How Global Fishing Watch protects global fisheries with Google Cloud Platform
More than one billion people rely on wild-caught fish for their main source of protein, but fish stocks across the globe are plummeting.The population of some fish species has plunged by 90 percent in the last generation. One in five fish in global markets are caught illegally, or are unreported or unregulated. But many countries lack the resources to crack down on unlawful fishing and don’t have adequate data to design and manage sustainable fishery programs. With so much of the ocean out of sight and out of mind, a global understanding of large scale fishing activities has never existed.
That’s the problem Global Fishing Watch set out to solve, by using satellite data to track the global movement of fishing boats. Historically, analysts would monitor vessels in a small region looking for patterns, but were always constrained by the time and people required, so most of the world was never looked at. But with more than 200,000 vessels continually broadcasting their locations using the Automatic Identification System (AIS) and dozens of satellites now in orbit to collect that data over the open ocean, new opportunities are being created for monitoring. Using machine learning to automate the patterns humans had been looking for, and combining that with massive cloud computing platforms, Global Fishing Watch has created the first global view of fishing over time and made it available to everyone in the world for free.
To do this, it chose Google Cloud Platform, because GCP can deal with the immense scale of data, and has tools such as Google BigQuery and Google Cloud Machine Learning for sophisticated data analytics and visualization.
“We have hundreds of thousands unique vessels and 40 billion AIS positioning messages in our data set — 10 terabytes of data and growing. We chose Google Cloud Platform because it is the only cloud service that offers highly scalable computing, machine learning and data analytics.”
— Paul Woods, Chief Technology Officer, Global Fishing Watch
Tracking fishing boats across the world
Global Fishing Watch uses data not just to identify potentially illegal fishing, but also to provide tools to countries and organizations around the world that help them to manage fisheries sustainably. Every day, Global Fishing Watch downloads 22 million AIS position messages from more than 200,000 vessels into Google Compute Engine, which processes the data to detect patterns that signify which vessel are fishing vessels and when and where they are fishing. The AIS information includes the vessels’ location, course, speed, type and nation under which it is flagged.The ability to easily run a complex model on a large dataset quickly by spreading the load over thousands of machines in parallel makes this possible.
Compute Engine sends the data to BigQuery for analysis by Global Fishing Watch and its partners, including university and government researchers. Researchers can perform queries across the full 5 year dataset, such as showing where vessels from China frequently fish, correlating fishing activities to reported catch, or clustering to identify potential transshipments between boats.
Researchers run queries iteratively, fine-tuning them each time to get the exact information they need. In the past when queries took hours or days to run, you couldn’t really “explore the data,” and could only take a tiny step each time, but with BigQuery now returning results in 60 seconds the process becomes interactive and the researcher stays fully engaged, enabling them to ask different types of questions.
Cloud Machine Learning models are built by feeding thousands of “training data tracks” that humans have manually classified so it can then analyze movement patterns to determine the class of each vessel (cargo, tugboat, or type of fishing gear such as long-line, purse seine, trawl), and when and where fishing activity occurs. These enable the team to build an estimate of the global level and distribution of fishing effort over time, as well as to automatically detect and flag instances of illegal fishing.
Google Earth Engine, a cloud based geospatial analytics platform, is used as a delivery system for providing researchers the processed detected fishing activity. The researchers can easily combine it with other layers they import or pull from the Earth Engine Public Data Catalog such as sea surface temperature and chlorophyll.
Providing critical data to protect the world’s fisheries
Data provided by Global Fishing Watch helps governments and agencies put sustainable fishery policies into effect. By seeing which parts of the ocean are heavily fished, governments can decide how much, if any, fishing should be allowed in specific areas. The Global Fishing Watch website and data is open to the public, increasing transparency about fisheries worldwide.
Global Fishing Watch’s efforts have already impacted fishing policies and resulted in the government cracking down on illegal fishing. For example, in 2015, Global Fishing Watch provided information to government of the island nation of Kiribati about a boat fishing illegally in its protected area. The boat’s owner paid the nation a $2 million fine. The United Nations is collaborating with Global Fishing Watch to develop transparency tools to improve the monitoring, control and surveillance of fishing activities. Seafood digital supply chain company Trace Register will use Global Fishing Watch to verify whether fish have been caught in a sustainable manner for customers including Whole Foods.
“With Global Fishing Watch, we’re showing for the first time the state of the world’s fishing activities on a daily basis, and helping make sure fishing is done in a sustainable way across the globe,” says Woods.