International Journal

on Marine Navigation

and Safety of Sea Transportation

Volume 2

Number 4

December 2008

415

The Model of Oil Spills Due to Ships Collisions

in Southern Baltic Area

L. Gucma & M. Przywarty

Maritime University of Szczecin, Poland

ABSTRACT: The paper presents implementation of probabilistic ships collision model to evaluation of

possible oil spills in the Southern Baltic Sea area. The results of the model is time, place and size of the oil

spill due to ships collision. The results could be used for oil spill response action plans. The paper will open

the discussion about validation of achieved results and will try to answer the question about verification of

Baltic Sea oil spills data in comparison to worlds statistics.

1 INTRODUCTION

The collision between ships, grounding and fire on

board are most contributing factor in ship accident.

The consequences of this three kind of accidents

especially on coastal waters due to possibility

of oil spill could be catastrophic. The paper

presents methodology of ships accident probability

evaluation in different conditions with use of

complex stochastic safety model and real statistical

data.

To achieve the aims of the paper stochastic

model of ships traffic was developed and applied

on the Southern Baltic Sea area. The model is

capable to assess the risk of large complex system

with consideration of human (navigator) behaviour

models, ship dynamics model, real traffic streams

parameters and external conditions like wind

current visibility etc (Fig. 1). The model works in

fast time and could simulate large number of

scenarios.

In the second step the statistical data of real

collision accidents was collected. With the number

of collision situations, assessed, in previous step, the

collision probability in different scenarios was

evaluated. This probability of collision in different

encounter situations was estimated.

In the third stage of researches calibrated model

was run with the traffic data estimated at level 2010

and routing schemes introduced in 2006. The output

from model as collision place, ships involved,

navigational conditions could be useful for further

risk assessment.

2 STOCHASTIC MODEL OF SHIPS

ACCIDENTS

One of the most appropriate approach to assess the

safety of complex marine traffic engineering systems

is use of stochastic simulation models [Gucma 2005,

Gucma 2003]. The model presented on Figure 1

could be used for almost all navigational accidents

assessment like collisions, groundings, collision with

fixed object [Gucma 2003], indirect accidents such

as anchor accidents or accidents caused by ship

generated waves [Gucma & Zalewski 2003]. The

model could comprise several modules responsible

for different navigational accidents.

416

Fig. 1. Diagram of fully developed stochastic model of

navigation safety assessment

This methodology was used already by several

authors before with different effect [Friis-Hansen &

Simonsen 2000, Merrick et al. 2001, Otay & Tan

1998]. In presented studies the model was used to

assess the probability of oil spills in the Baltic Sea

area.

2.1 Traffic data

There are several sources of data necessary for the

running of simulation model. The data of traffic was

obtained by analysis of AIS records [Assessment

2005] Polish national AIS network studies, and

statistics of ships calls to given ports. The weather

data was obtained from Polish meteorological

stations and extrapolated in order achieve open sea

conditions [Risk 2002]. The navigation data was

obtained from navigational charts, guides and own

seamanship experience.

10000

20000

30000

40000

50000

60000

70000

East of

Gotland

West of

Gotland

Nord of

Bornholm

South of

Bornholm

Drodgen Kadet

Fairway

Great Belt Kiel

AIS traffic 2005

Yearly increase of the traffic

[%]

[ships/year]

Fig. 2. Traffic of ships and its increase on analyzed part of

the Baltic Sea

2.2 Collision accident models

To model the collisions simplified statistical model

is used. The model neglects several dependencies

but because it is based on real statistical data the

achieved results are very close to reality. The most

unknown parameter necessary for collision

probability assessment on large sea areas is number

of ships encounter situations. In complex systems

with several traffic routes this number could be

evaluated only by traffic streams simulation models

such as the one presented in this study.

The traffic of ships is modelled by

nonhomogenous Poison process where actual

intensity of traffic is evaluated on the basis of real

AIS (Automatic Identification System) data from the

Helcom network which is operated since mid 2005.

The collected AIS data is used also for determination

of ships routes, types, length and draught

distribution. The variability of mean ships routes is

modelled by two-dimensional normal distribution

which parameters were estimated by AIS data and

expert-navigators opinion. New ships routings were

considered. Routes are presented on Figure 7. Ships

traffic and its annual increase in different Baltic Sea

regions is presented on Figure 2.

After collecting necessary input data the simula-

tion experiment was carried out and the expected

number of encounter situation was calculated. The

critical distance where navigators perform anti-

collision manoeuvre was assumed on the base of

expert opinion separate for head on (heading

difference ±170°), crossing and overtaking situations

(heading difference ±10°). These distances called

minimal distances of navigators reaction were

estimated by expert opinion and real time non-

autonomous simulation experiment performed on

ARPA simulator. The mean distances of reaction

were estimated to 0,35; 1,0; 0,45 Nm for head on

collision avoidance, crossing and overtaking

[Kobayashi 2006].

The overall number of encounter situation

estimated by simulation model are around 140000

per year where 30% of them are head on situations,

40% of crossing and 30% of overtaking (Figure 3).

10000

20000

30000

40000

50000

60000

Head on Cross Overtake

2005

2000

Fig. 3. Simulated number of encounter situations N

(the influence of traffic increase)

417

Statistical data from southern part of the Baltic

Sea accidents were used for evaluation of mean

intensity of ship collision accidents in the Southern

Baltic. The number of accidents significantly

increase mostly due to traffic increase. Only the

accidents on the open sea area was considered.

Figure 4 presents number of accidents per year on

the investigated area.

y = 0.3143x - 627.52

0.5

1.5

2.5

3.5

1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

Fig. 4. Number of collision accidents per year (A) located at

the open sea of the Southern Baltic

Data presented on Figure 4 and the simulation

results of ships encounter situation (Figure 3) have

been used for estimating the number of collisions.

To simplify the calculations it was assumed that

probability of collision is equal in all considered

situations. The existing databases of real accidents

scenarios justify this assumption.

The calculated probability of collision in single

encounter situation (Figure 5) is higher than 1*10

-5

which is the typical mean value of probability used

in safety of collision assessment. The difference

between probability in 2000 and 2005 can be

justified only by the error of estimation and

simplicity of applied model. Normally minor

decrease of collision probability could be expected

due to better navigational and positioning systems,

traffic regulations, better training of navigator.

1.00E-06

1.00E-05

1.00E-04

Head on Cross Overtake

2005

2000

Fig. 5. Probability of collision accident (P) in different

encounter situations in 2000 and 2005

3 OIL SPILLS

The collision, grounding and fire on ship accident

could be followed by the oil spills. The conditional

probability is used and finally the probability of oil

spill accident (P

) is calculated as follows:

OSAAS

PPP

where: P

= probability of accident; P

A/OS

= condi-

tional probability of oil spill if accident occur.

Several databases [MEHRA 1999, ITOPF 1998,

MAIB 2005, LMIS 2004, HECSALV 1996, IMO

2001] was used to find the conditional probability of

oil spills if given accident occurs. Fig. 6 shows

conditional probability of oil spills in different

accidents.

Oil spills due to collision is estimated with the

double bottom tankers with relation to ships size

expressed in DWT. Identical procedure was

followed to find probability of oil spill after

grounding and fire. Fire on ships is highly unlike to

be the result of oil spill. Only about 10% of such

accident is ended with oil spill.

0.1

0.2

0.3

0.4

0.5

0.6

0-2000 2000-5000 5000-20000 20000-50000 50000-

BaltMax

Collision

Grounding

Fire

Ships size [DWT]

Fig. 6. Conditional probability of oil spill if given kind of

accident occurs

3.1 Estimation of size of oil spill after collisions

To evaluate the probability oil spill size after ships

collision several databases and another study results

was used [MEHRA 1999, ITOPF 1998, MAIB 2005,

LMIS 2004, HECSALV 1996, IMO 2001

]. The

simplified statistical model is used. The model

assumes that the size of oil spill is dependant only of

ships size expressed in DWT in tons. The results are

presented on Figure 7.

418

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0-500 500-1000 1000-

10000

10000-

50000

500000-

100000

100000-

BaltMax

0-2000

2000-5000

5000-20000

20000-50000

50000-BaltMax

Tanker size [DWT]

Spill size [tons]

Probability

Fig. 7. The probability of given oil spill size for different size

of tankers in collision accident

4 BUNKER SPILLS

Bunker spills was estimated with use of accident

databases [MEHRA 1999, ITOPF 1998, MAIB

2005, LMIS 2004, HECSALV 1996, IMO 2001].

The following formula is used for finding the bunker

spill accident probability:

BSAABS

PPP

where: P

= probability of accident; P

A/BS

conditional probability of bunker spill if accident

occur.

It was found that the conditional probability of

bunker spill is dependant only of kind of accident

and for collision equals P

BS/C

=0.125 for grounding

BS/G

= 0.12 and the for fire accidents P

BS/F

= 0.017.

To find the size of bunker spill the mean capacity

of bunker tanks in different ships was used. It was

assumed that only 50% to 30% of bunker could be

spilled after accident. This value is dependant of

ships size. The results as mean bunker spill can was

fitted to exponential function (Fig. 8).

v = 152 exp(3*10

-5

DWT)

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

0 20000 40000 60000 80000 100000 120000

Ships size in DWT [tons]

Bunker spill [tons]

Fig. 8. The model of bunker spill size

5 RESULTS

New routing measures in South Baltic are adopted in

July 2006. Traffic separation scheme established

north of Bornholm significantly changed the traffic

layout in the Southern Baltic Sea. The most reliable

accident database is carried out by Maris database of

Helcom. The accident statistics between 2000 and

2005 are presented on Figure 9.

Fig. 9. Statistical data of collision accidents on the Southern

Baltic Sea (2000-2005) [Helcom 2006]

The real oil spill accident are presented in Table 1.

Mean intensity of oil spill accidents in Southern

Baltic Sea equals 0.20 oil spill per year which gives

mean time between accidents equals 5.0 years.

Table 1. The major oil spill accidents on whole Baltic Sea

Year

Name of ship

Oil spill [t]

Location

2003 Fu Shan Hai 1,200 Bornholm,

Denmark/Sweden

2001

Baltic Carrier

2,700

Kadetrenden,

Denmark

1998 Nunki 100 Kalundborg Fjord,

Denmark

1995

Hual Trooper

180

The Sound, Sweden

1990

Volgoneft

1,000

Karlskrona, Sweden

In the further step the presented simulation model

was applied for the Southern Baltic Sea region

to assess the expected number of collision. The

following assumptions have been made:

− traffic of ships with new routing measures

applied;

− traffic on estimated level in 2010 year applied

according to Figure 2;

− time of simulation: >1000 years until stabilization

of parameters is achieved;

− the mean encounter reaction distances same as for

probability of collision evaluation;

419

− no influence of weather for simplification

reasons.

The results of simulation are presented on Figure

10. The simulated intensity of collision is 2.33 per

year (mean time between accidents 0.42 years) and

the intensity of collision ended with oil spill is 0.36

per year (mean time between accidents 2.76 years).

5900000

6000000

6100000

6200000

6300000

6400000

200000 300000 400000 500000 600000 700000 800000 900000 1000000

0-200

200-300

300-1000

1000-10000

10000-50000

365 years

size [t]

Fig. 10. Simulated collision accidents with constant traffic

estimated at 2010 year (within 365 years of simulation 77

collision with oil spill observed)

6 CONCLUSIONS

Stochastic model of navigational safety was applied

to assess the safety of Southern Baltic in respect of

oil spills. The traffic on expected at 2010 level and

new routing schemes on the Baltic Sea was applied.

As it was expected the number of accidents will

increase significantly.

The collision probability in different conditions

(meteorological, traffic, navigational) evaluated in

this researches will be used in the further step as the

input value in navigational risk assessment models

on large costal areas. The evaluation of ships traffic

influence on environment due to possible oil spills

after collision is also presented.

The comparison of simulation results with real

data of oils spills are little surprising. As it was

presented the simulated time between oil spills

accidents is almost twice as high as real data. It

should be clearly stated that oil spill accidents are

very rare events and high uncertainty in presented

simulation should be considered. The results

achieved in this paper should be considered as

hazardous and all necessary precautions against

accident should be taken into account.

REFERENCES

Assessment of the traffic in the Baltic Sea. 2005. Study

submitted on 51 session of Sub-Committee on Safety of

Navigation by Denmark and Sweden.

Friis-Hansen, P. & Simonsen, B.C. 2000. GRACAT: Software

for Grounding and Collision Risk Analysis. Collision and

Grounding of Ships; Proc. Intern. Conf. Copenhagen.

Gucma L. & Zalewski P. 2003. Damage probability of offshore

pipelines due to anchoring ships. Polish Maritime Research.

Gucma L. 2003. Models of ship's traffic flow for the safety of

marine engineering structures evaluation. Proc. of ESREL

Conf. Rotterdam: Balkema.

Gucma L. 2005. Risk Modelling of Ship Collisions Factors

with Fixed Port and Offshore Structures (in polish).

Szczecin: Akademia Morska.

Helcom. 2006. Report on shipping accidents in the Baltic Sea

area for the year 2005. Helsinki Commission Baltic Marine

Environment Protection Commission (HELCOM). Draft.

Kobayashi H. 2006. Standard Mariners Behavior in Avoiding

Collision Maneuver. Proc. of Marsim Conf. Terschelling.

Merrick J.R.W. et al. 2001. Modelling Risk in Dynamic

Environment of Maritime Transportation. Proc. of Intern.

Conf. of Winter Simulation. Washington.

Otay E. & Tan B. 1998. Stochastic Modelling of Tanker Traffic

Through Narrow Waterways Proc. of Intern. Conf. on Oil

Spills in the Mediterranean and Black Sea Regions.

Risk assessment of undersea gas pipeline Baltic-Pipe in Polish

EEZ with consideration of shipping and fishery activities.

2002. Maritime University of Szczecin (draft).

Severity, probability and risk of accidents during maritime

transport of radioactive material Final report of a

co-ordinated research project 1995-1999 Int. Atomic

Agency 2001 Viena.

MEHRA 1999, Factors Influencing Vessel Risks in UK Waters,

Appendix 6 1999.

ITOPF 1998 (The International Tanker Owners Pollution

Federation Limited), Database of Oil Spills from Tankers,

Combined Carriers and Barges, 1974-1998.

LMIS 2004. Lloyds Maritime Information Services The

database of maritime incidents and accidents reported to

Lloyd.

MAIB 2005. Maritime Accident Investigation Branch. UK

registered ships accident database.

HECSALV1996. Herbert Engineering Corporation. Salvage

Engineering Software 5.08 Huston.

IMO 2001 International Maritime Organisation, Fire Casualty

Database.