Search

Vessel Monitoring System (VMS) data provide information about speed and position of fishing vessels. This opens the door to methods of estimating and mapping fishing effort with a high level of detail. To addess this task, we propose a new method belonging to the class of hidden Markov models (HMM) that accounts for autocorrelation in time along the fishing events and offers a good trade-off between model complexity and computational efficiency. We carry out an objective comparison between this method and two competing approaches on a set of VMS data from Denmark for which the true activity is known from on-board sensors. The DMKMG approach proposed outperformed the competitors approach with 6% and 15% more accurate estimates in the vessel-by-vessel and trip-by-trip case, respectively. In addition, these better performances are not paid in terms of computation time. We also showcase our method on an extensive dataset from Sweden. A quick (real-time) data processing has the potential to change how fisheries can be better harmonized to other utilisation of the seas and fill the gap between the local scales at which fishing pressure and stock depletion occurs with the large temporal and spatial scales of traditional fisheries assessment and management. The computer code developped in this work is made publicly available as an R package from http://www2.imm.dtu.dk/~gigu/HMM-VMS, 22 pages, 3 figures, 1 table