They are a species synonymous with the Pacific Northwest and even have names like Star and Tahlequah, but with only 75 Southern Resident killer whales left in the wild, we must accelerate our ability to monitor and protect this iconic and endangered species. 

For more than a decade Dr. Holly Fearnbach (Sealife Response + Rehabilitation + Research) and Dr. John Durban (Oregon State University) have been using aerial photogrammetry to assess the health of endangered Southern Resident killer whales. 

Processing aerial photogrammetry images has relied on manual analysis done by humans, and datasets take anywhere from four to six months to fully process. A lot can happen in those six months.  

Vulcan’s Machine Learning team partnered with SR3 to develop the Aquatic Mammal Photogrammetry Tool (AMPT), which uses machine learning to help researchers analyze large numbers of photos faster, and it is the first machine learning tool of its kind to combine individual recognition with photogrammetry for killer whales. 

AMPT uses machine learning and an end-user tool to dramatically decrease the time needed for image analysis, from around 6 months down to a matter of weeks, or even days.

Early use of the new tool by SR3 researchers show a decrease of 25% in time taken for individual image processing. Additional time savings are expected at the batch level, and as users increase their familiarity with the tool. 

Faster turnaround on health metric data will help guide management decisions directed towards the recovery of Southern Resident killer whales.

Such decisions could include adaptive management actions like targeting specific salmon runs or limiting disturbance by vessels, with an aim to increase both the abundance and accessibility of Southern Resident killer whale’s primary prey, Chinook salmon, throughout the year.