December 2020
Crow Research
Transforming observations through technology
Transforming observations through technology
Yes, crows! As part of a partnership that began in 2018, I’ve been working with volunteer researchers observing and studying the thousands of crows that roost in Lawrence, MA each winter. I provide insight and guidance when it comes to how we can use new and emerging technologies to enhance our ability to observe these birds in exciting new ways.
Both of the below reports are treated as living documents, which we edit and update as time goes on, and share publicly with others in the community.
In collaboration with Craig Gibson of wintercrowroost.com, last winter we continued our observation of the roosting crows in Lawrence, Massachusetts. Throughout the 2020/21 season we explored new documentation techniques, including 4K videography and advanced aerial imaging, and we are proud to present Winter Crow Roost, our 2021 video recap.
Comprised of clips captured from July of 2020 to March of 2021, our video recap tells the story of the tens of thousands of American crows (Corvus brachyrhynchos) and Fish crows (Corvus ossifragus) that arrive in Lawrence throughout the winter months, and roost in the trees along the banks of the Merrimack River.
Winter Crow Roost: 2021 Video Recap premiered at the 2021 Northeast Natural History Conference Video Festival, where we are proud to announce we received a first place audience choice award. Thank you to all the attendees who screened and selected our film!
To follow along with our continued activities at the Lawrence crow roost, visit wintercrowroost.com.
Beginning in mid 2020 our observations shifted to focus more on being able quantify the bird population in the Lawrence roost. While previous numbers were based on visual estimates, we sought to develop data-driven methods that could not only quantify the population but also produce proof of our count that other researchers can audit. Below is a sample of the report I prepared on how photographic technologies and automatic counting systems can be used to determine the size of a crow roost with a high degree of accuracy.
Updated 2025: This partnership has continued throughout the last 5 years as we explore more data gathering techniques. One ongoing project is to improve automated counting methodologies by training a deep learning model to detect and count crows in images gathered from drones and ground based imaging. The model was trained on a dataset of images gathered from the Lawrence roost, and is able to identify crows with a much higher degree of accuracy than previous automated methods. The model is constantly being refined and tested as more images are gathered and labeled.
Early excerpts of CrowNN inference results.
Much earlier in my work with the Lawrence roost, I developed a process for recording and analyzing the acoustic activity of the crow roost. The following is a description of the process and custom software I developed.
