Accidents during marine transport of dangerous goods. Distribution of fatalities

Accidents during marine transport of dangerous goods. Distribution of fatalities

Accidents during marine transport of dangerous goods. Distribution of fatalities Hans [email protected]*j_, Palle Haastrup* and H. J. Styhr Petersen? *European Co...

630KB Sizes 1 Downloads 17 Views

Accidents during marine transport of dangerous goods. Distribution of fatalities Hans [email protected]*j_, Palle Haastrup* and H. J. Styhr Petersen? *European Commission, Joint Research Centre Italy tTechnica1 University of Denmark, Department Building 229, DK-2800 Lyngby, Denmark

Ispra,

t.p. 650, I-21020 Ispra,

of Chemical

Engineering,

Received 30 June 1994 On the basis of 2781 accident case histories, the consequences measured by the number of fatalities from marine accidents (n = 1780) during transport of dangerous goods have been investigated and compared with those from other transport modes (n = 1001). Accidents from marine transport of dangerous goods have been found to have a larger proportion of accidents with fatalities in the range of IO-50 than other transport modes. Therefore, f/N curves for marine accidents are not similar to straight lines as usually seen but have a hump. This is probably due to the size of the potentially affected population, which is often in the range of IO-50 during marine transport, reflecting the number of crew members on one or two vessels; further, the population potentially affected is placed in a limited area. Almost all accidents with more than 40 fatalities were collisions, and accidents with more than 100 fatalities were due to collisions between tankers and ferries, which significantly increases the population at risk. In these accidents, the dangerous goods were oil. The high number of fatalities is not surprising, as oil at sea has the potential for surrounding a vessel and catching fire. Differences have been found for the distribution of fatalities between different local surroundings and transport phases. Similarity has been found for the distribution of fatalities for type of cargo, tank type, geographical location and year of accident. Keywords:

transport;

dangerous

goods; human consequences

have previously shown that Haastrup and Brockhoff’ the distributions of the number of fatalities from accidents with hazardous materials related to transport and those related to fixed installations are similar. Further, similarity was shown between the different transport modes: road, rail and pipeline. In 1992, a research project on the marine transport of dangerous goods was initiated by a pilot studyz. The project involves cooperation between the Department of Chemical Engineering, Technical University of Denmark, and the Joint Research Centre, Ispra, Italy. The further studies presented here are based on a newly established database on marine and inland waterway accidents occurring during transport of dangerous goods. It contains 1780 accident records, each described by 34 fields. The accidents occurred worldwide between 1945 and 1993, and details have been found in the open literatureS12. The results concerning transport over land are based on a previously established database. In total, 2781 accident case histories have been investigated. The only criterion for including an accident in the marine database is that dangerous goods either: (i) were transported 0950-4230/95/01002%06 0 1995 Butterworth-Heinemann

by any kind of marine

vessel,

(ii) were transferred to/from any kind of marine vessel, or (iii) had been transported by a marine vessel, and the vessel had a fire or explosion in ballast or during tank cleaning or repair It is a condition that the dangerous substances were carried/transferred as cargo and not for instance as bunker fuel or for refrigeration (e.g. ammonia on fishing vessels). The parameters examined in this paper are shown in Table 1. flN curves were chosen as the standard way of expressing the results. Visual comparison of the distributions of fatalities for the parameters given was then used to judge the influence of each parameter. Whether vessels carrying dangerous goods have higher frequencies of accidents than others is not examined in this article. However, a rough analysis of data in COST 30114 and DAMAl indicates that there is no reason to assume significantly different frequencies for the incident types collision, grounding and foundering, whereas there might be a higher frequency of fire/explosions for ships carrying dangerous goods.

Ltd

J. Loss Prev. Process Ind., 1995, Volume 8, Number

1

29

Marine

transport

of dangerous

goods:

Table 1 The parameters

examined

Parameter

Options examined

Involvement of cargo Local surroundings Tank type Time period Geographical location of accident Transport phase Type of cargo Accident type Transport mode

H. [email protected] et al.

Port, inland waterways, restricted waters, coastal waters, open sea Tanks, others 1946-1972,1973-1993 Europe and USA, rest of the world Sailing, transfer, empty tanks Oil, LPG and similar, other than oil Collisions, others Marine/inland waterways, road/rail/ pipeline

Results The results of the analysis are discussed the nine parameters shown in Table 1.

according

to

Cargo influence on fatalities Each accident has been evaluated with respect to whether the dangerous goods had any influence, direct or indirect, on the number of fatalities (N) in the given accident. An example of an accident in which the dangerous goods were not involved is the sinking of a vessel with the cargo intact, in which death of the crew is by drowning. In case of doubt, it has been assumed that the cargo did have an influence. This has often been the case in fires and explosions. Figure 1 shows the f/N curves for those accidents where the cargo did or did not have an influence on N. The distributions are similar up to a certain point. The

Conditional probability P[x>NIN>

11

---~Cargoinfluence I - - - - - No cacao influence

curve for accidents where N is not influenced by the cargo seems to fall strongly at approximately N = 30, whereas the curve for accidents influenced by the cargo reaches almost 5000. Hence, the probability of involving more than one crew in the accident seems to be larger when dangerous goods are involved than when they are not. However, the latter conclusion is based on only a few observations, and it is possible that an oil tanker may penetrate a ferry such that the ferry goes down and the oil cargo is intact. An example of this occurred in 1980 when the ferry Don Juan went down after a collision in the Tablas Strait, Philippines, with the oil tanker Tacloban City which was in ballast (i.e. with emptied tanks). The tanker was almost intact, but the overcrowded ferry went down, and, according to the local press, a thousand passengers lost their lives and a further thousand were rescued. Local surroundings The accidents have been divided into five categories: inland waterways, restricted waters, coastal port, waters and open sea. Restricted waters are defined as narrow waters with heavy traffic, e.g. straits and sounds (less than 50 miles between the two nearest shores). Coastal waters are defined as less than 50 miles to the nearest shore. The distribution of the accidents is shown in Table 2. Since restricted waters, coastal waters and open sea transport (marine transport) are assumed to be of a similar type, and the f/N curves have proved to be very similar, they have been pooled as ‘marine transport’. In Figures 2 and 3, marine transport accidents are compared with inland waterway and port accidents, respectively. The curves are slightly different in both cases. There is some over-representation of accidents with N between 5 and 30 for marine transport. The curves for ports and inland waterways do not have this characteristic hump, but are more like the usual straight lines seen for accidents on land. Tank type, time period and geographical location Comparisons have been made of accidents that occurred on tankers and those that occurred on vessels with other forms of containment, e.g. bags, dry bulk, containers and drums. Further, accidents for 1945-1972 have been compared with those for 1973-1993, and

Table 2 Accident case histories from the open literature (n = 1780) were subdivided according to local surroundings. Marine includes open sea, coastal waters and restricted waters

1

10

100

1000

Number of fatalities, N Figure 1 Distribution of the number of fatalities from accidents in which the cargo influenced the number of fatalities W) (169 accidents) and accidents in which it did not (293 accidents)

30

J. Loss Prev. Process

Ind.,

1995, Volume

Local surroundings

Number of accidents

Accidents with NZl

Port Inland waterways Marine Unknown

439 224 1070 47

133 36 281 13

Total

1780

463 (26%)

8, Number

1

(30%) (16%) (26%) (28%)

Marine transport of dangerous

goods:

H. Romer et al.

Conditional probability P[xrNIN>

Table 3 The accident transport phase

l] I

I.I_., I ..._I Mar&

1

10-l

case histories

subdivided

according

to

Phase

Number of accidents

Accidents with N>l

Sailing Cargo transfer Empty tanks Other

1188 270 130 192

240 86 96 41

Total

1780

463 (26%)

(20%) (32%) (74%) (21%)

1o-2 under-reporting of accidents fatalities in the earlier period.

with

low

numbers

of

1o-3

Phase Table 3 shows the accidents

Number of fatalities, N Figure2 Distribution of the number of fatalities (N) from accidents that occurred at sea (marine, 281 accidents) compared to those that occurred on inland waterways (36 accidents)

Conditional probability P[xrNIN>

11

1

I

_.m. . s-,-sMarine . . . . . po*

subdivided according to the transport phase. The sailing phase is defined as the actual transport of the goods, and empty tanks include ballast, repair, tank cleaning and gas freeing. The high percentage of accidents with fatalities for empty tanks is probably due to the criteria of selection for the database, as these accidents are only included if there is a fire or explosion. The reason for the higher percentage of accidents with more than one fatality for cargo transfer compared to sailing might be that these accidents often include a release (49% of the accidents), whereas the corresponding value for the sailing phase is 34%. The f/N curves for empty tanks and cargo transfer were similar and have been pooled. The pooled curve is compared with the sailing phase in Figure 4. The

Conditional probability

10-l

P[x>NINr

11 I

1o-2

.lll...l-l-l Sailing

1

10-l

1o-2

Number of fatalities, N Figure3 Distribution of the number of fatalities (N) from accidents that occurred at sea (marine, 281 accidents) compared to those that occurred in ports (133 accidents)

those from the geographical area of Europe and the USA have been compared with those from the rest of the world. The f/N curves in these comparisons were almost identical. Differences might have been expected between the time period curves due to an assumed

1

10

100

1000

Number of fatalities, N Figure4 Distribution of the number of fatalities (N) from accidents that occurred during sailing (240 accidents) compared to those that occurred during cargo transfer and on vessels with empty tanks (184 accidents)

J. Loss Prev. Process Ind., 1995, Volume 8, Number

7

31

Marine

transport

of dangerous

goods:

H. [email protected] et al.

curves are not similar. The one for empty tanks and cargo transfer is almost a straight line, whereas the one for sailing has a hump, which reflects the characteristically high occurrence of accidents with 10-50 fatalities during the sailing phase.

Conditional probability P[x>NIN>

l] I

....-.......,,.~. Oils ----- Otherthanoils

1

Accident type The accident type collisions is especially interesting since this is the only one that has to include more than one vessel. Figure 5 shows that accidents other than collisions usually give below 40 fatalities whereas collisions have a higher potential for more than 40 fatalities.

10-l

1o-2

Cargo The effect of the cargo was initially assumed to be high, due to the fact that, for example, the potential number of fatalities is assumed to be much higher for liquefied petroleum gas (LPG) and similar substances than for oil. The analysis does not confirm this hypothesis. LPG and similar substances includes ethylene, propylene, propane, butylene and butane. To create large enough samples for the sailing phase, we have chosen to compare all types of oil with all other cargoes (Figure 6). The curves are close to identical while N s 40. That the curve for oil goes to higher values of N is probably because oil has the potential for floating on the water and surrounding other vessels with fire, e.g. a ferry with which the tanker has collided. A comparison has been made between f/N curves for oil and for LPG and similar substances for road,

rail and pipeline transport and transfer accidents (Figure 7). They both resemble straight lines and are very close to each other. This leads to the conclusion that the distributions of fatalities from accidents during

Conditional probability

Conditional probability

P[x>NINr

P[x>NIN>

11

1

-----

Other than

100

1000

Number of fatalities, N Figure 6 Distribution of the number of fatalities (N) from marine and inland waterway accidents that occurred during the sailing phase. Accidents that occurred during transport of oil (106 accidents) are compared to those where the cargo was other than oil (129 accidents)

Land transportation

11

-.-.-.-.-I (&llisiom I

10

I

collisions

-.-.-,-....oils ----- LPGandsimilar

1

1

10-l

1o-2

1

10

100

1000 Number of fatalities, N

Number of fatalities, N Figure 5 Distribution of the number of fatalities (IV) from marine and inland waterway accidents that occurred during the sailing phase. Those in which the first registered event was a collision (65 accidents) are compared to those in which it was not (168 accidents)

32

J. Loss Prev. Process

Ind.,

1995, Volume

Figure 7 Distribution of the number of fatalities (IV) from accidents during road, rail and pipeline transport (including cargo transfer accidents). Those that occurred during transport of oil (52 accidents) are compared to those that occurred during transport of LPG and similar substances (72 accidents)

8, Number

1

Marine carriage cargo.

of the goods

are similar

for different

types of

transport

of dangerous

goods:

H. [email protected] et al.

Conditional probability P[x>NIN>30] ...u”...l~,

Comparison of transport modes Since it has been shown above that there are differences in the distribution of the number of fatalities from accidents during the sailing phase and other phases such as cargo transfer and empty tanks, it has been chosen to restrict the comparison of transport modes to the sailing phase for marine and inland waterway transport and the corresponding phase for road, rail and pipeline transport. In Figures 8 and 9, the differences in the distributions of N from marine and inland waterways and road/rail/pipeline transport are examined. In Figure 8, the curves are different. It appears that the probability of escalation to up to 40 fatalities is higher for marine transport. This can be explained by the fact that the crew size of one or two vessels, and thereby the normal potential number of fatalities, is between 10 and 50 persons. In Figure 9, it is seen that the risk of escalation of the number of fatalities, given at least 30 fatalities, is higher for accidents on land. This is due to the relatively high number of reported marine accidents in the interval 30-40. For catastrophes with more than 50 fatalities, the probability of escalation of the number of fatalities was shown to be similar for transport by land and water. However, this is based on only very few observations, and there are in fact differences in the number of land and water transport accidents with more than 50 fatalities. Table 4 gives an overview of the types of cargo involved in these accidents. Oil products dominate for transport by water whereas

Conditional probability P[x>NIN>

.

.

.

.

.

1

10-l

1o-2

1o-3

/

I

I

1

10

(

I

I

I

I

100

I

1000

Number of fatalities, N Figure 9 Distribution of fatalities from accidents that occurred during the movement of goods by marine and inland waterways (29 accidents) and road, rail and pipeline (10 accidents) and which resulted in 30 or more fatalities

Table 4 Accidents with 50 or more fatalities subdivided according to type of cargo. Only accidents that occurred during movement of the goods are shown Number of fatalities in single accidents with N 2 50 Marine and inland waterways

Cargo

Road, rail and pipeline

11 .,.*.,.,., Marine and inland waterways . . . . . Road, rail, and pipeline I , I

I

Oil Explosives LPG, butane, propylene, LNG

I

70, 84, 4386

60, 508 50, 73 60, 63, 100, 216, 500

50 Unknown Total LNG = liquefied

4590

1630

natural gas

flammable gases dominate for transport by land. The importance of oil at sea might be explained by the floating properties of oil on water, and that only few accidents with loss of liquefied gases happen at sea. The accidents in Table 4 with 500 and 508 fatalities are from transport by pipeline. I

I

1

10

I

I

I

100

1000

Number of fatalities, N Figure 8 Distribution of fatalities from accidents that occurred during the movement of goods by marine and inland waterways (240 accidents) and road, rail and pipeline (186 accidents)

Discussion The analysis presented here shows that the distributions of fatalities are different for transport of dangerous goods by land and by water. Focusing on marine accidents, differences in the f/N curves were found between local surroundings.

J. Loss Prev. Process Ind., 1995, Volume 8, Number

1

33

Marine

transport

Table 5 Number

of dangerous

goods:

of accidents in different

H. [email protected] et al

intervals of N for different transport

N

% of total marine accidents

% of total land transport accidents

Marine

1-9

61

83

Relatively

few accidents

IO-29

27

12

Relatively

many accidents

30-4s

10

1

Relatively

many accidents

2

4

Relatively

few accidents

350

versus land transport

The curve for ports was close to a straight line like the one for transport on land. Further, slight differences were shown when comparing actual transport and other phases such as cargo transfer. This leads to the conclusion that it is in fact the population density around the accident site that determines N, since this factor is certainly different at sea and on land. It is reasonable to believe that accident scenarios with between 10 and 50 vulnerable persons (the crew of one or two vessels) are more frequent at sea than on land. No effect was shown for the type of cargo, type of containment, geographical location or time period. It is surprising that it has not been possible to show clear effects for the type of cargo. This could be due to lack of data or the fact that the regulations are more strict for the transport of extremely dangerous goods (e.g. LPG) than for goods considered to be less dangerous (e.g. fuel oil). The fatalities from marine and inland waterway transport accidents (not in ports) can be characterized as shown in Table 5. The number of accidents seems to be relatively high in the interval of l&49 fatalities compared with the number of accidents in this interval occurring during transport on land. This reflects the total crew size of one or two marine vessels (10-50 persons) and the fact that it is relatively difficult for these persons to escape if an accident occurs. The two known accidents with more than 100 fatalities at sea are oil tanker-ferry collisions. This emphasizes the need for double hull oil tankers, which is obligatory for oil tankers ordered from now on. Hopefully this will reduce the risk of accidents where

34

J. Loss Prev. Process

Ind.,

1995, Volume

modes

Comments

on marine accidents

There are often more than 10 persons in vicinity of the accident centre The normal single crew size is between 10 and 29 persons In collisions, the potential number of persons in the vicinity of the accident centre is usually raised to this interval The population near the accident seldom exceeds 50 persons, i.e. the crew size of two vessels

ferries are surrounded by oil and engulfed in flames, although the initiative for this change in international regulations was primarily concern for the marine environment16.

References L. J. Loss Prev. Process lnd. 1990, 3. 395-405 Romer, H., Brockhoff, L., Haastrup. P. and Styhr Petersen, H. J. .I. Loss Prev. Process lnd. 6, 219-225 Hooke, N. ‘Modern Shipping Disasters’, Lloyd’s of London Press Ltd, London, 1989 Incident log. Hazardous Cargo Bull. 1986 to 1993 List of incidents 1981 to 1987. Loss Prev. Bull. 1982 to 1989 Blything, K. W. and Lewis, R. C. E. ‘Incident probabilities on liquid gas ships’, UKAEA Technical Report SRD R340. 1985 Blything, K. W. and Edmondson, J. N. ‘Fire/explosion probabilities on liquid gas ships’. UKAEA Technical Report SRD R292, 1984 ‘Oil tanker design and pollution prevention’, International Chamber of Shipping, London, 1990 Sicurezza nel trasporto di oh minerali e rinfuse. La Marina Italiana 1992, 90, 27-31 ‘A report by officials on oil spills and clean-up measures’, Pollution Paper 8, HMSO, London, 1976 Aldwinckle, D. S. and Pomeroy, R. V. ‘Reliability and Safety Assessment Methods for Ships and Other Installations’, Lloyd’s Register of Shipping, London, 1983 Marine Pollution Bull. 1985 to 1990 Brockhoff, L. Collection of Transportation Accidents involving Dangerous Goods’, Technical Report EUR 14549 EN, Commission of the European Communities, Luxembourg, 1992 ‘COST 301. Shore-based marine navigation aid systems’, Technical Report EUR 11304 EN, Commission of the European Communities, Luxembourg, 1988 ‘DAMA. Statistik for sjoulykker’, Internal Report, Det Norske Veritas, Oslo, Norway, 1990 Rawson, K. ‘The carriage of bulk oil and chemicals at sea’, Institution of Chemical Engineers, Rugby, Warwickshire, 1994

1 Haastrup. P. and Brockhoff, 2 3 4 5 6 7

8 9 10 11

12 13

14

15 16

8, Number

7