Offshore Infrastructure

Human pressure on the world’s ocean—known as the blue acceleration—is increasing at an unprecedented rate. Global Fishing Watch is mapping all offshore infrastructure around the globe with satellite-based synthetic aperture radar (SAR) imagery and deep learning to assess the global growth of industrial infrastructure in the ocean. We detect and classify human-made stationary objects at sea such as oil platforms, wind turbines and fish cages, bringing new light to impacts on our marine world. 

Activity without measure

The rapid industrialization of the ocean mirrors the early stages of activities that have triggered mass extinctions on land. The need to keep up with the increasing demand for oil production, and the search for clean and green energy sources, has led to an accelerated expansion of offshore infrastructure. Development of oil rigs, wind farms, wave power systems and open water aquaculture is ramping up around the world with enormous ecological consequences in some cases. The global extent of marine and air pollution associated with these activities is poorly known, partly because records of the spatial distribution of most of these structures are incomplete, and for the most part this data is held privately. This limits access to information by those who manage the ocean and hinders our ability to take action.

“Seeing” with radar

We are developing a global ocean detection system using radar satellites to map the evolution of fixed infrastructure at sea. Our approach is based on the powerful Google Earth Engine platform to analyze petabytes—millions of gigabytes—of Sentinel-1 synthetic aperture radar imagery and detect stationary objects at sea, combined with a state-of-the-art convolutional neural network to classify the type of structure. A unique feature of our detector is the ability to construct time series of fixed infrastructure presence, allowing us to monitor offshore structures around the world as they are being developed, as well as monitor the seasonal variation of human activities such as aquaculture. Our data pipeline automates the processing and analysis of massive streams of SAR imagery. We detect tens of thousands of structures per month and will analyze the full archive of Sentinel-1 SAR imagery to make this information freely available. This unique spatiotemporal dataset of objects at sea will allow us to quantify the development of human-made oceanic structures at global scale for the first time. We will soon open source our detection system and freely distribute this novel data ready for analysis.