The demand for maritime transport has increased with the growing demand for worldwide trade.This has led to a major increase in maritime traffic and ship sizes over the last decades, whichraises the probability of accidents. The methods used in maritime risk assessments today arebased on old hypotheses that do not include all data available today. The main objective of thisthesis is to develop numerical models and methods for the analysis of what is considered asnormal navigation behaviour at sea today and improve the analysis of probability for shipbridge allisions.The first part of the thesis describes what is considered as normal meeting distance at sea today.This information is later used while identifying failure events to ensure that the event behaviourwas not caused by other ships. These few cases are excluded from the methodology since thecommunication and situational awareness in the situations are not known. However, whilestudying the probability of ship-bridge accidents, it is also important to understand howwaterway restrictions may affect the probability of ship-ship collisions. Therefore, this thesisalso includes a study of how the improved knowledge concerning meeting distance could beused in a near ship-ship collision identification model. One of the main findings consideringnormal meeting distance is that small and large ships meet each other at a similar distance atsea.In the second part of the thesis, a methodology is proposed to estimate the probability of shipbridge allision. The presented methodology uses Automatic Identification System (AIS) dataand a ship manoeuvring simulator to simulate and analyse marine traffic with regards to risksfor accidents, such as ship-bridge allisions. A failure event identification method is alsopresented, which is needed to determine the frequency, duration and behaviour for the accidentscenarios. The three events that were modelled and simulated in the simulator were: driftingship, sharp turning ship and missing turning point. The probability of the different failure eventscorresponded to previous statistics confirming the AIS-based methodology. This means themethods to obtain the probability and duration of the failure events could be utilised in otherareas. The simulation methodology was confirmed with the probability of grounding in theGreat Belt VTS area.This thesis firstly contributes to a better understanding of the modelling of probability for shipbridge allisions. This will support bridge-building engineers who need to take into accountaccidental loads from ship-bridge allision while designing bridges. Secondly, this thesis alsocontributes to a better representation of normal beha