Skip to content
Technology — Science — Ecosystem — Sustainability — Health — Funding — Engineering — Mental — Environment

Optimizing data gathering strategies for enhanced salmon population monitoring

Researchers Sara Beery (MIT Professor) and Justin Kay (PhD student) are constructing an autonomous data gathering mechanism, designed to surveillance salmon populations in the northwestern region of the U.S., specifically in the Pacific.

,   and  Administrator
3 min read
Researchers Sara Beery and Justin Kay, both at MIT, are creating an autonomous data collection...
Researchers Sara Beery and Justin Kay, both at MIT, are creating an autonomous data collection apparatus to track salmon populations in the Pacific Northwest region of the U.S.

Optimizing data gathering strategies for enhanced salmon population monitoring

Salmon Spotting with a Tech Twist: Professor Sara Beery Automates the Migration Monitoring Process at MIT

In the heart of MIT, Assistant Professor Sara Beery steps up her game to focus on environmental challenges, intent on putting her skills in computer vision, machine learning, and data science at the forefront of conservation and sustainability efforts. With a passion for making a real-world impact, she's drawn to MIT's commitment to "computing for the planet."

Salmon in the Pacific Northwest hold a unique power over the health of their ecosystems. Their complex reproductive cycle, stretching from freshwater streams to the ocean and back, offers a perfect playground for Beery's research. The salmon's journey is not just a continuous loop, but an essential chain linking thousands of organisms in the ecosystems they traverse. By providing nutrients like carbon and nitrogen from the ocean up to the rivers, they help sustain a diverse and interconnected web of life.

Salmon's significance, however, extends beyond their ecological role. Economically, commercial and recreational fisheries contribute significantly to the local economy, while for many Indigenous peoples in the region, salmon hold deep cultural value.

Minimizing the Count: Making Salmon Monitoring Easier

As human activities threaten salmon populations through overfishing, habitat loss, climate change, and hydropower development, keeping tabs on their migration and ensuring balanced interests becomes increasingly important.

Historically, monitoring efforts relied on eyesight, but with the use of underwater sonar systems in the past few decades, counting salmon has become slightly more manageable. However, human effort is still required to set up sonar cameras and count the salmon based on their output in a laptop. Automating the process is necessary to cope with the demanding task of understanding and managing these intricate ecosystems.

"We need these technological tools. Without automation, we can't keep up," says Beery. It's this conviction that led her to implement cutting-edge computer vision methods for automating salmon counting.

Currently, her research project, led by PhD student Justin Kay, collects data in the form of videos from sonar cameras at different rivers. They annotate the data to train the computer vision system, enabling it to spot and count salmon as they make their way upstream. In rivers where the team has created the necessary training data, the system has produced impressive results, boasting only a 3-5% counting error, well below the target of no more than a 10% error.

Adaptation and Growth: Scaling the System

However, when deployed in unfamiliar environments like the newly restored Klamath River, the system's performance degrades, increasing its error rate to around 15-20%. In response, the researchers at MIT have developed an automatic adaptation algorithm within the system to self-calibrate to the new conditions and environment, ensuring accuracy in detecting the migrating fish.

Real-Time Decisions and Community Collaboration

The researchers also faced the challenge of creating an efficient data infrastructure. Conventional methods like cloud-based approaches and shipping hard drives have notable drawbacks. In response, they introduced the "Fishbox" - a power-efficient computer that can be used in the field to perform the necessary processing, allowing managers to make hour-by-hour decisions for real-time, responsive salmon population management.

Finally, the team is actively fostering collaboration with a range of stakeholders in the Pacific Northwest, including nongovernmental organizations, tribes, and agencies, to pool expertise, insight, and resources to advancing the field of salmon fisheries management. By empowering the community and forging stronger connections, Beery's research project promises to make meaningful strides in the study and conservation of these vibrant and ecologically essential creatures.

  1. In the environmental science realm, Professor Sara Beery at MIT concentrates on research, using her expertise in computer vision, machine learning, and data science to aid conservation and sustainability.
  2. The salmon's migratory cycle, encompassing both freshwater and ocean environments, presents an opportunity for Beery's research, as it affects numerous organisms within ecosystems and plays a crucial ecological role.
  3. The economic significance of salmon is still substantial, with both commercial and recreational fisheries and their cultural value to Indigenous peoples contributing to the region's economy.
  4. To ease the monitoring of salmon migration and address issues resulting from factors like overfishing, habitat loss, climate change, and hydropower development, automation in the form of computer vision methods is essential.
  5. Beery's research project, under the leadership of PhD student Justin Kay, gathers data through sonar camera videos and trains the computer vision system to accurately identify and count salmon.
  6. In familiar environments like carefully calibrated rivers, the system can achieve a counting error of only 3-5%, demonstrating its effectiveness.
  7. When deployed in new environments, however, the system's accuracy minimally decreases, prompting MIT researchers to develop an automatic adaptation algorithm for self-calibration to ensure precise salmon detection.
  8. To address the challenge of efficient data infrastructure, the researchers introduced the "Fishbox" – a power-efficient computing device capable of handling on-site data processing for real-time decision-making in salmon population management.
  9. The team at MIT is working closely with various stakeholders in the Pacific Northwest, including nongovernmental organizations, tribes, and agencies, to share expertise, insights, and resources for advancements in salmon fisheries management.
  10. Research funding for technological innovations in engineering, computing, science, and environmental science is vital to expanding the applications of automation and other technologies, ultimately fostering a more sustainable and healthier planet.

Read also:

    Related

    Latest