Adventures with plankton on the North Sea
Published:
Noushin Eftekhari on life aboard the RV Cefas Endeavour, and how plankton monitoring could revolutionise marine research.
This May, I had the unique opportunity to participate in the Centre for Environment, Fisheries and Aquaculture Science (Cefas) Nephrops Underwater Video Survey (UVS NEP) aboard the RV Cefas Endeavour. The survey took place from 16 to 23 May, starting in Lowestoft and concluding in Newcastle, and aimed to record the burrow densities of Nephrops (small lobsters whose larvae make up part of ocean plankton) in the Farn Deeps area of the North Sea.
Being at sea, despite the occasional seasickness and the terrifying experience of a real fire alarm, was a unique and unforgettable adventure. Life on the ship offered a precise daily routine and the chance to spend time with people from diverse backgrounds, each sharing their expertise and experiences. It was a refreshing change, living without constant internet access, which allowed me to be more present and fully appreciate the beauty of the calm North Sea during the survey. During the survey, I had the unique opportunity to use our cutting-edge real-time plankton monitoring software. This software, designed to leverage computer vision technology, was integrated directly into the ship’s systems. Here’s a closer look at how this innovative system operates and its potential impact on marine research.
Collecting Marine Particle Images with the Plankton Imager
Our primary tool for plankton data collection was the Plankton Imager (version Pi-10), a high-speed color line scan imaging instrument. Connected to the clean seawater underway system at a depth of 4 meters, the Pi-10 continuously supplied water at a flow rate of 34 liters per minute, capturing high-resolution (10 µm) RGB images of passing particles within a size range of 180 µm to 3.5 cm.
Images are processed in real-time using an EPIX E8 frame store, which extracts regions of interest and saves them in TIFF format with timestamps and unique identifiers. Raw images are stored in 12-bit resolution and converted to 8-bit for easier viewing and analysis. Given the operational constraints and skewed distribution of plankton, only mesozooplankton (200 µm–2 cm) images are processed and saved.
The real-time marine particle monitoring system in action during the survey. This image shows the live dashboard, displaying species distribution changes as they occur, providing immediate insights into the marine environment.
Developing a Real-Time Classification Pipeline
We developed a real-time classification pipeline to enhance the efficiency of plankton monitoring, building on work done by a previous Turing Data Study Group project,. This system bypasses the lengthy delays traditionally associated with image collection and analysis. Images captured by the Pi-10 instrument are processed on-site using the ‘Edge-AI’ NVIDIA Jetson AGX Orin, which categorises images into copepods (small crustaceans), non-copepods, or detritus using advanced deep learning models.
The summarised data is then transmitted via the ship’s broadband satellite communication systems to a cloud-hosted digital dashboard, accessible remotely for analysis. This real-time processing capability not only stores images for further analysis but also allows for immediate classification and transmission of data, significantly enhancing the responsiveness of our monitoring efforts.
The extensive data collected by the Pi-10 during the survey provided a robust dataset for training our classification model. The model’s evaluation occurs in two stages. Initially, it is tested against a designated test set to ensure accuracy and reliability. The model is then integrated into the real-time pipeline for active monitoring.
Reflections and Future Directions
Testing the real-time plankton monitoring software during the Nephrops survey was a significant step forward in marine research technology. Classifying and analysing plankton in real-time offers numerous advantages, including quicker response times to environmental changes and more efficient data processing. This opens the door to a new era of monitoring, where measured variables can be seen in real-time thus allowing for sampling to be adapted according to visible changes.
The potential applications are vast as we continue to refine and develop this technology. Improvements to real-time monitoring offer powerful new tools to marine scientists, and have the potential to hugely enhance our understanding of marine ecosystems.
Life at Sea
Living aboard the ship, despite occasional seasickness and the unexpected experience of a real fire alarm, was a unique adventure. Life at sea offered a precise daily routine, collaboration with individuals from diverse backgrounds, and a refreshing break from constant internet access, allowing me to appreciate the beauty of the calm North Sea. I am deeply grateful to the Cefas team, especially James Scott and Sophie Pitois, for their support throughout the survey.
Find out more about other collaborations between Cefas and the Turing, including the AI for biodiversity and ecosystem health project. As part of the project, the Turing help fund Remote Electronic Monitoring (REM) equipment onboard the RV Cefas Endeavour as part of a project improve sustainable fishing practices, detailed in this Cefas blog post.