Abstract: This disclosure provides a data-driven approach to departure detection. In particular, the disclosed approach uses location data, such as GPS, which is readily available for many vessels globally and historically through the automatic identification system (AIS). The approach clusters this data and then constructs a convex hull around these clusters for each port to define a port area. A vessel is then determined as departed when it leaves that port area. Clustering is computationally efficient as a number of optimised algorithms exist, which means even a very large dataset like historical locations of thousands of vessels can be processed relatively quickly. Further, the departure detection enables accurate estimation of arrival times. Further, actions can be triggered by the departure determination, such as automatic control of cranes, trucks, trains and other port equipment.