundary

layer meteorology, 3, pp 47-58, 1972.
GOLDWIRE et al., Desert Tortoise series data report – 1983 pressurized ammonia
spills, UCID 20652, December 1985.
GOLDWIRE, Large-Scale Ammonia Spill Tests. A series of ammonia spill tests

found that improved simple models are needed for predicting the consequences of
accidental releases. Chemical Engineering Progress, pp 35-41, April 1986.
As above: article containing comments, published in ammonia plant safety, Vol.

26, pp 203-211, 1986.
HANNA, DAVID, CHANG, Evaluation of fourteen hazardous gas models with

ammonia and hydrogen fluoride field data, Journal of hazardous materials, Vol. 26,
pp 127-158, 1991.
HANNA, CHANG, STRIMAITIS, Hazardous gas model evaluation with field

observations, Atmospheric Environment, Vol. 27A, No. 15, pp 2265-2285, 1993.
HODIN, Modélisation du debit à la brèche et du jet géneré par une fuite

d'ammoniac en phase liquide ; état des connaissances et modélisation des rejets
diphasiques, EdF, Centre Lyonnais d'Ingénierie, Note d'étude, 1995.
IANELLO, ROTHE, WALLIS, Aerosol Research Program: improved source term

definition for modeling the ambient impact of accidental release of hazardous
liquids, Paper presented at the 6th International Symposium on Loss Prevention
and Safety Promotion on the Process Industries, Oslo, Norway, 1989.
INRS, Réaction chimiques Dangereuses, ammoniac, note 1024-84-76, 1976.
INRS, Les Mélanges Explosifs, édition de 1989, A. CLEUET.
INRS, Fiche Toxicologique n° 16, gaz ammoniac et solutions aqueuses, édition

révisée, 1992.

Page 106 sur 130
KIRK-OTHMER, Encyclopedia of Chemical Technology, 3rd ed., vol. 2, p. 470-516,
1978.
KOOPMAN, McRAE, GOLDWIRE, ERMAK, KANSA, Results of Recent Large-

Scale NH3 and N2O4 Dispersion Experiments, S. Hartwig (ed.), Heavy Gas and
Risk Assessment – III, p 137-156, 1986.
KRAMER, Minimum ignition energy of ammonia-air mixture, Gas discharges,

Their Appl. Proc. Int. Conf. 8th, 1985.
LANGARD, ANSTETT, MAVROTHALASSITIS, La prise en compte des

phénomènes biphasiques dans l'analyse des risques, Rapport INERIS, October
1995.
LANTZY, Vapor cloud source modeling workshop, Chairman’s report,

Plant/Operations Progress, Vol. 11, No. 1, pp 41-45, 1992.
LEUNG, Two-phase flow discharge in nozzles and pipes, a unified approach, J.

Loss Prevention Ind., vol.3, pp 27-32, January 1990.
MAY, Ecoulements diphasiques lors de la vidange de gaz liquéfiés initialement

sous saturés, validation par l’eau et le CFC11, Thèse de l’école des Mines de
Saint-Etienne et de Paris, 1996.
MEDARD, Les explosifs occasionnels, vol. 1, p. 210-211-213-261, 1979.
NFPA (National Fire Protection Association), No. 325 M, 1990.
NFPA (National Fire Protection Association), Fire Protection Guide to Hazardous

Materials, 11th edition, 1994.
NIOSH (National Institute for Occupational Safety and Health), pocket guide to
chemical hazards, 1987.
NIELSEN and OTT, Fladis Field Experiments, Final Report, RISØ-R-898(EN),

July 1996.
NYRÉN, WINTER, Two-phase discharge of liquefied gases through pipes, field

experiments with ammonia and theoretical model, 4th international symposium on
loss prevention and safety promotion in the process industries, No. 33, vol. 1, p E
1-E 11, 1983
NYRÉN et al., Discharge trials with liquefied ammonia at Simpra, Landskrona,

Sweden, Swedish National Defence Research Institute, FOA report E40009, 1983,
quoted by Wheatley (April 1987).
PASQUILL, The Estimation of the Dispersion of Windborne Material, Meteorol.

Mag., vol. 90, pp. 33-49, 1961.
PASQUILL, Atmospheric diffusion, 2nd edition, Ellis Howood, London, 1974.

Page 107 sur 130
REED, Containment of leaks from vessels containing liquefied gases with
particular reference to ammonia, 1st international symposium on loss prevention
and safety promotion in the process industries, The Hague, pp. 191-195, Elsevier,
1974, quoted by Wheatley (April 1987).
RESPLANDY, Etude des prescriptions à préconiser pour la sécurité du voisinage

des dépôts d’ammoniac liquide. Compte-rendu des expériences effectuées le 28
juin 1967. Comité Consultatif des Etablissements Classés. September 1967.
RESPLANDY, Etude des mesures à préconiser pour la sécurité du voisinage des

stockages d’ammoniac liquéfié. Comité Consultatif des Etablissements Classés.
December 1968.
RESPLANDY, Etude expérimentale des propriétés de l’ammoniac conditionnant

les mesures à prendre pour la sécurité du voisinage des stockages industriels.
Chimie et Industrie - Génie chimique, Vol. 102, n°6, October 1969.
RIOU, Comment estimer la dispersion des gaz lourds?, EdF bulletin de la direction
des études et recherches, série A, Nucléaire, hydraulique, thermique, 1989.
SAX’S Dangerous properties of industrial materials, 9th edition, 1996.
S.E.I. (Service de l’Environnement Industriel, du Ministère de l’Aménagement du

Territoire et de l'PEnvironnement), fiches techniques, courbes de toxicité par
inhalation, 1998.
S.N.I.E. (Syndicat National de l’Industrie des engrais), Aide-mémoire sur les

risques présentés par l’ammoniac, 1991.
TISSOT, PICHARD, Seuils de toxicité aiguë, Ammoniac, 2003. (see

www.ineris.fr).
TURNER, Workbook of Atmospheric Dispersion Estimates, US, Department of

Health Education and Welfare, NTIS, Cincinnati, Ohio, 1970.
U.I.C. (Union des Industries Chimiques), Emission à la brèche d’une capacité,

débit en phase gaz, débit en phase liquide, formation et vaporisation des flaques,
Cahiers de sécurité n°11, September 1987.
U.I.C. (Union des Industries Chimiques), Dispersion atmosphérique, Groupe de

travail Prévention des risques majeurs, Cahier de Sécurité n°6, June 1995.
ULMANN, Encyclopedia of lndustrial Chemistry, 5th edition, vol. A2 p. 143-242,

1985.
VAN DEN HAKKER, SNOEY, SPOELSTRA, Discharges of pressurized liquefied

gases through apertures and pipes, 4th international symposium on loss prevention
and safety promotion in the process industries, No. 33, vol. 1, pp E23-E35, 1983.

Page 108 sur 130
WEISS, Hazardous Chemicals Data Book, NDC (Noyes Data Corporation),
2nd edition, 1985.
WHEATLEY, A theoretical study of NH concentrations in moist air arising from

accidental releases of liquefied, using the computer code TRAUMA, Health and
Safety Executive; SRD/HSE/R 393, February 1987.
WHEATLEY, Discharge of liquid ammonia to moist atmospheres - Survey of

experimental data and model for estimating initial conditions for dispersion
calculations, Health and Safety Executive; SRD/HSE/R 410, April 1987.
WHO (World Health Organisation), IPCS, Environmental Health Criteria 54,

Ammonia, 1986.
WOODWARD, Expansion zone modeling of two phase and gas discharges, J. of

Hazardous materials, 33, pp 307-3 18, 1993.
WOODWARD, PAPADOURAKIS, Modeling of droplet entrainment and

evaporation in dispersing jet, in CCPS (1971), pp 147-167.
YAN, Modélisation de l’autoévaporation en écoulements subcritiques et critiques,

thesis, U. Catholique de Louvain, Belgium, 1991, quoted by May (1996).
ZAKAZNOV, KURSHEVA, FEDINA, Determination of Normal Flame Velocity and

Critical Diameter of Flame Extinction in Ammonia-Air Mixture, Combustion,
Explosion and Shock Waves, No. 6, vol. 14, p. 710-713, 1978.

Page 109 sur 130
9. LISTE DES ANNEXES
Repère Désignation précise N°pages
Appendix 1 Thermocouple indication 7
Appendix 2 Main direct measurement techniques 6

Page 111 sur 130
APPENDIX 1
THERMOCOUPLE INDICATION

Page 113 sur 130
The thermocouples referred to in this appendix are those which were installed:
in the ammonia tank,
on the release system,
on the release slab,
and, if applicable, on the wall or retention dike.
A description of the location of the thermocouples is given in the table below.
For all the tests, the following thermocouples were installed:
Thermocouple
Caption
location
in tank 1450 mm
Tank 1450 mm
from base
in tank 1100 mm
Tank 1100 mm
from base
in tank 750 mm from

Tank 750 mm
base
in tank 400 mm from

Tank 400 mm
base
in tank 50 mm from
Tank 50 mm
base
at end of release
OUTLET – th6
system
between two release

th 7
system valves
at inlet of release
th 8
system

Page 115 sur 130
Test 4:
For this test, four additional thermocouples were installed on the ground. Two in
the direction of the release 3 and 5 metres from the orifice and two on either side
at 3 metres.
Thermocouple
Caption
positions
on ground, in direction of
release (27), 3 metres ground dir 27 at 3m
from orifice
from orifice
from orifice
from orifice
Tests 5 and 6:
With respect to these tests, a 3 metres wide, 3 metres high and 1 metre thick wall
was constructed at a distance from the orifice of 3 metres and 1 metre,
respectively. For both configurations, 19 thermocouples were installed, 4 on the
wall and 15 on the ground.
On the wall, the 4 thermocouples referenced 11 to 14 were installed as described

in figure 32 below.
figure 32

Page 116 sur 130
For test No. 5, the thermocouples were installed on the ground as described in
figure 33.
figure 33

Page 117 sur 130
The caption used for each of these thermocouples is given in the table below.
Caption
th11 – wall 1 m th21 – ground

centre centre
th12 – wall 1 m th22 – ground

right centre
th23 – ground
th13 – wall 1 m left
centre
th 14 – wall 2 m
th24 – ground centre
centre
th16 – ground 60° th25 – ground 30°

right left

right left

right left

right left

right left
th30 – ground 60°

left

Page 118 sur 130
For test 6, the wall was located one metre from the release system orifice.
Therefore, the thermocouples were installed on the ground as shown in figure 34
below.
figure 34
Tests 7 and 8: These tests were instrumented as follows (see figure 35):
figure 35

Page 119 sur 130
Test 9:
For this test, the end of the release system impacted into a 2m x 2m x 1m
retention dike. The thermocouples were installed along the direction 20, as
described in the figure below.
$ $
-
height /
thermocouples
ground (mm)
0 th11 th27 th28 th29
30 th12
60 th13
100 th14
150 th15
200 th16
300 th17
400 th18
500 th19 th26
600 th20
800 th21
1000 th22 th23 th24 th25

Page 120 sur 130
Tests 10, 11 and 12:
For these three tests, 6 thermocouples were installed on the ground along the axis
of the release system. From the vertical line from the release point, there was one
every metre. We indicated them as follows:
#
#
#
#
#
#
figure 36
Tests 2b and 8b:
For these two tests, in addition to the 6 thermocouples described above, we

installed 12 on two masts, and 1 to 50 cm above the release orifice. The latter is
indicated in the caption: "point 0 – h=1.5m". The other thermocouples are
indicated as follows on masts 1 and 2, located 2 and 4 metres from point 0 in the
direction 7, respectively.
! !
! !
! 0 ! 0
! !
! 0 ! 0
! 0 ! 0
figure 37

Page 121 sur 130
APPENDIX 2
MAIN DIRECT MEASUREMENT TECHNIQUES

Page 123 sur 130
The purpose of appendix 2 is to give a brief presentation of the main ammonia
concentration measurement techniques. This chapter is particularly based on the
work by A. Accorsi's team [1,3] and the documentation supplied by the
manufacturers, particularly [2,4].
1 ELECTROCHEMICAL CELL DEVICES
This section is essentially based on document [1].
Device to measure gas concentrations in the air using electrochemical cells

emerged approximately twenty years ago.
Measurement principle
Each electrochemical cell is relatively specific for the detection of a single gas. The
gas to be measured passes through a semi-permeable membrane, i.e. permeable
for gases but impermeable for liquids. The gas comes into contact with a sensitive
electrode coated in a catalyst in the presence of which, it is oxidised, while a
counter-electrode reduces the oxygen in the air. These oxidation-reduction
phenomena are accompanied by an ion flow through the electrolyte and an
electron flow to the outside. (See Figure 2.1)
The reaction to the sensitive electrode is (according to MARTEC documentation):
2 NH3 --------> N2 + 6 e- + 6 H+
The reaction to the counter-electrode is (according to the CECA report):
2 e- + 0.5 O2 + 2 H+ -------> H2O
Overall:
2 NH3 + 3/2 O2 -------> N2 + 3 H2O

Page 125 sur 130
To avoid capacitive limit layer phenomena at the electrode/electrolyte interface, a
reference electrode maintains the potential of the sensitive electrode at a fixed
value, which makes it possible to keep a signal stable over time.
Figure 2.1: Functional description of an electrochemical cell

(diagram p 44 fig. 5 in CECA report [1])
Limitations
Selectivity is obtained, firstly, with the value of the reference potential and,
secondly, with the oxidation catalyst. In practice, absolute selectivity does not
exist. For example, the table below, given by Dräger, shows some compounds
inducing interference on PAC II NH3:
Deviation in ppm
Test gas
Gas Symbol NH3 from
concentration (V/V)
measurement value
Chlorine Cl2 10 ppm <1
Cyanhydric acid HCN 25 ppm <3
-5 < and < 0 (default
Carbon dioxide CO2 1%
signal)
Carbon monoxide CO 115 ppm <3
Methane CH4 30% <3
Methanol CH3OH 170 ppm < 35
Sulphur anhydride SO2 20 ppm <1
Hydrogen sulphide H2S 20 ppm < 75
Nitrogen dioxide NO2 20 ppm <3
Hydrogen H2 1.53% < 115
Table 2.1: Compounds inducing interference on PAC II NH3, based on Dräger'
s technical
documentation for PAC II NH3
In this way, it is difficult to know exactly what is being measured and, in particular,
how accurate the measurement is, in the case of an environment in which several
gases are present (for example, in industrial atmospheres).

Page 126 sur 130
The measurement range of the majority of electrochemical type devices, for
measuring ammonia concentrations, is from a few ppm to 300, 500 or 1000 ppm.
Cells capable of measuring around one thousand ppm are very rare. The limit of
detection is of the order of 20 to 30 ppm. The response time is in excess of 20 s
and may be up to one minute, and depends on the type of membrane used to
protect the sensitive electrode. The service life of this type of device is
approximately one to two years. The average price of an electrochemical cell
alone is approximately 1000 FF and varies from 500 to 7000 FF.
2 CATALYTIC OXIDATION DEVICES (PELLISTOR)

This section proposes a summary of the EEV [2] and INRS [3] technical
documents.
Principle
The principle consists of inducing the catalytic oxidation of the flammable gas, in
this case, ammonia, and measuring the heat produced.
In fact, the catalytic detector is equipped with two components: an active catalytic
detector and a non-active compensator component. Each component consists of a
platinum filament in a bead of heat-resistant material. For the detector component,
a catalytic mixture is applied to this bead, while for the compensator component,
the bead is treated such that catalytic oxidation cannot occur.
The two components are connects in a Wheatstone bridge as shown in figure 2.2
taken from the document [3].
Figure 2.2: Principle diagram of catalytic oxidation measurement device

(INRS diagram)

Page 127 sur 130
The direct current flowing through the beads brings the detector to a suitable
temperature for the gas to be detected. The detector filament works like a
thermometer: when the combustible gas is oxidised on the detector bead, the
resulting temperature rise increases filament resistance. This unbalances the
bridge, giving a signal that can be read, either as a gas percentage, or as a
percentage of the Lower Explosibility Limit (LEL), or that triggers an audio or visual
alarm.
The rate at which the gas will burn on the detector component depends on the gas
diffusion rate in the detector. The diffusion rate is controlled by enclosing the
components in a unit of precise dimensions.
In addition, EEV explains that there are two assembly arrangements. In the first,
the gases circulate at a given flow rate of 0.5 litre/minute. The second
arrangement is designed for simple gas diffusion in a unit. In both cases, a frit is
included and a paper filter is placed in the diffusion head to prevent dust from
penetrating into the cavity. Another possibility consists of replacing the frit with a
disk of tiny bronze or stainless steel beads (recommended for ammonia).
Limitations
The service life of the detector is limited. It is associated with the operating
temperature, and also with a loss of efficiency over time, which depends on
prolonged operation, exposure to excessively high concentrations. Finally, the
detector may be damaged ("poisoned") by some gases or vapours. To protect the
detector from the effects of some organic or silicone-based vapours, an
interchangeable activated charcoal or carbon filter may be used.
The bridge output signal is of the type represented in figure 2.3. It rises above the
LEL (Lower Explosibility Limit), up to the stoichiometric value which is
approximately 10% in air for methane (example represented below) and 21% for
ammonia. For higher concentrations, the signal declines in a linear fashion due to
the lack of oxidant.
It can be noted that, when the combustible gas concentration exceeds the
stoichiometric value, it is possible to use another type of detector based on the
thermal conductivity principle to measure high concentrations and thus eliminate
any doubt with respect to ambivalence. As such, the compensator components of
some catalytic oxidation detectors may be used. However, according to the
information supplied by EEV, the thermal conductivity of ammonia is very similar to
that of air, so much so that it may be impossible, or very difficult, to make an
accurate, detailed measurement.

Page 128 sur 130
Figure 2.3: Pellistor type sensor output signal for exposure to methane
(diagram from EEV brochure)
Concentrations greater than the stoichiometric value may be measured by diluting

with air according to a known dilution ratio, e.g. 10. In this way, 100% v/v ammonia
would be lowered to 10% and the detection sensitivity would be of the order of
1% v/v.
3 SEMICONDUCTORS
The documents [3,4] were used to prepare this section.
Principle
A semiconductor is a material in which the electricity conducting properties are

intermediary between those of a conductor and those of an insulator. The
presence of a gas induces a variation in the electrical resistance of the
semiconductor components of the detector by means of physical adsorption
(reversible phenomenon) and catalytic combustion phenomena on the
semiconductor surface; the adsorbed gas is oxidised in the presence of surface
oxygen and the combustion product is desorbed. Modifying the oxygen content on
the surface induces a modification in the electron-hole equilibrium (the hole being
the adsorption site) in the semiconductor. This variation is directly linked with the
gas concentration and modifies the electrical equilibrium of the circuit in which the
detector is placed.
In the case of the Figaro detector, the semiconductor material used is tin dioxide
(SnO2-X) which is heated to a high temperature (e.g. 400°C).
Limitations
Semiconductor detectors are very temperature-sensitive: the conductibility

increases with the temperature. In addition, semiconductor operation is disturbed
by interfering parameters such as humidity, dust, grease of other preferentially
adsorbed gases.
Page 129 sur 130
Bibliographic References
[1] Accorsi. Rapport final CECA – Faisabilité de campagnes de mesures

d'atmosphères de travail par des appareils multigaz transportables. 1992
[2] EEV. Combustible Gas Detectors – Product Data
[3] A. Accorsi, J. C. Laforest, P. Huré. Détection de mélanges air-ammoniac.

INRS. Cahiers de notes documentaires n°127, p 217-222, 2e trimestre 1987.
[4] FIGARO – Products Catalog.

Page 130 sur 130

Work Study 20/12/2005 #10072

Uploaded by

Copyright:

Available Formats

Work Study 20/12/2005 #10072

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Work Study 20/12/2005 #10072

Uploaded by

Copyright:

Available Formats

WORK STUDY 20/12/2005

Large-scale atmospheric dispersion tests

Réf. : INERIS - DRA - DRAG – 2005 - 10072 – RBo – Ammonia

Large-scale atmospheric dispersion tests

INERIS – Accident Risks Division

Translation of the French report:

AMMONIAC : Essais de dispersion d’ammoniac à grande échelle –

Réf. : INERIS - DRA - DRAG – 2005 - 10072 – RBo – Ammonia

INERIS cannot be held liable if the information received was incomplete or

Any findings, recommendations, suggestions or equivalent that are recorded by

Author Verificateur Approbateur

Visa Signed Signed Signed

Réf. : INERIS - DRA - DRAG – 2005 - 10072 – RBo – Ammonia

2.1 Physical properties ........................................................................................9

2.1.1 General points ........................................................................................... 9

2.1.2 Thermodynamic data ................................................................................. 9

2.1.3 Solubility .................................................................................................. 10

2.2 Explosibility and flammability.......................................................................11

2.2.1 Explosibility limits..................................................................................... 11

2.2.2 Self-ignition temperature.......................................................................... 11

2.2.3 Minimum ignition energy.......................................................................... 11

2.2.4 Extinguishing agents................................................................................ 11

2.3 Reactions with contaminants.......................................................................12

2.3.1 Halogens and inter-halogens................................................................... 12

2.3.2 Heavy metals ........................................................................................... 12

2.3.3 Oxidants and peroxides ........................................................................... 12

2.3.5 Other aspects .......................................................................................... 13

2.3.6 Stability .................................................................................................... 14

2.4 Toxicity ........................................................................................................14

2.4.2 Acute toxicology....................................................................................... 15

2.4.3 Toxicology at the workplace .................................................................... 16

Réf. : INERIS - DRA - DRAG – 2005 - 10072 – RBo – Ammonia

2.4.5 Toxicology of fauna and flora................................................................... 20

2.4.5.2 Aquatic life ......................................................................................... 20

2.4.5.3 Plant life............................................................................................. 21

3.1 Context ....................................................................................................... 23

3.2 Atmospheric dispersion tests...................................................................... 24

3.2.1 Characteristics of accident ammonia releases......................................... 24

3.2.2 Tests by A. Resplandy ............................................................................. 26

3.2.2.1 Context .............................................................................................. 26

3.2.2.2 Test conditions .................................................................................. 27

3.2.2.3 Results .............................................................................................. 29

3.2.3 Tortoise Desert tests................................................................................ 30

3.2.3.1 Context .............................................................................................. 30

3.2.3.2 Test conditions .................................................................................. 31

3.2.3.3 Results .............................................................................................. 32

3.2.4 Unie van Kunstmest Fabrieken bv tests, Netherlands, 1972 ................... 33

3.2.5 Imperial Chemical Industries tests, United Kingdom, 1974...................... 34

3.2.6 Unie van Kunstmest Fabrieken bv tests, Netherlands, 1980 ................... 34

3.2.7 .Landskrona tests, Sweden, 1982............................................................ 35

3.2.8 FLADIS programme tests ........................................................................ 35

3.2.9 Tests conducted at Ecole des Mines, Alès .............................................. 36

3.2.10 Other atmospheric dispersion tests ................................................... 36

3.2.10.1 Summary ........................................................................................... 39

3.3 Modelling of atmospheric ammonia dispersion........................................... 39

Réf. : INERIS - DRA - DRAG – 2005 - 10072 – RBo – Ammonia

3.3.2 Three-dimensional models ...................................................................... 40

3.3.3 Integral models ........................................................................................ 40