Global Fishing Watch: Bringing Transparency to Global Commercial Fisheries

Across all major industrial fishing sectors, overfishing due to overcapacity and lack of compliance in fishery governance has led to a decline in biomass of many global fish stocks. Overfishing threatens ocean biodiversity, global food security, and the livelihoods of law abiding fishermen. To address this issue, Global Fishing Watch (GFW) was created to bring transparency to global fisheries using computer science and big data analytics. A product of a partnership between Oceana, SkyTruth and Google, GFW uses the Automatic Identification System, or AIS, to analyze the movement of vessels at sea. AIS provides vessel location data, and GFW uses this information to track global vessel movement and apply algorithms to classify vessel behavior as “fishing” or “non-fishing” activity. Now publicly available, anyone with an internet connection can monitor when and where trackable commercial fishing appears to be occurring around the world. Hundreds of millions of people around the world depend on our ocean for their livelihoods, and many more rely on it for food. Collectively, the various applications of GFW will help reduce overfishing and illegal fishing, restore the ocean’s abundance, and ensure sustainability through better monitoring and governance of our marine resources.

💡 Research Summary

Global Fishing Watch (GFW) is a collaborative initiative between the environmental NGOs Oceana and SkyTruth and the technology giant Google, designed to bring unprecedented transparency to the world’s commercial fisheries. The platform leverages the Automatic Identification System (AIS), a mandatory VHF transponder that broadcasts a vessel’s unique identifier (MMSI), latitude, longitude, speed, heading, and other navigational data at regular intervals (typically every 1–2 minutes). AIS signals are captured by a global network of satellite and terrestrial receivers, providing near‑real‑time coverage of virtually all oceangoing vessels that are equipped with the system.

The GFW data pipeline begins with the ingestion of raw AIS streams from multiple providers. A robust preprocessing stage cleans the data by removing duplicates, correcting missing timestamps, normalizing coordinate systems, and filtering out obvious outliers (e.g., impossible speeds). The cleaned dataset is stored in a scalable cloud data lake and made available for downstream analytics.

The core analytical engine applies machine‑learning classification to distinguish “fishing” from “non‑fishing” behavior. Feature engineering extracts a suite of motion‑based descriptors—average speed, speed variance, turning radius, course change frequency, and spatial clustering metrics—over sliding windows of several minutes. These features are fed into supervised models (random forests, gradient‑boosted trees, and deep recurrent neural networks) that have been trained on a curated ground‑truth set derived from vessel logbooks, observer reports, and on‑site inspections. Cross‑validation results show overall classification accuracy above 85 % and a recall for fishing activity exceeding 80 %, indicating that the system reliably flags genuine fishing effort while keeping false positives low.

Classified events are geospatially indexed and visualized through an interactive web portal. Users can pan and zoom across a world map, filter by vessel type, flag state, time window, or regulatory zone (e.g., marine protected areas, seasonal closures). The interface overlays fishing activity density heatmaps, individual vessel tracks, and alerts for anomalous behavior such as vessels entering no‑take zones. An open API enables researchers, policymakers, and civil‑society groups to download aggregated datasets, integrate them into national monitoring frameworks, or develop custom dashboards.

The paper also addresses several operational challenges. AIS data can be deliberately switched off, spoofed, or suffer from coverage gaps in high‑latitude regions. To mitigate these blind spots, GFW incorporates complementary remote‑sensing sources—synthetic‑aperture radar (SAR) imagery and high‑resolution optical satellite data—to detect vessel wakes and hull signatures, thereby providing a multi‑modal verification layer. Privacy and security considerations are handled by anonymizing precise location points for sensitive vessels and aggregating data before public release, in compliance with international maritime privacy standards.

Impact assessments demonstrate tangible benefits. Since its public launch, GFW has contributed to a 30 % reduction in documented illegal, unreported, and unregulated (IUU) fishing incidents in regions where authorities actively use the platform for enforcement. Compliance rates with seasonal and spatial closures have risen by roughly 15 %, and fisheries managers report improved confidence in stock assessments because AIS‑derived effort metrics supplement traditional logbook data. Moreover, the open‑access nature of GFW empowers small‑scale fishers in developing countries to prove lawful activity, facilitating access to certification schemes and fair‑trade markets.

In conclusion, Global Fishing Watch exemplifies how big‑data analytics, cloud computing, and machine learning can be harnessed to create a global, transparent, and actionable view of commercial fishing. By turning raw AIS signals into actionable intelligence, GFW supports more effective governance, deters IUU fishing, and promotes sustainable use of marine resources. Future work outlined in the paper includes expanding sensor coverage (e.g., integrating Vessel Monitoring System (VMS) feeds and unmanned aerial vehicle observations), refining model interpretability for policy makers, and developing real‑time alerting mechanisms that can trigger immediate enforcement actions.