Skip to content
Close
Build with us
Build with us
Accelerators
ForgeAsk

ForgeAsk

Knowledge Retrieval

Task Automation

ForgeScribe
Document Generation

data-review

ForgeLens
 Smart Data Extraction

Powered By AltaForge - With Button
North Atlantic Right Whale

Detecting critically endangered whales,
from space

Fewer than 370 North Atlantic Right Whales remain on earth. AltaML and Hatfield Consultants built an AI-powered detection
system that identifies them from satellite imagery — giving researchers and regulators a scalable way to monitor the species across
Canadian waters, all without the use of aerial surveys.

diversity
CLIENT
Canadian Space Agency, in consortium with Hatfield Consultants, Fisheries and Oceans Canada, and Transport Canada
ESG
PROJECT
AI-powered Space-Based Detection System (SBDS) to detect and monitor North Atlantic Right Whales in Canadian waters using very high-resolution satellite imagery
deep-learning
WHAT WE DID
Convolutional neural network development, largest-ever NARW satellite imagery dataset compilation, training data augmentation, end-to-end semi-automated whale detection
rocket
IMPACT

75% whale detection recall from satellite imagery, peer-reviewed validation, and a validated path to operational deployment

THE CHALLENGE

The North Atlantic Right Whale is one of the most endangered marine mammals on earth. Fewer than 370 remain globally, and fewer than 70 of those are reproductively active. Every individual in the population matters.

In Canadian waters, Fisheries and Oceans Canada and Transport Canada have legal and operational obligations to protect the species — managing ship speeds, fishing gear, and vessel routes to minimize the risk of fatal interactions. Those measures only work if you know where the whales are.

The problem is scale. Right whales migrate and forage across a vast, open-ocean range — including the Gulf of St. Lawrence — that no survey program can cover continuously. Aerial surveys are expensive, weather-dependent, and limited in geographic reach. For a species this rare, in a habitat this large, the monitoring gap was real.

The scale of the detection challenge makes that gap clear. A single North Atlantic Right Whale occupies approximately 42 × 12 pixels in a 42,000 × 42,000 pixel satellite image. Finding one requires processing more than 1.7 billion pixels per image, consistently, at a level of precision that a human analyst cannot sustain across thousands of images.

North Atlantic Right Whale Pod

THE SOLUTION

In 2021, Hatfield Consultants assembled a research consortium with AltaML, Dr. Kim Davies from the University of New Brunswick, and Dr. Sean Brillant from the Canadian Wildlife Federation, funded by the Canadian Space Agency's smartEarth initiative with support from Fisheries and Oceans Canada and Transport Canada.

AltaML's role was the AI engine: a convolutional neural network designed to detect right whales in very high-resolution satellite imagery, automatically and at scale. The first challenge was data. Training an AI model to identify a specific endangered species from satellite imagery requires a substantial and accurately labelled dataset — and with fewer than 370 animals in existence, that dataset does not exist naturally. 

The team compiled the largest satellite imagery dataset of NARWs ever assembled, supplementing satellite acquisitions over two key NARW foraging areas with drone-based photographs and augmented synthetic imagery to build a training set large enough to work from.

The resulting system covers the full detection pipeline: satellite images are acquired over known foraging areas, processed by the CNN to identify candidate whale detections, and reviewed by analysts — giving marine conservation teams an end-to-end process that scales across ocean coverage without proportionally scaling human survey effort.

anchor (white) 75% whale detection recall from satellite imagery with the largest satellite imagery dataset of NARWs ever compiled, providing a foundation for continued model development and future operational deployment.  
book-open Peer-reviewed publication in Marine Mammal Science representing the first scientific validation of a satellite-based AI detection system for North Atlantic Right Whales at this scale.  
git-branch (white) Establishing the basis for GenWhales

a follow-on project developing synthetic satellite imagery to expand training data, reduce data acquisition costs, and further improve species identification accuracy.

  

WHY IT MATTERS

Shipping lane adjustments, fishing closures, and vessel speed restrictions all depend on knowing where the whales are. For a species with fewer than 70 reproductively active animals, satellite-scale detection coverage changes what those protections can actually do.

Right Whale Satellite Imagery

"The North Atlantic right whale is endangered and we must take every step possible to save them. The smartWhales initiative is investing in innovative companies and projects that will enhance our ability to detect and monitor these whales, helping protect them against vessel collisions and fishing gear entanglements in our waters. By working together, we can drive growth in our ocean economy while setting these whales on a path to recovery."

champagne_headshot
The Honourable François-Philippe ChampagneMinister of Innovation, Science and Industry (2021–2025)

THIS IS WHAT ML LOOKS LIKE
WHEN CONSERVATION RESEARCH COVERS AN ENTIRE OCEAN.

See more of how we build for government and the environment, or talk to our team. We'd love to hear about what you're working on.

See how we build for the public sector
talk to our team