Uncertainty in AIS/VMS-based Classification of Fishing Activity: Behavioural and Bathymetric Constraints in Demersal Trawl Fisheries

Resumen

Accurate interpretation of fishing activity from vessel tracking systems remains a key challenge in spatially regulated fisheries. Although Automatic Identification System (AIS) and Vessel Monitoring System (VMS) data provide detailed information on vessel movement, fishing activity must be inferred from indirect behavioural proxies, introducing uncertainty, particularly at fine spatial scales.
This study analyses the demersal trawl fleet operating from the port of Llançà (Gulf of Lion, GSA-07, northwestern Mediterranean) to evaluate how  fishing activity is characterised in relation to depthbased regulatory frameworks. Using high-resolution AIS data, vessel trajectories are reconstructed and classified through movement-based indicators, primarily vessel speed. Spatial analysis is conducted by combining vessel positions with multiple bathymetric datasets, including hydrographic chart data and digital bathymetric models.
The results show that a substantial proportion of low-speed vessel activity cannot be unambiguously cassified as fishing, reflecting overlap between operational states. In parallel, spatial classification of vessel activity relative to depth-based regulatory limits varies depending on the bathymetric dataset used, leading to inconsistent interpretations near regulatory boundaries. These effects are particularly pronounced in areas where fishing activity is concentrated close to depth limits.
Overall, the findings demonstrate that the interpretation of fishing activity in spatially regulated fisheries is strongly influenced by the combined effect of behavioural and spatial uncertainty. This highlights the need for uncertainty-aware monitoring approaches to improve the consistency and robustness of activity classification in regulatory and scientific contexts.
These findings demonstrate that fishing activity classification in spatially regulated fisheries cannot be treated as deterministic and must explicitly account for both behavioural and spatial uncertainty.