51
1 INTRODUCTION
The North Sea is one of the busiest shipping traffic
areas in the world. By monitoring AIS signals
transmitted byships, the NetherlandsCoastguard is
abletomonitorthecurrenttrafficsituationandassist
ships. For the Dutch Ministry of Infrastructure and
theEnvironment(I&M)itisalsoimport
anttomonitor
any changes in the safety levels of traffic at various
locations,particularlyatbusyjunctions.
Traditionally, the safety levels of the shipping
traffic and the impact of new developments and
measures,canbeassessedwithriskmodels,suchas
theSAMSONmodel(VanderTak&DeJong, 1996)
tha
t was developed at MARIN. In the SAMSON
model, risk is a combination of accident probability
and consequences. For the risk of collisions, the
probability is modeled by estimating the number of
encountersbetweenshipswitha statictrafficmodel,
andmultiplyingthisbytheprobabilityofacollision
givenanencounter.
The t
raffic model is used to predict routes and
shippingintensitiesinfuturesituations,butitcannot
be used to monitor the safety levels that occurred
withtheactualtraffic.
Ariskindexwasdeveloped(Koldenhofetal.2009)
toapplytheriskmodelfromSAMSONtotheact
ual
real‐time trafficinformation that isprovided byAIS
data. This index can be used to monitor the safety
levels. In the risk index, the encounters are
determinedfromactualshipspositionsinsteadofthe
statictrafficmodel.
Both in the SAMSON model and the risk index,
theprobabilityofanencounterresult
inginacollision,
is still estimated from accident statistics. These
statisticsare,fortunately,ratherscarce.Theamountof
detail that can be incorporated in the probability
model,istherefore very limited. Moreover,formost
Classifying Ship Encounters to Monitor Traffic Safety
on the North Sea from AIS Data
W.H.vanIperen
M
aritimeResearchInstituteNetherlands(MARIN),Wageningen,TheNetherlands
ABSTRACT:InstudiesfortheDutchMinistryofInfrastructureandtheEnvironment,MARINhasdeveloped
methodstoclassifyshipencountersontheNorthSeafromAISdata.ThemethodsusetheDistanceatClosest
PointofApproach(DCPA),TimetoClosestPointofApproa
ch(TCPA),andanestimateofshipdomains,to
determineforeachcrossing,head‐on,andovertakingencounter,whetherthesefollowabnormalpatterns.On
august 1 2013, the route structure on the North Sea, was rearranged to improve safety and efficiency. The
encounterclassificationmethodswereappliedtotwoyearsofAISdata.Heatma
psofencountersshowhowthe
junctionshaveshifted.Forthesejunctions,thenumbersofencounterswerecompared.Thispaperdiscussesthe
methodstoclassifyencounters,andtheresultsofthecomparisonoftheroutestructures.Theresultsshowa
decreaseofthenumberofexceptionalhead‐onandcrossingencountersinthenewroutestructure.
http://www.transnav.eu
the International Journal
on Marine Navigation
and Safety of Sea Transportation
Volume 9
Number 1
March 2015
DOI:10.12716/1001.09.01.06
52
statistics, the exact tracks of the ships are not
available. Characterizations of encounters therefore
only distinguish between head‐on, crossing, and
overtaking encounters. As a result it is still hard to
estimate theprobabilityof a collisionfromAIS data
directly.
Instead of using the probability ofa collision for
monitoring the safety level, I&M therefore asked
MARIN to develop methods to distinguish between
normal and exceptional encounters, or even near
misses, and thus get an indication of dangerous
locations. The methods use the progress of the
Distance at Closest Point of Approach (DCPA) and
Time to Closest Point of
Approach (TCPA) during
encounters, as well as an estimate of the normally
maintained ship domains, to determine whether an
encounterfollowsanormalorabnormalpattern.
Onaugust12013,theroutestructureontheDutch
part of the North Sea, was radically changed to
improve safety and efficiency regarding offshore
platforms,portapproachandanchorageareas,andto
allocatesafeareasthatcanbeusedforwindenergy.
MARIN applied the classification methods for
encounterstotwoyearsofAISdatafortheNorthSea:
oneyearundertheoldroutestructure,andoneyear
under the new structure. Resulting
heat maps of
encounterlocationsshowtheareaswheremanyship
encounters take place. By comparing the heat maps
andthetrafficflowsfortheoldandnewsituation,the
shiftingofjunctionsbetweenvarioustrafficflowswas
described.Anumberofspecificareasweredefinedin
both structures, and the
types of encounters were
countedandcomparedpermonth.
This paper describesthe developedmethods and
theresultsofapplyingthesetocomparetheoldand
new route structure. First, Section 2 describes the
available data. The underlying principles to the
methodsaredefinedinSection3.Section4gives
an
overview of the various criteria that are used to
classifytheencounters.Section5describestheresults
of the encounter comparison between the old and
newroutestructure.
2 AVAILABLEDATA
Each month, MARIN receives AIS data from the
Netherlands Coastguard forthepurpose of riskand
safety studies regarding
shipping traffic. Not all
tracksofrecreationalandfishingshipsareavailable,
sincesomeshipsarenotyetequippedwithAIS.
The AIS data has already been processed by
Coastguard software upon reception at its base
stations along the Dutch coastline and on various
platformsintheNorthSea.The
processingmeansthat
forexample all overlappingsignals at differentbase
stationshavebeenmergedandsometracksmayhave
been extrapolated after signal loss. Also, where
available, positions have been checked with radar
tracks.
The main information that is used from the AIS
signalsaretherecordedmomentsatwhich
thesignals
are parsed on board of the ship (‘parse times’), the
positions at those moments (tracks), speed over
ground,heading,courseoverground,shiptype,and
dimensionsoftheship.Thetrackpositionsrepresent
themiddleoftheship.
Inthepre‐processingofthedataforthisproject
at
MARIN, information is derived and stored only for
fixed intervals of 1 minute. For these ‘plot times’
relativepositionsandspeedsofshipsarecalculated.
Giventhespeeds,accelerationandmaneuverabilityof
mostships,thechoicefor1minuteintervalsisenough
toavoidmissinganycloseencounters.
Theaccuracy
oftracksisnotalwaysoptimal.The
merging process with radar positions, for example,
maycausesomefalsepositionsinthetracksinsome
areas.Suchpositionsarenottakenintoaccountinthe
encounter classification process. False positions are
detected based on peaks in the calculated average
speedbetween
twoconsecutivepositions.
3 DEFINITIONS
Theencounterclassificationmethodsuseinformation
about relative positions between ships. It is
summarized by distance, relative bearing, closest
point of approach and maintained ship domains.
Definitionsanddescriptionsofthesearegivenhere.
3.1 Encounters
An encounter is defined as the tracks of two ships
having a speed of at least 1 knot, that at certain
momentsduringtheirapproach,areexpectedtopass
eachotherwithin3nauticalmileswithin20minutes,
basedontheirspeedandcourse.Theencounterstarts
at the first of such moments, and ends 20 minutes
afterthelast
ofsuchmoments.The20minutesextra
areusedtoalsobeabletostudythetrajectoriesafter
theshipshavepassedeachother.
An encounter always takes place between two
shipsAandB,andisstudiedfromtheperspectiveof
both ships. If a third ship C is present,
this is
consideredinthreeencounters:anencounterbetween
A and B, between A and C, and between B and C
separately.
3.2 Distance
The distance between ships A and B at time t is,in
principle, the distance between the center points of
the two ships. However, sincethe
two ships do not
necessarily parse signalsat the same time, the exact
distanceattcannotbecalculated.Thedistanceofship
AtoshipB,denotedasd(A,B)(t),iscalculatedforthe
moments that signals are parsed at ship A by
extrapolatingthe position ofship B atthat
moment.
Extrapolation is based on the position, speed and
course over groundat the lastparse time of ship B.
Giventhehighfrequencyofsignals,thisissufficiently
accurateforthepurposesofthisresearch.
53
3.3 Absoluteandrelativebearing
ThepositionofshipBfromtheperspectiveofshipA
can be uniquely described by distance and relative
direction. In navigation, direction is measured in
relation to a reference direction. This is called the
bearing. The referencedirection can beabsolute(for
examplenorth)
orrelative(courseofship),resulting
inanabsoluteorrelative bearing.Forencounters,the
course over ground of the own ship is taken as
referencedirection.
TherelativebearingofshipBasseenfromshipA
is defined as the angle between the line connecting
the centers of
ships A andB, and the course line of
ship A. This is denoted as rb(B,A)(t). This angle is
measuredasthedifferenceinclockwisedirection,and
takes values between 0° and 360°. In general,
rb(B,A)(t)≠rb(A,B)(t),unlesstheshipssailinexactly
theoppositedirection.
3.4 Ship
directionsduringencounters
Theaveragedirectionofashipduringanencounter,
denotedasθ,isdeterminedastheabsolutebearingof
thelinebetweenthefirstandlastpositionoftheship
duringanencounter.
ThedifferenceofdirectionofshipBasseenfrom
shipAisdefinedas:
BA B A

 (1)
BA
takes values between ‐360° and 360°, and
moreover
AB BA
 (2)
The difference of direction of the encounter,
regardlessoftheship‐perspective,isdefinedas:
AB BA

 (3)
3.5 Distanceatclosestpointofapproach(DCPA)
Theclosestpointofapproach(CPA)isthepositionof
ashipduringanencounterwherethedistancetothe
other ship is minimal. The distance at that point is
denoted as DCPA. During the encounter, the CPA
and DCPA can
be estimated from thecurrent speed
andcourseovergroundofbothships.Theprediction
of DCPA therefore progresses over time and is
denotedasDCPA(t),expressedinnauticalmiles.
There is an important difference regarding the
passing distance for a give‐way ship A between
passingastand‐onship
Batthesternoratthebow.
ThiscanbeexpressedbydefiningasignforDCPA:
DCPA>0: the give‐way ship from portside
passesatthebowofthestand‐onship;therelative
bearingincreases;
DCPA<0: the give‐way ship from
portside
passesatthesternofthestand‐onship;therelative
bearingdecreases.
The sign of DCPA is thus determined by the
changeofrelativebearing.
3.6 Timetillclosestpointofapproach(TCPA)
Foreach estimated DCPA‐value,the time it takes to
reachtheCPA,isdenotedas
TCPA(Timetillclosest
pointofapproach).TCPAalsohasasign:
TCPA>0:the distance between the ships
decreasesandtheCPAisstillahead;
TCPA=0:theCPAisreached;
TCPA<0:the CPA is passed; the distance
betweentheshipsincreases.
3.7 Shipcoordinatesandshipdomains
For safeand comfortable navigation, ships prefer to
maintain a certain minimal distance to other ships.
Theresultingfreezonearound theshipiscalledthe
shipdomain.Theminimaldistances canbeexpressed
in miles, but also in number of ship lengths (and
breadths).
Depending on this, domains will be
referred to here as absolute (distance) or relative
(distance)shipdomainsrespectively.
The absolute ship domain can be observed from
tracks of encounters by applying a coordinate
transformationthatputseachshipintheorigin,after
which all tracks of encountering ships can be
superimposed.Thistransformationusesthe absolute
distanceandrelative bearingbetweentheships.The
tracksofshipBasseenfromshipAinabsoluteship
coordinatesarecalculatedas:
B,A
B,A
( ) d(A,B)( ) sin(rb(B,A)( ))
( ) d(A,B)( ) cos(rb(B,A)( ))
x
tt t
tt t
 (4)
The relative ship domain can be observed by
additionally scaling the absolute ship coordinates
accordingtothelengthofshipA,L
A:
B,A B,A A
B,A B,A A
() ()/L
() ()/L
ltxt
bt yt
(5)
Figure 2 shows all tracks of encounters (mainly
overtakingandcrossingencounters)inabsoluteship
coordinates that occurred during one month at the
busy junction in thetraffic separation schemeabove
VlielandIslandintheNetherlands.
The plot clearly suggests the ship domain where
few ship tracks are observed,
and an increased
densityoftracksaroundit.Alsovisibleinthedomain
are thetracks of (two)vessels being towed by tugs.
The clustered tracks above the origin represent the
tug, and the clustered tracks below the origin
representthetowedvessel.
The size of the ship domain (either absolute
in
miles,orrelativeinshiplengths)canbemeasuredby
determiningthedistributionoftracksandtakingfor
examplethe0.5%percentile.Forthis,adistinctionis
made in sectorsof11.25°, since distancesinfront of
theshiparelargerthanatthesideoftheship.