The overall aim of this work is to explore the possibility of automatically detecting Search And Rescue (SAR) activity, even when a distress call has on yet been received. For this, we exploit a large volume of historical Automatic Identification System (AIS) data so as to detect SAR activity from vessel trajectories, in a scalable, data-driven supervised way, with no reliance on external sources of information (e.g. coast guard reports). Specifically, we present our approach which is based on a parallelised, nonparametric statistical method (Random Forests), which has proved capable of achieving prediction accuracy rates higher than 77%.