Professional Documents
Culture Documents
Alexander 1996
Alexander 1996
Mid-1990s
DAVID ALEXANDER
This paper gives an overview of the global pattern of casualties in earthquakes which
occurred during the 30-month period from 7 September 1993 to 29 Februay 1996. It
also describes some of the behavioural and logistical regularities associated with
mortality and morbidity in these events. Of 83 earthquakes studied, there were
casualties in 49. Lethal earthquakes occurred in rapid succession in Indonesia, China,
Colombia and Iran. In the events studied, a disproportionate number of deaths and
injuries occurred during the first six hours of the day and in earthquakes with
magnitudes between 6.5 and 7.4. Ratios of death to injury varied markedly (though
with some averages close to 1:3), as did the nature and causes of mortality and
morbidity and the proportion of serious to slight injuries. As expected on the basis of
previous knowledge, few problems were caused by post-earthquake illness and disease.
Also, as expected, building collapse was the principal source of casualties: tsunamis,
landslides, debris flows and bridge collapses were the main seconday causes. In
addition, new findings are presented on the temporal sequence of casualty estimates
after seismic disaster. In synthesis, though mortality in earthquakes may have been
low in relation to long-term averages, the interval of time studied was probably typical
of other periods in which seismic catastrophes were relatively limited in scope.
North and Central America, Europe and excludes the two greatest events, the Latur
central Asia are under-represented, but earthquake of September 1993 in India and
this depends on where earthquakes the Kobe (Japan) earthquake of January
caused deaths during the study period, 1995,' the figure for deaths remains high at
not where they occurred, which gives a 69 per cent. However, the figure for
different aggregate pattern. injuries is affected by the evening-time
Seven of the events studied generated earthquake that occurred in Yunnan Pro-
tsunamis, with a maximum recorded wave vince, China, in which 16,925 people were
height of 10-15 m. Six of the tsunamis non-fatally injured in the collapse of a very
occurred in the eastern Pacific basin and large number of weak-walled masonry and
each led to fatalities, while the seventh adobe buildings. Therefore, if this event is
tsunami affected Guatemala and did not excluded, the percentage of injuries that
cause deaths. Five earthquakes occurred occurred between midnight and 6 a.m.
with landslides (four in the east Pacific and rises to 96, but, surprisingly, only 43 per
one in Colombia). Fatalities occurred in cent of the earthquakes on the list occurred
three of these: due to soil slips in Taiwan during the first six hours of the day. This
in February 1995 and East Timor in May suggests that the risks of injury may be
1995, and as a result of debris avalanches significantly higher at night in many
in Colombia in June 1994. The latter killed different physical and architectural
villagers and five Red Cross workers. settings.
Previous studies examined the
hypothesis that there is an average of one
BASIC FINDINGS
death for every three significant injuries in
Figure 2(a) shows that 86 per cent of the seismic disasters (PAHO, 1981). On the
deaths and 97 per cent of the injuries listed other hand, long-term data collected by
in Table 1 were caused by earthquakes in the Red Cross suggest a ratio of 1:1.3,
the magnitude range 6.5-7.4, although though this probably refers to a count of
these constituted only 45 per cent of the only the more serious injuries (IFRCRCS,
events listed. The most plausible explana- 1994, p. 150). Some years ago, in a study
tion for this discrepancy is that smaller of two decades of basic data on earthquake
events were too weak to cause major loss casualties, Alexander (1985) observed that
of life and that larger events were rare and there is no standard definition of injury,
tended to occur in sparsely populated no specific injury typology for earth-
areas. However, it is unwise to speculate quakes, and no threshold for injury sever-
too freely without data on, for example, ity to define the level at which victims are
the duration and acceleration of shaking, classified as injured. Earthquake-induced
the quality and state of maintenance of injuries are not internationally notifiable
housing stock and the occurrence of and governments tend not to have
secondary hazards such as fire: in the 1972 standard procedures for documenting
earthquake in Nicaragua, these factors them. Despite this, Alexander concluded
combined to create a death toll of more that the ratio had some limited validity for
than 4,000 on the basis of an event with a earthquakes in the magnitude range
magnitude of only 5.6 (Whittaker et al., 6.5-7.4.
1974). In the present enquiry, death-injury
Figure 2(b) shows that 94 per cent of ratios varied by several orders of magni-
the deaths and 77 per cent of the injuries tude. In part this is because events with
occurred between the hours of midnight very few casualties did not generate stable
and 6 a.m., local time. Even if one ratios. In addition, there are anomalous
TABLE 1
Earthquakes that caused deaths and injuries, 1 September 1993-29 February 1996
TABLE 1
(Continued)
*Does not include Sakhalin (Russia) earthquake of 28 May 1995 (deathslinjuries = 4.89)
(a)
Deaths: n = 39, Injuries: n = 41
70,000
u)
a, 60,000
2
'q
-550,000
40,000
r.
(0 30.000
u)
s, 20,000
& 10,000
(D
n
"
5.25 5.75 6.25 6.75 7.25 7.75 8.25
Magnitude cfass mid-point
(b)
Deaths: n = 37, Injuries, n = 34
70,000
Q, 60,000
u)
*g50,000
*Z
U 40,000
e
@ 30,000
u)
s(D 20,000
8 10,000
n
U
01.30 04.30 07.30 10.30 13.30 16.30 19.30 22.30
Time period (3-hour blocks)
FIGURE 2 (a) Deaths and injuries by magnitude; (b) Deaths and injuries by time of day
data, such as the very high mortality to before the victim could be removed from
morbidity ratio in the Sakhalin Russian the rubble. Severe and fatal burn injuries
earthquake of May 1995, where victims were also widely reported, especially as a
had little opportunity to survive the catas- result of events in India, Indonesia, Japan
trophic collapse of buildings. Neverthe- and Russia: reports were about equally
less, if Sakhalin is excluded, the mean distributed between the effects of domestic
ratio is very close to 1:3, which implies fires (burns while cooking, scalding with
either that the hypothesis has some resi- boiling water and so on) and more general
dual validity or that it is a very enduring urban post-earthquake fire. The latter was
spurious correlation. Moreover, the mean a particularly important cause of death and
ratio is 1:3.22 for earthquakes in the injury in the Kobe earthquake of January
magnitude range 6.0-7.4, which tends to 1995, which gave rise to 531 urban fires
confirm my earlier findings (Alexander, (AGU, 1995).
1985). In synthesis, despite the allure of After the Kobe earthquake, full statis-
the death to injury ratio, it could not be tics on injury were published by the
correlated more precisely with magnitude. Medical Examiner’s Office of Hyogo Pre-
Nor, for the earthquakes studied, did it fecture. About 71 per cent of 3,649 victims
correlate with time of day or number of autopsied died within 14 minutes of the
buildings damaged. earthquake, and a further 10.7 per cent
may have died within six hours of the
tremors. Some 54 per cent died as a result
THE NATURE, CAUSES AND SETTINGS
of crush injuries or other physical trauma
OF INJURY
sustained in the collapse of buildings,
Not all reports gave details of the nature or while 15.3 per cent suffered severe burns
severity of injuries sustained, and hence a (Anon., 1995). More comprehensive statis-
full injury epidemiology could not be tics published a year later reduced this
constructed. However, the nature of injur- figure to 10 per cent (Disasfer Research,
ies was amply referred to, which enabled a 1996).
fairly comprehensive picture of injury In the earthquakes studied, death also
types to be assembled. Physical trauma resulted from dehydration, choking and
included head and back injuries, leg frac- suffocation. The last of these was a very
tures, broken ribs, multiple fractures of significant cause of mortality in the Kobe
limbs, clavicle fractures, spinal damage, earthquake (ibid.). In the Pereira (Colom-
paraplegia, cuts, bruises, lacerations, bia) earthquake of February 1995, dust
burns and crush injuries. Alexander (1993, generated by building collapse led to the
p. 470) described similar injury types in death of some trapped victims, while
other earthquakes but noted that the others suffered dehydration before they
proportions varied considerably from case could be rescued from under the rubble of
to case. In the Latur (India) earthquake of their homes. In nine other events, and in a
1993, upper limb injuries accounted for wide variety of settings, a small number of
nearly one-quarter of the very small deaths occurred as a result of acute myo-
number of people who were hospitalised cardial infarction (heart attack); forexam-
(IFRCRCS, 1994, p. 120). Crush injuries ple, five such fatalities were recorded in
were widely reported in earthquakes in the January 1994 Northridge, California,
the Philippines, Japan and Russia. In earthquake. Victims were predominantly
addition, at least one case during the over the age of 50 (cf. Katsouyani et al.,
Sakhalin disaster of May 1995 involved 1986). Three sets of reports gave details of
crush effects that necessitated amputation suicides that were apparently linked with
TABLE 2
Casualties in relation to building damage for selected earthquakes, 1994-1996
(Table 2). Hence, in a field study in Italy, criteria used in the compilation of statis-
De Bruycker et al. (1985) found that being tics. However, 10 of the disasters studied
trapped under rubble increases a victim’s did end with the dissemination of a crude
chances of being injured fivefold. Data on classification of injury severity into serious
damage were reported for 16 of the earth- (requiring hospitalisation or otherwise
quakes studied, but little pattern is evident requiring protracted medical attention)
in the ratio of total casualties to the total and slight (needing only limited outpatient
number of buildings damaged (last col- status). Table 3 shows that for these
umn, Table 2, mean 0.178, standard devi- events the proportions varied widely from
ation 0.182). The number of casualties 84 per cent serious to 99 per cent slight
varied between one and 45 per 100 build- (though, of course, the number of casual-
ings damaged, with a mean of 18 per 100. ties varied by about 2 orders of magni-
In contrast, there is slightly more regular- tude). With regard to the latter figure, the
ity in the average ratio of deaths to tiny proportion of people who remained in
buildings destroyed, which suggests a hospital after the Latur earthquake of 1993
fatality rate of 10-16 people per 100 build- in India (only 125 out of 10,500 injured, see
ings suffering total or partial collapse. IFRCRCS, 1994, p. 120) recalls the 1968
Apart from the causes of mortality and Khorasan tremors in Iran, in which only
morbidity, one other question concerns 3.3 per cent of 11,000 injured people
the classification of non-fatal injury. required prolonged inpatient care (Rennie,
Official statistics rarely differentiate 1970).
between injuries sustained according to In sum, no pattern is evident in the
their seriousness and, if they do, it is small number of available data, which is
highly unusual to find any explanation of perhaps hardly surprising given the wide
-
DISASTERS VOLUME 20 NUMBER 3
240 David Alexander
TABLE 3
Proportion of serious and slight injuries in selected earthquakes, 1993-1996
Date Locat ion Local time Magnitude Total injured Seriously Slightly
of day injured (TO) injured (%)
people were rescued alive nearly 48 hours (Figure 4) that eventually converges upon
after the event, but by and large the the official totals. The latter rose with a
earthquakes studied confirm the well- regularity not found in other earthquake
known rule that most living victims are disasters (for example, see Alexander
rescued within the first 24 hours following 1982b, p. 82), which is testimony either to
an earthquake (Noji, 1991). Rescue ope- the orderliness of the Russian authorities’
rations then largely concentrate on reco- handling of the emergency, or to the
vering bodies and accounting for people unvarying difficulty of recovering the
who have been reported missing. The bodies. As it is common to find that the
Kobe earthquake of January 1995 is a numbers of casualties are over- or under-
typical case (Figure 3). Although it took 25 estimated in the period immediately after
days to establish preliminary totals, and an earthquake, it would be helpful to
final totals were eventually published a study this aspect in more disasters so as to
full year later, most of the dead, missing arrive at better early predictions of death
and injured were accounted for within a and injury tolls. Such estimates have a
week of the catastrophe and thereafter the considerable bearing on the number and
curves home in asymptotically upon their quality of medical and logistical resources
final values. By contrast, the total of that are sent into disaster areas.
houses damaged and destroyed rose
steadily for weeks after the disaster as
EVALUATION
buildings and their sites were methodi-
cally surveyed. However, it should be According to the International Federation
noted that over the year following the of Red Cross and Red Crescent Societies,
disaster, official death tolls were increased mortality and morbidity in earthquake
by 15 per cent and injury totals by 30 per disasters averaged 22,956 and 30,003, res-
cent as the definitions of casualties were pectively, over the period 1968-92
gradually altered to include people who (IFRCRCS, 1994, pp. 148-50). Though the
had died of their injuries and those who figure for morbidity (35,560) is slightly
were treated some time after the higher in this study, the rather lower
earthquake. figure reported here for mortality (7,960)
In the light of previous studies, Alex- results from the fact that the period stu-
ander (1993, p. 18) observed that the died was free from major catastrophes
official total of casualties in a disaster rises such as the magnitude 7.8 earthquake that
steadily for some days on the basis of struck Tangshan, China, in 1976 and killed
head- and body-counts, but officials who 242,469 people (Coburn and Spence, 1992,
are involved in search-and-rescue ope- p. 5), and the magnitude 7.3 disaster in
rations tend to overestimate the totals. The Iran in 1990 that killed 40,000. In such
‘overshoot’ of estimated figures then con- events, the scale and severity of destruc-
verges on the official figures until a final tion were so great that mortality tended to
total is arrived at about 2-3 weeks after the be high in relation to morbidity, hence the
disaster. Data from the Sakhalin earth- discrepancy in death to injury ratios
quake of May 1995, derived from both between the Red Cross totals and those
official Russian government communiques reported in this study (1:1.31 and 1:3.22,
and interviews with rescue workers, con- respectively). It should be noted also that
firm this model for non-fatal injuries but overall mortality in earthquakes is concen-
suggest that the death tolls were not only trated heavily in the largest events and in
overestimated, but also underestimated. particular places (Coburn and Spence,
The result is an ‘envelope’ of estimates 1992). The Red Cross’s long-term data
6,000 I
5,000 -
Final totals published
Dead
one year later:
6,308dead,
2 missing
17 18 20 22 24 26 28 30 I 3 5 7 9
JanlFeb 1995
"17 18 20 22 24 26 28 30 1 3 5 7 9
JanlFeb 1995
Q 1 ~ ' ~ ' l ' i ' ~ ' ~ ' ' ' ~ ' ~ ' " ' ' ~
17 18 20 22 24 26 28 30 1 3 5 7 9
JanlFeb lSg5
FIGURE 3 Temporal pattern of accounting for deaths, missing persons, injuries and damage to housing in
the Kobe, Japan, earthquake of 17January 1995. Data were supplied by the lapanese Government and
disseminated widely on the Internet.
(IFRCRCS, 1994) suggest that 35 per cent events each year from 1900 until 1992,
of earthquake disasters occur in Asian which suggests that perhaps only half the
countries, which makes the 63 per cent former total may have involved fatalities.
recorded during this study period (and the This is corroborated by Alexander (1985),
96 per cent of deaths) anomalously high, who studied earthquakes over 21 years
though perhaps less so when one takes and found that seven to 11 of them
into account the observation (Coburn et resulted in significant numbers of deaths.
al., 1989) that half of all recorded deaths in In terms of mortality per earthquake, the
earthquakes have occurred in China. Red Cross's annual average figure of
In synthesis, the period studied here 22,956 is higher than that quoted by Alan
may well be typical of intervals of time Feuerbacher in an Internet communication
during which earthquake casualties are dated 1994. He suggested a mean of 18,672
relatively limited. However, this assertion seismic deaths per annum for the first nine
is difficult to back up with quantitative decades of the present century; however,
information, in particular because histori- it appeared that this figure was based on
cal earthquake data tend to be under- an average of only 2.8 events each year,
reported and unreliable, and hence long- which implies considerable under-report-
term averages may be suspect. While the ing of earthquakes that involved relatively
Red Cross gave an average of 25.7 signifi- few deaths. Feuerbacher's data for the last
cant earthquakes per annum over the five centuries showed a drop of nearly
period 1968-92 (IFRCRCS, 1994), Coburn three-quarters in seismic mortality per
and Spence (1992) listed only 12.7 lethal million of the world population in the
'.. ....
, ...-....
: o -........ -..._ Envelope of estimates
-.. .-.._.....................
.......... / of number of dead
....
.K-... .
.-..........
_ _ _ _.-O...---o--
_- 0 ___. U". 0. .I -5, -- ...........
0 0 0
. . .. . .. .
_-....
_.__.-
-.
----.-.- . . .
.. ,-_._-..
0 -.......... -.,_.-.- 0 -
0. ~ ~
. -.4. -\
,._.-..
....__.-
'., /..'
,.."
_/-
Official figures
'. ,. -..'_,_.*'. - - b
-i. (no. of dead)
28 29 30 31 1 2 3 4 5 6 7 16
May June 1 995
FIGURE 4 Estimates and official figures for the number of deafhs and ifijuries in f h e Sakhalin, Russia,
earthquake of 28 M a y 1995
twentieth century, but as the number of disasters that were separated from one
deaths remained stable over the years, this another by only a few weeks, and in
implies that the fall is largely accounted for Yunnan Province, China, by only a few
by increasing numbers of people, not by months. But despite the frequency of
decreasing death tolls. As the present events, only about three-fifths of the
century’s 20 most lethal earthquakes damaging earthquakes that occur around
involved a mean mortality of 52,900 the world generate mortality or morbidity.
(Coburn and Spence, 1992, and other Not only do these vary widely in earth-
sources), all that can be said with certainty quake disasters, but so do the proportions
about the mid-1990s is that no event then of serious to slight injuries. The pattern of
led to loss of life on such a scale. earthquake casualties in single events can
Discrepancies between long-term data vary from a few people with minor injur-
and the results of the present study call ies, or one or two heart-attack victims, to
into mind a series of criticisms that can major disasters in which loss of life is
possibly be levelled against macroscopic heavier than the toll of non-fatal injuries.
studies such as the present one. For Leaving aside the question of very large
instance, it is well known that death and earthquake catastrophes (with death tolls
injury during earthquakes depend on a of at least 30,000), none of which occurred
wide variety of factors (Coburn and during the study period, casualties may be
Spence, 1992), including physical variables greatest in the early hours of the day and
(hypocentral depth, duration of shaking, in events of intermediate magnitude
seismic acceleration, geology of the local (6.5-7.4).
terrain, distance from epicentre), architec- In earthquakes that generate only
tural variables (construction type, state of moderate numbers of casualties (no more
maintenance of buildings, enactment and than a few score victims), mortality and
enforcement of anti-seismic building morbidity tend to be clustered in particular
codes) and social variables (patterns of locations, such as individual schools,
human activity, level of popular aware- hotels, apartment complexes or groups of
ness of earthquakes, counter-disaster vernacular houses. Despite the publicity
training, efficiency of rescue operations), given to landslides and tsunamis, these
none of which is explicitly considered tend to cause far fewer injuries than does
here. Given these complexities, one may building collapse. Moreover, the data ana-
ask whether global surveys have any lysed here suggest that seismic effects
validity at all. However, the presence of upon transport (bridge collapse and vehi-
regularities in the data analysed here cular crashes) may be much less important
indicates that the exercise was indeed sources of casualty than the crushing and
worthwhile, and some broad generalisa- trapping of victims under collapsed
tions can be made about earthquakes, buildings.
even if these tend to be more akin to Physical trauma tends to receive more
partially tested hypotheses than to phy- attention than psychological impairment,
sico-social laws. though it appears that Wallace’s ‘disaster
Accordingly, this article ends with a syndrome’ (Wallace, 1956 - and see
series of provisional conclusions that may above) could well be a universal result of
serve as hypotheses for further investi- seismic disaster, at least among a certain
gation. First, earthquakes are highly repe- proportion of survivors. Internationally,
titive events that tend to cluster in both the phenomenon is not as well studied as
time and space. In Indonesia and Iran, for panic, though it may be at least as wides-
instance, casualties occurred in seismic pread. The present data suggest that a
Armenian, H.K., E.K. Noji and A.P. Oganes- of Social Order: Popular Myth, Social
sian (1992) Case Control Study of Injuries Theory, Empirical Evidence. Sociological
Due to the Earthquake in Soviet Armenia. FOCUS 29(3), 171-83.
Bulletin of the World Health Organization 70, Jones, N.P., E X . Noji, F. Krimgold and G.S.
251-57. Smith (1990) Considerations in the Epide-
Blake, P.A. (1989) Communicable Disease miology of Earthquake Injuries. Earthquake
Control. In M.B. Gregg (ed.) The Public Spectra 6, 507-28.
Health Consequences of Disasters. Federal Katsouyani, K., M. Kogevinas and D. Tricho-
Centers for Disease Control, Atlanta, 5-20. poulos (1986) Earthquake-related Stress and
Coburn, A.W., A. Pomatis and S. Sakai (1989) Cardiac Mortality. International Journal of
Assessing Strategies to Reduce Fatalities in Epidemiology 15, 326-30.
Earthquakes. lnternational Workshop on Earth- Noji, E.K. (1989) Medical and healthcare
quake l nj uy Epidemiology for Mitigation and aspects of the 1988 earthquake in Soviet
Response. Johns Hopkins University Press, Armenia. In L.A. Wyllie, Jr and J.R. Filson
Baltimore, 107-32. (eds.) Armenia Earthquake Reconnaissance
Coburn, A. and R. Spence (1992) Earthquake Report. Earthquake Spectra 5 (supplement),
Protection. Wiley, New York. 101-7.
De Bruycker, M., D. Greco and M.F. Lechat Noji, E.K. (1991) The Medical Consequences of
(1985) The 1980 Earthquake in Southern Earthquakes: Co-ordinating the Medical and
Italy: Mortality and Morbidity. lnternational Rescue Response. Disaster Management 4(1),
Journat of Epidemiology 14, 113-17. 3-11.
Degg, M. (1993) The 1992 Cairo Earthquake: Olson, R.S. and R.A. Olson (1987) Urban
Cause, Effect and Response. Disasters 17(3), Heavy Rescue. Earthquake Spectra 3(4),
226-38. 645-58.
Disaster Research, 31 January 1996. Internet Page, R.A ., J.A. Blume and W .B. Joyner (1975)
Bulletin of the Natural Hazards Research Earthquake Shaking and Damage to Build-
and Applications Information Center, ings. Science 189, 60148.
University of Colorado, Boulder. PAHO (1981) Emergency Health Management
Durkin, M.E. (1987) Casualties, Search and After Natural Disasters. Scientific Publication
Rescue, and Response of the Health Care 407, Pan American Health Organization,
System. (The San Salvador earthquake of Washington.
October 10, 1986.) Earthquake Spectra 3(3), Rennie, D. (1970) After the Earthquake. The
621-34. Lancet 2, 704.
Durkin, M.E., C.C. Thiel Jr, J.E. Schneider and Roces, M.C., M.E. White, M.M. Dayrit and
T. De Vriend (1991) Injuries and Emergency M.E. Durkin (1992) Risk Factors for Injuries
Medical Response in the Loma Prieta Earth- Due to the 1990 Earthquake in Luzon,
quake. Bulletin of the Seismological Society of Philippines. Bulletin of the World Health Orga-
America 81(5), 2,143-66. nization 70, 509-14.
Ganse, R.A. and J.B. Nelson (1981) Catalog of Sapir, D.G. (1993) Health Effects of Earth-
Significant Earthquakes, 2000 B.C.-1979. quakes and Volcanoes: Epidemiological and
National Geophysical Data Center, National Policy Issues. Disasters 17(3), 255-62.
Oceanographic and Atmospheric Administ- Titchener, J.L. and F.T. Kapp (1976) Family and
ration, Boulder. Character Change at Buffalo Creek. American
Glass, R.I., J. J. Urrutia, S. Sibony, H. Smith, B. Journal Of PsychiatT 133(3),295-9.
Garcia and L. Rizzo (1977) Earthquake Injur- Wallace, A.F.C. (1956) Human Behaviour
ies Related to Housing in a Guatemalan During Extreme Situations. Disaster Study
Village. Science 197, 638-43. No. 1, National Academy of Sciences,
IFRCRCS (1994) Wortd Disasters Report 1994. Washington.
International Federation of Red Cross and Whittaker, R., D. Fareed, P. Green, P. Barry
Red Crescent Societies, Geneva, and and A. Borge (1974) Fletes-Barrios Earth-
Kluwer, Dordrecht. quake Disaster in Nicaragua: Reflections on
Johnson, N.R. (1987) Panic and the Breakdown the Initial Management of Massive Casual-