Professional Documents
Culture Documents
REVIEWS Further
Quick links to online content
James A. Fay
Annu. Rev. Energy. 1980.5:89-105. Downloaded from www.annualreviews.org
Access provided by 180.190.146.56 on 02/08/23. For personal use only.
INTRODUCTION
Public dispute about LNG and LPG safety revolves around the question
of whether government regulation of the industry, together with its own
self-regulation, limit the risk of harm to affected populations to a level that
is politically acceptable. This review is concerned principally with the status
of risk analysis for LNG and LPG systems and only tangentially with the
more general problem of the relationship of formal risk analysis to public
acceptability of such systems.
LNG is a form of natural gas obtained at atmospheric pressure by re
frigerating it to -260oP (-162°C) which permits long-term storage in tanks
(rather than in underground reservoirs) and transportation by marine, road,
Annu. Rev. Energy. 1980.5:89-105. Downloaded from www.annualreviews.org
bution of natural gas energy from well to end consumer that parallels that
of liquid petroleum. But because of its very low temperature (halfway
between absolute zero and the freezing point of water), special and com
paratively expensive technology is required to store and transport LNG
which deters its more widespread use. Because the fire hazard from
accidental release of LNG is perceived to be more serious than that of
petroleum, public controversy regarding risk acceptance and facility siting
abounds (2).
Transforming natural gas to LNG serves two distinct purposes: (0) It
permits the transoceanic shipment of LNG from distant gas fields in Africa
and the Middle East to markets in Western Europe, the United States, and
Japan, thus supplementing indigenous supplies of natural or synthetic gas
in these industrial nations, and (b) In the United States, especially in the
northeast, LNG stored in peak-shaving facilities is regasified on cold winter
days to help satisfy peak demand for space heating when domestic pipeline
capacity is insufficient. In either case, regasified LNG is distributed by
pipeline to customers.
Modem LNG marine tankers typically carry 125,000 m3 of cargo in five
separate holds (3). Because of its low temperature, the liquid cargo is
contained in cryogenic tanks thermally insulated and, in some instances,
structurally isolated from the hull of the vessel. Special alloys and insulation
are needed for this purpose. Because of the low density of LNG (40% of
that of sea water) and the need for its containment separate from the ship's
structure, these tankers are much bulkier and more expensive than their
crude oil tanker counterparts. Relative to the energy content of their
cargoes, LNG tankers are nearly ten times as costly as oil tankers. Only
about eighty such tankers have been built worldwide to date (2).
Marine importation terminals are sized to match LNG tanker capacity,
commonly providing a total storage volume more than twice that of the
largest tanker supplying the terminal. Because of the economies of scale,
individual storage tanks at LNG terminals are larger than is usual at oil
LNG/LPG RISKS 91
terminals, having typical storage capacities o f 105 m3• Like the tanks aboard
ships, they are constructed of special materials and require a complex
configuration. Three such terminals are currently in use in the United
States, on the east coast, and similar terminals are located in Japan and
several European nations.
Peak-shaving terminals are smaller in size, although a few have storage
tanks of more than 50,000 m3 capacity. At these terminals pipeline gas is
liquefied during summer months and stored for wintertime use. (In New
England, some terminals are supplied directly with LNG trucked from the
import terminal in Everett, Massachusetts.) In addition, peak-shaving ter
Annu. Rev. Energy. 1980.5:89-105. Downloaded from www.annualreviews.org
minals ship LNG by truck to nearby satellite terminals for use during
Access provided by 180.190.146.56 on 02/08/23. For personal use only.
periods of peak demand. More than 100 such terminals are located in 3 1
states of the United States (2).
The LNG industry is comparatively new; most of its growth occurred in
the last decade. Spurred by the declining production of gas in the United
States and its scarcity in Europe and Japan, and, of course, by the availabil
ity of natural gas in Africa, the Middle East, and the Far East, expansion
has been limited only by the necessity to develop new technology and the
difficulty of acquiring capital, which is so intensively used in this industry.
But two major accidents in the United States-a tank failure and fire in
Cleveland in 1944 (4) and an explosion and fire in an empty LNG tank in
Staten Island in 1973 (5), both involving substantial loss of life-have raised
questions about the safety of LNG systems, which has affected both the rate
of construction and the selection of sites for LNG terminals in the United
States in recent years (6).
Liquefied petroleum gas (LPG) is the liquid form of a fuel or petrochemi
cal feedstock, consisting mostly of propane, which is separated from oil or
gas streams at the wellhead or produced during the petroleum refining
process. If stored at atmospheric pressure it must be cooled to -44°F
(-42°C) but, unlike LNG, it can also be stored as a liquid at normal
temperatures under an absolute pressure of about 8 atm. When stored under
pressure, LPG tank volumes usually do not exceed that of railway tank cars,
about 40 m3• Much larger terminal storage tanks are refrigerated, as are
most transoceanic LPG tankers.
LPG is stored and distributed like any other petroleum product, except
for the special structural and handling requirements imposed by its low
temperature or high pressure. But unlike LNG, LPG is not gasified for
distribution by pipeline to the end consumer (except when used as a natural
gas supplement) and therefore requires many small-scale shipments by rail
or road tank vehicles. Transoceanic tanker shipments are currently fewer
than those of LNG, and LPG tankers are, on average, smaller than their
LNG counterparts.
92 FAY
attention and inquiry. LNG is not used by end consumers, who are familiar
Access provided by 180.190.146.56 on 02/08/23. For personal use only.
LNG RISK
LNG Hazards
In a series of predominantly experimental investigations, Burgess and co
Annu. Rev. Energy. 1980.5:89-105. Downloaded from www.annualreviews.org
workers (9-11) first showed the dominant features of LNG spills: rapid
Access provided by 180.190.146.56 on 02/08/23. For personal use only.
boiling of the liquid whether spilled on land or the surface of water, horizon
tal spreading of the evolved vapor because of its negative buoyancy, dilution
of the vapor as it drifts downwind by mixing with ambient air, ignitability
of the vapor cloud or plume, and the possibility oftlameless vapor explosion
of the LNG liquid under certain circumstances. Subsequent field tests of
LNG spills up to 10 m3 in volume on land and water confirmed these earlier
findings, adding more specific information on boiling rates and dispersive
properties of the vapor cloud (12, 13). [For a comprehensive review of past
and current test programs, see USDOE (14).]
This experimental research became the basis for a qualitative, generalized
description of the physical behavior of spilled LNG. When suddenly spilled
on water, the less dense LNG spreads laterally across the water surface,
boiling at a constant mass rate per unit of contact area. The evolved vapor,
being more dense than air, also spreads laterally, simultaneously mixing
with air. As a consequence, the vapor cloud depth is much less than its
width. The vapor cloud drifts in the direction of the wind and atmospheric
turbulence dilutes its vapor concentration. Within the limit of travel where
ignitable proportions are encountered, the cloud remains at ground level.
When spilled on land within a confining enclosure, the LNG boiling rate
decreases with time because of heat conduction into the soil, but the vapor
cloud or plume behavior is not significantly different in character from that
for a water spill.
The difficulty and expense, to say nothing of the danger, of testing LNG
spills of more than a few cubic meters in volume led to the development of
analytical models for predicting the behavior of spills of volumes up to
lOs m3 on land or water (12, 13, 15-19). [For a comparative review of the
water spill models, see Havens (1).] While there is general agreement among
the models on the time to evaporate an LNG spill, there is a wide disparity
in the prediction of the distance of travel of the vapor cloud or plume
beyond which no significant portion of the cloud is flammable (1). As this
distance defines an outer limit to the region surrounding a spill where a fire
94 FAY
extending downwind from the boiling pool, a flame may propagate upwind
Access provided by 180.190.146.56 on 02/08/23. For personal use only.
and therefore a minimum relative speed between the colliding vessels. This
Access provided by 180.190.146.56 on 02/08/23. For personal use only.
Among the possible causes of storage tank failure that have been consid
ered are natural events (e.g. earthquake, hurricane, tornado, tidal wave, and
flood) and man-made events (e.g. aircraft or missile impact and sabotage).
Only for proposed California terminals (18, 30) has earthquake rupture
been considered as an event having a nonzero probability, although all US
terminals would have to meet some seismic standards. One study estimates
the probability of a catastrophic tank failure, given an earthquake that
exceeds the design value, as 10-2 (30). The annual probabilities of failure due
to aircraft striking a storage tank are typically estimated to be about
10-5 per year. As a consequence, land storage tank failure rates from all
causes except sabotage are estimated not to exceed about 10-5 per year
(30, 33).
Risk from LNG systems has been aggregated under two categories,
societal risk and individual risk. Given the set of all events, their conse
quences and probabilities, total societal risk is the sum over all events of the
product of the probability and the consequences (as measured by fatalities
or casualties) of each event. Individual risk, on the other hand, is specific
to a chosen geographic location. It is the sum of the probabilities of all
events of the set, that cause a casualty or fatality at the given location.
Societal risk is often presented as a cumulative probability distribution
of casualties or fatalities, i.e. the yearly probability of an accident with
consequences exceeding a given value, as measured in casualties or fatalities.
The area under such a curve is then the expected value of annual casualties,
which is a measure of societal risk.
A simple single measure of societal risk, which is useful for the purpose
of comparing various risk analyses, is the annual probability, P, of an
accident causing one or more casualties. For all studies, the major contribu
tors to P appear to be a marine spill followed by a vapor cloud drifting
downwind from the spill site and a marine spill at or near the terminal that
ignites at the spill site. Land storage tank failures seem not to be important
98 FAY
probability of release from the LNG tanker while at the unloading jetty is
Access provided by 180.190.146.56 on 02/08/23. For personal use only.
assumed to be 3 X 10-5 per berthing, while for the other studies this proba
bility is assumed to be zero. The contribution to P from a vapor cloud
drifting downwind from the spill site is directly dependent on the assumed
probability that the cloud is not ignited at the source. Values of 1% (31),
10% (32), and 30% (33) have been used for this probability. None of these
values is based upon any empirical evidence such as the frequency of
ignition of oil spills. There are also different criteria used for defining a
casualty or fatality as a function of thermal radiation intensity or exposure
to a flame, e.g. 4.7 kw/m2 in (30) and 17 kw/m2 in (31, 32).
In addition to differences in various event probabilities, which will affect
the calculated value of P, the studies conducted so far explicitly disregard
certain events by assigning zero probability to them. It is claimed that
sabotage of storage tanks should not be difficult (8), but no methodology
for assessing such risks has been advanced. Unconfined vapor cloud explo
sions (as distinct from detonations), which are not uncommonly encoun
tered in the release of other combustible materials (7), have not been
considered despite the significant amount of experience with this phenom
ena. Except for some aspects of the Canvey Island study (33), no large scale
releases resulting from a sequence of events starting with a minor spill or
malfunction have been considered.
LNG import terminal has been examined in two studies (33, 35). The
Canvey Island study (33) considers only truck accidents in the vicinity of
the terminal and concludes that their risks are negligible compared with
other sources. The second study (35) reviews the societal risks associated
with the entire system of landside transport, including up to 20,000 ship
ments per year averaging about 90km of travel per shipment, from an LNG
import terminal located in Everett, Massachusetts. It concludes that the
societal risk probability, P, is 5 X 10-3 accidents per year causing one or
more casualties, or 2.5 X 10-7 such accidents per year per truck shipment.
This overall conclusion apparently depends directly upon an assumption
that the probability of a catastrophic spill of the full contents of a truck (40
m3) is 3 X 10-9 km-I, or 2.7 X 10-7 per average truck shipment of 90 km,
and implies that catastrophic spills are the principal cause of one or more
casualties per accident. It is interesting to note that the societal risk, as
measured by P, for the posited 20,000 annual shipments from the Everett
terminal, is 200 times the value of P estimated for the marine transportation
and land storage system for that same terminal (see Table 1).
The methodology used in this study (35) for analyzing the risk from road
tankers parallels closely that used for marine spills. The major hazard arises
from a fire at the spill site or from a combustible vapor cloud drifting
downwind from the spill. Because of the much smaller maximum spill
volume from a road tanker, compared with a marine vessel, the areas
affected by fire hazard are much smaller. Also, the probability of a major
spill per transit is much smaller (3 X 10-7 for a truck compared with 1�
to 10-5 for a tanker). Nevertheless, the much larger number of annual truck
shipments offsets these factors in determining the value of P.
The very substantial differences in calculated risk among the several
studies examining similar systems reinforces the Fairley criticism that judg
mental bias and lack of experiential substantiation of both probabilities and
consequences greatly vitiates the value of these studies. There does not
100 FAY
appear to be any trend in the later studies toward meeting the Fairley
criteria for substantive adequacy.
LPG RISK
LPG Hazards
As explained in the introduction, LPG hazards are similar in kind to those
encountered with LNG and hence the methods used in analyzing the conse
quences of accidental LPG spills (35, 36) have been patterned after those
described above in the section on LNG hazards. The number of studies
Annu. Rev. Energy. 1980.5:89-105. Downloaded from www.annualreviews.org
regulation of LPG safety is less prominent than that of LNG; also there is
a much smaller scientific base for LPG hazard assessment. Major attention
has been given to vaporization rates in water spills (37) and to the possibili
ties of fiameless vapor explosions (28) described above.
But not all LPG effects have their LNG counterpart. When LPG is stored
as a pressurized liquid at normal temperature, a sudden major failure of the
storage vessel will result in the rapid expulsion of its contents. A portion
of the liquid rapidly vaporizes as its pressure is reduced. The violent motion
of the discharge disperses the remaining liquid, forming a cloud of droplets
and vapor mixed with air. If ignition occurs simultaneously with the dis
charge, a fireball is formed. Such an event has been termed a "boiling liquid,
expanding vapor explosion," or BLEVE (38). A three-stage model describ
ing the formation of a BLEVE cloud, corroborated by small-scale experi
ments, has been suggested by Hardee & Lee (38). Their model predicts that
mixing of the contents of an LPG truck with air in a typical accident can
be accomplished in about 30 sec, a process that proceeds independent of
atmospheric turbulence or gravitational spread as is usually considered in
LNG or refrigerated LPG spills on water or land. Hardee & Lee (39) also
analyze the resulting fireball, calculating that a railroad tank car of LPG
could bum in less than 10 sec and form a fireball whose maximum radius
would be 400 ft. Again this analysis is consistent with a few smaller scale
experiments.
The violence of BLEVEs has prompted Reid (40) to propose that, for a
sudden depressurization of an LPG tank, an unstable superheated liquid
may be formed that explosively releases energy, aiding the fracturing of the
pressure vessel. This may account for the tank fragment projectiles which
sometimes are generated in LPG BLEVE accidents.
Explosion and, possibly, detonation of LPG vapor clouds have been
observed at accident sites (7). While these observations are consistent with
the detonability of LPG air mixtures (26), there is no quantitative model
that can be used to predict the explosive yield or the conditions that will
LNG/LPG RISKS 101
The risk analyses of LNG and LPG systems have been undertaken predom
inantly for the purpose of supplying information on the safety aspects of
proposed facilities as a component of a regulatory or permitting process
which often involves public review. They may be described fairly as objec
tive analyses, in the sense that the methodology invokes accepted scientific
principles, although subjective choice of values of various parameters may
102 FAY
addressed in the LNG and LPG risk analyses reviewed above, attention has
been focused on a comparison of individual risk (as measured by the annual
probability of a fatality or casualty to an individual from a proposed project
or facility) with that from natural hazards (e.g. lightning stroke) and man
made hazards (e.g. automobile accidents and fires). Not only does this
approach fail to address the full range of possible criteria that have been
shown to possess some value, but it fails to account for some elements of
the risk problem that are especially salient to LNG and LPG, such as the
equity issue (lack of proportionate distribution of risks and benefits among
all individuals at risk) and the "zero-infinity dilemma" (very high conse
quence, low probability accident).
There is also some empirical evidence that objective risk, as measured by
historic fatality or casualty rates, is not equivalent to perceived risk as
determined by attitudinal surveys of the general public (42). While some of
the difference can be ascribed to public misconception regarding the relative
sizes of fatality rates, there is also evidence that other factors than fatality
rate affect the perception of risk (42). While none of these studies has
directly considered public perception of LNG or LPG risks, the available
evidence clearly points to a significant distinction between risk perception
and objective risk assessment for all types of risks to which the public is
exposed.
There is a view that risk analysis can play but a very limited role in the
adjudication of controversies related to public risk acceptance. Nelkin &
Pollak (43) claim that "recent disputes have taught us a lesson-that deci
sions about risk are not simply matters of sufficient technical evidence or
adequate information. Substantive issues are at stake which embody highly
controversial political and social values." Bazelon (44) notes that an impor
tant role for the courts is to force a more explicit acknowledgment of where
and how value jUdgements have been used in decision-making process and
to guard against their incorporation into apparently objective scientific
assessments. The history of LNG controversies (2) illustrates well the value
of these cautionary comments on the limits of the usefulness of risk analysis.
LNG/LPG RISKS 103
CONCLUSION
Since the use of LNG and LPG as fuels is likely to increase and will
certainly persist for some time to come, assessment of the safety of
LNG/LPG systems will continue to draw attention. Even in the absence
of major accidents in this industry in the future, continuing public concern
about large hazardous facilities (e.g. petrochemical plants and nuclear reac
tors) is quite likely to force continuing review of operating and design
standards for LNG/LPG facilities. Are we adequately equipped to evaluate
Annu. Rev. Energy. 1980.5:89-105. Downloaded from www.annualreviews.org
the risk of existing systems and those that may be proposed in the future?
Access provided by 180.190.146.56 on 02/08/23. For personal use only.
ACKNOWLEDGMENT
Literature Cited
Annu. Rev. Energy. 1980.5:89-105. Downloaded from www.annualreviews.org
Access provided by 180.190.146.56 on 02/08/23. For personal use only.
16th, pp. 1397-1405. Pittsburgh, Penn: 33. Health and Safety Executive. 1978.
Combust. lnst. Canvey: an investigation of potential
22. Hardee, H. C., Lee, D.O., Benedick, W. hazards from operations in the Canvey
B. 1978. Thermal hazard from LNG IslandlThurrock area. London: HMSO
fireballs. Combust. Sci Tech. 17:189-97 34. Fairley, W. B. 1977. Evaluating the
23. Fay, J. A., Desgroseilliers, G. J., Lewis, "small" probability of a catastrophic ac
D. H. Jr. 1979. Radiation from burning cident from the marine transportation
hydrocarbon clouds. Combust. Sci. of liquefied natural gas. In Statistics and
Tech. 20:141-51 Public Policy, ed. W. B. Fairley, F.
24. Strehlow, R. A. 1973. Unconfined va Mosteller, pp. 331-53. Reading, Mass:
por-cloud explosions-an overview. Addison-Wesley. 397 pp.
Symp. (Int.) on Combustion, 14th, pp. 35. Arthur D. Little, Inc. 1978. Assessment
1189-1200. Pittsburgh, Penn: Combust. of risks and risk control options asso
Inst. ciated with liquefied natural gas truck
ing operations from the Distrigas termi
Annu. Rev. Energy. 1980.5:89-105. Downloaded from www.annualreviews.org