ABSTRACT

However most of these methods rely on scarce accident statistics, predictive traffic flow methods (Fowler & Sørgård 2000) or simulation techniques (Goerlandt & Kujala 2011), which necessitate assumptions on what constitutes a vessel encounter. This may lead to an unreliable identification of risk areas, as discussed in (Goerlandt & Kujala 2014). Therefore, the method presented in this paper utilizes non-collision information from actually observed traffic to assess system risk and avoid assumptions, which may question the reliability and validity of the method.