Ports are located in low-lying coastal and riverine areas making them prone to the physical impacts of natural disasters. The consequential disruptions can potentially propagate through supply chains, resulting in widespread economic losses. Previous studies to quantify the risks of port disruptions have adopted various modelling assumptions about the resilience of individual ports and marine network logistics. However, limited empirical evidence is available to validate these modelling assumptions or to provide deeper understanding of the ways in which operations are adapted during and after disruptions. Here, we use vessel tracking data to analyse past port disruptions due to natural disasters, evaluating 141 incidences of disruptions across 74 ports and 27 disasters. Results show a median disruption duration of six days with a 95th percentile of 22.2 days. All analysed events show multiple ports being affected simultaneously, challenging some of the studies that only focus on single port disruptions. Moreover, we find that the duration of the disruption scales with the severity of the event, with an increment of 1.0 m storm surge or 10 m/s wind speed associated with a two day increase in disruption duration. In contrast to commonplace assumptions in model studies, substitution between ports is rarely observed during short-term disruptions. On the other hand, production recapture happens in practice in many cases of port disruptions. In short, empirical vessel tracking data provides valuable insights for future modelling studies in order to better approximate the extent of the disruption and the potential resilience of the port and maritime network.