International Journal
on Marine Navigation
and Safety of Sea Transportation
Volume 6
Number 3
September 2012
337
1 BACKGROUND INFORMATION
1.1 Ship Grounding
Ship grounding accounts for about one-third of
commercial ship accidents all over the world [1,2],
and has the second rank in frequency, after ship-ship
collision, in global perspective [3]. The consequenc-
es of ship grounding could be devastating for both
humans and the environment. In less grave acci-
dents, ship grounding might result in only minor
damages to the hull; however, in more serious acci-
dents, it might lead to the total loss of the vessel, oil
spills and human casualties, in which the compensa-
tion would be either highly costly or even impossi-
ble. Therefore it would be wise to think about tools
that can prevent ships to be involved in such acci-
dents.
1.2 Ship Domain
One of the methods that have never been tried for
grounding candidate detection is using the ship do-
main. The concept of ship domain has been first in-
troduced by Fujii [4] in maritime transportation as
an imaginary area around a ship, where the naviga-
tors try to keep it clear from other ships. Later on,
Goodwin [5] redefined the concept as the effective
area around a ship where navigators try to keep it
clear from other ships and stationary objects. Since
then, many other authors [6-16] have tried to define
the size and shape of ship domain with different
methods. However, the main common issue in be-
tween all introduced domains is that all are suitable
for ship-ship collision accidents, as the used meth-
ods are ruled by the nature of this type of ship acci-
dent. This fact is also recently highlighted by Wang
[15,16]. Although some authors have mentioned
their domains are suitable for grounding scenarios as
well [5,6,13,14], the affecting factors that they have
used to define the size and shape of the domain and
also the application of the domains are more useful
for ship-ship encounter situations. This is the main
courage for the present research in defining a ship
domain proper for ship grounding scenarios, in order
to be used as a decision support tool in VTS (Vessel
A Decision Support Tool for VTS Centers to
Detect Grounding Candidates
A. Mazaheri, F. Goerlandt & P. Kujala
Aalto University School of Engineering, Espoo, Finland
J. Montewka
Aalto University School of Engineering, Espoo, Finland
Maritime University of Szczecin, Poland
ABSTRACT: AIS (Automatic Identification System) data analysis is used to define ship domain for ground-
ing scenarios. The domain has been divided into two areas as inner and outer domains. Inner domain has clear
border, which is based on ship dynamic characteristics. Violation of inner domain makes the grounding acci-
dent unavoidable. Outer domain area is defined with AIS data analyzing. Outer domain shows the situation of
own ship in compare with other similar ships that previously were in the same situation. The domain can be
used as a decision support tool in VTS (Vessel Traffic Service) centers to detect grounding candidate vessels.
In the case study presented in this paper, one type of ship, which is tanker, in a waterway to Sköldvik in the
Gulf of Finland is taken into account.
338
Traffic Service) centers to detect the ships that are
grounding candidates.
2 SHIP DOMAIN FOR GROUNDING
Some factors that could affect the shape and size of
a domain useful for grounding scenarios are ship
main characteristics (length, breadth, draft, speed,
and type), her maneuverability, navigator experience
and his familiarity to the area, shape and depth of the
waterway, engine and rudder characteristics, and
weather condition; which some of them are not easy
to consider and to model. In addition, the 3
rd
dimen-
sion (depth) is vital for defining the ship domain for
grounding since the grounding is defined as the
event that the bottom of a ship hits the seabed, in
compare with stranding, which is defined as the
event that a ship impacts the shore line and strands
on shore [17]. Moreover, since normally ship has
forward speed while goes aground, the domain for
grounding could not be longitudinally symmetric.
For the same reason lateral dimension of the domain
should be always smaller than longitudinal dimen-
sion of the domain, when is defined for grounding
and stranding cases.
One additional point about ship domain either for
grounding or collision is that a domain should have
two areas as they can be called inner and outer do-
mains. Inner domain is the area, which is defined
based on the dynamic of the ship. Because of the
ship inertia, the ship’s course cannot be altered in a
moment. Inner domain defines the last/latest possi-
ble point/time that evading maneuver is possible for
the ship by the most possible aggressive but safe
maneuver, in order to avoid the accident. It means if
the inner domain is violated by a shoal, even though
the ship has not run aground yet, there is no way for
her to survive an accident. Outer domain, on the
other hand, can be defined as such that describes the
area of different levels that mariners are advised to
keep clear from any shoals or other stationary obsta-
cles. Failing to do so, makes the vessel a grounding
candidate with a certain degree. In contrary of the
inner domain, the outer domain does not have clear
border. Outer domain should be defined as such that
if a ship does not do any evasive maneuver by cer-
tain time/distance, it is considered, by some degree,
odd or unsafe for that particular ship with specific
characteristics in specific situation and location.
It is worthwhile to mention, depends on the rea-
son of the accident, ship grounding can be catego-
rized into two major groups as powered and drift
groundings. Nevertheless, drift grounding is a kind
of accident that occurs as a consequence of an inci-
dent like engine or rudder failures, which makes the
ship domain concept not applicable for this type of
grounding.
3 METHODS TO DEFINE SHIP DOMAIN FOR
GROUNDING
3.1 Inner Domain
The shape of the domain in this paper is taken as an
imaginary half-elliptical prism. The ellipse is chosen
to just explain the procedure of defining the size of
the domain. To define a proper shape for the do-
main, in order to be rational for grounding accident
analysis, more detailed data analysis and modeling
are needed, which will be addressed in future stud-
ies.
The size of inner domain should be defined based
on ship maneuverability, which is based on the dy-
namic of the ship. The length of the inner domain is
defined to be equal to the summation of overall
length of the ship (LOA), influence region of ship-
shore interaction (bank effect), and stopping distance
or the advance in turning circle maneuver, whichev-
er is shorter. To define the length of the inner do-
main in this paper, it is assumed that length of the
advance in turning circle is smaller than the stopping
distance, which is a valid assumption for ships mov-
ing with speed more than 12 kn [18]. The advance in
turning circle in this paper is estimated with a quasi-
linear modular hydrodynamic model of the vessel in-
plane motion. For detail explanation of the used hy-
drodynamic model, the readers are referred to [19].
The width of the inner domain is taken equal to
twice of the width of the influence region of bank ef-
fect. The influence region of bank effect (y_infls) in
this paper is estimated based on a formula suggested
by [20]. It should be mentioned that for defining the
width of the inner domain it is assumed the ship
does not comply with the given commands if she en-
ters the influence region of bank effect. Therefore,
controlling the ship will not be possible with ordi-
nary skills, which makes the ship eventually hitting a
channel bank. Although this assumption is not far
from reality, it should be considered that some ex-
pert mariners might still be able to control the ship
in that condition and therefore be able to survive
from an accident. However, to define the inner do-
main, rare situations are neglected and it has been
tried to define it as such to be suitable for majority
of the cases.
The depth of the inner domain is taken equal to
the maximum squat plus the draft of the vessel. The
maximum squat in this paper is estimated based on a
formula suggested by [21]. The schematic figure of
the defined inner domain is shown in Figure 1.
339
Figure 1: Three dimensions of the inner domain
3.2 Outer Domain
In this paper, outer domain is not defined by a
unique imaginary shape; but as points in different
waterway legs, in where the position and situation of
the vessel is analyzed based on extensive AIS (Au-
tomatic Identification System) data analysis in re-
spect to being a grounding candidate. The used algo-
rithm is shown in Figure 2. The general idea is to
choose a specific shoal/obstacle and analyze availa-
ble AIS data transmitted by ships similar to the sub-
ject (own) ship, which have previously approached
to the shoal, in order to find distribution for the lon-
gitudinal distance between ships and the shoal, in
where ships start to turn to either evade the shoal or
follow the fairway [Action Distance (AD), the point
where it happens is named Action Point (AP)].
Thereafter, use the obtained probability density
function (PDF) of AD to analyze the situation of
subject ship in respect to the shoal, in regard to
grounding accident. The PDF of action point will
help the VTS operators to relate the present location
of subject ship to the percentile of similar ships,
which have chosen that specific location to start
their maneuvers. In this regard, the appropriateness
of the present location of the subject ship to start the
turning maneuver can be judged by the safe maneu-
vers previously performed by ships similar to the
subject ship. Similarity can be identified by indexes
such as ship type, length, width, draught, speed, and
even environmental conditions. The more indexes
are defined, the more resembled cases can be re-
trieved and therefore the more reliable support for
decision can be provided. However, more indexes
need bigger and more complete databases to be used,
in order to retrieve sufficient data for creating useful
PDFs. Due to the scarce of data, the similarity in this
paper is identified just by type and length of the
ships.
Figure 2: Algorithm to define outer domain
It should be borne in mind that because of the
ship inertia, the ship’s course cannot be altered in a
moment; therefore it takes time between when the
command is given to the controlling devices till
when the command is started to be obeyed by the
vessel, in where is defined to be Action Point. None-
theless, this difference is neglected in this paper.
The action point detection process is based on a
pattern matching algorithm shown in Figure 3. The
pattern matching is based on course-over-ground
(COG) of ships. The idea is to visualize COG of the
ship in her path and then use the algorithm to detect
the performed maneuvers based on the visualized
COG. Here, visualizing means making the sequence
of COGs smooth in order to not have any disruption
in between. To explain visualizing and the algo-
rithm, part of waypoints in a trajectory of a tanker in
route from Sweden to Sköldvik in the Gulf of Fin-
land (GOF) for year 2007 is shown in Figure 4-Left
as an example. The history of COG of the shown tra-
jectory of the tanker is shown in Figure 4 -Right.
340
Figure 3: Pattern matching algorithm
The normal method being used for recording
COG is to mark the heading to the North as 0
o
, to the
East as 90
o
, to the South as 180
o
, and to the West as
270
o
(turning clockwise). As a result, COG can nev-
er get negative values; and if, for instance, the ship
is turning clockwise and COG value passes 359.99
o
,
the COG will be registered again as 0
o
. Therefore, if
the graph of the history of COG be drawn, there
might be some jumps in the graph (Fig. 4). To re-
move the disruptions and making the sequence of
COGs smooth (visualizing), the COGs are trans-
ferred to another discipline that is shown in Figure 5.
In the new discipline COG can get negative values
as well as values more than 360
o
. The history of
COG after visualizing is shown in Figure 4, which
shows the jumps are disappeared. The visualized
COGs of ships navigating in a fairway are somehow
unique for the fairway, and can act as fingerprint of
the fairway, which the pattern matching algorithm
can recognize. By knowing the position of
turns/shoals in a fairway and having the visualized
COGs of the ships navigating in the same fairway,
the evasive maneuvers that have been done to follow
the turn/avoid the shoal can be identified. The start-
ing point of the associated maneuver is stored as AP
and the shortest distance between AP and the shoal
is reported as AD. It should be added that for de-
creasing the margin of error for pattern matching,
the visualized COGs are coarse-grained in order to
remove the small changes in COG, which are nor-
mally appears due to course adjustment. Moreover,
to minimize the possibility of choosing a collision
avoidance maneuver, the presence of ship traffic in
instance time domain in an area around the vessel,
which is defined by the domain proposed by [15] for
collision scenarios, is also investigated and taken in-
to account.
Figure 4: Left: Part of a trajectory of a tanker in route from
Sweden to Sköldvik in GOF for year 2007, Right: COG and
visualized COG of the trajectory of the tanker
Figure 5: Discipline used for visualizing the history of COG
4 CASE STUDY AND RESULTS
The analyzed AIS data in this paper is for the year
2007 of ship traffic in the Gulf of Finland, which
was gathered by the Finnish Transport Agency. The
Gulf of Finland is used for the study due to availa-
bility of data, and also because of the importance of
grounding accident in the area. The studied area is a
waterway in GOF with approximate length of 40
km, in where the ships have to navigate in between
shoals in order to reach to Sköldvik. The waterway
is located in a rectangle which end points of one of
its hypotenuses have positions of 60.0
o
N 025.4
o
, E,
and 60.4
o
N, 025.7
o
E in WGS-84 reference system
(Fig. 6-Left). The majority of the traffic in this area
belongs to tanker traffic. Therefore, the other types
341
of ships are eliminated from the analyzed data due to
data scarce. In total 850 tankers navigated in that ar-
ea in 2007 with the shortest length of 75 m and the
longest length of 265 m. The AIS data analysis is
done with Mathwork’s MATLAB. Thus, for the sake
of coding, the shoals in the area are defined as poly-
gons. In total, five shoals in the area are defined and
taken into account for data analysis. The shoals and
vertices of the polygons are shown in Figure 6-
Right.
Figure 6: Left: The waterway to Sköldvik in the Gulf of Fin-
land with the traffic in 2007- Right: The same waterway with
the analyzed shoals as polygons. The vertices of the polygon
shoals are shown in dots.
To define the domain in order to be used for VTS
operators, PDF of AD for the ships in each leg of a
waterway should be extracted. Based on the extract-
ed distributions, inner domain, and speed of the ves-
sel, the VTS operator can have a good analysis of
the present position of the vessel. By way of illustra-
tion, it is assumed that the subject ship is a most
common tanker for this harbor, with the dimensions
of L=145 m, B=17 m, T=10 m navigating in the
studied waterway with speed of 15 kn. Using ship
type and ship length as indexes, the related PDF for
AD can be extracted from the database. The PDFs
for shoals 1 and 2 are shown in Figure 7 as exam-
ples. With the help of the extracted PDFs, the per-
centile of the similar ships that have started to turn
by specific point in the same leg of the waterway
can be estimated. In addition, the defined inner do-
main gives the remained time to go aground on ap-
proaching shoals. The inner domain for the studied
tanker is estimated based on the advance of turning
circle in maximum rudder angle, which is assumed
to be 35
o
. Example of analysis of nine positions of
the chosen tanker in the studied area (Fig. 8) is
shown in Table 1 as a way of illustration.
Figure 7: PDF of Action Distances for tankers with LOA of
145 m approaching shoals 1 and 2
Figure 8: The subject ship (L=145 m, B=17 m, V=15 kn) in a
way to Sköldvik shown with her inner domain. The dark areas
are the inner domains. The tanker is seen as a small black dot
in this scale
It can be seen in Table 1, wherever the outer do-
main shows that the majority of the similar ships,
342
previously navigated in the same waterway leg, had
started their turning maneuver in that specific posi-
tion to either evade the shoal or follow the water-
way, the inner domain shows less available time for
maneuvering in order to avoid grounding. Infor-
mation as such will help the VTS operators to detect
those ships that their remaining chance to survive
from a grounding accident are getting less and less,
with the aim of marking them as the ship that her ac-
tions should be monitored more closely. Later one,
the VTS operator may decide to contact the ship to
find if the officer on watch is aware of the situation.
In this way, the VTS operators are capable of being
more proactive.
5 DISCUSSION AND CONCLUSION
A new approach to define ship domain for ground-
ing scenarios based on AIS data analyzing and ship
maneuverability is presented in this paper. The in-
troduced domain is suggested to be used as decision
supports tool in VTS centers. It is shown the intro-
duced domain is capable of providing useful infor-
mation, like remaining time to point of no-return and
going aground, based on the vessel maneuverability.
In addition, the proposed method is able to provide
the ground for judging the safeness/oddness of the
performing maneuvers. Since the method uses pre-
viously performed maneuvers to analyze the current
maneuvering action, it can be argued the method is
providing expert opinions as a support for decision
making process.
The turning circle and stopping distance are used
in the definition process of the inner domain for
grounding in this paper. Since those concepts are
unique for every single vessel in unique conditions,
this method neutralizes the effects of type and num-
ber of controlling devices in hand. Nonetheless, it
makes hard to estimate the area of inner domain pre-
cisely, as the available hydrodynamic models for
predicting the ship motion are not completely flaw-
less. However, the quasi-linear modular hydrody-
namic model used in this paper can predict the turn-
ing circle of vessels precisely enough for the scope
of this paper [19]. In addition, using turning circle to
define inner domain area limits the usability of the
suggested domain to when all reserved maneuvera-
bility of the ship is available, which means when the
vessel is moving straight. The maneuvering task is
somehow different while the ship is in turning pro-
cess, as she does not have all the reserved maneu-
verability in hand. Due to this fact, grounding ship
domain in complex turns might be different than
what has been introduced in this paper.
The analyzed AIS data used for defining outer
domain in this paper are indexed based on ship type,
ship length and the location. This has been done due
to the scarce of the data. By increasing the size of
the used database and also using data about weather
and sea conditions, the indexes can be expanded to
other characteristics of the vessel and also to envi-
ronmental conditions, in order to provide more reli-
able supports for decision making process. Moreo-
ver, the analyzed data are limited to year 2007.
Analyzing more data from other years will help the
used algorithm to be more precise in providing the
grounds for decision making. In addition, the algo-
rithm can be made smarter if a learning loop be add-
ed, in order to teach the algorithm by new perform-
ing maneuvers.
The introduced domain is proposed as a decision
support tools for VTS centers. Nevertheless, it is
possible that the introduced domain be used as a de-
cision support tools onboard the vessels, in order to
provide expert opinions for officer on watch to per-
form maneuvers.
AKNOWLEDGMENT
The authors wish to thank the financial contribution
of the European Union and the city of Kotka. This
research is carried out within SAFGOF project and
in association with Kotka Maritime Research Center.
Table 1: Situation analysis of the subject tanker in nine positions shown in Figure 8
Position
Percentile of similar
ships that have started to
turn
Time to breach inner do-
main, maintaining COG and
speed [min]
Time to ground on the approaching
shoal, maintaining COG and speed [min]
1
0%
54
56
2
0%
36
38
3
5%
15
17
4
38%
10
12
5
6%
15
18
6
4%
15
17
7
83%
6
8
8
75%
7
9
9
83%
6
8
* Position numbering is started from left-down corner of the Figure 8. The first position in left-down corner is position 1, next
position is 2 and so on.
343
REFERENCES
1. Kite-Powell, H.L., et al., Investigation of Potential Risk
Factors for Groundings of Commercial Vessels in U.S.
Ports. International Journal of Offshore and Polar Engi-
neering, 1999. 9(1): p. 16-21.
2. Jebsen, J.J. & V.C. Papakonstantinou, Evaluation of the
Physical Risk of Ship Grounding, in Department of Ocean
Engineering. 1997, Massachusetts Institute of Technology.
p. 239.
3. Samuelides, M.S., N.P. Ventikos, & I.C. Gemelos, Survey
on grounding incidents: Statistical analysis and risk as-
sessment. Ships and Offshore Structures, 2009. 4(1): p. 55-
68.
4. Fujii, Y. & K. Tanaka, Traffic Capacity. The Journal of
Navigation, 1971. 4: p. 543-552.
5. Goodwin, E.M., A statistical study of ship domains. The
Journal of Navigation, 1975. 28(3): p. 328-344.
6. Davis, P.V., M.J. Dove, & C.T. Stockel, A computer simu-
lation of marine traffic using domains and arenas. The
Journal of Navigation, 1980. 33: p. 215-222.
7. Coldwell, T.G., Marine Traffic Behaviour in Restricted Wa-
ters. The Journal of Navigation, 1983. 36: p. 430-444.
8. Zhao, J., Z. Wu, and F. Wang, Comments on Ship Domains.
The Journal of Navigation, 1993. 46(3): p. 422-436.
9. Lisowski, J., A. Rak, & W. Czechowicz, Neural netwrok
classifier for ship domain assessment. Mathematics and
Computers in Simulation, 2000(51): p. 399-406.
10. Zhu, X., H. Xu, & J. Lin, Domain and its model based on
neural networks. The Journal of Navigation, 2001. 54(1): p.
97-103.
11. Pietrzykowski, Z. & J. Uriasz, Methods and criteria of nav-
igational situation assessment in an open sea area, in 4th
International conference on computer and IT applications
in the maritime industries, Compit'05. 2005: Hamburg,
Germany.
12. Pietrzykowski, Z. & J. Uriasz, Ship domain in navigational
situation assessment in an open sea area, in 5th Interna-
tional conference on computer and IT applications in the
maritime industries, Compit'06. 2006: Oud Poelgeest, Lei-
den, Netherlands.
13. Pietrzykowski, Z., Ship's Fuzzy Domain - a Criterion for
Navigational Safety in Narrow Fairways. The Journal of
Navigation, 2008. 61: p. 499-514.
14. Pietrzykowski, Z. & J. Uriasz, The Ship Domain - A Crite-
rion of Navigational Safety Assessment in an Open Sea Ar-
ea. The Journal of Navigation, 2009. 62(1): p. 93-108.
15. Wang, N., An intelligent spatial collision risk based on the
quaternion ship domain. The Journal of Navigation, 2010.
63: p. 733-749.
16. Wang, N., et al., A unified analytical framework for ship
domains. The Journal of Navigation, 2009. 62: p. 643-655.
17. Mazaheri, A., Probabilistic Modeling of Ship Grounding -
A review of the literature, in TKK-AM-10. 2009, Helsinki
University of Technology: Espoo. p. 73.
18. Mandel, P., Ship Maneuvering and Control, in Principles of
Naval Architecture, J.P. Comstock, Editor. 1967: New
York. p. 463-606.
19. Montewka, J., et al., Probability modelling of vessel colli-
sions. Reliability Engineering and System Safety, 2010. 95:
p. 573-589.
20. Lataire, E., et al., Navigation in confined waters: Influence
of bank characteristics on ship-bank interaction, in Interna-
tional Conference on Marine Research and Transportation.
2007: Ischia, Italy. p. 135-143.
21. Derrett, D.R., Ship Squat, in Ship Stability for Masters and
Mates, C.B. Barrass, Editor. 1999.