Statistical study of IOPCF data with relevance to

establishing CATS criteria

Yasuhira Yamada
National Maritime Research Institute (Japan)
FSA-WS-2009 , 27th Feb in Athens, Greece

1/27

Contents
1 Background
2 IOPCF data
3 Regression Analysis
- Regression
R
i formula
f
l
- CATS / CATScr
4 Conclusion

FSA-WS-2009 , 27th Feb in Athens, Greece

2/27

Background (1)
- Oil spill from Tankers -

Sea bird with oil spilt

FSA-WS-2009 , 27th Feb in Athens, Greece

3/27

Background (1)
- Oil spill from Tankers -

Nakhodka in Japan, 1997

Oil Spill from D/H Tanker

Baltic Carrier, Denmark, 2001
It is important to protect maritime environmental from oil spill from
ships (especially oil tankers).
FSA-WS-2009 , 27th Feb in Athens, Greece

4/27

Background (2)
Objective of environmental FSA (oil spill)
To reduce oil spill accidents from ships (tankers) in order to preserve
maritime environment
Are Triple hull, Quadruple hull practical ?
A balance between Costs of risk reduction measures (RCO) and
Benefits (risk reduction) is important.
Cost-Benefit Assessment ((CBA)) is necessary
y ((Step
p 4 in FSA))
Phase 1: FSA study and its verification
Phase 2: Discussions in IMO start as one of agenda
Phase 3: Revision of standards/rules
Assessment of each accident is not objective of FSA
FSA-WS-2009 , 27th Feb in Athens, Greece

5/27

Background (3)
Formal Safety Assessment
Safety FSA

Environmental FSA

Environmental FSA
(A)Accidental/Acute Risk
e.g. oil spill, spill of hazardous substances
(B)Regular/Chronic Risk
e.g. Exhaust Gas (CO2,NOx, SOx) emission,
sewage, garbage, ballast water, VOC

As a first step of Env. FSA oil spills are considered

FSA-WS-2009 , 27th Feb in Athens, Greece

6/27

Formal Safety Assessment

(MSC 83/INF.2)
5 Steps in FSA
Step 1 Identification of Hazard (IH)
Step 2 Risk Analysis (RA)
Step 3 Risk Control Options (RCO)
Step 4 Cost Benefit Assessment (CBA)
Step 5 Recommendation for Decision Making (RDM)

Present study would contribute to discussions with regard to CostBenefit Assessment (Step 4)
FSA-WS-2009 , 27th Feb in Athens, Greece

7/27

1 Background
2 IOPCF data
3 Regression Analysis
4 Conclusion

FSA-WS-2009 , 27th Feb in Athens, Greece

8/27

IOPCF

Data (1)

International Oil Pollution Compensation Fund (IOPCF)

98 States are the members of 1992 Conventions (by 2007).
IOPCF compensates to spills of persistent oil from oil tankers that
cause pollution damage in the territory (including EEZ) of a State
Party to the respective Convention
Data used in the present study include accidents from 1970 to 2007
(under 1972, 1992 Conventions).
Most of major oil spill accidents in Member states are included such
as Braer(1993), Nakhodka(1997), Erika (1999), Prestige(2001) and
Baltic Carrier (2001)
FSA-WS-2009 , 27th Feb in Athens, Greece

9/27

IOPCF Data (2)

MEPC58/17/1 based on IOPCF-2005.
Present based on IOPCF-2007 (Slightly updated)
Database includes 135 oil spill accidents in member States
Weight (W) known: 113 accidents
Cost (C) known: 129 accidents
W&C available for 99 accidents
Regression analysis was carried out using 99 accidents data
Exchange rate: Average rate (2002-2007) in IOPCF report is used.

FSA-WS-2009 , 27th Feb in Athens, Greece

10/27

Causes (Lower Level) of Oil Spill from Tankers

Fire, 3%

Others, 1%
Corrosion, 1%

Breaking, 5%
Collision, 29%
Unknown, 5%

Collision is the most probable

cause (lower level) of oil spill
from tankers
More than 50% of causes (lower
level) consist of Collision &
Grounding

Discharge, 7%
Sinking, 12%
Mishandling,
14%

Grounding,
22%

Cause of oil spills from tankers in 1970-2007

IOPCF (2007)

Higher level causes of collision

& grounding might be
machinery failure or lack of
watch keeping

RCOs to prevent higher level causes should be also taken into account in risk
analysis (Step 2 in FSA) and in considering RCOs (Step 3 in FSA)
FSA-WS-2009 , 27th Feb in Athens, Greece

11/27

Ranking of Oil Spill Accidents

(IOPCF 2007)
Rank

Ship name

Year

1
2
3
4
5
6
7
8
9
10

Prestige
Osung N3
Nakhodkha
Erika
Nissos Amorgos
Aegean Sea
Braer
Sea Empress
Haven
Sea Prince

2002
1997
1997
1999
1997
1992
1993
1996
1991
1995

Cost
[million US$]
$1,101
$301
$231
$168
$115
$108
$95
$70
$56
$55

Rank

Ship name

Year

1
2
3
4
5
6
7
8
9
10

Haven
Braer
Aegean Sea
Sea Empress
Prestige
Evoikos
Globe
Seki
Erika
Tanio

1991
1993
1992
1996
2002
1997
1981
1994
1999
1980

Cost of Oil Spill

Oil spill weight

[ton]
144,000
84,000
73,500
72,360
63,272
29,000
16,001
16,000
14,000
13,500

Oil Spill Weight

Ranks of oil spills in terms of cost do not correspond to those in

terms of oil spill weight.
FSA-WS-2009 , 27th Feb in Athens, Greece

12/27

Components of Costs in IOPCF data

Item
Clean-up
Indemnification
Fishery-related
Property Damage
Tourism-related
Loss of income
Preventive Studies
Others
Environmental Damage

Number
of incidents
115
59
57
18
14
11
10
10
2

Indemnification: Paying somebody an

amount of money because of the
damage or loss.

Details of indemnification are not

described in the report

FSA-WS-2009 , 27th Feb in Athens, Greece

In addition to clean-up costs,

compensation for fishery, tourism, loss
of income are also included.
Environmental damage is included in 2
cases. Adjustment might be necessary
t take
to
t k into
i t accountt environmental
i
t l cost.
t
Important! Not to double-count costs of
compensation, if combining with
another database.

13/27

1 Background
2 IOPCF data
3 Regression Analysis
4 Conclusion

FSA-WS-2009 , 27th Feb in Athens, Greece

14/27

Costs of oil spill in Log

Regression Analysis
12

Log 10 C [US$]

10

Nakhodkha
(1997)
Plate Princess
(1997)

IOPCF(2007)
Erika
(1999)

Prestige
(2002)

8
6

Haven
(1991)

y = 0.66
0 66 x + 4.59
4 59

R = 0.46

Weight ofLogoil
spill in Log
W [Ton]

10

Positive correlation between LogC and LogW can be seen although

deviation is relatively large (Friis-Hansen & Ditlevsen, 2001)
Regression formula is obtained (least square method)
FSA-WS-2009 , 27th Feb in Athens, Greece

15/27

Regression Analysis
- Comparison with existing (60,000) 12

Nakhodkha
(1997)
Plate Princess
(1997)

Erika
(1999)

IOPCF(2007)
Prestige
(2002)

1 scale: difference 10 times

Log 10 C [US$]

CATScr=60,000 overestimate costs.

10
The differences increase as W increase

In case of Haven, 60,000

overestimates about 100
times than
larger10costs
more
times

Haven
(1991)

6
4
2

ALL

Linear regression

SAFEDOR

0
0

3
4
Log10W [Ton]

differences
Environmental costs are
assumed to be 1.5 times
of cleaning up costs
(SAFEDOR)
2.5 times < 100 times

FSA-WS-2009 , 27th Feb in Athens, Greece

16/27

Nonlinear regression formula

Log10C = aLog10W + b

a=0.66, b=4.59

Log 10 C = aLog 10W + b

C = W a 10 b

C = W 0.66 10 4.59 = 38735 W 0.66 = C0 W a

C = C0 W 0.66

C0 W 2 / 3 = C0

2/3

V 2/3

Dimension [m2] Area

Cost of oil spill is proportional to the Area (contaminated )

Oil spill

Contaminated
Area
Spilled oil spread over as oil slick

FSA-WS-2009 , 27th Feb in Athens, Greece

17/27

CATS
CATS: Cost to Avert a Ton of oil Spilt
Japan basically support the concept of CATS and CATScr proposed
by SAFEDOR (Consistent with Safety FSA)
RCO is judged as cost-effective if following formula is satisfied

CATS < CATS cr

C
CATS =
R

Risk analysis/Cost estimation

Cost Increase for RCO [US$]
Oil Spill Risk Reduction RCO [Ton]

CATScr is a threshold value to judge cost-effectiveness of RCO

It is important to distinguish CATS and CATScr
Present study is carried out mainly for establishing CATScr
FSA-WS-2009 , 27th Feb in Athens, Greece

18/27

Smaller spills - Larger spills

W: Weight

10
Log 10 C [US$]

C: Costs

12

Nakhodkha
(1997)
Plate Princess
(1997)

Erika
(1999)

IOPCF(2007)
Prestige
(2002)

8
Haven
(1991)

6
4

C/W
2

ALL

0
0

W1

C Linear regression
2

3
4
Log10W [Ton]

SAFEDOR

Slopes to points represent C/W

Tendency: slopes decrease as W increases according to IOPCF data
Smaller spills have larger C/W, larger spills have smaller C/W
FSA-WS-2009 , 27th Feb in Athens, Greece

19/27

Comparison of C/W
Formulation
SAFEDOR

C = 60000 W
dC
d
(60000 W ) = 60000
CATS cr =
=
dW dW

Constant

Present

C = 38735 W 0.66

dC
d
CATS cr =
=
38735 W 0.66 = 25441 W
dW dW

0.34

Power function of W

FSA-WS-2009 , 27th Feb in Athens, Greece

20/27

Comparison of C/W
Small spill: High C/W
Large spill: Low C/W

CATS
TScr (=dC/dW) [US$ / ton]

70000
60000
50000

SAFEDOR

40000

Sames&Hamann(2008)

30000

Etkin (2000)

20000

Present

Mainly due to initial cost of

launching cleaning up ship
and setting up oil
fence/boom (Etkin, 2000)

10000
0
0

2000

4000
6000
Oil spill weight [Ton]

8000

10000

Present tendencies corresponds well to the previous studies (Etkin,

2000; Sames & Hamann, 2008).
FSA-WS-2009 , 27th Feb in Athens, Greece

21/27

Possible main factors for the difference

(1) The effect of oil amount CATS cr = 60000
on the costs.
CATS cr = Cclean + Cenv + Cassurance
(2) Environmental costs
CATS cr = (Cclean Fe ) Fa
(3) Assurance costs
CATS cr = (16,000 2.5) 1.5
(Skjong et al, 2005)
= 16,000 3.75
= 60,000
SAFEDOR
FSA-WS-2009 , 27th Feb in Athens, Greece

22/27

The effect of environmental costs and

assurance costs
CATS
S (=dC/dW) [US$ / ton]

70000
60000

SAFEDOR
Present x 10
Present x 3.75
Present
Etkin(2000)

50000
40000
30000
20000
10000
0
0

2000

4000

6000

8000

10000

Oil spill weight [Ton]

Even considering the effect of environmental costs and assurance

costs (facor=3.75- 10), 60,000 [US$/ton] is higher than present
results especially in larger oil spills.
FSA-WS-2009 , 27th Feb in Athens, Greece

23/27

Possible Grouping of CATScr

CATS
TScr (=dC/dW)
(=dC/dW)[US$
[US$/ /ton]
ton]
CATS
ATS

Small
oil spill
70000
10000
60000
8000
50000

Present x 3.75
Present
SAFEDOR

Medium Spill
Etkin
(2000)
Etkin(2000)

6000
40000

30000
4000

Present

20000
2000
10000

Large Oil Spill

00
00

50000
150000 6000
200000 250000
300000
2000100000
4000
8000
10000

Oilspill
spillweight
weight[Ton]
[Ton]
Oil

It is not reasonable to use large CATScr for small spillage in

estimating costs of large oil spill.
CATScr should be defined depending on amount of oil spill risk
FSA-WS-2009 , 27th Feb in Athens, Greece

24/27

1 Background
2 IOPCF data
3 Regression Analysis
4 Conclusion

FSA-WS-2009 , 27th Feb in Athens, Greece

25/27

Conclusion (1)
A nonlinear formula to estimate oil spill costs from weights of oil spill
is derived based on regression analysis of IOPCF data.
The formula provides larger C/W for smaller spill, and provides
smaller C/W for large spill mainly due to the effects of initial costs of
cleaning process. These tendencies corresponds well to the previous
studies
t di (Etki
(Etkin, 2000
2000, Sames
S
& Hamann,
H
2008)
In carrying out cost-effective analysis, it is reasonable to estimate
costs of oil spill depending on the corresponding weight of oil spilt.

FSA-WS-2009 , 27th Feb in Athens, Greece

26/27

Discussion
In order to take into account costs of environmental damage
environmental damage factor (Fe) can be used in combination
with the regression curve although further study to quantify
reasonable value of Fe is highly required.
As for the assurance factor (Fa), this factor might be explicitly put
outside
t id CATScr
CATS since
i
thi factor
this
f t is
i a kind
ki d off factor
f t decided
d id d by
b
decision makers as pointed out by Skjong et al (2005).

CATS < CATS cr Fa

FSA-WS-2009 , 27th Feb in Athens, Greece

27/27

Thank you for your attention

FSA-WS-2009 , 27th Feb in Athens, Greece

28/27