While autonomy offers a solution to many issues facing the maritime and naval industry, assessing the safety and reliability of autonomous shipping is one of the key challenges. Established methods and standards for conventional ships are still relevant but do not account for the challenges that come with the application of Artificial Intelligence (AI) and cyber-physical systems for Maritime Autonomous Surface Ships (MASS). Such systems require specialised safety assessment techniques. Scenario-based safety assessment for autonomous systems is one of the potential approaches and represents the state of the art. In this research, we conduct a feasibility study on identification and classification of seafaring scenarios, as described in COLREG, from Automatic Identification System (AIS) data. Furthermore, the statistical distributions for the parameters of these scenarios are empirically determined. We illustrate the utility of our methods using real-life AIS data from the Strait of Dover. Results indicate that AIS data can be a rich data source for identifying real-life scenarios that can be used for safety assessment of MASS.