Bias in Global Fishing Watch AIS data analyses results in overestimate of Northeast Atlantic pelagic fishing impact

Abstract The ability to pinpoint fishing activity in the world’s oceans has greatly improved over the past decades, a period in which both satellite-based vessel monitoring systems (VMS) and automatic identification systems (AIS) were introduced for fisheries control and maritime safety purposes, respectively. These data have been used extensively for fisheries research and have brought new insights into the spatial and temporal activities of many different fishing fleets. More recently, data products from Global Fishing Watch (GFW), derived from AIS data analyses, have boosted research. This is because GFW data resulting in identified fishing events is reported globally at high spatial and temporal resolution. However, working with pre-processed data comes with a risk because data scientists who may rely on GFW data products are unable to change the underlying assumptions used by GFW to define fishing events. In this study, we compare the fishing events identified by GFW with fishing events defined from self-sampling programmes on board two large pelagic fleets in the Northeast Atlantic. Within these self-sampling programmes, the exact position and time of hauls are meticulously reported, allowing for a comparison in both the number of hauls identified and the haul duration. Results reveal that the assumptions made by GFW to define fishing events lead to an overestimated duration of gear deployment within a range of 30%–380%, depending on the target species and vessel type. In addition, by comparing the self-sampling data with unprocessed VMS data, we demonstrate that it is likely that the activity in which vessels search for fish using sonar and echosounder equipment is mistaken for gear deployment. We recommend that authorities and GFW allow scientists free access to the unprocessed AIS data or that organizations such as GFW work closer with the fishing sector and scientific community to improve their data products.