Some Recent Advances in Liquefied Natural Gas (LNG) Production, Spill, Dispersion, and Saf

Review
pubs.acs.org/EF
Some Recent Advances in Liqueed Natural Gas (LNG) Production,

Spill, Dispersion, and Safety
Walter Chukwunonso Ikealumba and Hongwei Wu*
School of Chemical and Petroleum Engineering, Curtin University, GPO Box U1987, Perth, Western Australia 6845, Australia
ABSTRACT: The global demand of liqueed natural gas (LNG) has risen rapidly in recent years for the reasons of energy
security and sustainable development. This has led to considerable recent research interests and eorts in the LNG production
chain and associated risks in handling, storage, and transport of LNG, largely driven by the intrinsic process safety issues of LNG,
potential terrorist threats, and public condence in LNG safety. This review presents an overview on some recent advances in the
LNG value chain, covering upstream gas production and gathering, liquefaction, shipping, and regasication processes. Recent
developments in the experimentation and modeling of LNG spills associated with the LNG value chain are then summarized,
covering the events following a LNG spill, including LNG pool formation, vapor dispersion, and combustion. The consequent
hazards and safety issues are also discussed, with a focus on the methods for improving the safety of personnel, facilities, and
ships. The key technical gaps in the related research areas have been identied, and future research directions are outlined.
1. INTRODUCTION
1.1. Global Energy Outlook and the Role of Natural
Gas. Global energy demand is rapidly rising at a time when the
world needs to respond to the threat of climate change by
reducing carbon dioxide emissions, many of which came from
the use of fossil fuels. On the basis of BP statistical review of
world energy 2013,1 Figure 1 illustrates the growth of the world
energy mix with a forecast to 2030. It clearly depicts the growth
in energy consumption by fuel source, with the energy
consumption by natural gas approaching that of coal and oil.
For the year 2012, with a total energy use of 521.97 EJ [494.73
quadrillion British thermal units (Btu)], oil is the leading source
of energy in the world at 33%, closely followed by coal at 30%,
natural gas at 24%, nuclear at 4%, and other sources (wind,
solar, etc.) at 9%. However, in Australia, the total energy use is
5.26 EJ (4.99 quadrillion Btu), with coal being the leading
source at 39%, followed by oil at 37%, natural gas at 18%, other

sources (wind, solar, etc.) at 6%, and without nuclear (0%).
Of the three largest energy sources (coal, oil, and natural
gas), natural gas continues to be favored as an environmentally
attractive fuel. Natural gas is arguably the cleanest fossil fuel.2 In
comparison to oil and coal, it emits virtually no sulfur, no solid
waste, far less nitrogen oxide, and signicantly less carbon
dioxide than oil or coal. Furthermore, according to the U.S.
Energy Information Administration,3 new gas power plants
required 50% of the levelized capital cost of coal per MWh,
less than 33% the cost of nuclear, and less than 20% the cost of
onshore wind, let alone oshore wind (Table 1). The direct
greenhouse gas emissions from the combustion of coal and oil
are also signicantly higher per unit of energy compared to that
of natural gas (Table 2) based on the carbon dioxide emission
coecients of the U.S. Energy Information Administration4 and
the review paper by Lim et al.5 Within the next 15 years, many
old coal-red power stations will likely be retired, and if they
are replaced with modern gas-red power plants, emissions can
be reduced by approximately 50% per unit of energy produced,
as depicted in Table 2.
1.2. Importance of Liqueed Natural Gas (LNG) for
Natural Gas Utilization. Although considered to be the most
favored fossil fuel, natural gas faces the challenge of transport to
the demand sites, because of its gaseous state. Pipelines and as
LNG are the two most widely used methods for natural gas
transport. An economic analysis by the Institute of Gas
Technology (Figure 2, included in a report by the Center for
Energy Economics in the University of Texas at Austin6) shows
that transport of natural gas as LNG becomes cheaper
compared to oshore pipelines for distances of more than
1290 km (800 miles) or 3540 km (2200 miles) in the case
of onshore pipelines. Table 3 compares the fuel properties of
Figure 1. Growth of the world energy mix from 1990 and extrapolated
to 2030 (data were extracted from the literature;1 1EJ = 0.9478
quadrillion Btu).
2014 American Chemical Society
Received: March 20, 2014

Revised: April 23, 2014
Published: April 23, 2014
3556
dx.doi.org/10.1021/ef500626u | Energy Fuels 2014, 28, 35563586
Energy & Fuels
Review
Table 3. Fuel Properties of Natural Gas at STP710

Compared to Those of LNG (162 C)2,1114
Table 1. Average Levelized Capital Costs of New Power

Generation Resourcesa
energy
source for
the plant
levelized
capital cost
($/MWh)
levelized
capital cost
($/MMBtu)
total system
capital cost
($/MWh)
total system
capital cost
($/MMBtu)
natural gas
coal
nuclear
wind
(onshore)
wind
(oshore)
15.844.2
65.788.4
83.4
70.3
4.613.0
19.325.9
24.4
20.6
65.6130.3
100.1135.5
108.4
86.6
19.238.2
29.339.7
31.8
25.4
56.7
221.5
64.9
193.4
Data were extracted and converted from the U.S. Energy Information
Administration.3
coal
oil
natural
gas
a
nitrogen oxides,
kg/GJ
(lbs/MMBtu)
sulfur dioxide,
kg/GJ
(lbs/MMBtu)
90 (210)
71 (164)
50 (117)
196 (457)
210 (488)
40 (92)
1114 (2591)
482 (1122)
0.3 (0.6)
natural gas (STP)
LNG (162 C)
19.5
0.70.9
0.0375
161
1.1 105
16.043
422.5
20.3
161
1.14 104
13.36
4.186
0.2015
57
2.215
0.033
LNG trade is expected to come in the time frame of 2015

2025 because of projects in Australia expected to be completed
in 2020, closely followed by the completion of projects in
North America.16 Some key LNG projects worldwide are
presented in Table 4.1725
Technological advances have signicantly increased the
protability of LNG value chains,26 leading to a substantial
reduction in the cost of LNG value chains by 3050% between
the 1980s and 2003.27,28 For the total capital cost, the average
breakdown is 21% for assets in upstream exploration and
production, 40% for liquefaction plant, 20% for shipping
facilities, and 18% for storage and regasication terminals.5,29
Concrete values depend upon dierent factors, including traded
volume, transportation distance, and/or employed technologies. Table 5 shows the typical investment of an 8 mtpa LNG
process.28,30 The upstream gas prices mainly depend upon the
reservoirs, while the liquefaction cost depends upon feedgas
composition and the liquefaction process in use. The shipping
costs are determined primarily by the distance between the
sellers and buyers, while regasication prices are dependent
upon the construction costs, regasication technology, and
storage capacity.28
1.3. Life Cycle Performance of Natural Gas and LNG.
The life cycle performance of coal (underground and surface
mining),3,31,32 natural gas (conventional and shale gas),3,3133
oil (crude and oil sands),3133 and biofuel3,3234 energy
systems for a life period of 20 year is presented in Table 6. It
can be seen that, of the total greenhouse gas (GHG) emissions
in the life cycle assessment (LCA) of the energy systems,
approximately 91, 95, and 79% for coal, oil, and natural gas,
respectively, are released during the downstream combustion.
Additionally, on average, emission during fuel production is 3%
lower for oil and 21% higher for coal when compared to natural
gas. This is mainly due to the high emissions during the fuel
combustion stages in the coal system. It is noted that biofuels
can be produced from a wide range of sources, and as a result,
obtaining an average comparison of emissions to that of the
fossil fuels is quite challenging.
A similar trend is also observed in the electricity production
analysis, in which coal and oil on average produce more
emissions than natural gas; however, fossil fuels, in general,
produce more emissions than biofuels. It is important to note
that the emissions from electricity production for biofuels
(Table 6) did not consider carbon emissions from land use/
management, and most studies do not include the CO2
emission from biomass plants, with the assumption of CO2
emitted being equal to CO2 absorbed in biomass growth
stages.32 Because the main area of use for biofuels is in
transport, biofuels have low GHG emissions per kilometer
Table 2. Emission of Air Pollutants from the Direct

Combustion of Coal, Oil, and Natural Gasa
carbon dioxide,
kg/GJ
(lbs/MMBtu)
properties
molecular weight (g/mol)
density (kg/m3)
energy density (MJ/L)
boiling point (C)
viscosity (kg m1 s1)
surface tension (N/m)
specic heat (kJ kg1 K1)
thermal conductivity (W m1 K1)
shelf life (day)
Data were extracted and converted from the literature.4,5
Figure 2. Transport cost of natural gas technologies relative to

distance (data were extracted from the literature;6 1 $/GJ = 1.055
$/MMBtu).
natural gas at standard temperature and pressure (STP)710

and LNG (162 C).2,1114 It can be seen that liquefaction of
natural gas into LNG substantially increases the energy density
of the fuel, enabling the economic long-distance transport of
the fuel. Because international trade of natural gas is typically in
distances exceeding 3540 km (2200 miles), transport of
natural gas as LNG is the preferred option. Therefore, LNG
oers a global commodity, and its supply chain has been
subjected to signicant development in recent years as a result
of concerns over increasing requirements on energy security,
rising energy production costs, increasing natural gas prices,
and rising gas import.15 The global LNG trade has evolved
from 73.62 billion cubic meters (bcm) [2.6 trillion cubic feet
(tcf)] in 1990 to 283.17 bcm (10 tcf) in 2010 and is estimated
to rise to 566.34 bcm (20 tcf) by 2040.16 Another surge in the
3557
Energy & Fuels
Review
Table 4. Some Key LNG Projects Worldwide1725

project
country
capacity (mtpa)
number of trains
status
Darwin LNG
Brass LNG
Venezuela LNG
Trinidad and Tobago
North West Shelf
Arzew
Bontang LNG
Nigeria LNG
Qatar gas 14
Gladstone LNG
Ichthys LNG
Yamal LNG
Australia
Nigeria
Venezuela
Trinidad
Australia
Algeria
Indonesia
Nigeria
Ras Laan
Australia
Australia
Yamal
3.7
10.0
14.1
15.7
17.1
17.3
22.2
22.2
41.2
7.2
8.4
16.5
1
2
3
4
5
3
8
6
7
2
2
3
operating
operating
operating
operating
operation
operating
operating
operating
operating
under construction, 2015
Table 5. LNG Value Chain Costs for a Typical 8 mtpa Process28,30

upstream
gas use
capital expenditure
(billion dollars)
miscellaneous
(operating cost,
maintenance cost
etc.)
unit cost per GJ
($/MMBtu)
nil
$26
liquefaction
shipping
1014%
$610
regasication
1.53.5%
$12.5
12%
$11.5
57% of capital cost
$0.952.84 ($13)
$2.844.27 ($34.5)
total
12.519.5%
$1020
34% of capital cost
$0.761.42 ($0.81.5)
$0.380.76 ($0.40.8)
$4.939.29 ($5.29.8)
Table 6. LCA of Dierent Energy Sectors for a 20 Yeara Period3,3134

coal
fuel producedb (g of CO2 equiv/MJ)
electricity producedb (g of CO2 equiv/kWh)
transportb (g of CO2 equiv/km)
total system levelized cost ($/MWh)
110130
6751689
100.1135.5
natural gas
oil
biofuel
90101
290930
155185
65.6130.3
97116
5101170
185220
68131
18360
15195
111.0
The 20 year period was chosen because of the global warming potential of the specied fuel sources within that time frame. bThe GHG emission is
the total GHGs released during the development of the infrastructure, production/procurement of the fuel, and combustion of the fuel.
emission as a result of leakages from the pipeline transport

method was neglected. It was further reported that, over the
years, various technologies used in LNG value chains result in
an energy eciency ranging from 60 to 90%.30,36 The two most
important methods of improving eciency are by improving
the eciency of the heat exchanger or that of turbomachinery.
1.4. Objectives and Outlines of This Review. Considerable recent research was conducted in developing ecient
LNG value chains and managing the associated risks, such as
those in handling, storage, and transport of LNG. This has been
largely driven by the rapid increases in production and use of
LNG, potential terrorist threats, and public condence in LNG
safety. The potential hazards include cryogenic tissue damage
and embrittlement of material. Such hazards can be potentially
caused by various physical and chemical interactions with LNG
via direct contact, pressure from rapid phase transitions
(RPTs), asphyxiation, deagrations, detonations, LNG vapor
cloud res, and LNG pool res. Upon a LNG spill, a LNG pool
forms on the substrate and vaporizes rapidly as a result of the
local mixing and heat transfer. Ignition of the LNG pool (LNG
pool re) or the LNG vapor cloud (vapor cloud re) both
aect the vaporization/dispersion process and create thermal
hazards. Table 8 summarizes some of the reviews in the
literature5,15,3741 on dierent aspects of the LNG value chains.
It can be seen that the existing reviews related to LNG
traveled in a passenger car, closely followed by natural gas, and

then oil with the highest GHG emissions. Overall, the LCA
analysis in Table 6 shows that natural gas on average produces
low GHG emissions for fuel production, electricity production,
and transport use and costs less as a system compared to other
fossil fuels, such as oil and coal. This makes natural gas a very
strong competitor in the world energy sector. While the GHG
emissions for biofuels are considerably lower than those of the
fossil fuels, the biofuel sector is in its infant stage and there is
still a long way for its growth to make a signicant contribution
to the world energy mix.
In comparison to pipeline gas, LNG results in a 50%
increase of GHG emissions as a result of additional liquefaction
and tanker transport, according to the LCA analysis by
Jaramillo et al.35 presented in Table 7. It is noted that the
Table 7. LCA of LNG versus Pipeline Transporta
emission intensity (g of CO2 equiv/kWh)
stages of life cycle
LNG
pipeline (natural gas)
fuel production
liquefaction
transport
regasication
109
68
27
5
109
N/A
N/A
N/A
Data were extracted from the study by Jaramillo et al.35

3558
Energy & Fuels
Review
Table 8. Some Published Review Papers Regarding Dierent Aspects of the LNG Value Chain
reference
literature covered
Wood15
outlook of the global LNG trade
Lim et al.5
LNG plant designs
Luketa-Hanlin37
studies on large-scale LNG spills
Cleaver et al.38
Koopman et al.39
summary of experimental data on

LNG safety
LNG safety research
Raj40
LNG res
Havens and Spicer41
problems with the United States

LNG siting regulations
years
covered
up to
2012
up to
2012
up to
2006
up to
2006
up to
2006
up to
2006
up to
2006
aspects covered
current status and future growth of global LNG
available LNG plant designs and implemented designs and their optimizations in practice
based on costs
behavior of LNG spill, including combustion and development of predictive models,
concluding that experiments on the order of 100 m2 are required
behavior of LNG spill, combustion, and modeling issues, concluding that more
experimental studies of RPT are required
behavior of LNG from research performed by Lawrence Livermore National Laboratory,
including dispersion model development
some LNG re experiments on land and water from the 1970s to 1980s are reviewed,
including re hazard prediction models
problems with determining and specifying exclusion zones for LNG spills
production and the events following a LNG spill were all done
up to 2006.
Therefore, this review focuses on some recent research
advances in LNG production, spill, dispersion, and safety from
2007, although the classic earlier literature will also be covered
to maintain a smooth connection to prior knowledge in a
logical ow. The structure of this review is organized as follows:
Section 1 is a brief introduction of LNG and the need of this
review. Section 2 presents an overview on the LNG production
process, covering upstream gas production and gathering,
liquefaction, shipping, and regasication processes. Some recent
developments in the optimization of LNG plants are also
included. Section 3 summarizes the recent advances in
experimentation and modeling of LNG spills, covering the
events following a LNG spill, including LNG pool formation,
vapor dispersion, and combustion. Section 4 presents the
methods of improving the safety of personnel, facilities, and
ships with a direct correlation to regulations, such as U.S.
Department of Transportation (DOT) Regulations, 49 Code of
Federal Regulations (CFR), National Fire Protection Association (NFPA) 59A standard, and International Gas Carrier
(IGC) code. The key ndings and future research perspectives
are then outlined in Section 5. The review is then concluded
with overall concluding remarks in Section 6.
Table 9. Typical Composition of Natural Gas Compared to

LNG44
component (mol %)
natural gas
H2S
CO2
N2
CH4
C2H6
C3H8
i-C4H10
n-C4H10
i-C5H12
n-C5H12
n-C6H14
others
total
0.96
2.45
3.97
82.62
4.84
1.78
0.39
0.67
0.29
0.27
0.34
1.42
100.00
LNG
0.001.00
84.5596.38
2.0011.41
0.353.21
0.000.70
0.001.30
0.000.02
0.000.04
100.00
could be wet, containing hydrocarbons, such as butane,

propane, and other condensates. Associated gas on the other
hand is very expensive to gather for sale, and as a result, it is
ared.28 Table 9 lists the typical composition of a sample of
natural gas taken from Qatars North Field.44
All forms of natural gas need to be treated to remove
impurities (e.g., carbon dioxide, hydrogen sulde, nitrogen, and
mercury) and/or heavier hydrocarbons (e.g., butane and
propane) before the gas can be sent to the downstream
liquefaction plant. The pretreatment may consist of three main
stages: acid gas removal, dehydration, and mercury removal
(Figure 3).45 Impurities in the gas stream are referred to as acid
gas, and removal of this acid gas can be achieved via using an
amine solvent mixed with water.46 If signicant hydrogen
sulde is present in the acid gas, then a separate stream for
sulfur recover is required.28 Gas leaving the acid gas removal
system is generally saturated with water vapor and, therefore,
passed through a dehydration system. This is to prevent
freezing when the gas reaches the main liquefaction exchanger
and attained by rst cooling the gas with water, air, or a
refrigerant and then passing it through a molecular-sized sieve.6
The nal stage of gas pretreatment is mercury removal.
Mercury can corrode aluminum, and because most equipment
in the heat exchanger is made from aluminum, it is crucial that
this process is carried out.46 Mercury removal is achieved by
passing the feed gas through a sulfur-impregnated carbon bed,
during which the mercury becomes a non-volatile mercury
sulde.46 The natural gas after this series of pretreatment is
then sent to the liquefaction plant.
2. LNG PRODUCTION CHAIN

A LNG production chain transforms the raw natural gas into
LNG as a carrier product for the transport and distribution of
natural gas to the end users. It consists of four crucial stages,
including upstream gas production and gathering, liquefaction,
shipping, and regasication. Below is a brief overview on these
four stages, including some recent research and development
(R&D) advances in these aspects.
2.1. Upstream Gas Production and Gathering. All fossil
fuels are formed when remains of plants, animals, and
microorganisms are compressed underneath the earth under
high pressures and temperature for long periods of time. The
suitable geological conditions breaks down the carbon bonds in
organic matter to form natural gas, oil, or coal, with more oil
produced at lower temperatures but more natural gas produced
at higher temperatures.42 Natural gas is normally trapped under
ground in a reservoir (conventional gas) or by sedimentary
rock (shale gas) and, at times, interacts and is absorbed by coal
(coal seam gas).43 It can be found onshore or oshore, as
associated gas or non-associated gas.28 The non-associated gas
can be dry, meaning that it is composed of purely methane, or it
3559
Energy & Fuels
Review
Figure 3. Natural gas pretreatment and liquefaction (adapted from the literature,45 NGL = natural gas liquid, LPG = liqueed petroleum gas).
Table 10. Commonly Used LNG Liquefaction Processesa

single train capacity
(mtpa)
eciency relative to
cascade
eciency relative to
C3MR
ConocoPhillips
Black and Veatch Pritchard
APCI
49
0.12.1
0.56.1
1.00
1.25
1.15
1.16
1.19
1.00
Shell and APCI

IFP/Axens
Shell
Gaz de France
0.58.4
N/A
69
N/A
N/A
N/A
N/A
N/A
1.03
N/A
N/A
N/A
5.89
46.6
N/A
N/A
N/A
N/A
process
developer(s)
Phillips optimized cascade (POC)

PRICO
APCI propane precooled mixed refrigerant
(C3MR)
Shell and APCI dual mixed refrigerant (DMR)
IFP/Axens Liquen
parallel mixed refrigerant (PMR)
Gaz de France integral incorporated cascade
(CII)
APCI AP-X
StatoilLinde mixed uid cascade (MFC)
a
APCI
Linde in collaboration with
Statoil
Data were extracted from the study by Lim et al.5 that was based on the literature.4970
Table 11. Comparison of Refrigeration Cycle Congurationsa

refrigerant
process
precooling
heat exchanger
liquefaction
subcooling
methane
precooling
liquefaction
plate n heat exchanger (PFHE) or

core-in-kettle
cold box (PFHEs)
PFHE
MR
core-in-kettle
MR
MR
propane or MR
MR
propane
MR
MR
parallel MR
CWHE
PFHE
coil-wound heat exchanger

(CWHE)
CWHE
PFHE
heat-exchanger line (two PFHEs)

core-in-kettle
CWHE
MR
MR
PFHE
CWHE
POC
propane
ethylene
PRICO
C3MR
mixed refrigerant
(MR)
propane
DMR
Liquen
PMR
CII
AP-X
MFC
MR
nitrogen
gas
MR
subcooling
PFHE
CWHE and
PFHE
CWHE
Data were extracted from the study by Lim et al.5 that was based on the literature.4970
2.2. Liquefaction and Storage. The natural gas leaving

the production facilities is piped to a liquefaction plant, which
can be located onshore or oshore (to date, most liquefaction
plants have been onshore), and may have a number of parallel
systems (called trains) of heat exchangers and gas or steam
turbines driving the compressor.28 The natural gas is liqueed
into LNG via cooling it to 112 K,47 reducing the volume of
natural gas by a factor of 600. The LNG product is then stored
in insulated double-walled tanks designed to maintain the low
temperatures of the LNG. These storage tanks contain an inner
cryogenic nickel/steel tank, surrounded by insulation and then
followed by a pre-stressed concrete or mild steel outer tank
(when mild steel is used as the outer tank, a berm is

constructed to contain the LNG in the case of a spill).28
The rising demand of LNG has led to the development of
the gravity-based structure (GBS) and the oating LNG
(FLNG) concepts aimed at exploiting oshore natural gas
reserves that are too remote or not economically viable for
onshore liquefaction facilities.48 Essentially, the GBS is an
articial island, where LNG production and storage will take
place, and is intended for shallow waters. In the case of FLNG,
the system is oating and mobile, unlike the GBS. The idea is
that the structure will be prefabricated onshore and sunk to the
seaoor at the required location. This technology has been
3560
Energy & Fuels
Review
applied to multiple oil production facilities.48 However, such

LNG facilities also suer from some undesired disadvantages,
including large footprint, large construction costs, constant
upgrading for existing onshore facilities, only applicable to a
smaller range of water depths, and potential safety risks
(requiring a safe mechanism for transferring LNG into LNG
tankers).28
Various liquefaction processes have been developed by
dierent companies, as reviewed in a recent publication5 and
also summarized in Tables 10 and 11.4970 The fundamental
principles of these processes can be found in other related
publications by Barron,71 Walker,72 or Timmerhaus and
Flynn.73 As shown in Tables 10 and 11, these processes have
dierent single train capacity and eciency (relative to cascade
or C3MR) and deploy a range of refrigerants and heat
exchangers for precooling, liquefaction, and/or subcooling.
The recent research advances on the topic has been focused
on the optimization of the LNG liquefaction process, via several
possible strategies:
Design or Selection of Optimized Refrigerant Mixture.
Alabdulkarem et al.74 use a generic algorithm (GA) method
from MATLAB in combination with HYSYS to design an
optimized refrigerant mixture for a C3MR LNG plant,
achieving a total reduction of over 9% in power consumption.
The method can be applied with condence for a global
optimum, unlike the previous method by Michelsen et al.53 As
shown in Figure 4, it is also more ecient than the previous
Optimization of the Precooling Cycle. Via HYSYS

simulations, Castillo et al.78 have recently concluded that a
three-stage propane precooled cycle can be more sucient than
a mixed refrigerant cycle. However, in cool climates (6 C), the
reduced power share from the three-stage propane precooled
cycle results in the mixed refrigerant cycle as a preferred
alternative.
Utilization of Process Energy. Conventional optimization of
the LNG liquefaction process focuses on component
replacement that requires signicant modications to existing
process licenses and becomes dicult and expensive for
integration into the existing facilities. Recent research progress
has been made on the smart use of process energy that can be
easily implemented into existing facilities for improving the
eciency of the LNG liquefaction process. For example, the
cold exergy of subcooled LNG can be used to recondense the
light component of LNG after it has passed through a phase
separator, leading to reduction in the pumping power of
vaporized natural gas and, hence, the whole liquefaction
process.79 The energy of the ash gas can also be used for
cooling the natural gas and refrigerant and, hence, reducing the
specic power to various extents, depending upon the ratio of
the natural gas feed rate to LNG ow rate.80 Another method is
to use gas turbine and/or waste heat power for the subcooling
of the propane cycle (after the condenser by 21 C), leading to
13 and 23% increases in the process performance coecient
and total cooling capacity, respectively.81
Development of Numerical Methods for Optimization.
Recent research advance demonstrates that quadratic sequential
programming (via the use of a Fortran code NLPQLP, coupled
with HYSYS) can drastically improve the calculation eciency
in the optimization of LNG processes (in particular, the
PRICO process).82 The method produced results within 1% of
the optimal value and carried out simulations in 26 min
compared to the typical 422 h of other methods.82 The
method is proven to be accurate and ecient for a simple LNG
process, although its application to more complex LNG
processes is yet to be assessed because of more diculties in
the coding process and the linking between NLPQLP and
HYSYS. In addition, thermodynamic analysis, mathematical
programming, and rigorous simulations were used by Wang et
al.83 to develop a new methodology for LNG liquefaction
process design and operation optimization. This method was
able to achieve a total power saving of 13% (higher energy
saving than that by Alabdulkarem et al.74) for a C3MR LNG
plant and was shown to be able to reach a global optimum.
Deployment of New Control Strategies. A recent study by
Michelsen et al.53 demonstrates that the use of a set of nonconstant control variables rather than the standard constantset-point policy can reduce the operation losses (by up to
98%) because of the changes in unknown disturbances and
other uncertainties. The disadvantages of such a non-constant
control variable method are the trial and error approach for
selecting the non-constant control variables, requirement of
recalculations upon changes in process constraints, and no
guarantee in reaching a global optimum.
Considerations of Thermal Interaction between Streams.
Multi-stream heat exchangers, consisting of two hot streams
(high-pressure mix refrigerant and natural gas feed) and a cold
stream (low-pressure mixed refrigerant), are usually optimized
by using commercial software, such as HYSYS and SIMSCI
PRO II.84 In process analysis, these commercial software
assume that the two hot streams are at the same temperature.
Figure 4. Optimized power consumption by Alabdulkarem et al.74

compared to the methods by Paradowski et al.75 and Venkattaratham76
methods (e.g., by Paradowski et al.75 and Venkattaratham76) for

propane precooled refrigerant optimization processes. However, the method provided by Alabdulkarem et al.74
incorporates numerical algorithms without including process
knowledge, such as knowledge of the boiling point dierences
of MR components, MR component distribution curves, and
specic refrigeration eect of bringing a MR system closer to
reversible operation. Khan et al.77 therefore proposed a
knowledge-based optimization (KBO) method for selecting
mixed refrigerant, which achieved a total reduction in power of
30 and 13% in the SMR and C3MR LNG processes,
respectively. Although a global optimum cannot be achieved
from the KBO,77 it is quite robust and can result in a higher
energy saving than other methods, such as that provided by
Alabdulkarem et al.74
3561
Energy & Fuels
Review
Table 12. Summary of Various Cargo-Containing Systems for LNG8992
Via energy balances, of dierential equation, Chang et al.84 was

able to show that the assumption made by commercial software
is unrealistic. The assumption can only be achieved if the
product of the heat-transfer coecient and heat-exchanger area
(UA) between a hot/hot stream (hot stream in contact with a
hot stream) is greater than that of a hot/cold stream (hot
stream in contact with a cold stream).84 Chang et al.84 was also
able to show that for any total UA, there is an optimal ratio of
UA between the streams to increase performance, and any
direct contact (heat-exchanger tubes) will deteriorate performance.

Improvements via LNG Plant/Process Evaluations. An
exergy evaluation of a LNG plant in Norway85 concludes that
37% of the overall exergy losses are due to the process plant,
52% of the overall exergy losses are from the combustors and
gas turbines, and 11% of the overall exergy losses are from the
exhaust gas (stack exhaust). Ambient temperature analysis also
shows that a cooler climate will reduce the total exergy required
for the liquefaction process. Improvement of the exergy
3562
Energy & Fuels
Review
Figure 5. LNG spill events.
more expensive. However, the moss tanker has a higher

resistance to sloshing and will most likely be considered for
future oshore storage, especially in areas of bad weather. Next
is the IHI SPB-type ships that also use rectangular tanks with
polyurethane insulation; however, unlike the membrane-type
tanks, the IHI SPB does not rely on the hull of the ship for
structural support.91 The IHI SPB ships are also resistant to
sloshing, but its expensive design and small capacity (87 500
m3) compared to that of the membrane or the moss systems
explain why only two ships are currently in-service. The last
type is the type C (cylindrical) system. The type C ships are
able to resist sloshing, use polystyrene for insulation, and were
designed specically for cases where small-cargo transportation
is required (250030 000 m3).90
During the shipping of LNG, boil-o occurs (approximately
0.15% of the cargo28). Because current LNG ships are not
constructed with an on-board regasication facility, the boil-o
gas is captured and used as fuel to assist in propelling the ship.
Once at the import terminal, the LNG is unloaded, leaving
behind a small amount of LNG, the heel, to keep the tank cool
in the ballast voyage.28 To date, there has been no LNG cargo
lost or spilled during the shipping phase, and it is believed that,
with acceptable maintenance, the LNG vessels can have a
working life of 40 years or more.28
2.4. Regasication Terminals. A LNG value chain ends
with regasication terminals. At each unloading berth, several
unloading arms, with articulated joints and emergency release
couplings, are generally employed for transferring LNG from
the LNG tankers to the storage tanks. The main purpose of
storage tanks is to hold the LNG until it is ready to be
vaporized. The storage tanks used at the regasication terminals
are exactly the same as those at the liquefaction plant.28
The LNG is then passed through vaporizers, which warms
the LNG to or above 5 C (41 F). The vaporizers use either
seawater, air, natural gas, or an external heat source in the
vaporization process of LNG and consume a signicant amount
of electrical energy. Using a new combined power cycle that
incorporates a gas turbine and an ammonia cycle with one
pressure step, the eciency can be improved from 44 to 46%
while decreasing capital cost.93 Because LNG is a cryogenic
liquid, boil-o occurs at about 0.05% per day of tank volume or
less. To decrease boil-o gas (BOG), it is believed that the
number of stages (compression and reliquefaction) in the
eciency can be achieved via a generic algorithm optimization

method developed by Li et al.86 While the exergy eciencies of
current large-scale processes are about 4145%, the exergy
eciency can be increased up to 58% for methane
liquefaction.86 Of the three main options for oshore associated
gas liquefaction resources, a comparative analysis87 showed that
the nitrogen expander process has higher power consumption
(5268%) and poorer economic performance than the C3MR
or MR process. The nitrogen expander process appears to be
more suitable for oshore processes because of its simplicity,
compact design, less sensitivity to oating production storage
and ooading, higher safety, and operability.87
In addition, after liquefaction, LNG is stored in storage tanks
before been transferred into a LNG tanker for transport. In
comparison to several conventional LNG storage models, a
recent model88 based on normal equations applied to nonlinear
parameter estimation has signicantly improved the modeling
accuracy for safe and economic managing of LNG storage sites
(prevent stratication and rollover of LNG). The model also
converges faster to a heat- and mass-transfer coecient and can
be used with leveltemperaturedensity proles (which can
easily be obtained from gauges on the storage tanks) to obtain
the required results.88
2.3. Shipping. After liquefaction and storage, the LNG is
then loaded onto specialized ships for its long-distance
transport. The ships are designed with double hulls for
structural rigidity and also to provide room for ballast water,
considering how light the cargo is. The cargo tanks are not a
part of the ship structure and are therefore installed separately
into the holds of the ship.
Table 12 lists the features and characteristics of four types of
LNG ships that have been developed since the late 1960s.8992
The membrane-type ships, ranging in size from 145 000 to 265
000 m3, use fully integrated rectangular tanks with reinforced
polyurethane or plywood/perlite insulation.90 With regard to its
close competitor, the moss-type ships, ranging in size from 138
000 to 255 000 m3, are designed with spherical tanks,
independent of the hull structure, and use polyurethane
insulation.90 There is a preference for the membrane tankers,
mostly because of their capacity eciency (use of the hull
shape), with little to no void space between the storage tanks
and the ballast tanks compared to the moss tanker.28 This will
make a moss tanker of similar capacity as the membrane tanker
3563
Energy & Fuels
Review
ity, wind speed, turbulence, solar heat ux, etc.) and gas
parameters (e.g., temperature, concentration, ground heat ux,
etc.).38 Spills of LNG, on either land or water, can range from
all sizes. At times, the anticipated spill size or conditions cannot
be replicated experimentally, and this has led to the use of
mathematical models to simulate, study, and understand the
characteristics of a LNG spill, leading to vapor production,
dispersion, and/or combustion. These mathematical models
would primarily need to be validated, with existing experimental
studies, to be condent in using them for further studies.
While the overall research progress is briey summarized,
this review emphasizes the recent experimental and modeling
development since 2007 on these important aspects. Section
3.1 presents a summary of pool formation and factors aecting
this process, followed by a recently developed mathematical
model, to allow for better prediction of the shape and behavior
(up to dispersion) of a LNG pool following a spill. Section 3.2
summarizes the vapor dispersion process, including what
parameters drive this process and the mathematical models
used in studying vapor dispersion. Section 3.3 contains a
summary of LNG res, including the modeling methods of
LNG pool/vapor cloud ignition, in eect of studying
overpressures and radiant heat ux that would arise.
3.1. Pool Formation. Upon spilling on water, LNG spreads
and boils because of the temperature dierence between LNG
and the water surface. Spills on conned, calm water can
become similar in behavior to spills on land, because of ice
formation, which will result in a decrease in the evaporation
rate over time. The vaporization rate is also important in that it
inuences the distance to the low ammability limit (LFL) and
the burn rate of pool res. Therefore, the vaporization rate and
the size and shape of the LNG pool are key parameters
inuencing LNG pool formation.37
3.1.1. LNG Spreading on Water. In the literature, most of
the experiments concerning LNG spreading on water were
conducted prior to 2007. Those experiments were on small
scales, and application of such experiments results to large-scale
applications are known to have the main technical uncertainties
in terms of the dynamics of the front of the spreading pool and
the heat-transfer rate.37 While eorts were made to deploy
models for simulating bubble formation at large scales based on
small-scale experimental data, the eect of waves and ice
formation can introduce some more uncertainties in such
predictions.38 Therefore, large-scale experiments are certainly
still in need.
3.1.2. Pool Boiling. Pool boiling occurs when LNG is spilled
on water through three stages, namely, nucleate boiling,
transition boiling, and lm boiling. The up-to-date R&D on
pool boiling of LNG leads to several major conclusions:37,108
(a) evaporation is a function of the molecular weight of the
material (starting from the lowest); (b) boiling does not take
place at a constant temperature; (c) the vaporization rate of
LNG is very dierent from that of pure methane, and the
addition of ethane or propane results in a more rapid
vaporization at increased boil-o rates by a factor of 1.52;
and (d) evaporation during the spill phase is also an important
consideration because it may contribute to up to 20% of the
overall evaporation.
3.1.3. RPTs. RPTs can be classied as physical or mechanical
expansion, in which high-pressure energy is released, and occur
when cold LNG comes into contact with water. RPTs occur
primarily during the experiments at laboratory scale.37 The only
large-scale RPTs observed were during the Coyote test series,
regasication system should be increased, because an increase

in the number of stages leads to a decrease in the BOG ow
rate.94 Although this will reduce the operation cost, it will also
increase the total capital cost; therefore, an optimum number of
stages will need to be determined (via optimization using
models, e.g., developed by Liu et al.94).
Cold is also an important energy source that is available
during regasication. Innovative technologies, such as a novel
cogeneration concept, have been proposed to recover exergy
from LNG for the generation of power (with an overall system
eciency of 52%).95,96 There is potential for this novel
cogeneration concept to be improved by focusing on the gas
turbine power system, to improve eciency, reduce environmental impact, and improve cost. 97 Other innovative
technologies using cold energy consider the use of helium
(49% eciency),98,99 ammoniawater mixture (48% eciency),100 and a combination of tetrauororomethane and
propane (23.5% eciency)101 as a working uid, with some
plants even able to achieve near-zero CO2 emission.102 The
feasibility of using cold energy from LNG regasication in other
industries, such as deep freezing agro-food industries and airconditioning facilities in supermarkets and hypermarkets, has
also been demonstrated in numerous literature, with carbon
dioxide as the preferred uid for transferring the cold
energy.103106
After LNG regasication and prior to any natural gas leaving
the regasication terminal, it is metered, during which its
pressure and quality is regulated to meet customer specications; at times, the natural gas is odorized to help identify leaks
in transportation systems or customer appliances.28
2.5. LNG Value Chain Potential Hazards. During the
LNG production processes, some simple failures/events can
lead to a series of catastrophic events, with the most common
been a LNG spill. The block ow diagram in Figure 5 shows
the likely events that can follow, after a LNG spill. The factors
and parameters that inuence and control the spread of these
events are not fully understood. The events in Figure 5 can be
categorized into three main areas of interest: (a) pool
formation, understanding is needed on the key factors that
aect the pool spreading, pool boiling, and occurrence of RPTs;
(b) vapor dispersion, understanding is needed on what eect
wind and obstacles have on the dispersion processes of the
vapor cloud and there other governing factors; and (c)
combustion, understanding is needed on the characteristics of
(and the key factors that inuence) the pool re and vapor
cloud re formation and spreading, and how the two dierent
res compare to each other. These types of questions have led
to the use of experimentation and mathematical modeling to
study LNG spills, with the aim of improving knowledge and
further understanding the likely case of events following a LNG
spill and how it can be prevented and/or managed. Such
knowledge and understanding are essential to maintaining and
managing the safety of LNG.
3. LNG SPILL, VAPOR PRODUCTION, DISPERSION,

AND COMBUSTION
Experimental studies of LNG vapor production, dispersion, and
combustion commenced from the early 1970s at both lab and
eld scales.38 These investigations were carried out with the
main objectives of collecting experimental data for benchmarking and the validation of computer models.107 Extensive
measurements were conducted considering various parameters,
including meteorological parameters (e.g., temperature, humid3564
3565
code accessibility
applications
scales
disadvantages
N/A
inability to account for mixing with air in a connement

inability to account for expansion of the vapor volume
because of heating above the boiling temperature
eld
laboratory
LNG hazard assessment
cannot model non-circular pools, channeled ows, or

connement by arbitrarily shaped dikes
heat transfer from the sun, re, and air are all neglected
cannot model a rough or slopping ground
can model LNG spreading on land, connement by a simple shape bund/dike, instantaneous or continuous releases
accounts for smooth and rough ground
can model LNG spreading on land, connement by a simple

shape bund/dike, instantaneous or continuous releases,
drainage and permeable ground
can also model LNG spread on water for both instantaneous
and continuous releases
accounts for heat conduction from substrate
N/A
LNG hazard assessment
eld
radiative heat from re is not included but can

be added as user-dened input
does not account for RPTs
heat transfer from the sun, substrate, and air are

all included
permeable ground, sloping ground, and drainage are not modeled
does not account for non-circular pools
can account for spills on land and water
output from SOURCE5 can be used as input to the

DEGADIS dispersion code
advantages
some validations were conducted during the

development of the program but most were
not published
unclear, although it might have been validated in the process

of validating DEGADIS
integral model
PHAST118122
hazard analysis (power plants, reneries,

chemical plants, petrochemical plants, pharmaceutical facilities, and oil rigs)
N/A
eld
can model heat conduction from the substrate,

convection from the air, and radiation
cannot model non-at ground
eects of bund walls
accounts for instantaneous and continuous

releases
wind tunnel experiments

Kit Fox
numerous eld experiments
can account for spills on land or water surfaces
initiated in 2002 and simulates discharge, pool

formation, evaporation, dense gas dispersion,
and res
algebraic equations and, in some cases, are
solutions of ordinary dierential equations
pseudo-component approach for mixtures
LNG pool formation models
initiated in the 1980s and simulates instantaneous or continues releases of a circular pool
and vaporization rate
algebraic equations and, in some cases, are
solutions of ordinary dierential equations
models bulk quantities as a function of time
integral model
GASP
114,115
validations
accuracy
algebraic equations and, in some cases, are solutions of

ordinary dierential equations
models bulk quantities as a function of time
method
integral model
initiated in the 1980s and simulates instantaneous or

continues releases, pool formation, and evaporation
principles
model
SOURCE5
41,111113
Table 13. Some Well-Known Pool Formation Models for LNG Pool Modeling
N/A
hazard analysis (LNG ships and oshore platforms)
eld
restricted to marine spills (spills on a water surface)
cannot account for RPTs
can model pool spreading, transport on water, evaporation

from water, and the transport in the atmosphere; either
constant or time dependent
accounts for thermal eects of conduction and radiation
initiated in 2005 and simulates instantaneous or continues

releases of a circular pool, vaporization rate, and res for
marine spills
discrete set of linked algorithms and a particle-based
approach
fully integrated geographic information system
can range from 1 to 38%; however, no large-scale
experiments are available to be used for validation
ABS Consulting and Sandia numerical simulation cases
integral model
LNGMAP11,126,127
Energy & Fuels

Review
Energy & Fuels
Review
ratio between the spill substances and the sea, the eects of the
pool leading edge, and the initial and spill conditions. When the
maximum radius and evaporation time of the supercritical
model were compared to the standard model, a ratio
(supercritical to standard) of 2.51 was obtained for the
maximum radius and 0.159 for the evaporation time.130 On
the basis of this, a supercritical model would result in a larger
maximum radius and shorter evaporation time than the
standard model. These ndings were also veried by a China
Lake experiment,184 in which the supercritical model had a
better t to experimental data.
3.2. Vapor Dispersion. In the case of unconned LNG
spills on water, the LNG cloud travels at the wind speed prior
to dispersion. Because the vapor forms at the boiling
temperature of LNG, it will initially be denser than air. For
land-based facilities, dense gas behavior is advantageous
because it can be easily controlled. However, this can also be
a disadvantage because it takes longer to disperse. Most largescale LNG dispersion tests of spills on water were performed
prior to 2007, with the key experimental conditions being
summarized in Table 14 by Luketa-Hanlin.37 The spill volume,
in which 6 of 13 spills carried out resulted in RPTs. The up-todate R&D on this aspect leads to the following observations:3739 (a) RPTs at large scales behave dierently from
their small-scale counterparts; (b) occurrence is signicantly
inuenced by the water temperature and depth of penetration;
(c) the strength of RPTs correlates with the spill rate; and (d)
RPTs can lead to an increase (up to 65%) in the distance to the
LFL. Several theoretical models exist for studying RPTs, such as
the superheat theory, the predictive empirical model, and the
computational uid dynamics (CFD) model.37,38
3.1.4. Modeling. In recent studies, there have been limited
experimental investigations on the formation of LNG pools.
Instead, the recent studies seem to be focused more on
modeling of this process. Numerical models for LNG pool
formation are mainly classied in two categories: the integral
model or the NavierStokes model. There are numerous
models for studying LNG pool formation, including Raj and
Kalenkar,109 Opschoor,110 SOURCE 5,41,111113 GASP,114,115
SafeSite3G,116,117 PHAST,118122 ALOHA,123125 ABS Consulting model,11 LNGMAP,11,126,127 and FLACS.128,129 Of
these models, the most commonly used are presented in Table
13.
Integral models originated in the 1980s are the simpler of the
two techniques. These models use algebraic equations to obtain
solutions and are usually limited to modeling of circular pools,
at substrates, and heat transfer only from the substrates.
Examples of some commonly used integral models for pool
formation modeling is shown in Table 13. The
LNGMAP11,126,127 seems to be the most robust because of
its ability to eectively incorporate real-time geographic
information, such as wind eects, current eects, atmospheric
conditions, etc., into the model.126 PHAST118122 is an older
model than LNGMAP,11,126,127 seems to be more widely used,
and will most likely continue to be because of its ability to
model spills on both land and water. PHAST118122 is also
more superior to SOURCE541,111113 and GASP114,115 models
because of its ability to account for non-circular LNG pool
formation and inclusion of heat convection/radiation from
sources other than the substrate. NavierStokes models are
more complex and the most complete models. Modeling pool
formation with NavierStokes models can be time-consuming
because of their complexity. As a result, researchers prefer to
model pool formation with integral models and then transfer
the data over to NavierStokes models for further analysis.111
A more in-depth description of the NavierStokes models is
given in section 3.2.4.
The current pool spread models (e.g., the standard model of
inertialgravity spreading) are based on oil pools spreading;
however, it is important to note that oil and LNG behave
dierently.130 This standard model is applicable for rapidly
formed pools; to account for this, many assume that the spill of
LNG from a tanker is due to quasi-steady gravity ow.130 The
pressure eld is also assumed to be hydrostatic, which is not the
case in the pool front. The spread of a LNG pool is treated as
inviscid by the standard model, which is a reasonable
assumption because of the occurrence of the Liedenfrost
eect.130 Fay130 carried out studies in which the properties of
LNG spills were examined to suggest that a dierent model
should be used and to compare the dierences that would
ensue from the use of an alternative model.
An alternative model, characterized by an asymptotic
spreading law, called the supercritical model was also
developed130 because of questions arising based on the density
Table 14. LNG Dispersion Test on Water37

downwind
distance to
LFL (m)
(maximum)
spill volume
(m3)
spill rate
(m3/min)
pool radius
(m)
ESSO
Shell
0.7310.2
27193
18.9
2.719.3
442
2250 (visual)
Maplin
Sands
Avocet
(LLNL)
Burro
(LLNL)
Coyote
(LLNL)
Falcon
(LLNL)
520
1.54
714
NA
(jettisoned)
10
4.24.52
6.827.22
220
2439
11.318.4
420
828
1419
not reported
310
20.666.4
8.730.3
not reported
380
experiment
190 20
spill rate, vaporization rate, presence of obstacles, and

atmospheric conditions are considered to be key parameters
in determining the LFL.66,68,71 The recent research advance in
LNG spill has been focusing on vapor dispersion, including
experimental studies on the eect of water curtains, underwater
releases, and turbulence on vapor dispersion and modeling
studies on the use and development of models for simulating
vapor dispersion.
3.2.1. Water Curtains. The water spray curtain is widely
used as an inexpensive technique for controlling and mitigating
many toxic and ammable vapors. Field tests131136 have
shown that water curtains can reduce the concentration of LNG
vapor clouds and are able to interact with vapor clouds by
imparting momentum, heat transfer, and air entrainment.
However, the eectiveness of dierent water curtains (with
dierent ow congurations, drop sizes, coverage heights, and
coverage widths) for LNG concentration reduction and
temperature increase is still largely unknown. Such knowledge
is essential to developing structured engineering guidelines for
the design of water curtains in practical applications.
Rana et al.137,138 investigated the eectiveness of two
dierent water curtains in dispersing LNG vapor cloud using
an experimental setup consisting of LNG supplied by a tanker
3566
Energy & Fuels
Review
installation congurations) for water curtain emergency system

design. Even though further investigations are required,
especially regarding turbulent eects of dierent sprays, there
were several important conclusions: (a) the heat-transfer rate
increases with the water ow rate; for droplet sizes ranging
from 0.58 to 1.43 mm, a 0.94 mm droplet was noted to have
the highest heat-transfer rate per water ow rate; (b) the higher
the droplet temperature, the better the dispersion; a 313 K
droplet temperature was the most optimal, and any droplet
temperature below this showed signs of potential hazards with
the LNG vapor cloud owing around the water curtain because
of insucient heat transfer; and (c) the installation
congurations have an optimal tilt angle with the wind (in
that case, 60 compared to other angles), and the closer the
nozzles are to the source, the better the interaction and forced
dispersion of the vapor cloud.
3.2.2. Underwater Releases. Little is known about behavior
of LNG following its release underwater, and current models
are not sucient to quantify the potential hazards from such a
spill. Using a concrete pit lled with water, Qi et al.143 carried
out tests at the Brayton Fire Training Field to understand the
phenomena that occur from an underwater LNG release,
including the behavior of the emanating vapor. LNG was
released from a 2.5 cm (9.84 in.) nozzle at a depth of 0.71 m
(2.33 ft) below the water surface. There was no notable LNG
pool formation on the water surface, most likely because of the
high rates of evaporation and gas release. It is also possible that
all of the LNG was not vaporized underwater but, instead, was
thrown up with the rest of the vaporized liquid. The vapor
cloud that emanated from the water surface was at a
temperature below the dew point of air, and as a result, the
vapor cloud was visible. The lowest temperature recorded for
the vapor cloud was 1 C, which explains why the vapor cloud
was buoyant because the neutral buoyancy temperature of
natural gas is 117 C.
3.2.3. Eects of Turbulence. The rate of evaporation is
necessary for the accurate prediction of the formation, growth,
and dispersion of vapor clouds. Typically, it is assumed that the
evaporation rate is constant; for example, the detailed CFD
study by Cormier et al.144 used a constant mean evaporation
rate. However, it is known that, at the spill point as well as the
leading edge of the pool, the turbulence and evaporation rate
are both elevated. Morse et al.145 focused on providing data
regarding the evaporation rate of LNG from a water surface by
controlling the turbulence. Small-scale experiments were
performed in a double-walled vertical cylinder, with a
submerged turbulent jet, as shown in Figure 7. The water
level was held at a constant level, and once the turbulence
surface was established, the LNG was poured onto the water
surface. The pressure in the cylinder as well as the temperature
of the escaping vapor were recorded and used to calculate the
evaporation rate. Morse et al.145 observed that, for every 1 cm
(0.39 in) increase in LNG thickness, there was approximately a
10% increase in the evaporation rate. On the other hand, if
turbulence intensity doubles, then the evaporation rate will also
double. On the basis of this, it is clear that the evaporation rate
depends upon turbulence intensity and, at a smaller level, the
thickness of the LNG layer. However, this eect could be
dierent at large scales.
3.2.4. Modeling. Various numerical models have been
developed for studying LNG vapor dispersion. The main
dierences among the models are in the completeness of
simulation for the dispersion process, the capabilities in
truck and discharged on a spill area (concrete pad), enclosed

with wooden frames. The nozzles selected were 60 full cone
spiral and 180 at-fan water curtain types, which were placed
downwind of the spill site. Forced dispersion from the water
curtains led to a reduction in the LNG vapor concentration.
However, when the strength and dilution ratios of both water
curtains were studied, it was evident that the full cone spray is
more eective at creating turbulence and, therefore, increasing
mixing with air. However, the at fan is eective in creating a
solid barrier and, hence, pushing the vapor cloud upward and
reducing the ground level concentration. A further study by
Rana et al.139 involved the LNG been spilled on a water surface.
A comparison between the vapor cloud motion from the two
studies suggests that water curtains disperse LNG vapor clouds
by air entrainment and mixing. From these studies, it is evident
that the full cone water curtain is the most ecient (as shown
in Figure 6, where the full cone spray resulted in a higher heat
Figure 6. Calculated heat loss by water curtain (data were extracted

from the study by Rana et al.139).
loss compared to the at fan). However, a combination of

dierent types of sprays will ensure sucient mixing, heat
transfer, and momentum impact to disperse various size of
LNG vapor clouds. Olewski et al.140 also carried out water
curtain dispersion experiments using a smoke trace in place of
LNG for safety reasons. A are was placed in front of an
improvised wind tunnel. Two types of at-fan sprays were
tested with several water ow rates and fan speeds, totalling 24
dierent tests. Experimental results reinforced the ndings by
other researchers as to the ability of water curtains to control
vapor clouds by mechanical- or momentum-transfer eects.
Kim et al.141 carried out CFD simulation (on a similar
experimental setup to that used by Rana et al.138) of water
curtains using ANSYS Fluent, because current models were not
enough to draw a nal conclusion on design parameters and the
fact that heat transfer and well-dened atmospheric dispersion
were not included in most water curtain models. With broadly
good prediction, the model simulation overpredicted the
experimental data, which was desirable when trying to nd
the vapor cloud exclusion zone.141 The added heat-transfer
eect was noted to induce a distinct airvapor mixture,
resulting from the natural circulation within the airvapor
mixture. An increase in heat and/or mass transfer lead to a
reduction in the vapor cloud exclusion zone. Further CFD
simulations by Kim et al.142 aimed at determining key
parameters (including droplet sizes, droplet temperature, and
3567
Energy & Fuels
Review
ZEPHYR.128,129,153,161 Table 16 lists the key features and

comparisons of four well-known NavierStokes models
(FEM3, FLACS, FLUENT, and CFX) for LNG vapor
dispersion. It can be seen that, recently, FLUENT and CFX
numerical models have been the main NavierStokes models
used for modeling. This is largely due to the key advantages of
these models, including robustness, multiple solving methods,
high levels of accuracy, and ability to add to the coding for
specic simulations.162 Although giving a more complete
description of the physical processes available and performing
better than box or top-hat models,37 the NavierStokes models
are more computationally expensive.146
The recent developments in modeling are on four aspects,
including developments in models, modeling of complex
geometries, modeling of complex scenarios (e.g., the eect of
wind and ship motion), and evaluation of mathematical models.
These advances are summarized below.
3.2.4.1. Developments in Models. The Federal Energy
Regulatory Commission (FERC) developed a box/top-hat-type
model that uses the model developed by ABS Consulting11 for
source term modeling and DEGADIS for vapor dispersion
modeling. The FERC model was successfully used for a
comprehensive sensitivity analysis on the eect of tank (25 000
m3) conditions, release scenarios, and environmental conditions
on LNG spill, spread, and dispersion.163 Table 17 summarizes
the main conclusions of the simulations, considering key
parameters, such as breach diameter, ullage pressure, weather
conditions, and surface roughness, on LNG spill and
dispersion.163 It was concluded that breach size and ullage
pressure are important parameters that have a signicant eect
on LNG spill duration, pool size, and dispersion. LNG spilled
from a spherical tank always has a higher spill rate.
The most recent SafeSite3G model is also a box/top-hat-type
model that was developed to predict the dynamic eect of LNG
discharge and pool ignition on the pool spread process.116 The
model was used for the dynamic analysis of LNG tank (25 000
m3) release conditions: LNG spreading, dispersion, and
combustion. It was found that the bevelled cross-section of a
membrane tank can be easily replaced with a rectangular crosssection and still obtain a high degree of accuracy (discharge rate
versus time). The analysis showed that, when the average pool
depth is above a minimum, both burning and non-burning
pools spread at the same rate but the evaporation ux of the
discharged LNG is dependent upon turbulence generated
during the spill.116 SafeSite3G was also able to capture and show
that, for a period of constant discharge, the discharge rate
would be greater than the evaporation rate; this eect was also
noted by Johnson et al.164 using the consequence analysis
methodology developed by FERC.
LNGMAP is another box/top-hat-type model that uses a
discrete set of algorithms within a geographic information
system framework for studying LNG releases and transport in
the case of marine spills,126 overcoming a series of limitations of
existing models.165,166 The model predicts time-dependent
release, spreading and transport of LNG on the water surface,
LNG evaporation from the water surface (eect of the current
and waves), LNG transport and dispersion in the atmosphere,
and the burning and associated radiant elds of LNG res.126
The model was validated by carrying out simulations presented
in the ABS Consulting11 and Sandia reports.127 In comparison
to the Sandia cases, LNGMAP predicted the distance to the
thermal radiation contours being larger for small hole sizes and
lower for the largest hole size (see Table 18). LNGMAP was
Figure 7. Small-scale experimental setup for measuring the

evaporation rate of LNG from a water surface by controlling the
turbulence (adapted from the study by Morse and Kytomaa145).
dierent release processes, the ability of the model to describe

processes, the completeness in elds and data used, and the
complexity of the terrain for which the model is situated.39
Other dierences that are considered when looking at
numerical models include the computational requirements,
such as power, speed, and memory. Since the 1980s, various
numerical models have been developed for the study of
atmospheric dispersion of denser than air clouds. Such
mathematical models can be classied as either box/top-hat
models or NavierStokes models.146
Box or Top-Hat Models. There are two types of box or tophat models: modied Gaussian models and similarity-prole
models, depending upon the complexity of conservation
equations that must be solved.39 The modied Gaussian
models are the simplest because the Gaussian equation is used
for the conservation of species while neglecting or simplifying
those for momentum and energy.146 The similarity-prole
models use simplied conservation equations with a
mathematical complexity of one dimension.37 Such simplicity
is achieved via averaging the LNG cloud properties across the
surface of the entire cloud or over the cross wind plane.37,39 To
regain the structural loss because of averaging, similarity proles
are used, therefore leading to quasi-three-dimensional solutions.
Examples of similarity-prole models include SCIPUFF,147
TWODEE,148,149 SLAB,150 HEGADAS,151153 DEGADIS,154
ALOHA,124,125 and GASTAR.155 Of these models, the most
commonly used are SLAB, HEGADAS, DEGADIS, and
ALOHA (Table 15).124,125,150152,154 ALOHA seems to be
the most widely used for safety engineering modeling
applications in industry because of its fast computational time
and reasonable accuracy.124 On the other hand, the ease of use
and fast computational time of DEGADIS and SLAB have led
to them been used in both the public and private sectors.37
NavierStokes Models. The NavierStokes models contain
the most physically complete description of the LNG
dispersion process and are constructed from three-dimensional
and time-dependent conservation equations of momentum,
mass, energy, and species.111,128,129,144,156160 Examples of
NavierStokes models that have been used for denser than air
modeling include FEM3, FEMSET, FLACS, HEAVYGAS, and
3568
3569
code
accessibility
applications
N/A
laboratory
eld
ground level evaporating pool, an elevated horizontal jet, a stack
or elevated vertical jet, and an instantaneous volume source
cannot model complex terrain or ow around obstacles
disadvantages
scales
typical dispersion simulations can be solved in a few minutes on

a IBM-AT class computer
advantages
laboratory
eld
prediction of concentrations in the low
ammability range
N/A
heat and water-vapor transfer from substrate

gravity slumping and dispersion
N/A
cannot model complex terrain or ow

around obstacles
simulation of instantaneous releases, time

varying releases, and continuous releases
laboratory
eld
dispersion downwind of a transient ground-level source or a vertical transition
plane with a near-eld jet model
cannot model elevated source ows
cannot model complex terrain or ow around obstacles
Burro and Coyote series tests

Eagle series
validations
lumped parameter approach
time-dependent model uses a quasi-steady-state description by using observers
downwind gas concentration decay is

consistent with experimental data
Burro and Coyote series tests
Maplin Sands
Thorney Island phase I trials
simulates a wide variety of denser than air

gas releases
DEGADIS154
simulates the steady-state or transient ground-level dispersion of a heavy gas

cloud
HEGADIS151153
similarity-prole models/modied Gaussian models
prediction of maximum downwind concentration is overpredicted; however, this

prediction then becomes more accurate as the downwind distance increases
wind tunnel and laboratory experiments
Maplin Sands eld tests
Thorney Island experiments
Goldsh 1 experiment
simulation of instantaneous releases, time varying releases, and continuous
releases
spatial averaging to treat the clouds as a steady-state plume,

transient pu, or a combination of both
Conservation equations of mass, momentum, energy and species
LFL was predicted to 15% for LNG spills
method
accuracy
simulates the atmospheric dispersion of denser than air releases
SLAB
principles
model
150
Table 15. Some Well-Known Similarity-Prole/Modied Gaussian Models for LNG Vapor Dispersion
N/A
designed for people

responding to chemical
accidents
emergency planning and
training
models pu and plume and

heavy gas dispersion
simulations can be
conducted within a few
minutes
cannot model
complex/changing terrain
cannot model particulate
dispersion
eld
performed well for all

validation tests
Prairie Grass
Kit Fox
simulates a wide variety of

denser than air gas
releases
Gaussian and heavy gas
dispersion models
ALOHA124,125
Energy & Fuels

Review
3570
predicts well the salient features in large

instantaneous releases
Thorney Island phase I trials
inappropriate K-theory submodel

cannot model in homogeneous vegetation cover
all in the same computational domain
cannot model jet releases, explosive sources, and
chemical reactions
eld
modeling of gravity slumping and spreading,

formation of doughnut-shaped cloud, and cloud
bifurcation in the case of LNG spill
submodel for treating aerosol eects in pressurized NH3 spills
phase-change model for humidity
tool for emergency response planning for liquid
Cl spills
N/A
disadvantages
applications
code accessibility
scales
3D
can model complex terrain and ow around
obstacles
can accommodate multiple instantaneous sources
advantages
accessible through user-dened functions
modeling of complex ows, including turbulence, heat

transfer and radiation, chemical reactions, combustion,
multiphase ows, and moving geometries
can model dispersion and ventilation in complex

geometries
N/A
eld
numerical
design and optimization of most industrial applications
can create user-dened functions

can model temporal and spatial gas dispersion, including
gravity slumping and time-dependent eects
more complex than other specic models, such as ADMS
long simulation times required
Burro series eld test

DEGADIS model
ADMS model
Falcon tests
3D
can model complex terrain and ow around obstacles
user interface that allows for customization and

automation using session les, scripting, and the CFX
expression language
modeling of complex ows, including turbulence, heat

transfer and radiation, chemical reactions, combustion,
multiphase ows, and moving geometries
eld
single solver, unlike FLUENT
more complex than other specic models

long simulation times required
user interface that allows for customization and

automation
3D
can model complex terrain and ow around obstacles
underprediction of concentrations at high elevations

because of over assumption of gas behavior
Coyote series trials
Brayton Fire Field tests
k for turbulence
other mathematical models available for modeling
processes, such as combustion or radiation
uses a coupled solver
overprediction of thermal impulses by a factor of 2,
however, still acceptable from a safety point of view
realized k for turbulence

multiple solving approaches
RSM predicted the turbulence kinetic energy to within

80% of the expected value, however, reduced to 50% at
ground level
gas concentration was within 80% of the experimental data
nite volume method

time-dependent equations of mass, momentum, and
energy
RANS equations for calculating the process of momentum
simulation of various ow scenarios
CFX144,156,160
nite volume method

time-dependent equations of mass, momentum, energy,
and species
RANS equations and Reynolds stress model (RSM) for
calculating the process of momentum
initiated in 1983
simulation of various ow scenarios
FLUENT111,158,159
eld
laboratory
primarily designed to model explosions in
oshore oil platforms
in the case of at terrain simulations, Gaussian

models can provide better accuracy than the
FLACS model
can accurately predict down wash from large

tanks
Prairie Grass
Kit Fox
EMU
MUST
3D
can model complex terrain and ow around
obstacles
distributed porosity concept is used to characterize the geometry while not reducing
calculation time as much as other models
underprediction of peak concentration values by a

factor of 2
accuracy
validations
70% of the time, its predicted values are within a

factor of 2 of the observed experimental data
modied Galerkin nite element method

time-dependent equations of mass, momentum,
energy, and species
Reynolds-averaged NavierStokes (RANS)
equations for calculating the process of
momentum
K-theory sub-model for turbulence
method
initiated in the 1980s

simulation of gas dispersion and subsequent
explosion
nite volume method
time-dependent equations of mass, momentum,
energy, and species
k for turbulence
FLACS
NavierStokes models
initiated in 1973
simulation of large heavier than air gas releases
FEM3
128,129
principles
model
157,162
Table 16. Some Well-Known NavierStokes Models for LNG Vapor Dispersion
Energy & Fuels

Review
Energy & Fuels
Review
Table 17. Eect of the Hole Size and Breach Diameter on Spill and Dispersions Using the FERC Modela
parameters
eect of hole size
time to empty
decreases dramatically with the

increases (asymptotes at 5 m
decreases dramatically with the
increases (asymptotes at 5 m
vapor dispersion
eect of ullage pressure

increase in the hole size; the pool radius also
hole size)
increase in the hole size; the pool radius also
hole size)
LFL
a
increases with the increase in the ullage pressure (not aect for
ullage pressure of >13.79 kPa)
Data were extracted from the study by Qiao et al.163
Table 18. Comparison of LNGMAP to Sandia Predictionsa
pool diameter (m)

spill duration (min)
distance to thermal radiation contours
(m)
a
decreases with the increase in the ullage pressure; the pool radius
also increases in size
hole size of
1 and 2 m2
hole size of
5 m2
1523% larger
1% larger
1020% larger
15% lower
0% dierence
1038%
lower
evolving LNG ow and vaporization rate along the trench. This

was then imported into CFD FLUENT to calculate the
dispersion of the vapors. For cloud dispersion from a trench
with perpendicular wind ow direction (relative to the trench),
the 1/2LFL extended 65 m (213.25 ft) downwind of the trench
and receded before the 10 min spill duration was over as a
result of the decreases in heat transfer from the trench. When
the wind direction was parallel to the trench, 1/2LFL was in the
order of 100 m (328.08 ft). However, the inclusion of a vapor
fence had little eect in containing the cloud; instead, it
increased turbulence and mixing, increased the dissipation, and
decreased the LFL by 30 m.
Considering LNG spill and dispersion into a sump or
impoundment, Ponchaut et al.168 solved the shallow water
equations (SWEs). Unlike the integral models, SWEs account
for the pool thickness distribution and the growth of the pool
with time, using a one-dimensional Fourier conduction
equation for modeling the heat transfer. The SWEs account
for pool thickness and hydraulic jump, which cannot be
modeled by integral models. The pool was predicted to reach
the sump wall in 6 s via SWEs, instead of 9 s via PHAST, which
is a common integral model. This resulted in a slightly lower
vaporization rate in the PHAST model. The eects of elevation
of the spill source and the eect of sump oor shape (conical
versus at) were also considered. An increase in the spill source
elevation led to a higher peak vaporization rate. However, a
conical sump oor would be more favorable in the case of
hazard mitigation in cases where the sump oor is close to
property boundaries, as peak vaporization was reduced.
3.2.4.3. Modeling of Complex Scenarios. The use of
SOURCE5 and DEGADIS is not accurate for predicting
ammable vapor dispersion from LNG spills into impoundments because vapor entrainment by wind is not considered.
The more accurate CFD package FLUENT was used for such
prediction, validated using Falcon test data.111 The comparison
between the CFD simulation results and the experimental data
clearly showed that the general shape of the cloud captures the
stable stratication that was measured in the experiments. Even
though there were some discrepancies, the CFD predictions
showed good agreement with the experimental data in terms of
the gas concentration prole. The eect of the impoundment
and source turbulence were also studied. As expected, the
impoundment partially contained the spill and limits its spread
in the downwind and lateral directions, while a decrease in
source turbulence led to highly stratied and undiluted with
vapor cloud within the impoundment. Therefore, mixing seems
to be dominated by the turbulence generated by the spill, and
the turbulent-driven entrainment aects the rate at which the
gas is dispersed and, as a result, the downwind concentrations.
Considering the eect of wind at a 9.3 wind direction,
Giannissi et al.171 simulated the Falcon test series to study the
eect of a two-phase jet release, using ADREA-HF, which is a
three-dimensional CFD code developed for pollutant and
Data were extracted from the study by Spaulding et al.126
successfully used to simulate three emergency LNG tanker

cases (both with and without ignition of the LNG pool): a
tanker continuing on course at 7.5 m/s (24.61 ft/s), a tanker
stopping within 3 min, and a tanker changing direction and
heading to the nearest location to ground.
Vilchez et al.167 deviced a dispersion safety factor (DSF)
based on the concept that, when LNG is spilled on land or
water, the visible cloud formed is due to the air reaching the
dew point temperature of water, resulting in condensation. DSF
can be calculated on the basis of a relationship between the
downwind length of the ammable cloud at LFL, XLFL (m), and
the downwind length of the visible cloud, XVIS (m): DSF =
XLFL/XVIS. With the help of DEGADIS, DSF was validated
against the experimental data of Maplin, Burro, and Coyote
tests. As humidity increases, DSF decreases, which leads to the
ammable region falling within the limits of the visible cloud
(safer situation). If humidity decreases, DSF increases, which
leads to the ammable region extending beyond the limits of
the visible cloud (more dangerous). If XVIS is known, XLFL can
be determined via gathering signicant sets of data.
3.2.4.2. Modeling of Complex Geometries. Most eld
studies, with the exception of the Falcon trials, of LNG
dispersion are in unobstructed conditions. In practice, in the
event of a LNG spill onshore, the receiving terminals are
designed to direct the spill toward a sump or impoundment
area where a LNG pool is formed.168 To study the dispersion of
natural gas clouds evolving from LNG pools in trenches,
Melton et al.169 deployed a re dynamics simulator, a CFD
model originally developed from modeling thermally driven
uid ow behavior during res in a 450 000 m3 capacity LNG
import terminal. The focus was on the eect of the substrate
material and the terrain, via varying the combination of
medium-density concrete, spill into trench and owing to
impoundment, and direct spill into impoundment. It was
evident that the inclusion of terrain eects increases turbulence
and mixing, resulting in shorter dispersion distances (but not
always being the case). It is important to dene where the LFL
is measured from because a pocket or isolated volume of
cloud can result in a larger LFL, an increase of 1030%
identied in the study.
Gavelli et al.170 also investigated LNG dispersion from the
trench, using a hydraulic model (not accounting for the
decrease in the mass ow rate) to analytically calculate the
3571
Energy & Fuels
Review
hazardous gas dispersion studies.172174 The simulations

considered two cases: the rst was a two-phase release, and
the second involved a mass ux from the water pond surface,
similar to that by Gavelli et al.111 Both source modeling
techniques were in good agreement with the experimental data.
However, the two-phase case was better at predicting the
maximum concentration and arrival time, and the mass ux case
resulted in a longer vapor cloud hold up within the
impoundment when compared to the two-phase jet case.
This is most likely due to the more realistic/accurate
representation of the eect of turbulence generated by the
two-phase source as the LNG impacts the water surface. The
eect of a 9.3 wind direction was noted to have a large
impact on the vapor cloud for the mass ux case, as compared
to the two-phase release case. This agrees with the study by
Gavelli et al.111 that, under low wind and stable atmospheric
conditions, source turbulence will dominate the mixing and
dispersion process. Modeling the spill as a two-phase jet release
appears to give a more realistic simulation and more accurate
results.
Qi et al.156 used the CFD code ANSYS CFX to simulate an
environmental setup similar to that used in the Falcon series
trials, with additional experimental data in the Brayton Fire
Training Field (BFTF), where no upwind obstacle was used. It
involved spilling LNG into a water surface in an impoundment.
The physical behavior of the LNG vapor cloud was wellcaptured. Unlike the studies conducted by Gavelli et al.111 and
Giannissi et al.,171 it was clear that both wind velocity and
turbulence were dominant in terms of vapor dispersion. This is
because, in the experimental work by Qi et al.,156 the wind
velocity was higher and the spill rate was lower (Table 19) than
changed the results from underestimated to actually matching

the experimental results.
Spaulding et al.126 found that, in comparison to the
stationary case, the moving motion of the vessel leads to
much lower concentrations of LNG vapor distributed over a
much wider area. This is because, as soon as the vessel stops,
the LNG pool will be restricted, reaching a maximum pool area
and for a longer duration, therefore resulting in a higher vapor
cloud concentration than if it continues moving. A similar
conclusion was also deduced for an ignited LNG pool, in which
an earlier stop of the vessel results in a more isolated thermal
radiated area, which lasts for a longer time. The modular
algorithm-based design of LNGMAP makes it able to
incorporate new algorithms. Currently, LNGMAP is more
computationally ecient than CFD models and, with further
advancement, might be condently used to provide more
realistic spill consequences.
There has been little focus on the eect of a substrate
temperature to the vaporization of LNG, especially for LNG
spills on water. Vesovic175 investigated the rate of vaporization
of LNG, by developing a model for heat transfer from water to
the LNG, with the assumption that LNG is released
instantaneously and the spreading pool forms a cylindrical
shape. The spreading occurs in the gravitationalinertial
regime, in which the height of the pool provides the driving
force, while the inertia of the ambient water provides the
resistance to spreading.175 Two scenarios were considered: rst,
the water temperature remains constant, and second, heat
arrives to the water surface by conduction only. It was found
that the maximum evaporation time is proportional to the
fourth root of the initial mass spilled. The derived expression is
similar to that by Raj and Kalelkar109 and Opschoor.110
However, the results are 10% higher, most likely because of the
dierent assumptions and models used. In the second case,
where heat transfer was assumed to be by conduction, the
results showed that the rate of vaporization of LNG depends
upon not only its composition but also the dynamics and
surface temperature of the water. Initially, the thermal inertia of
the thin vapor lm governs the heat transfer, and 1015 s after
the spill, the transition boiling will start, leading to the thermal
inertia of the growing ice layer governing the heat transfer.
3.2.4.4. Evaluation of Models. Various models for
simulating LNG dispersion relied on dierent assumptions
and empirical constants, which can cause uncertainty. The
uncertainty because of the lack of knowledge (i.e., epistemic
uncertainty) can be reduced over time as more data are
available, while the uncertainty because of variability within the
variables of a model (i.e., aleatory uncertainty) cannot be
reduced. Siuta et al.176 showed that two respective methods can
be used to evaluate the uncertainty of models, i.e., the fuzzy sets
(FS) theory for analyzing epistemic uncertainty and the Monte
Carlo (MC) simulation for analyzing the aleatory uncertainty.
These techniques have been successfully applied to evaluate the
prediction from the classical model (developed by ABS
Consulting),11 Gaussian dispersion (GD) model, and Britter
McQuaid177 (BMQ) model for the Maplin Sand and Coyote
eld experiments (see Table 21).
Ivings et al.178 proposed a comprehensive model evaluation
protocol (MEP) for evaluating the accuracy of various models.
The MEP was devised for the NFPA 59A standard. Four main
pillars were considered: a model evaluation questionnaire for
collecting essential information needed for model assessment, a
model validation database with sucient information for
Table 19. Test Conditions for Spills into Impoundments

Followed by Dispersion
156
Qi et al.
Gavelli et al.111
Giannissi et al.171
wind speed (m/s)
spill rate (m3/min)
1.82.2
1.7
1.7
0.2650.75
28.7
28.7
Table 20. Comparison of the Distance to LFL for LNG

Dispersion in BFTFa
LFL [5% (v/v) concentration of LNG]
experiment
ANSYS CFX simulation
a
at 0.3 m (height)
at 1.22 m (height)
8.6913.53 m
9.80 m
6.0913.47 m
13.47 m
Data were extracted from the study by Qi et al.156
that by Gavelli et al.111 and Giannissi et al.171 This would result

in lower source turbulence, allowing the wind to dominate the
ow process. The simulation results from ANSYS CFX were in
reasonable overall agreement with the experimental data (as
shown in Table 20); however, it underpredicts the ground level
gas concentration while underestimating downwind gas
concentrations at higher elevations. This suggests that CFX
overly assumes the slumping behavior of the gas and takes the
buoyancy change less into account. Qi et al.156 believed that
this might be due to an incomplete description of the heattransfer process during the simulation setup and other sources
of uncertainty, such as mesh size and source term turbulence.
The cruciality of a sensitivity study in model validation was
demonstrated by decreasing the maximum mesh spacing, which
3572
Energy & Fuels
Review
There has been limited recent research on LNG combustion

tests, which focused on the properties/governing factors of
pool res and vapor cloud res. Lowesmith et al.193 investigated
the nature of two-phased oil and gas jet res, considering the
eects of connement on jet res and their behavior with water
deluge, for simplifying hazard assessment. Correlations and
guidance for assessing jet res were successfully developed on
the basis of experiments from various studies, including those
conducted by Advantica Limited, SINTEF, and Shell, since the
1980s.
Studer et al.194 studied the properties (such as ame visible
length, radiation ux, and blowout) of large-scale methane/
hydrogen jet res. The CAST3M code was in good agreement
with the experimental results and later used to calculate some
safety-related quantities. It was found that, right after ignition, a
ame ball formed and the ame length was noted to decrease
with time because of a decrease in the mass ow rate. A set of
correlations was then developed for estimating the ame
length.194 However, the predictions using these correlations are
15% lower than those predicted by the correlations developed
by Lowesmith et al.,193 mainly because the released power of
the experiments in the study by Studer et al.194 is at the lower
limit of the correlation.
Advances have also been made to develop blowout stability
(ame stability) diagrams for dierent fuels. For example, Wu
et al.195 carried out a study that developed a blowout stability
diagram of the hydrogen ame diameter versus atmospheric
pressure. Lowesmith et al.193 developed one for natural gas
ames. When the blowout of the Studer et al.194 experiment
was imposed on the ame stability diagrams by Wu et al.195 and
Lowesmith et al.,193 good agreements were reached. It is clear
that the phenomenological model developed by Studer et al.194
is capable of predicting the ame length, blowout velocities, and
radiant uxes, while Lowesmith et al.193 presented good
background and correlations for jet ame hazard assessment.
3.3.2. Deagration/Detonation Experiments. Explosion
from fuel combustion can be classied as deagration or
detonation. In deagration, the fuelair mixture burns slowly at
Table 21. One of the Test Cases Comparing the Addition of

the Fuzzy and MC Methods to the Source Term Modeling of
the Classical Modela
fuzzy/MC compared to classical model
20% lower
15% higher
12.520.7% lower
8.5% lower
maximum release rate

spill duration
maximum pool radius
maximum evaporation time
a
Data were extracted from the study by Siuta et al.176
verication and validation (following the method developed

during the SMEDIS project179181), a set of qualitative and
quantitative model assessment criteria (stated in MEP182) for
qualitative assessment against statistical performance measures,
and a comprehensive model evaluation report as key output of
the MEP. MEP has been fully used and published for a CFD
model by Hansen et al.183 and for the evaluation of the integral
model PHAST that has now been approved for use in LNG
siting applications in the U.S.A.
3.3. Combustion. 3.3.1. LNG Pool Fire and Vapor Cloud
Fire Experiments. A variety of LNG re tests were conducted
over the past 3 decades. These re tests can be classied as res
of LNG pool and vapor cloud on land or res of LNG pool and
vapor on water, with the dierences mainly in the burn rate and
resulting ame height.37 In comparison to land LNG pool res,
res on water have a higher burn rate and ame height by as
much as a factor of 2 as a result of additional heat ux from the
water.37 On the other hand, surface emissivity power for both
water and land tests has been found to be similar for LNG pool
diameters up to 15 m (49.21 ft).37,40 Table 22 lists the up-todate large tests for LNG pool and vapor cloud res,
summarized by Luketa-Hanlin37 and extracted from previous
studies.184192 It was also found that pool res are aected by
the wind and can be inuenced by the shape of any
surrounding bunds, whereas surface emissivity is aected by
the pool size.38 The area of burn is also dependent upon the
type of ignition.40
Table 22. LNG Fire Tests on Water and Landa
surface emissive power

(kW/m2)
pool
diameter
(m)
ame length (m)

(L/D)
1.26.6
15
(eective)
2555 (2.84.4)
520
3.25.8
30
(eective)
80 (2.6)
water
14.628
13.517.1
not
measured
not measured
Maplin
Sands188
land
not
reported
NA
20
43 (2.15)
Montoir192
land
238
NA
35
77 (2.2)
spill
terrain
spill volume
(m3)
U.S. CG China
Lake
tests184187
water
35.7
Maplin
Sands189,191
water
Coyote190
study
spill rate (m3/

min)
pool re
vapor cloud
re
burn rate, 104 m/s

(kg/m2 s)
ame speed for

vapor cloud res
(m/s)
210 30
(narrow)
220 50
(wide)
178248
203 (avg)
not
measured
220 30
(narrow)
200 90
(wide)
137225
174 (avg)
150340
411 (0.180.495)
(calculated)
817 (relative to
cloud)
2.1 (0.0945)
(calculated)
4.56.0
not measured
153 (avg)
219 (max)
290320
(narrow
angle)
257273
(wide
angle)
350 (max)
NA
2.37 (0.106)
(measured)
3050 (near
ignition
sources)
NA
NA
3.1 (0.14) (measured)
NA
Data were extracted from the study by Luketa-Hanlin37 that was based on the literature.184192
3573
semi-empirical approach
inverse square law of radiation
provides a reasonably good agreement with experimental data, which is variable

because of the assumption made
dierent correlations and related factors have been developed on the basis of
ndings from experimental work and are selectively used on the basis of the
modeling taking place
simple and can easily produce results
method
accuracy
eld
re hazard assessment
N/A
N/A
applications
code
accessibility
ames with complex shapes, especially those arising from irregular-shaped pools,
cannot be modeled
cannot account for ame zones with object interaction
provides a reasonably good agreement with experimental data, which is variable

because of the assumption made
dierent correlations and related factors have been developed on the basis of
ndings from experimental work and are selectively used on the basis of the
modeling taking place
can account for interactions with water vapor and carbon dioxide in the
atmosphere
eects because of wind, such as ame tilt, can be modeled
energy radiated depends upon a number of factors and is not an intrinsic

property of combustion
assumes that all radiant energy freed at the ground level
assumes that the re is a small element at the ground level
cannot account for wind and obstacle eects
eld
scales
solid ame196202
initiated prior to the 1970s and simulates res as a geometric shape, usually on
the basis of wind condition
semi-empirical approach
representation of re as a geometric shape, usually cylinder
LNG re models
disadvantages
advantages
validations
initiated prior to the 1970s and simulates pool re as a point at the ground level
point source
principles
model
196
Table 23. Some Well-Known Models for LNG Fire Modeling
eld
laboratory
some NavierStokes models, e.g., FLUENT
and CFX, have accessible codes
initiated in the 1970s and simulates any type

of re with any shape
nite volume method
time-dependent equations of motion
k for turbulence
combustion and soot models can be
incorporated
provides good agreement with experimental
data if the correct physics is applied
via experiments, such as the Esso tests,
China Lake tests, Maplin sands tests, Gaz
de France, etc.
can capture complex ame shapes and
interaction of ames with objects
can model pool res, vapor cloud res, and
reballs
main disadvantage is the computational
requirement compared to other models
more complex compared to the other
models
NavierStokes111,160,196
Energy & Fuels

Review
3574
Energy & Fuels
Review
speeds in the order of 1 m/s, whereas in detonation, the ame

front travels at a shock wave at speeds in the order of 2000
3000 m/s (6561.689842.52 ft/s), followed by a combusting
wave that supplies it with energy.37,38 Since the reviews
published in 2007, no signicant deagration/detonation
experiments have been performed. As concluded previously,
detonations are very dicult to achieve in LNG res because
the primary component methane has a low reactivity and
deagrations can transition into a detonation explosion if there
is a connement or blockage.37,38
3.3.3. Modeling. A LNG pool re can transmit signicant
radiant heat to an object outside the re, with the heat ux
strongly dependent upon various parameters, including the
properties of re (e.g., size, shape, and geometry), surrounding
atmosphere (e.g., transmissivity), and object (e.g., location and
orientation).40 Technical issues that arise when modeling LNG
pool res are usually due to the scale of the re. As pool res
become larger, physical phenomena, such as oxygen starvation
in the center of the pool re, smoke generation, and reduction
in the emissivity power, become more important.38 Extrapolation from small res can lead to misleading results; therefore,
developing modeling techniques are in great need for studying
large-scale LNG pool res.
Typically, there are three approaches to modeling LNG res:
the point source method (simplest), solid ame method (next
level of complexity), and eld (or NavierStokes) method
(most complex/complete), as shown in Table 23. The point
source model can easily produce results; however, the
assumptions taken, such as neglecting wind and obstacle
eects, assuming that all heat is radiated at the ground level, can
lead to questionable accuracy.196 On the other hand, the solid
ame model accounts for wind and atmospheric condtions;
however, the cylindrical ame modeling approach can, at times,
lead to inaccurate results.196202 NavierStokes models as
previously discussed in section 3.2.4 are the most complete and
robust models that are able to provide the most accurate
results.111,160,196 The same conclusion stated in section 3.2.4
applies here. Fire modeling can only be improved if further
evaluation studies are carried. In doing so, issues, such as the
eects of radiation, surface emissivity, ame interaction with
objects, and pool re size, need to be studied extensively at the
eld scale to generate data for model evaluation/validation.
3.3.3.1. Overpressures from LNG Vapor Cloud Ignition.
Previous studies for LNG facilities focused on overpressures
from ignition of vapor clouds or pool res from a LNG spill.
These analyses were usually performed with simple methods,
such as the TNO multi-energy method (MEM) or the Baker
StrehlowTang (BST) method.203 However, these models are
inadequate for near-eld calculations, are not sensitive to the
degree of congestion, and cannot account for obstacles. Gavelli
et al.203 used the advanced modeling tool FLACS to study LNG
vapor cloud explosions in two spill scenarios (a worst case and
a realistic case) occurring from a 140 000 m3 moss-type LNG
vessel moored at an ooading pier. In the worst-case scenario,
23 000 m3 of stoichiometric gas cloud was spilled, and in the
realistic scenario (one of many possible encounters), LNG was
spilled at the loading rate of 5000 m3/h for 10 s and then
dispersed for 3 min by wind. In both cases, the resulting
overpressure was well below the minimum threshold for human
injury (16.5420.68 kPa) or facility damage (101.35 kPa).203
Clearly, the low reactivity of methane results in overpressures,
although such overpressures appear to occur within a safe
range, even at an applied safety factor of 2 to account for

uncertainties in modeling.
Abe et al.204 studied the blast eects of the liquid oxygen/
LNG fuel mixture using the hypercode ANSYS AUTODYN,
considering four cases, including two large-scale liquid fuel
explosion of 17 and 5 tonnes, two equivalent scale gas
explosions (13 and 3.9 tonnes), and 5.8 g of gas explosion.
Assumptions, such as wind-free and a steady-state environment,
were assumed. From the large-scale simulations, it is clear that
overpressures are generated. A secondary shock wave was also
noted (believed to be the reection of the shock wave from the
ground). As the overpressure moved further away, the constant
volume combustion pressure prole formed is nearly identical
to that of a detonation. For the liquid case, blast eects were
noted to be 4050% higher at any distance than those for the
gas cases. The properties of the blast wave at relatively long
distances were also believed to be independent of ignition
conditions. The maximum overpressure recorded was 65 kPa
[9.42 pounds per square inch gauge (psig)] at 100 m from the
explosion point, which is well above the safety thresholds for
human (16.5420.68 kPa) but below that for facilities (101.35
kPa). However, it is important to note that the study did not
consider wind disturbances and weather conditions that can
have a signicant eect on the energy of the overpressure.
3.3.3.2. LNG, LPG, and Gasoline Modeling. The consequence analysis methodology developed by the FERC has
become the standard method of modeling LNG release, spread,
and pool res.164 Vaporization can occur from two mechanisms: heat transfer from the spill surface and heat transfer
from a ame. In both cases, FERC methods assume constant
heat ux, i.e., 85 kW/m2 (0.167 kg m2 s1) for a non-burning
LNG pool and 143 kW/m2 (0.282 kg m2 s1) for a burning
LNG pool. On the other hand, radiation from a ame is
modeled by means of the solid ame model.
Johnson and Cornwell164 performed studies to evaluate the
eectiveness of the FERC method in simulating spills,
vaporization, and pool res of material other than LNG
(such as LPG and gasoline), considering two release scenarios.
In the rst scenario of an equal release volume of all three
liquids, the burning gasoline pool reached the largest diameter,
mainly because of its low vaporization rate, although the
impacts from an expanding burning pool were nearly identical
for all liquids. In the second scenario of the respective ship
containment system of the dierent liquids, the gasoline pool
was still larger than the LNG pool, although the total volume of
LNG spilt was 5 times larger. Therefore, LNG release and
ignition will not necessarily produce a signicantly larger
radiant impact than a smaller release of a less volatile material,
such as gasoline, for typical cargo containers under similar
release conditions. The FERC spill, vaporization, and burning
modeling method for LNG can also be used for other material,
such as LPG and gasoline, if the correct material properties and
physical data are used.
4. SAFETY
Accidental release to the environment poses a risk and requires
special care when handling. Marine vessels, unloading facilities,
land storage tanks, and processing facilities are considered to be
the key areas where risks are to be quantied.205 The key issues
of siting LNG terminals include (a) no exclusion zone siting for
spills on water, (b) misleading or erroneous specications of
input parameters for exclusion zone modeling, and (c) the use
of unreliable models for exclusion zone calculations.41 While
3575
Energy & Fuels
Review
Table 24. Comparison of Dosage, Heat Flux, and Duration Needed To Attain First-, Second-, and Third-Degree
Burns206,207,209212
rst degree
dose
[(kW/m2)4/3 s]
Perez et al.
207
115
Raj206
TNO209
130
others210212
118
heat ux (kW/m2)
0.971.63 (60120 s)
2.733.14 (2530 s)
up to 4 (60120 s)
up to 5 (2530 s)
1.061.78 (60120 s)
3.003.44 (2530 s)
0.981.66 (60120 s)
2.793.2 (2530 s)
second degree
dose
[(kW/m2)4/3 s]
third degree
heat ux (kW/m2)
dose
[(kW/m2)4/3 s]
heat ux (kW/m2)
250
1.732.92 (60120 s)
4.905.62 (2530 s)
421
2.564.31 (60120 s)
7.258.32 (2530 s)
390
2.424.07
6.857.85
1.662.82
4.695.44
421
2.564.31 (60120 s)
7.258.32 (2530 s)
236239
(60120 s)
(2530 s)
(60120 s)
(2530 s)
The thermal eects of res and explosions were studied by

Perez et al.,207 during which new Probit equations were also
derived to assist in consequence analysis. Existing Probit
equations for thermal eects were devised by TNO209 in 1989
for rst, second, and third degree burns where TNO used
assumptions to derived the Probit equation for second degree
burns due to insucient data. Perez et al.207 derived new Probit
equations for calculating thermal dose, benchmarking against
the values from TNO and existing literature.210212 As shown
in Table 24, the Probit equations derived by Perez et al.207 are
in better agreement with the empirical data in the literature
than those by TNO. In addition, the results suggest that a heat
ux of 5 kW/m2 will lead to second-degree burns within 30 s.
Such values (0.971.63 kW/m2 for 60120 s and 2.733.14
kW/m2 for 2530 s) are apparently not in agreement with
those (4 kW/m2 for 60120 s and 5 kW/m2 for 2530 s) by
Raj.206 This suggests the need to carry out more experimental
testing than can lead to more conclusive results.
4.2. Facilities. In the past, the relatively low demand for
natural gas storage made it possible to locate such installations
away from densely populated areas. However, it has become
clear that there are advantages in bringing storage facilities
closer to points of use. When studying the risks involved in a
LNG facility, it is therefore crucial to deal with scenarios that
are213 (a) feasible in the course of operation, (b) feasible by
interaction with the surroundings, (c) highly improbable but
theoretically possible, and (d) intentional damage to the
installation with a view of restricting the functions or
endangering the surroundings. However, most safety and
security documentation barely cover these scenarios. Scenarios
ac can be simulated using typical mathematical models, while
for scenario d, modeling of the entrance, including the inuence
of the protective measure should be taken into account. As a
result, Bernatik et al.213 devised a method adapted from the
critical evaluation for national monuments under the National
Infrastructure Protection Plan214 and applied it to a LNG
facility. This safety assessment method allows for a simple
mathematical equation that considers the type of infrastructure,
casualties, economic impact, length of outage, impact on other
sectors, and environmental impact that would ensue from a
LNG hazard. Coupled with a logical ow process (which
involves setting security goals, identifying assets, risks to assets,
and implementing protective programs), this method can
determine the eectiveness of a protective measure for a LNG
facility.
4.2.1. Accident Modeling/Mitigation. Continuous monitoring and implementation of appropriate actions are essential to
prevent, control, and mitigate unfavorable consequences of
there have been no major incidents in this industry with the use
of the current system, recent studies have focused on safety
measures for humans, LNG facilities, and ships, because of
rapid increases in production and use of LNG, potential
terrorist threats, and public condence in LNG safety.
4.1. Human. The recent studies regarding human safety
focus on thermal hazards from LNG res rather than the
cryogenic eects of a LNG pool/vapor cloud or the
asphyxiation from a LNG vapor cloud. Current regulations,
such as the U.S. DOT Regulations, 49 CFR, part 193, and the
NFPA 59A standard,173 require that a safety distance is set at a
radiant heat ux of 5 kW/m2. Unfortunately, there has been no
signicant research to determine whether this criterion is
correct. On the basis of this regulation, Raj206 conducted some
studies on the eectiveness of ordinary civilian clothing to
withstand a 30 s exposure of 5 kW/m2. It was found that
ordinary civilian clothing provides a reduction of radiant heat
ux by a factor of 2 and any object that intervenes (even a
newspaper held in front of the person) can lead to a reduction
by a factor of 4. A person with ordinary civilian clothing can
withstand a heat ux of 5 kW/m2 for 2530 s and 4 kW/m2 for
60120 s without severe pain or suering any injuries.
However, repeated exposure without cooling will reduce this
tolerance. Raj206 then concluded that the public should be safe
with the current hazard denition, which opposes the idea
posed by Havens and Spicer41 that a heat ux of 5 kW/m2 will
lead to second-degree burns within 30 s. It is important to note
that the weather, number of clothing layers, age, sex, and health
of people can all play a role in determining the radiant heat ux
that can be withstood.
People can easily run from a LNG pool re, but in the case of
a jet re, ash, or vapor cloud explosion (VCE), the possibility
of humans protecting themselves decreases. Probit equations
are dimensionless numbers that correlate directly to the
percentage of the population aected and considered as a
simple and fast method for consequence analysis.207 VCE
produces two damaging properties: pressure waves and thermal
eects. The damage because of pressure waves was investigated
by Alonso et al.,208 who uses Probit equations and explosion
characteristic curves for human damage consequence analysis,
via comparisons between the Probit results and the explosion
characteristic curves. Damages, such as eardrum rupture, death
because of skull fracture, and lung damage, can then be
evaluated and determined for certain distances from the
explosion center. This method proposed by Alonso et al.208
provides a simple and fast consequence analysis that allows for
damage to be linked to explosion energy and distance from
explosion.
3576
Energy & Fuels
Review
coupled with intuitionistic fuzzy set can be used to assess the

reliability and nd the component(s) of the ESD that require
the most attention. Even though human expertise and
knowledge of operations and maintenance of the ESD is
required to collect failure information, this method has proven
to be able to locate equipment whose improvements will greatly
increase the reliability of the ESD and resulting safety of the
facility.
Looking from a dierent perspective, shutdowns are also
periods during which maintenance takes place. To prevent the
loss of potential revenue during routine shutdowns, Keshavarz
et al.224 determined a risk-based shutdown strategy to minimize
the number of shutdowns while maintaining minimal risk for
the expected lifetime of the plant. This method is superior to
traditional methods because it uses the Weibull distribution
rather than linear failure probability methods and can,
therefore, accurately and easily use failure history data. The
method can be used on both standby and redundant systems
and takes into account numerous parameters, including the
current costs of the product. It was successfully tested on the
PPMR process of APCI, showing that active redundancy
reduces the risk for most short goal times. However, as the goal
time increases, there is a shift toward the standby strategy for
reducing risk, suggesting that standby, active redundancy, and
preventative maintenance are crucial to preventing operational
risks.224
In the case that a pool re occurs, expansion foam,
particularly high-expansion (HEX) foam, can be eective in
controlling the LNG pool re225 by blanketing the LNG pool
surface, as a result preventing oxygen from reaching the re and
also acting as an insulator by reducing re radiation from the
pool re. A HEX foam application rate of 10 L min1 m2 is the
most applicable, and the re control time can be reduced with
an increase in the application rate. The location of foam
generation units and design are all crucial factors that should be
considered. It is also important that the units are available and
operational at all times. Further investigation226 showed that
Foamglas is also able to signicantly suppress LNG pool res
and is comparable (if not more eective) to expansion foams.
However, Foamglas is not eective at LNG vapor mitigation.
Mitigation is also possible via CFD-based dynamic
simulations of hazard consequence reduction. Sun et al.227
examined both vapor dispersion by water curtains and re
mitigation by HEX foam using CFD validated against the wellknown Falcon tests for vapor dispersion and the Montoir tests
for pool re radiation. Water curtains can reduce vapor cloud
distance by approximately 6184%, while HEX foam
application leads to a re control time (time to reduced heat
radiation by 90%) of 130 and 55 s, which are comparable to
100 and 60 s from the respective experiments.225 This clearly
demonstrates the great capability and applicability of CFD to
hazard mitigation studies.
4.2.2. Other Safety Considerations. The historical safety
record of LNG facilities has been excellent, mainly because of
the international standard NFPA 59A design codes followed by
the designers, constructors, and operators. Taylor228 presents a
summary of the processes (NFPA 59A) that would be followed
for site location and equipment placement. On the other hand,
Raj and Lemo229 discuss the risk evaluation approach of the
2009 edition (latest edition) of the NFPA 59A, compare it to
the risk process of Europe, and present an example of how the
NFPA 59A can be used. However, the NFPA 59A is still
lacking, especially on criteria regarding radiant heat. It is clear
LNG production and use. LNG facilities have a good safety

record, and as a result, failure data on LNG systems are sparse.
Because of this limitation in failure data for risk analysis, Yun et
al.215 devised a Bayesianlayer or protection analysis (LOPA)
method for obtaining risk with less eort and time compared to
most other methods. The Bayesian estimation allows for
generic failure data (re or explosion) from other industries, to
be coupled with likelihood information from LNG industry, to
determine failure data to be used in the risk analysis. A hazards
and operability (HAZOP) study was incorporated, including
and Poisson distribution for prior and likelihood information.
This methodology was validated against existing data and
studies to show that it is a powerful method, especially for an
industry (such as the LNG industry) where failure data are
sparse.
Considering the cryogenic properties and ammable/
explosive behavior of LNG, Rathnayaka et al.216 recently
developed and validated a new accident modeling approach
that also incorporates a HAZOP based on system safety
identication, protection, and prevention (i.e., SHIPP) for
applications to a LNG facility. SHIPP is a safety assessment
methodology that describes the steps of process safety
assessment and provides a guide to possible improvement at
every step of the accident process. Following the process
accident scenarios identied using HAZOP, the accident
process follows three steps: initiation, propagation, and
termination. To develop a predictive model for the occurrence
of abnormal events, a Poisson probabilistic distribution was
incorporated. This method was tested to be reasonably eective
using realistic data from a LNG facility,216 with an overestimation of up to 14% for the rst three time intervals and
then underestimations (with decreasing accuracy) as time
intervals increase.
On the other hand, Parihar et al.217 devised a method for
consequence analysis at deep-water facilities. The method
consists of the use of analytical models to describe the
dynamics of unconned spills, semi-empirical models for pool
res, and CFD for vapor cloud dispersion (this information was
previously discussed). Such information is routinely needed for
independent risk assessment studies of proposed facilities, and
the methodology has been validated and scrutinized against test
data.
Li and Huang218 developed a more mathematical-based
approach for assessing the level of risk and damage to facilities
and humans at dierent radii from a LNG re or explosion. The
analysis consists of using the DOW method for re risk analysis
and a VCE model. It is generally accepted that the likelihood of
a boiling liquid expanding vapor explosion (BLEVE) occurring
in a LNG tank during a LNG pool re is low because of safety
measures currently in place.219 However, BLEVE should not be
disregarded because one BLEVE event did take place in Spain
(June 2002), involving a road LNG tanker accident.220 The
BLEVE event is usually followed by a reball, from which
approximately one-fourth of its energy is released as
radiation.221 Li and Huang218 included BLEVE for risk analysis
by making use of the BLEVE model proposed by the
International Labour Organization.
In the event of a leakage or emergency, emergency shutdown
(ESD) systems in the LNG plants can be used for stopping
pumps, etc., and isolating the leakage automatically.222
Therefore, it is crucial that this ESD system is reliable and
always operational to mitigate hazards and prevent escalation.
Cheng et al.223 devised a method in which fault tree analysis
3577
Energy & Fuels
Review
Table 25. Distribution of Known Relevant LNG Accidents232
that, with the inclusion of consequence modeling, the process

of determining the most cost-eective and safest LNG plant can
be accelerated.
Downstream safety in design is for chemical and renery
plants and follows a deterministic approach, while upstream
safety in design is for oshore platforms and follows a riskbased approach considering all risks. LNG plants are
categorized as midstream safety in design. Because of the
increasing number of developments, the modularization
concept is being widely applied to onshore LNG plants,
leading to the plant design features becoming similar to that of
oshore plants. Tanabe and Miyake230 compare the two
approaches and propose a safety design approach for LNG
plants, concluding that the most ecient way to ensure the
safety integrity of the design is to have a good design basis and
minimize uncertainty. Because of less exibility of the module
design, a deterministic approach appears to be more applicable
for an onshore modularized LNG plant and the approaches of
using a safety criteria design basis matrix and a hazard-design
logical relation tree are eective.230
Safety measures that are not usually fully developed at the
early stages in current design practices are the emergency
systems, the modularized plant and layout, and the tank
selection. Because of this, Tanabe and Miyake231 discuss an
approach to enhance safety design application at the concept
denition phase. The proposed method is a combination of a
deterministic and risk-based approach, which can overcome
diculties and restrictions because of limited information at the
early stages of a project and will yield schedule and cost
benets.
4.3. Ships. Ships are the major carrier of LNG, and the
current world eet consists of 400 ships in size of 120 000
250 000 m3.232 There has been considerable interest in the use
of LNG for ship propulsion,233 such as the dual-fuel steamturbine mechanical (DFSM) propulsion, dual-fuel gas-turbine
electric (DFGE) propulsion, dual-fuel diesel electric (DFDE)
propulsion, and diesel mechanical propulsion with reliquefaction (SFDM+R).234,235 Traditionally, the DFSM propulsion has
been the main system used but has a low eciency,234 with the
increase in LNG carrier size, more suitable propulsion systems,
such as the DFGE propulsion system, are required. However,
assessment on the applicability, availability, and safety of these
systems requires the considerations of numerous factors. For
example, in terms of LNG vapor hazards for a gas leak in the
DFGE compressor room, CFD simulations (using FLACS)
showed that moving gas detectors to within the compressor
room will eliminate a 50 s delayed detection time.234 It was also
concluded that there is no benet in having a second ventilation
fan within the compressor room for leaks larger than 1.27 cm,
and equipment outside the compressor room will not be
aected by any re that breaks out, unless the equipment is on
the roof of the compressor room.234
The major contributors of risk associated with LNG shipping
can also be determined via high-level risk assessments for LNG
carriers. One approach is based on historical data of LNG
accidents, published damage statistics, and expert judgments for
events, such as collision, grounding, contact, re and explosion,
and other events during loading and unloading.232 It was found
that collision, grounding, and contact were noted together to
account for 90% of the total incidents, while LNG containment
failure appears be among the highest contributors to risk (Table
25).232 Therefore, given the good safety record of LNG
shipping, it is reasonable to focus more on ensuring that the
accidents
(number)
accident category
collision
grounding
contact
re and explosion
equipment and machinery failure
heavy weather
events while loading/unloading
failure of cargo containment
system
total
19
8
8
10
55
9
22
27
158
frequency
(per shipyear)
6.7
2.8
2.8
3.5
1.9
3.2
7.8
9.5
103
103
103
103
102
103
103
103
5.6 102
LNG containment systems are designed to an acceptable

standard, with cryogenic reliability and high thermal insulation
performance for safe and ecient transport of LNG.
4.3.1. Cryogenic Performance of Insulation and Joints.
The International Maritime Organisation236 and the U.S. Coast
Guard237 specify requirements for pressure relief valve sizing on
liqueed gas carriers. An example for the adequacy of such
requirements is the pressure relief systems to prevent
overpressure failure of a current design, polystyrene-foaminsulated, liquid-full moss sphere exposed to a LNG re. A re
from a 1 m2 hole size could have a diameter of more than the
length of the ship, could exceed the maximum height of the
ship (above the water line), and might burn for 40 min. Havens
and Venart238 conducted a one-dimensional thermal analysis of
the moss sphere (steel weather cover > air gap > aluminum foil
> polystyrene insulation > aluminum tank wall) using
MATLAB for ame temperatures of 1300, 1400, and 1500 K.
The failure times of the dierent parts are shown in Table 26.
Table 26. Thermal Analysis of the Moss Spherea
steel cover
aluminum foil
polystyrene insulation
a
melting temperature
(K)
failure time (s)
810
873
510
100180
150330 after steel cover
95225 after aluminum foil
Data were extracted from the study by Havens and Venart238
At a melting rate of 3 cm/min (1.18 in./min), the polystyrene

insulation would be fully melted in less than 10 min. Under
such conditions, a moss sphere with non-re-resistance
polystyrene foam protected from a re by only the steel
weather cover could rupture and lead to overpressure if the
ames remained for periods longer than 10 min. It was also
found that the relief valve equation (by the U.S. Coast Guard)
would result in an underprediction of the required relieving
capacity by a factor of 9.
The membrane-type LNG tank has two insulation barriers to
prevent LNG boil-o and two tightness barriers to protect the
hull from low-temperature brittle damage as a result of LNG
leaks.239 The IGC code suggests that the secondary barrier
should be designed in a way to contain any LNG spills for at
least 15 days.236 As a result, Bang et al.239 investigated the
cryogenic performance of the adhesively bonded secondary
barrier that is composed of an aluminum sheet and a stainlesssteel foil (Figure 8) considering the eect of surface treatment,
such as mechanical treatment, ame treatment, and sulfuric acid
etching on the stainless-steel foil. The sulfuric-acid-etching
method showed the highest bond strength (Figure 9), because
3578
Energy & Fuels
Review
Figure 8. Schematic diagrams of the LNG cargo containment system composed of primary and secondary barriers: (a) overall drawing, (b) enlarged
view of the corner of the top bridge pad, and (c) insulation panels with level dierence (adapted from the study by Bang et al.239).
incomplete llet formation), more time and amperage would be

required.240
Improper bonding or defects from bonding of containment
system layers can lead to the formation of cracks and/or leaks.
Leakage may also occur if the primary barrier fails and the
adhesive joining of the secondary barrier is not thoroughly
sealed. Kim et al.241 investigated the leakage characteristics of
composite barriers to devise an improved curing method while
limiting residual thermal stresses. It was found that the leakage
rate decreases with an increase in the curing temperature and
with an increase in the curing pressure from the conventional
curing temperature of 25 C to a high curing temperature of 60
C. However, increasing the curing temperature also increases
residual thermal stresses. An adaptive curing method was
created, in which the curing temperature varies based on the
solidication of the adhesive, and was shown to decrease the
leakage rate by up to 53% without increasing the thermal
residual stress, as compared to the conventional 25 C curing
temperature.241
Eorts were also taken by other researchers242 to investigate
the crack retardation ability of the composite insulation panel
(Figure 8) with glass reinforcement. For ecient transportation, the insulation foam should have very low thermal
conductivity for high thermal insulation performance, which
limits the density of the foam to less than 120 kg/m3
[polyurethane foam (PUF)]. Cracks can propagate from this
foam because of cyclic thermal loads and low fractural
toughness and can lead to failure of the aluminum sheet,
resulting in a leak and possibly an explosion. Estimation based
on the simulation using nite element software ABAQUS 6.10
and experimental data suggest a maximum strain of 3.5% that
would lead to failure of PUF. E-glass fabric (thickness, 0.23
mm; density, 0.25 kg/m2) was found to be superior as a
reinforcement of PUF because it improved the failure strain to
14.3% (over 4 times larger than without the reinforcement).242
Yu et al.243 also investigated the thermal eect of reinforced
polyurethane foam (RPUF), including its eect on the
Figure 9. Eect of the surface treatment on bond strength (data were

extracted from the study by Bang et al.239).
of nanosize pores that may improve interlocking at the joint.

However, the ame treatment method appeared to be the most
applicable and productive in the industry, with the application
of an optimum treatment time of 57 s. After the bonds were
subjected to varying thermal-shock loadings, Bang et al.239
suggested that a 0.15 mm adhesive would be a good control to
work from, to create robust adhesive joints for the LNG ship
lifetime of 20 years.
Bonding of supporting plates in the primary barrier is usually
performed using spot pacing, drilling and tapping, and riveting.
Stud welding is a technique that can reduce cost by saving labor
and material, but improper processes may cause signicant
defects and failure of the containment system. Lee et al.240
carried out numerous experiments to device an optimal stud
welding condition for LNG ship containment systems. Via the
analyses of bead shape, metallography of the weld, and tensile
and bending tests, an optimal condition for stud welds would
be a weld current of 520540 A, a lift of 1.5 mm (gap between
the stud bolt and base metal surface), and weld time of 0.4 s,
and in the case of a cold weld (because of low llet height and
3579
Energy & Fuels
Review
production cost. The insulation foam is limited to a density of

less than 120 kg/m3; hence, it is also limited to a glass ber
volume fraction of 0.5% because of thermal properties of the
glass ber. Reinforcing the PUF increases the production cost
while decreasing thermal insulation (because of the high
thermal conductivity of the glass ber).243 Instead of
reinforcing PUF with glass bers of random sizes, Yu et al.243
compared a normal PUF to RPUF and to others reinforced
with single-sized glass bers of 0.3, 0.7, and 1 mm. As shown in
Table 27, selective reinforcement, especially with 1 mm glass
bers, will increase reliability, thermal insulation performance,
and manufacturing eciency at an optimum cost.
Table 27. Tensile Failure Strength and Safety Factor of
Insulation Boards with Respect to Dierent Types of
Reinforcementsa
PUF reinforcement
failure strength (MPa)
safety factor
none
0.3 mm glass bers
0.7 mm glass bers
1 mm glass bers
random glass bers
1.8
4.5
7.4
11.4
4.4
1.0
1.4
1.6
2.4
1.7
Data were extracted from the study by Yu et al.243
4.3.2. Structure. Transportation of LNG is highly dependent

upon LNG tanker technology. As discussed in section 2.3, the
two mainstream tank systems are the membrane and spherical
tank systems (see Table 12), and recently, the membrane tank
system has been widely adopted because of its large capacity.
The membrane system has a thermal-insulating layer and a
metallic membrane covering to maintain liquid tightness
without any leakage (as shown in Figure 8). The load of the
LNG cargo is transmitted to the thermal insulating layers
through the membrane (primary barrier). It is likely that
increased sloshing load in large-capacity LNG cargo containers
may lead to failure of the corrugated section of the membrane.
Chul Kim et al.244 employed both experimental methods and
nite element analyses (ABAQUS, version 6.7) to evaluate the
pressure resistance of the conventional stainless-steel membrane, via applying pressure loads to a 304L stainless-steel
corrugated membrane that is in most membrane tanks of MarkIII. Chul Kim et al.244 found the local yielding and plastic
buckling loads of the three types of corrugations (large
corrugation, small corrugation, and large corrugation with
reinforcing ribs) under the conditions of both symmetric
loading and asymmetric loading (Figure 10). The addition of
reinforcing ribs only seems to move the stress concentrations
toward the ribs and results in larger deformations than the
standard corrugations without ribs. Therefore, a new reinforcement method was then developed that inserts 2 mm thick
6061-T6 aluminum pipes into the corrugations. With testing, it
was conrmed that this new reinforcement method resulted in
no permanent deformation of the corrugations or membrane,
as a result increasing the pressure resistance of the membrane
without compromising exibility.
In addition to the membrane with rigid supports (rigid
membrane), it is also important to model the membrane with
exible supports in both static and dynamic cases. Finite
element analysis (software ABAQUS) for such cases showed
that the responses from the dynamic case signicantly exceed
those of a rigid-supported insulation, demonstrating the need to
include both the exibility and dynamic loading conditions in
Figure 10. Yielding and buckling pressures of corrugations under

symmetric loading: (a) symmetric loading and (b) asymmetric loading
(data were extracted from the study by Chul Kim et al.244).
analyses for LNG membrane systems.245 This is particularly

important because sloshing ow in ship containment systems is
excited by ship motion and the sloshing motion itself can also
aect the ship motion. However, up to date, most studies are
limited to linear studies of both ship motion and/or sloshing.
An example is the study by Pistani and Thiagarajan,246 in which
impact pressures from a sloshing experiment was measured for
a two-dimensional tank, with discussion on the problems that
would be encountered in sloshing experimentation and how to
overcome those problems. Mitra et al.247 developed an analysis
method that coupled both nite dierence and time domain
panel approaches for three-dimensional sloshing and ship
motion capable of predicting both sloshing heights and the
hydrodynamic pressures on the containment walls for further
analysis. The method was veried against the data from existing
scaled experiments. As expected, the sloshing height increases
with wave height, and the waves, current, and wind acting at
90 to the direction of the ship motion result in the highest
sloshing values. However, the eects of the waves were
signicantly more dominant (by up to 33%) than that of the
wind and currents. The study further illustrates the need to
include dynamic eects on studies of sloshing within LNG ship
containment systems.
3580
Energy & Fuels
Review
result in a decreased lower ammability limit, and terrain eects

can lead to an increase in turbulence. Therefore, the use of
vapor fences for LNG dispersion control can be eective.
However, new hazards can develop because of the buildup of
vapor within the vapor fence. The past studies were also on
small scales; therefore, future experimental investigations are
required at medium/intermediate-scale experiments (200 m3)
for reducing epistemic and aleatory uncertainties. In addition,
the scientic basis of many models has been questions,
especially integral models, so that vapor dispersion simulation
using eld models (NavierStokes models) is highly desired
because these models provide the most complete representation of the fundamental uid dynamics.
Combustion. There have been several studies on LNG res
with the aim of determining the thermal radiation, burn rate,
and ame speed. Recent studies have shown that, in the case of
LNG vapor cloud ignition, the low reactivity of methane (the
main component of LNG) results in overpressures within a safe
range, but the likelihood of high overpressures should not be
neglected. It is also concluded that LNG releases and ignition
will not produce a signicantly larger impact than a smaller less
volatile material, such as gasoline or LPG. A key gap in the eld
is the lack of large-scale LNG re experiments, which should be
well-designed and conducted with correct instrumentation in a
controlled environment, enabling easier determination of the
key parameters aecting the behavior of the ame. As such,
more experimental data are essential to the validation of the
ame model.
5.3. Safety. The majority of studies on LNG deal with
safety to some extent, with some considering the eects of
LNG hazards on personal and facilities. Various modeling
techniques have been developed for safety assessment of both
personnel and LNG facilities, and these techniques may be
eective for studying realistic scenarios. Mitigation techniques,
such as ESD, routine shutdowns, and the use of HEX and
Foamglas, have also been studied, with the aim of improving
safety in the most cost-eective way. With regard to LNG
carriers, the ame treatment of the aluminum sheets in the
insulation barrier of LNG ship containment systems is now
known to lead to best bond strength. With the assistance of
stud welding of support plates, adaptive curing of the adhesive,
and glass reinforced bers, the insulation foam is believed to
provide the best crack retardation ability, leakage prevention,
and thermal insulation.
There are at least several aspects that require future R&D.
First, it is still unclear whether the current criteria for exposure
to a radiant heat ux of 5 kW/m2 is suitable to prevent seconddegree burn injuries. Therefore, further research that also
considers the duration of exposure and the physical, thermal,
and physiological properties of the receptor need to be
conducted to determine whether this radiant heat ux range
is suitable. Second, assessments of LNG safety issues need to
integrate the credible inputs from the predictions of modeling
with relevant complex conditions and geometries tailored to the
underlying applications. Third, further improvements are
required in the methodology of studying the LNG containment
systems of ships. Particularly, the use of dynamic simulations/
studies is highly recommended because the eect of dynamic
loading (such as sloshing) can exceed those of static-loaded
cases.
5. FUTURE RESEARCH PERSPECTIVES

With the increasing demand on LNG as an energy source,
increasing interest has led to numerous studies regarding the
risks involved in LNG production, storage, handling, and
transportation. The future research perspectives are briey
summarized below.
5.1. LNG Production. Recent studies regarding the LNG
production process focused on the LNG liquefaction and
regasication processes, with dierent congurations proposed
and several implemented. Of those that were implemented,
optimization was focused on improving eciency. Future
research for maximizing economic prots is required via capital
cost reductions to obtain a more complete, economically
sound, and ecient LNG plant. In addition, the optimization in
the past studies was largely made via the use of a control
variable of the regulator control layer, a method of selecting the
control variables based on trial and error. Future research is
required to develop a self-optimizing complete control
structure design for exibility and economic considerations,
although the method requires increasing computational load.
The use of dynamic models for eective control design
structures is highly recommended. Variations in design
constraints can aect the whole design structure. The use of
dynamic models will be especially benecial for complex
processes. In the meantime, further research is needed for the
improvement of major equipment, such as multi-stream heat
exchangers, drivers, and expanders, and their integration for
overall process synthesis and optimization are desired for
delivering optimal solutions.
The recent developments include various advances in various
aspects of the LNG production process, e.g., the liquefaction
processes of improved eciency. Together with advances in
process synthesis and integration, a renewed life cycle analysis
of the LNG production process is highly desired to reect the
new advances in the carbon and energy footprints of the fuel.
5.2. LNG Spill, Vapor Production, Dispersion, and
Combustion. Despite signicant research advances, the
dynamics and hazards from a LNG spill require further
research. Particularly, there is a lack of experimental data from
large-scale spills for determining the possible hazards. Under
the circumstances where large-scale experiments are not
feasible, specially designed small-scale experiments need to be
carried out for validating mathematical models.
Pool Formation. The eect of sea waves on LNG pool
formation and spread is not well-understood. A major
assumption thus far is that the surface of the sea is at, while
in reality, it is dynamic and turbulent. The eect of currents
associated with the waves can also have a large impact on the
dynamics of the LNG pool and should be considered in future
studies. Future research is also need to investigate the inuence
of the surface temperature of the substrate on LNG pool
formation, especially ice formation from LNG spill on water
because of the cryogenic temperatures of the LNG.
Vapor Dispersion. There are various factors aecting LNG
vapor clouds, such as the method of release (above water verse
under water), spill rate, wind speed, atmospheric stability,
RPTs, obstructions, and terrain. Recent studies have added to
our knowledge by showing that water curtains are eective
methods of controlling LNG vapor dispersion by imparting
heat and momentum to the vapor cloud. Any increase in the
turbulence of the substrate (spill on water) will lead to an
increase in the evaporation rate of the LNG. Obstacles also
3581
Energy & Fuels
Review
6. CONCLUDING REMARKS
Signicant research advances have been made in the LNG value
chain and the understanding and management of the associated
risks/safety issues during the handling, storage, and transport of
LNG. Recent developments in the LNG production chain were
focused on optimizing the process for eciency improvement,
particularly those in the liquefaction and regasication
processes, of which several have been implemented. Extensive
research has improved our understanding of the fundamental
mechanisms that control the dynamics of a LNG spill and the
pool formation, vapor dispersion, and potential combustion/
re following such a spill. Such knowledge discovery has
allowed for establishing and validating mathematical models for
hazard prediction and developing methods for improving the
safety of personnel, facilities, and ships. There are still various
important technical gaps, and future research is warranted in
these important areas for addressing the challenges arising from
the rapid increases in production and use of LNG, potential
terrorist threats, and public condence in LNG safety. While
there have been no major incidents in the LNG industry with
the use of current systems, with further studies and continuous
improvements, this safety record can be maintained.
AUTHOR INFORMATION
Corresponding Author
*Telephone: +61-8-92667592. Fax: +61-8-92662681. E-mail: h.

[email protected].
MFC = mixed uid cascade

MM = million
MR = mixed refrigerant
N/A = not applicable or not available
NFPA = National Fire Protection Association
NGL = natural gas liquid
PFHE = parallel n heat exchanger
PMR = parallel mixed refrigerant
POC = Phillips optimized cascade
psig = pounds (force) per square inch gauge
PUF = polyurethane foam
RANS = Reynolds-averaged NavierStokes
RPT = rapid phase transition
RPUF = reinforced polyurethane foam
RSM = Reynolds stress model
SFDM+R = diesel mechanical propulsion with reliquefaction
SHIPP = system hazard identication, prediction, and
prevention
SWE = shallow water equation
tcf = trillion cubic feet
VCE = vapor cloud explosion
XLFL = downwind length to the ammable cloud at the low
ammability limit (m)
XVIS = downwind length to the visible cloud (m)
= turbulence dissipation rate
REFERENCES
(1) British Petroleum (BP). BP Statistical Review of World Energy

2013; BP: London, U.K., 2013.
(2) International Energy Agency (IEA). World Energy Outlook 2012;
IEA: Paris, France, 2012.
(3) United States Energy Information Administration (U.S. EIA).
Annual Energy Outlook 2013With Projections to 2040; U.S. EIA:
Washington, D.C., 2013.
Carbon Dioxide Emissions Coecients; U.S. EIA: Washington, D.C.,
2013; http://www.eia.gov/environment/emissions/co2_vol_mass.cfm
(accessed Nov 2013).
(5) Lim, W.; Choi, K.; Moon, I. Ind. Eng. Chem. Res. 2012, 52, 3065
3088.
(6) Foss, M. M. Introduction to LNG; Center for Energy Economics,
Bureau of Economic Geology, The University of Texas at Austin:
Austin, TX, 2003.
(7) Starling, K. E.; Savidge, J. L. Compressibility Factors of Natural Gas
and Other Related Hydrocarbon Gases; Operating Section, American
Gas Association: Washington, D.C., 1992.
(8) Shaw, B.; Shaw, G. Liqueed Petroleum Gas (LPG), Liqueed
Natural Gas (LNG) and Compressed Natural Gas (CNG); http://www.
envocare.co.uk/lpg_lng_cng.htm (accessed Nov 2013).
(9) Katz, D. L. Handbook of Natural Gas Engineering; McGraw-Hill
Book Company, Inc.: New York, 1959.
(10) National Institute of Standards and Technology (NIST). NIST
Chemistry WebBook; NIST: Gaithersburg, MD, 2011; http://webbook.
nist.gov/chemistry/ (accessed Nov 2013).
(11) ABS Consulting, Inc. Consequence Assessment Methods for
Incidents Involving Releases from Liqueed Natural Gas Carriers; ABS
Consulting, Inc.: Houston, TX, 2004; p 128.
(12) Murdock, J. W. Fundamental Fluid Mechanics for the Practicing
Engineer, 1st ed.; CRC Press: Boca Raton, FL, 1993.
(13) Wegrzyn, J.; Powars, C. Best Practices To Avoid Liqueed Natural
Gas Fueling Station Venting Losses; Brookhaven National Laboratory:
(14) Howell, J.; Harger, J. CNC and LNG: Whats Best for Your Fleet;
Westport: New York, 2013.
(15) Wood, D. A. J. Nat. Gas Sci. Eng. 2012, 9, 1627.
Notes
The authors declare no competing nancial interest.
NOMENCLATURE
avg = average
bcm = billion cubic meters
BFTF = Brayton Fire Training Field
BLEVE = boiling liquid expanding vapor explosion
BOG = boil-o gas
Btu = British thermal units
CFD = computational uid dynamics
CFR = Code of Federal Regulations
CWHE = coil wound heat exchanger
DFDE = dual-fuel diesel electric
DFGE = dual-fuel gas-turbine electric
DFSM = dual fuel steam-turbine mechanical
DMR = dual mixed refrigerant
DOT = Department of Transportation
DSF = dispersion safety factor
ESD = emergency shutdown
FERC = Federal Energy Regulatory Commission
FLNG = oating liqueed natural gas
FS = fuzzy sets
GA = generic algorithm
GBS = gravity-based structure
HAZOP = hazards and operability
IGC = International Gas Carrier
LFL = low ammability limit
LLNL = Lawrence Livermore National Laboratory
LNG = liqueed natural gas
LOPA = layer of protection analysis
LPG = liqueed petroleum gas
MC = Monte Carlo
MEM = multi-energy method
MEP = model evaluation protocol
3582
Energy & Fuels
Review

International Energy Outlook 2013; U.S. EIA: Washington, D.C., 2013.
(17) Western Australian State Government. LNG Industry Prole
Strategic Policy; Department of State Development, Western Australian
State Government: Perth, Western Australia, Australia, 2012; p 5.
(18) Boskalis. The Brass LNG Project (Nigeria); Boskalis:
Papendrecht, Netherlands, 2013.
(19) Castillo, L.; Nadales, R.; Gonzalez, C.; Dorao, C. A.; Viloria, A.,
Technology selection for liqueed natural gas (LNG) on baseload
plants. Proceedings of the XIX International Gas Convention; Caracas,
Venezuela, May 2426, 2010.
(20) Humphrey, G. Current State and Outlook for the LNG Industry;
http://www.forum.rice.edu/wp-content/uploads/2011/06/RT_
110909_Humphries.pdf (accessed Nov 2013).
Algeria Overview; U.S. EIA: Washington, D.C., 2013; p 16.
(22) Prasetyo, N. H. Maintaining Excellence in a Declining Production;
Process and SHE Engineering, PT Badak NGL: Bontang, Indonesia,
2013.
(23) Ukpohor, E. T. O. Nigerian gas masster plan: Strenthening the
nigeria gas infreastructure blueprint as a base for expanding regional
gas market. Proceedings of the 24th World Gas Conference; Buenos Aires,
Argentina, Oct 59, 2009.
(24) Qatargas. Inovation in Global Energy; Qatargas Operating
Company, Ltd.: Qatar, 2012.
(25) Total. Yamal LNG: Harnessing the Arctics Gas Resources; http://
total.com/en/energies-expertise/oil-gas/exploration-production/
projects-achievements/lng/yamal-lng (accessed Nov 2013).
(26) Okamura, T.; Furukawa, M.; Ishitani, H. Appl. Energy 2007, 84,
11361149.
(27) Foss, M. M. An Overview on Liqueed Natural Gas (LNG), Its
Properties, the LNG Industry, and Safety Considerations; The University
of Texas at Austin: Austin, TX, 2012.
(28) Tusiani, M.; Shearer, G. LNG: A Nontechnical Guide; PennWell
Corp.: Tulsa, OK, 2007; p 436.
The Global Liqueed Natural Gas Market: Status and Outlook; U.S. EIA:
(30) Barnett, P. J. Life cycle assessment (LCA) of liqueed natural
gas (LNG) and its environmental impact as a low carbon energy
source. B.Eng. Research Thesis, University of Southern Queensland,
Toowoomba, Queensland, Australia, 2010.
(31) Burnham, A.; Han, J.; Clark, C. E.; Wang, M.; Dunn, J. B.;
Palou-Rivera, I. Environ. Sci. Technol. 2011, 46, 619627.
(32) Lavray, H. Life Cycle Assessment of Electricity Generation;
Eurelectric: Brussels, Belgium, 2011.
(33) Cherubini, F.; Bird, N. D.; Cowie, A.; Jungmeier, G.;
Schlamadinger, B.; Woess-Gallasch, S. Resour., Conserv. Recycl. 2009,
53, 434447.
(34) Ou, X.; Zhang, X.; Chang, S.; Guo, Q. Appl. Energy 2009, 86,
197208.
(35) Jaramillo, P.; Griffin, W. M.; Matthews, H. S. Environ. Sci.
Technol. 2007, 41, 62906296.
(36) Bridgwood, P. Improved LNG Process; Liqueed Natural Gas,
Ltd.: Perth, Western Australia, Australia, 2013.
(37) Luketa-Hanlin, A. J. Hazard. Mater. 2006, 132, 119140.
(38) Cleaver, P.; Johnson, M.; Ho, B. J. Hazard. Mater. 2007, 140,
42938.
(39) Koopman, R. P.; Ermak, D. L. J. Hazard. Mater. 2007, 140,
41228.
(40) Raj, P. K. J. Hazard. Mater. 2007, 140, 444464.
(41) Havens, J.; Spicer, T. J. Hazard. Mater. 2007, 140, 43943.
(42) Natural Gas Supply Association (NGSA). Natural Gas
Background; http://naturalgas.org/overview/background/ (accessed
Nov 2013).
(43) Cook, P. P.; Beck, D. V.; Brereton, P. D.; Clark, P. R.; Fisher, D.
B.; Kentish, P. S.; Toomey, M. J.; Williams, D. J. Engineering Energy:
Unconventionl Gas Production; Australian Council of Learned
Academies (ACOLA): Melbourne, Victoria, Australia, 2013.
(44) Al-shobi, S. A. Simulation and integration of liqueed natural gas

(LNG) processes. M.Sci. Thesis, Texas A&M University, College
Station, TX, 2007.
(45) Poten & Partners and Merlin Associates. LNG Cost and
Competition, 2 ed.; Poten & Partners and Merlin Associates: New
York, 2004.
(46) Arrow LNG. Environmental Impact Statement; Arrow Energy:
Brisbane Queensland, Australia, 2012.
(47) Mokhatab, S.; Economides, M. J. Onshore LNG production
process selection. Proceedings of the SPE Annual Technical Conference
and Exhibition; San Antonio, TX, Sept 2427, 2006.
(48) Marin. ReportA Newsletter of MARIN; Marin: Wageningen,
Netherlands, 2003; Vol. 81, pp 119.
(49) Andress, D. L. The Phillips Optimized Cascade LNG Process: A
Quarter Century of Improvements; Phillips Petroleum Company:
Bartlesville, OK, 1996.
(50) Martin, P. Y.; Fischer, B. Natural gas liquefaction processes
comparison. Proceedings of the 14th International Conference and
Exhibition on Liqueed Natural Gas; Doha, Qatar, 2004.
(51) Shukri, T. Hydrocarbon Eng. 2004, 9, 7176.
(52) ConocoPhillips. Optimized Cascade Process; http://lnglicensing.
conocophillips.com/EN/Documents/ConocoPhillipsLNG_Brochure.
pdf (accessed Nov 2013).
(53) Michelsen, F. A.; Lund, B. F.; Halvorsen, I. J. Ind. Eng. Chem.
Res. 2010, 49, 86248632.
(54) Stebbing, R.; OBrien, J. An Updated Report on the PRICO
Process for LNG Plant; LNG & LPG Technology Congress: Paris,
France, 1975; GASTECH 75.
(55) Singh, A.; Hovd, M. Dynamic modeling and control of the
PRICO LNG process. Proceedings of the AIChE Annual Meeting; San
Francisco, CA, March 30April 3, 2006.
(56) Swenson, L. K. Single mixed refrigerant, closed loop process for
liquefying natural gas. U.S. Patent 4,033,735, 1977.
(57) Schmidt, W.; Kennington, B. LNG J. 2011, 812.
(58) McKeever, J.; Pillarella, M.; Bower, R. LNG Ind. 2008, 4449.
(59) Pillarella, M.; Liu, Y.-N.; Petrowski, J.; Bower, R. The C3MR
liquefaction cycle: Versatility for a fast growing, ever changing LNG
industry. Proceedings of the 15th International Conference and Exhibition
on Liqueed Natural Gas; Barcelona, Spain, April 2427, 2007.
(60) Longsworth, R. C. Cryostat with serviceable refrigerator. U.S.
Patent 4,277,949, 1981.
(61) Liu, Y.-N.; Newton, C. L. Feed gas drier precooling in mixed
refrigerant natural gas liquefaction processes. U.S. Patent 4,755,200,
1988.
(62) Martin, P.-Y.; Pigourier, J.; Fischer, B. Hydrocarbon Eng. 2004, 9,
7579.
(63) Dam, W.; Ho, S.-M. Oil Gas J. 2001, 99, 5858.
(64) Newton, C. L. Dual mixed refrigerant natural gas liquefaction
with staged compression. U.S. Patent 4,525,185, 1985.
(65) van de Graaf, J. M.; Pek, B. Large-Capacity LNG Trains: The
Shell Parallel Mixed Refrigerant Process; LNG and Gas Processing
Group, Shell Global Solutions International BV: The Hague,
Netherlands, Oct 2005.
(66) Flesch, E.; Raillard, J.-C.; de France, G. CII liquefaction process:
2 cascades into 1. Proceedings of the 12th International Conference and
Exhibition on Liqueed Natural Gas; Perth, Western Australia,
Australia, (May 47, 1998; pp 47.
(67) Roberts, M. J.; Agrawal, R. Hybrid cycle for the production of
liqueed natural gas. U.S. Patent 6,308,531, 2001.
(68) Roberts, M.; Liu, Y.; Bronfenbrenner, J.; Petrowski, J. Reducing
LNG capital cost in todays competitive environment. Proceedings of
the 14th International Conference and Exhibition on Liqueed Natural
Gas; Doha, Qatar, March 2124, 2004.
(69) Bauer, H. A novel concept for large LNG baseload plants,
Proceedings of the 2001 AIChE Spring National Meeting; Houston, TX,
April 2226, 2001.
(70) Berger, E.; Forg, W.; Heiersted, R.; Paurola, P. The MFC (Mixed
Fluid Cascade) Process for the First European Baseload LNG Production
Plant; Linde Technology: Munich, Germany, 2003; p 12.
3583
Energy & Fuels
Review
(71) Barron, R. F. Cryogenic Systems; Oxford University Press: New

York, 1985.
(72) Walker, G. Cryocoolers; Plenum Press: New York, 1983; p 392.
(73) Timmerhaus, K. D.; Flynn, T. M. Cryogenic Process Engineering;
Plenum Press: New York, 1989.
(74) Alabdulkarem, A.; Mortazavi, A.; Hwang, Y.; Radermacher, R.;
Rogers, P. Appl. Therm. Eng. 2011, 31, 10911098.
(75) Paradowski, H.; Bamba, M.; Bladanet, C. Propane precooling
cycles for increased LNG train capacity. Proceedings of the 14th
International Conference and Exhibition on Liqueed Natural Gas; Doha,
Qatar, March 2124, 2004; pp 107124.
(76) Venkatarathnam, G. Cryogenic Mixed Refrigerant Processes;
Springer: New York, 2008.
(77) Khan, M. S.; Lee, S.; Rangaiah, G. P.; Lee, M. Appl. Energy 2013,
111, 10181031.
(78) Castillo, L.; Majzoub Dahouk, M.; Di Scipio, S.; Dorao, C. A.
Energy Convers. Manage. 2013, 66, 4147.
(79) Baek, S.; Hwang, G.; Lee, C.; Jeong, S.; Choi, D. Energy Convers.
Manage. 2011, 52, 28072814.
(80) Lim, W.; Lee, I.; Tak, K.; Cho, J. H.; Ko, D.; Moon, I. Ind. Eng.
Chem. Res. 2014, 53, 19731985.
(81) Rodgers, P.; Mortazavi, A.; Eveloy, V.; Al-Hashimi, S.; Hwang,
Y.; Radermacher, R. Appl. Therm. Eng. 2012, 48, 4153.
(82) Wahl, P. E.; Lvseth, S. W.; Mlnvik, M. J. Comput. Chem. Eng.
2013, 56, 2736.
(83) Wang, M.; Zhang, J.; Xu, Q. Comput. Chem. Eng. 2012, 39, 84
95.
(84) Chang, H.-M.; Lim, H. S.; Choe, K. H. Cryogenics 2012, 52,
642647.
(85) Rian, A. B.; Ertesvag, I. S. Energy Fuels 2012, 26, 12591267.
(86) Li, Y.; Wang, X.; Ding, Y. Appl. Energy 2012, 99, 484490.
(87) Li, Q. Y.; Ju, Y. L. Appl. Therm. Eng. 2010, 30, 25182525.
(88) Lukaszewski, M. W.; Zimmerman, W. B. J.; Tennant, M. T.;
Webster, M. B. Chem. Eng. Res. Des. 2013, 91, 457463.
(89) MAN Diesel & Turbo. Propulsion Trends in LNG Carriers; MAN
Diesel & Turbo: Copenhagen, Denmark, 2009.
(90) Woehrling, J.-M.; Cotterell, C. D. International Safety Guide for
Inland Navigation Tank-barges and Terminals (ISGINTT); Central
Commission for the Navigation of the Rhine: Strasbourg, France,
2010.
(91) Peebles, M. W. H. Natural Gas Fundamentals; Shell International Gas Limited: The Hague, Netherlands, 1992.
(92) Rudan, S.; Ascic, B.; Visic, I. Proceedings of the 6th International
Conference on Collision and Grounding of Ship and Oshore Structures;
Trondheim, Norway, June 1719, 2013; pp 331337.
(93) Querol, E.; Gonzalez-Regueral, B.; Garca-Torrent, J.; Ramos, A.
Appl. Energy 2011, 88, 23822390.
(94) Liu, C.; Zhang, J.; Xu, Q.; Gossage, J. L. Ind. Eng. Chem. Res.
2010, 49, 74127420.
(95) Tsatsaronis, G.; Morosuk, T. Energy 2010, 35, 820829.
(96) Morosuk, T.; Tsatsaronis, G. Energy 2011, 36, 37713778.
(97) Morosuk, T.; Tsatsaronis, G.; Boyano, A.; Gantiva, C. Int. J.
Energy Environ. Eng. 2012, 3, 19.
(98) Dispenza, C.; Dispenza, G.; La Rocca, V.; Panno, G. Appl.
Therm. Eng. 2009, 29, 380387.
(99) Dispenza, C.; Dispenza, G.; Rocca, V. L.; Panno, G. Appl.
Therm. Eng. 2009, 29, 388399.
(100) Shi, X.; Che, D. Energy Convers. Manage. 2009, 50, 567575.
(101) Liu, Y.; Guo, K. Energy 2011, 36, 28282833.
(102) Zhang, N.; Lior, N. Energy 2006, 31, 16661679.
(103) Dispenza, C.; Dispenza, G.; Rocca, V. L.; Panno, G. Appl.
Therm. Eng. 2009, 29, 35953608.
(104) La Rocca, V. Energy 2010, 35, 20492058.
(105) La Rocca, V. Energy 2011, 36, 48974908.
(106) Messineo, A.; Panno, G. J. Nat. Gas Sci. Eng. 2011, 3, 356363.
(107) Koopman, R. P.; Baker, J.; Cederwall, R. T.; Goldwire, H. C. J.;
Hogan, W. J.; Kamppinen, L. M.; Kiefer, R. D.; McClure, J. W.;
McRae, T. G.; Morgan, D. L. LLNL/NWC 1980 LNG Spill Tests. Burro
Series Data Report; Lawrence Livermore National Laboratory (LLNL):

Livermore, CA, 1982; p 290.
(108) Conrado, C.; Vesovic, V. Chem. Eng. Sci. 2000, 55, 45494562.
(109) Little, D.; Raj, P. P.; Guard, U. S. C. Assessment Models in
Support of the Hazard Assessment Handbook; Arthur D. Little, Inc.:
Cambridge, MA, 1974.
(110) Opschoor, G. Cryogenics 1977, 17, 629634.
(111) Gavelli, F.; Bullister, E.; Kytomaa, H. J. Hazard. Mater. 2008,
159, 158168.
(112) Atallah, S.; Sirdesai, M.; Jennings, D. M. Source1: A Model for
Creating a DEGADIS Input File; Gas Research Institute (GRI): Des
Plaines, IL, 1993.
(113) Raj, P. K.; OFarrel, P. M. Development of Additional Hazard
Assessment Models; Arthur D. Little, Inc.: Cambridge, MA, 1977.
(114) Webber, D. M.; Jones, S. J. A model of spreading vaporising
pools. In International Conference on Vapour Cloud Modelling;
Woodward, J., Ed.; American Institute of Chemical Engineers
(AIChE): New York, 1987.
(115) Netherlands Organisation for Applied Scientic Research
(TNO). Methods for the Calculation of Physical Eects (TNO Yellow
Book); TNO: The Hague, Netherlands, 1997.
(116) Woodward, J. L. J. Hazard. Mater. 2007, 140, 478487.
(117) Woodward, J. L.; Pierorazio, A. J. SS3GAn integrated risk
analysis program using engineering principles for building damage.
Proceedings of the Mary Kay OConnor 2001 Annual Symposium;
College Station, TX, Oct 3031, 2001; pp 365382.
(118) Witlox, H. W. M.; Oke, A. Verication and Validation of
Consequence and Risk Models for Accidental Releases of Hazardous
Chemicals to the Atmosphere; DNV Software: London, U.K., 2007.
(119) McQuaid, J. Some Experiments on the Structure of StableStratied Shear Flows; Safety in Mines Research Establishment, Health
and Safety Executive: Liverpool, U.K., 1976.
(120) Petersen, R. L. J. Wind Eng. Ind. Aerodyn. 1990, 36, 643652.
(121) Pitblado, R.; Baik, J.; Hughes, G.; Ferro, C.; Shaw, S. Process
Saf. Prog. 2004, 24, 108114.
(122) Pitblado, R.; Baik, J.; Raghunathan, V. J. Hazard. Mater. 2006,
148154.
(123) Thoman, D. C.; OKula, K. R.; Laul, J. C.; Davis, M. W.;
Knecht, K. D. J. Chem. Health Saf. 2006, 13, 2033.
(124) Bernatik, A.; Libisova, M. J. Loss Prev. Process Ind. 2004, 17,
271278.
(125) Mazzoldi, A.; Hill, T.; Colls, J. J. Atmos. Environ. 2008, 42,
80468054.
(126) Spaulding, M. L.; Swanson, J. C.; Jayko, K.; Whittier, N. J.
Hazard. Mater. 2007, 140, 488503.
(127) Hightower, M.; Gritzo, L.; Luketa-Hanlin, A.; Covan, J.;
Tieszen, S.; Wellman, G.; Irwin, M.; Kaneshige, M.; Melof, B.;
Morrow, C.; Ragland, D. Guidance on Risk Analysis and Safety
Implications of a Large Liqueed Natural Gas (LNG) Spill over Water;
Sandia National Laboratories: Albuquerque, NM, 2004; p 168.
(128) Dharmavaram, S. Process Saf. Prog. 2005, 24, 316272.
(129) Hanna, S. R.; Hansen, O. R.; Dharmavaram, S. Atmos. Environ.
2004, 38, 46754687.
(130) Fay, J. A. J. Hazard. Mater. 2007, 140, 54151.
(131) Dandrieux, A.; Dusserre, G.; Thomas, O. Environ. Modell.
Software 2003, 18, 253259.
(132) Hald, K.; Buchlin, J.-M.; Dandrieux, A.; Dusserre, G. J. Loss
Prev. Process Ind. 2005, 18, 506511.
(133) McQuaid, J.; Fitzpatrick, R. D. J. Occup. Accid. 1983, 5, 121
133.
(134) McQuaid, J.; Fitzpatrick, R. D. The Uses and Limitations of
Water-Spray Barriers; Health and Safety Executive: Liverpool, U.K.,
1981.
(135) Moodie, K. Plant/Oper. Prog. 1985, 4, 234241.
(136) Moore, P. A. C.; Rees, W. D. Forced Dispersion of Gases by
Water and Steam; North Western Branch Papers, Institution of
Chemical Engineers (IChemE): London, U.K., 1981.
(137) Rana, M. A.; Cormier, B. R.; Suardin, J. A.; Zhang, Y.; Mannan,
M. S. Process Saf. Prog. 2008, 27, 345353.
3584
Energy & Fuels
Review
(138) Rana, M. A.; Guo, Y.; Mannan, M. S. J. Loss Prev. Process Ind.
2010, 23, 7788.
(139) Rana, M. A.; Mannan, M. S. J. Loss Prev. Process Ind. 2010, 23,
768772.
(140) Olewski, T.; Nayak, S.; Basha, O.; Waldram, S.; Vechot, L. J.
Loss Prev. Process Ind. 2011, 24, 798804.
(141) Kim, B. K.; Ng, D.; Mentzer, R. A.; Mannan, M. S. Ind. Eng.
Chem. Res. 2012, 51, 1380313814.
(142) Kim, B. K.; Ng, D.; Mentzer, R. A.; Mannan, M. S. J. Loss Prev.
Process Ind. 2013, 26, 16701678.
(143) Qi, R.; Raj, P. K.; Mannan, M. S. J. Loss Prev. Process Ind. 2011,
24, 440448.
(144) Cormier, B. R.; Qi, R.; Yun, G.; Zhang, Y.; Mannan, M. S. J.
(145) Morse, T. L.; Kytomaa, H. K. J. Loss Prev. Process Ind. 2011, 24,
791797.
(146) Ermak, D. L.; Rodean, H. C.; Lange, R.; Chan, S. T. A Survey of
Denser-than-Air Atmospheric Dispersion Models; Lawrence Livermore
National Laboratory (LLNL): Livermore, CA, 1988.
(147) Sykes, R. I.; Cerasoli, C. P.; Henn, D. S. J. Hazard. Mater. 1999,
64, 223247.
(148) Hankin, R. K.; Britter, R. E. J. Hazard. Mater. 1999, 66, 211
226.
(149) Hankin, R. K. S. J. Hazard. Mater. 2003, 103, 110.
(150) Ermak, D. L. Users Manual for SLAB: An Atmospheric
Dispersion Model for Denser-than-Air Releases; Lawrence Livermore
National Laboratory (LLNL): Livermore, CA, 1990.
(151) Witlox, H. W. M. Atmos. Environ. 1994, 28, 29172932.
(152) Witlox, H. W. M. Atmos. Environ. 1994, 28, 29332946.
(153) Blackmore, D. R.; Herman, M. N.; Woodward, J. L. J. Hazard.
Mater. 1982, 6, 107128.
(154) Spicer, T. O.; Havens, J. A. J. Hazard. Mater. 1987, 16, 231
245.
(155) Hanna, S. R.; Chang, J. C.; Strimaitis, D. G. Atmos. Environ.,
Part A 1993, 27, 22652285.
(156) Qi, R.; Ng, D.; Cormier, B. R.; Mannan, M. S. J. Hazard. Mater.
2010, 183, 5161.
(157) Chan, S. T.; Ermak, D. L.; Morris, L. K. J. Hazard. Mater. 1987,
16, 267292.
(158) Sun, B.; Utikar, R. P.; Pareek, V. K.; Guo, K. J. Loss Prev. Process
Ind. 2013, 26, 117128.
(159) Riddle, A.; Carruthers, D.; Sharpe, A.; McHugh, C.; Stocker, J.
Atmos. Environ. 2004, 38, 10291038.
(160) Rigas, F.; Sklavounos, S. Chem. Eng. Sci. 2006, 61, 14441452.
(161) Hanna, S. R.; Chang, J. C. Atmos. Environ. 2001, 35, 2231
2242.
(162) Luketa-Hanlin, A.; Koopman, R. P.; Ermak, D. L. J. Hazard.
Mater. 2007, 140, 50417.
(163) Qiao, Y.; West, H. H.; Mannan, M. S.; Johnson, D. W.;
Cornwell, J. B. J. Hazard. Mater. 2006, 130, 155162.
(164) Johnson, D. W.; Cornwell, J. B. J. Hazard. Mater. 2007, 140,
535540.
(165) Fay, J. A. J. Hazard. Mater. 2003, 96, 23.
(166) Lehr, W.; Simecek-Beatty, D. J. Hazard. Mater. 2004, 107, 12.
(167) Vlchez, J. A.; Villafane, D.; Casal, J. J. Hazard. Mater. 2013,
246247, 181188.
(168) Ponchaut, N. F.; Kyto maa, H. K.; Morrison, D. R.;
Chernovsky, M. K. J. Loss Prev. Process Ind. 2011, 24, 870878.
(169) Melton, T. A.; Cornwell, J. B. J. Loss Prev. Process Ind. 2010, 23,
762767.
(170) Gavelli, F.; Chernovsky, M. K.; Bullister, E.; Kytomaa, H. K. J.
Hazard. Mater. 2010, 180, 332339.
(171) Giannissi, S. G.; Venetsanos, A. G.; Markatos, N.; Bartzis, J. G.
J. Loss Prev. Process Ind. 2013, 26, 245254.
(172) Wurtz, J.; Bartzis, J.; Venetsanos, A.; Andronopoulos, S.;
Statharas, J.; Nijsing, R. J. Hazard. Mater. 1996, 46, 273284.
(173) Bartzis, J. G. ADREA-HF: A Three-Dimensional Finite Volume
Code for Vapour Cloud Dispersion in Complex Terrain; Institute for
Safety Technology, Joint Research Center, Commission of the

European Communities: Brussels, Belgium, 1991.
(174) Venetsanos, A. G.; Papanikolaou, E.; Bartzis, J. G. Int. J.
Hydrogen Energy 2010, 35, 39083918.
(175) Vesovic, V. J. Hazard. Mater. 2007, 140, 51826.
(176) Siuta, D.; Markowski, A. S.; Mannan, M. S. J. Loss Prev. Process
Ind. 2013, 26, 418426.
(177) Britter, R. E.; McQuaid, J. Workbook on the Dispersion of Dense
Gases; Health and Safety Executive: Liverpool, U.K., 1988.
(178) Ivings, M. J.; Lea, C. J.; Webber, D. M.; Jagger, S. F.; Coldrick,
S. J. Loss Prev. Process Ind. 2013, 26, 153163.
(179) Carissimo, B.; Jagger, S. F.; Daish, N. C.; Halford, A.; SelmerOlsen, S.; Riikonen, K.; Perroux, J. M.; Wurtz, J.; Bartzis, J. G.; Duijm,
N. J. Int. J. Environ. Pollut. 2001, 16, 614629.
(180) Daish, N. C.; Britter, R. E.; Linden, P. F.; Jagger, S. F.;
Carissimo, B. Int. J. Environ. Pollut. 2000, 14, 3951.
(181) Duijm, N. J.; Carrissimo, B. Evaluation Methodologies for Dense
Gas Dispersion Models; McGraw-Hill: New York, 2002.
(182) Ivings, M. J.; Jagger, S. F.; Lea, C. J.; Webber, D. M. Evaluating
Vapour Dispersion Models for Safety Analysis of LNG Facilities; Fire
Protection Research Foundation: Quincy, MA, 2007.
(183) Hansen, O. R.; Gavelli, F.; Ichard, M.; Davis, S. G. J. Loss Prev.
Process Ind. 2010, 23, 857877.
(184) Raj, P. P. K.; Moussa, A. N.; Aravamudan, K. Experiments
Involving Pool and Vapor Fires from Spills of Liqueed Natural Gas on
Water; Arthur D. Little, Inc.: Cambridge, MA, 1979; p 252.
(185) Schneider, A. L. US Coast Guard liqueed natural gas research
at China Lake. Proceedings of the Gastech 78 LNGLPG Conference;
Monte Carlo, Monaco, Nov 710, 1978; pp 58.
(186) Raj, P. K.; Moussa, N. A.; Aravamudan, K.; Lind, C. D. Oper.
Sect. Proc.Am. Gas Assoc. 1979, 246251.
(187) Schneider, A. L. J. Fire Flammability 1980, 11, 302313.
(188) Mizner, G. A.; Eyre, J. A. EFCE Publ. Ser. 1982, 25, 147161.
(189) Hirst, W. J. S.; Eyre, J. A. Maplin Sands experiments 1980:
Combustion of large LNG and refrigerated liquid propane spills on the
sea. Heavy Gas and Risk AssessmentII; Springer: Dordrecht,
Netherlands, 1983; pp 211224.
(190) Rodean, H. C.; Hogan, W. J.; Urtiew, P. A.; Goldwire, H. C., Jr;
McRae, T. G.; Morgan, D. L., Jr Vapor Burn Analysis for the Coyote
Series LNG Spill Experiments; Lawrence Livermore National
Laboratory (LLNL): Livermore, CA, 1984; p 89.
(191) Mizner, G. A.; Eyre, J. A. Combust. Sci. Technol. 1983, 35, 33
57.
(192) Nedelka, D.; Moorhouse, J.; Tucker, R. F. The Montoir 35m
Diameter LNG Pool Fire Experiments; Midlands Research Station: Nice,
France, 1989; p 16.
(193) Lowesmith, B. J.; Hankinson, G.; Acton, M. R.; Chamberlain,
G. Process Saf. Environ. Prot. 2007, 85, 207220.
(194) Studer, E.; Jamois, D.; Jallais, S.; Leroy, G.; Hebrard, J.;
Blanchetiere, V. Int. J. Hydrogen Energy 2009, 34, 96119619.
(195) Wu, Y.; Lu, Y.; Al-Rahbi, I. S.; Kalghatgi, G. T. Int. J. Hydrogen
Energy 2009, 34, 59405945.
(196) Schneider, A. L. J. Fire Flammability 1980, 302313.
(197) Brown, L. E. Hydrocarbon Process. 1974, 141143.
(198) Raj, P.; Atallah, S. Adv. Cryog. Eng. 1975, 143150.
(199) Lautkasi, R. J. Loss Prev. Process Ind. 1992, 5, 175180.
(200) Johnson, A. D. A model for predicting thermal radiation
hazards from large-scale LNG pool fires. Inst. Chem. Eng. Symp. Ser.
1992, 507524.
(201) Moorhouse, J.; Pritchard, M. J. Thermal radiation hazards from
large pool fires and fireballsA literature review. Inst. Chem. Eng.
Symp. Ser. 1982, 397428.
(202) Society of Fire Protection Engineers (SFPE). The SFPE
Handbook of Fire Protection Engineering, 2nd ed.; SFPE: Bethesda, MD,
1995.
(203) Gavelli, F.; Davis, S. G.; Hansen, O. R. J. Loss Prev. Process Ind.
2011, 24, 908915.
(204) Abe, A.; Katayama, M.; Kim, D.; Usuba, S.; Castillo, M.;
Wakatsuki, M.; Kakudate, Y.; Tanaka, K.; Watanabe, Y. Numerical
3585
Energy & Fuels
Review
prediction of blast eecs caused by large-scale exlosion of LOX/LNG

fuel. Proceedings of the International Symposium on Structures under
Earthquake, Impact, and Blast Loading (IB08); Osaka, Japan, Oct 10
11, 2008.
(205) Fay, J. A. Annu. Rev. Energy 1980, 5, 89105.
(206) Raj, P. K. J. Hazard. Mater. 2008, 157, 247259.
(207) Perez, J. F. S.; Ferradas, E. G.; Alonso, F. D.; Garca, D. P.;
Cano, M. V. M; Cotorruelo, J. A . B. Process Saf. Environ. Prot. 2010, 88,
109113.
(208) Alonso, F. D.; Ferradas, E. G.; Sanchez, T. d. J. J.; Aznar, A. M.;
Gimeno, J. R.; Alonso, J. M. J. Hazard. Mater. 2008, 150, 146152.
(209) Netherlands Organisation for Applied Scientic Research
(TNO). Methods for the Determination of Possible Damage; TNO: The
Hague, Netherlands, 1989.
(210) Pitblado, R. M.; Burns, D. J. Chem. Eng. Prog. 1989, 85, 6975.
(211) Stoll, A. M.; Chianta, M. A. Trans. N. Y. Acad. Sci. 1971, 33,
649670.
(212) Lihou, D.; Maund, J. Thermal radiation hazard from fireballs.
Inst. Chem. Eng. Symp. Ser. 1982, 191224.
(213) Bernatik, A.; Senovsky, P.; Pitt, M. J. Loss Prev. Process Ind.
2011, 24, 1924.
(214) U.S. Government. National Infrastructure Protection Plan; U.S.
Department of Homeland Security: Washington, D.C., 2009.
(215) Yun, G.; Rogers, W. J.; Mannan, M. S. J. Loss Prev. Process Ind.
2009, 22, 9196.
(216) Rathnayaka, S.; Khan, F.; Amyotte, P. J. Loss Prev. Process Ind.
2012, 25, 414423.
(217) Parihar, A.; Vergara, C.; Clutter, J. K. Saf. Sci. 2011, 49, 686
694.
(218) Li, J.; Huang, Z. Procedia Eng. 2012, 45, 7076.
(219) Pitblado, R. J. Hazard. Mater. 2007, 140, 527534.
(220) Zhang, Q.-x.; Liang, D. Procedia Eng. 2013, 52, 602606.
(221) Abbasi, T.; Abbasi, S. A. J. Hazard. Mater. 2007, 141, 489519.
(222) Grossel, S. S. J. Loss Prev. Process Ind. 2008, 21, 492.
(223) Cheng, S.-R.; Lin, B.; Hsu, B.-M.; Shu, M.-H. Expert Syst. Appl.
2009, 36, 1191811924.
(224) Keshavarz, G.; Thodi, P.; Khan, F. J. Loss Prev. Process Ind.
2012, 25, 159165.
(225) Suardin, J. A.; Wang, Y.; Willson, M.; Mannan, M. S. J. Hazard.
Mater. 2009, 165, 612622.
(226) Suardin, J. A.; Qi, R.; Cormier, B. R.; Rana, M.; Zhang, Y.;
Mannan, M. S. J. Loss Prev. Process Ind. 2011, 24, 6375.
(227) Sun, B.; Guo, K. J. Loss Prev. Process Ind. 2013, 26, 12461256.
(228) Taylor, D. W. J. Hazard. Mater. 2007, 142, 776785.
(229) Raj, P. K.; Lemoff, T. J. Loss Prev. Process Ind. 2009, 22, 820
829.
(230) Tanabe, M.; Miyake, A. J. Loss Prev. Process Ind. 2010, 23, 507
514.
(231) Tanabe, M.; Miyake, A. J. Loss Prev. Process Ind. 2012, 25, 809
819.
(232) Vanem, E.; Antao, P.; stvik, I.; de Comas, F. D. C. Reliab.
Eng. Syst. Saf. 2008, 93, 13281344.
(233) Vandebroek, L.; Berghmans, J. Procedia Eng. 2012, 45, 2126.
(234) Moon, K.; Song, S.-R.; Ballesio, J.; Fitzgerald, G.; Knight, G. J.
(235) Chang, D.; Rhee, T.; Nam, K.; Chang, K.; Lee, D.; Jeong, S.
Reliab. Eng. Syst. Saf. 2008, 93, 18771885.
(236) International Maritime Organisation (IMO). International Code
for the Construction and Equipment of Ships Carrying Liqueed Gases in
Bulk: IGC Code; IMO: London, U.K., 1993.
(237) United States Government Printing Oce (U.S. GPO).
Minimum relief capacities for cargo tanks containing compressed or
liqueed gas, U.S. Coast Guard. Code of Federal Regulations (CFR);
U.S. GPO: Washington, D.C., 2013.
(238) Havens, J.; Venart, J. J. Hazard. Mater. 2008, 158, 273279.
(239) Bang, C. S.; Kim, J. G.; Lee, D. G. J. Adhes. Sci. Technol. 2012,
26, 969986.
(240) Lee, J.-S.; Ryu, Y.-S.; Kim, N.-I.; Kim, B.-J.; Kim, Y.-K.; Kim,
M.-H. Trans. Nonferrous Met. Soc. China 2009, 19 (Supplement1),
s271s275.
(241) Kim, B. G.; Lee, D. G. Compos. Struct. 2008, 86, 2736.
(242) Yu, Y. H.; Kim, B. G.; Lee, D. G. Compos. Struct. 2012, 94,
462468.
(243) Yu, Y. H.; Kim, B. G.; Lee, D. G. Compos. Struct. 2013, 95,
547556.
(244) Chul Kim, B.; Ho Yoon, S.; Gil Lee, D. Ocean Eng. 2011, 38,
592608.
(245) Graczyk, M.; Moan, T. J. Offshore Mech. Arct. Eng. 2010, 133,
2110321103.
(246) Pistani, F.; Thiagarajan, K. Ocean Eng. 2012, 52, 6074.
(247) Mitra, S.; Wang, C. Z.; Reddy, J. N.; Khoo, B. C. Ocean Eng.
2012, 39, 113.
NOTE ADDED AFTER ASAP PUBLICATION

There were errors in values given in Table 3 of the version of
this paper published May 16, 2014. The correct version
published June 9, 2014.
3586

Some Recent Advances in Liquefied Natural Gas (LNG) Production, Spill, Dispersion, and Saf

Uploaded by

Copyright:

Available Formats

Some Recent Advances in Liquefied Natural Gas (LNG) Production, Spill, Dispersion, and Saf

Uploaded by

Document Information

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

Some Recent Advances in Liquefied Natural Gas (LNG) Production, Spill, Dispersion, and Saf

Uploaded by

Copyright:

Available Formats

What are some of the recent advances in LNG production discussed in the paper?

What are some of the recent advances in LNG production discussed in the paper?

What methods are discussed for transporting natural gas and what are the advantages of LNG transport?

What methods are discussed for transporting natural gas and what are the advantages of LNG transport?

Review

Some Recent Advances in Liqueed Natural Gas (LNG) Production,

source at 39%, followed by oil at 37%, natural gas at 18%, other

Received: March 20, 2014

dx.doi.org/10.1021/ef500626u | Energy Fuels 2014, 28, 35563586

Energy & Fuels

Table 3. Fuel Properties of Natural Gas at STP710

Table 1. Average Levelized Capital Costs of New Power

natural gas (STP)

LNG trade is expected to come in the time frame of 2015

Table 2. Emission of Air Pollutants from the Direct

Data were extracted and converted from the literature.4,5

Figure 2. Transport cost of natural gas technologies relative to

natural gas at standard temperature and pressure (STP)710

dx.doi.org/10.1021/ef500626u | Energy Fuels 2014, 28, 35563586

Energy & Fuels

Table 4. Some Key LNG Projects Worldwide1725

Table 5. LNG Value Chain Costs for a Typical 8 mtpa Process28,30

57% of capital cost

34% of capital cost

Table 6. LCA of Dierent Energy Sectors for a 20 Yeara Period3,3134

emission as a result of leakages from the pipeline transport

traveled in a passenger car, closely followed by natural gas, and

stages of life cycle

pipeline (natural gas)

Data were extracted from the study by Jaramillo et al.35

dx.doi.org/10.1021/ef500626u | Energy Fuels 2014, 28, 35563586

Energy & Fuels

outlook of the global LNG trade

LNG plant designs

studies on large-scale LNG spills

summary of experimental data on

Havens and Spicer41

problems with the United States

Table 9. Typical Composition of Natural Gas Compared to

could be wet, containing hydrocarbons, such as butane,

2. LNG PRODUCTION CHAIN

dx.doi.org/10.1021/ef500626u | Energy Fuels 2014, 28, 35563586

Energy & Fuels

Table 10. Commonly Used LNG Liquefaction Processesa

Shell and APCI

Phillips optimized cascade (POC)

Table 11. Comparison of Refrigeration Cycle Congurationsa

plate n heat exchanger (PFHE) or

coil-wound heat exchanger

heat-exchanger line (two PFHEs)

2.2. Liquefaction and Storage. The natural gas leaving

(when mild steel is used as the outer tank, a berm is

dx.doi.org/10.1021/ef500626u | Energy Fuels 2014, 28, 35563586

Energy & Fuels

applied to multiple oil production facilities.48 However, such

Optimization of the Precooling Cycle. Via HYSYS

Figure 4. Optimized power consumption by Alabdulkarem et al.74

methods (e.g., by Paradowski et al.75 and Venkattaratham76) for

dx.doi.org/10.1021/ef500626u | Energy Fuels 2014, 28, 35563586