Abstract:
The Strait of Istanbul has a great strategic importance in the world since it is the key component of the unique waterway system between the Black Sea and the Mediterranean. Around 160 transit vessel passages are daily accomplished, many of them carrying hazardous cargo. Due to its narrow and winding geography, adverse meteorological conditions (current, fog, wind and rain which affect the visibility conditions in the Channel) and very high local traffic density, maritime traffic becomes very complex and risky. This study, which is based on statistical data analyses of several datasets, consists of two core sections. In the first part, in order to support and provide input to simulation-based risk studies of the Strait, transit vessel arrival process and visibility conditions (which are the key factors affecting the maritime transit traffic of the Strait and hence, the accident potential of the vessels) are investigated and analyzed. In this regard, distribution fitting and parameter estimations are performed via detailed analyses of historic transit and visibility data. For the distribution fitting study of the arrival process, Input Analyzer function of the ARENA software is used. For visibility conditions, the modeling of fog start times and their durations are done with the aid of empirical and “Mixtures of Generalized Erlang” distributions. The second part of the study is based on maritime accidents occurred in the Strait of Istanbul in recent years. The first sub-section contains descriptive analyses of the accidents that happened in the Strait: these analyses can be cited as frequency histograms, chi-square tests (in order to identify the statistical relationships between different accident factors), discriminant and cluster analyses to classify the accident causes. In the second sub-section, the likelihood of maritime accidents involving transit vessels is investigated through logistic regression modeling. The related logistic regression models are generated from several data sources covering the years 2005-2007 and are aimed at identifying the most significant factors influencing the probability of a vessel to play a role in an accident. Six different models are generated based on various assumptions of human error occurrence in the vessel during its transit. As a result, significant factors triggering an accident are found to be the pilot status, daytime/nighttime status, season and existence of bad visibility conditions, failure and human error. Moreover, the accidents occurring in the anchoring areas are also discussed in this section. Due to the lack of information about the anchoring areas, the weaknesses of the model on these areas are discussed.
