Mapping risks of IUU fishing using vessel trajectories and Species Distribution Models

Illegal, Unreported, and Unregulated (IUU) fishing is a widespread problem that undermines global efforts to manage and protect marine resources. IUU fishing involves activities that violate conservation laws, occur without proper reporting, or happen in areas with inadequate management. This practice depletes fish stocks, disrupts ecosystems, and threatens the livelihoods of communities that rely on sustainable fishing. According to recent estimates, IUU fishing accounts for up to 20% of the global seafood catch, resulting in an economic loss of billions of dollars each year (FAO 2016). Additionally, it complicates efforts to ensure food security, conserve biodiversity, and enforce international maritime regulations. Thus, addressing IUU fishing is critical to preserving marine resources for future generations. Limiting IUU fishing is challenging for states due to several factors. First, the vastness of ocean areas makes it difficult to monitor and enforce regulations effectively. Some Exclusive Economic Zones (EEZs) span millions of square kilometres, making it nearly impossible to track every vessel without significant resources and advanced technology. High-seas fishing also complicates jurisdiction, as these international waters are not under any single country’s control. The costs of surveillance, the complexity of international laws, and limited cooperation among countries further hamper efforts to control IUU fishing effectively.

A key challenge in combating IUU fishing is to target specific areas for monitoring to reduce surveillance costs and increase efficiency. To address this, we propose a predictive approach that identifies areas at high risk for illegal vessel presence. Our method trains artificial intelligence algorithms to detect fishing activities along the tracks of legally monitored vessels using VMS (Vessel Monitoring System) or AIS (Automatic Identification System) data. For each fishing activity detected by our algorithms, we extract co-located environmental conditions, such as water temperature at various depths, chlorophyll-a concentration, and current strength.

These environmental factors are then compared to overall conditions in the area using a Species Distribution Model (SDM). Once trained, the SDM predicts the likelihood of fishing activity based on current environmental conditions at any given point within the study area. SDM have proven effective in modelling the distribution of natural predators but also fishing fleets ((Chen et al., 2024; Griffin et al., 2021; Torres-Irineo et al., 2021), as fishermen often function as marine predators, targeting areas with high fish density.

This approach allows us to apply the SDM to regions where fishing is restricted, such as Marine Protected Areas (MPAs) and certain EEZs, to predict areas with the highest fishing potential. These areas are likely to attract IUU vessels, making them prime targets for surveillance efforts.

One limitation of this approach is the impossibility to validate the model in regions where legal vessels do not operate. However, we can perform cross-validation within areas where we have legal vessel position data, making it reasonable to assume that predictive performance will hold in areas lacking position data but with similar oceanographic conditions.

This approach offers an innovative way to provide authorities with relevant insights, helping them focus IUU fishing control efforts on the highest-risk areas.