Back

Drone shorebird surveys with Carleton University

Featured in Raven Magazine, January 2020

Read the full article

Every summer, volunteers travel to the roadless coast of James Bay to count the shorebirds that stop there to rest and feed during migration. The counts are done on foot, walking transects of about 10 km a day. It is slow, demanding work, and it can disturb sensitive species such as the endangered Red Knot.

Researchers at Carleton University are testing another way: drones to photograph the shoreline, a neural network to count the birds, and 3D-printed decoys to put it all to the test.

A summary of “Birds and a Plane” by Tyrone Burke, Raven Magazine, Carleton University (January 22, 2020). Photos by Fangliang Xu.

Two people on the rocky shore of the Ottawa River among fall colours at Morris Island Conservation Area
Morris Island Conservation Area
Jeremy Laliberté and Gerhard Bruins walking a forest trail carrying survey gear
Jeremy Laliberté (left) and Gerhard Bruins

Where it started

Gerhard Bruins, a former Canadian Wildlife Service scientist who has made the James Bay trip three times, asked researchers at Carleton whether the counts could be more complete, accurate and efficient. His question grew into a project that brings together engineering, computer science and biology.

A drone hovering over the shoreline beside its landing pad while the team watches
Test flight at Morris Island

Flying the survey

Mechanical and Aerospace Engineering professor Jeremy Laliberté and master’s student Brendan Ooi fly an off-the-shelf DJI Mavic 2 quadcopter over decoy birds, comparing altitudes and grid versus line flight paths. The goal is a flight that can be set up in advance, so volunteers can run it in the field with little or no drone experience.

Close-up of a drone controller showing the live camera view
The drone’s live camera view

Counting with AI

Systems and Computer Engineering instructor Cheryl Schramm, the first person Bruins approached, proposed training a neural network to find shorebirds in aerial images. Computer science students Simon Dewilde and Rachel Myrah are building tools that stitch overlapping drone photos into panoramas, then detect and classify the birds in them.

A painted 3D-printed shorebird model standing among reeds at the water’s edge
3D-printed shorebird model

Lifelike decoys

A neural network needs thousands of training images, and aerial photos of rare shorebirds are scarce. Biology professor Jeff Dawson and student Skyler Bruggink designed 3D-printed Red Knots and Greater Yellowlegs, checking their dimensions against specimens, and wildlife artist Ed Bruggink painted them.

Placing a painted 3D-printed shorebird model among reeds at the water’s edge
Placing a model shorebird

Testing close to home

Field trials took place at Morris Island Conservation Area, about an hour west of Ottawa, standing in for James Bay. Even if every technical hurdle is cleared, one question remains: how will live birds react to the drone?

“If they stopped flying, we wouldn’t fly either.”

Brendan Ooi, master’s student, on keeping the drone from disturbing real birds

Brendan Ooi unpacking 3D-printed shorebird models from a bin in a car
Brendan Ooi
A researcher flying the drone with a handheld controller on the shoreline
Drone pilot on the riverbank with the drone hovering low over the water
Read the full story in Raven Magazine

Photos by Fangliang Xu for Raven Magazine, Carleton University.