AI cuts wildlife tracking time from months to days
Researchers at Washington State University and Google have shown that artificial intelligence can process camera trap images in days instead of months, producing scientific conclusions that match human expert analysis in roughly 85-90% of cases.
The team tested a fully automated system on hundreds of thousands to millions of images collected in Washington, Montana's Glacier National Park, and Guatemala's Maya Biosphere Reserve. For most species, the AI-generated models aligned closely with those produced by human experts on key measures such as where animals occur and what environmental factors influence them.
The findings appear in the Journal of Applied Ecology.
From months to days
Camera traps generate enormous datasets. A single project can produce hundreds of thousands or even millions of images that researchers must review to identify which species appear in each frame.
Daniel Thornton, the study's lead author and a wildlife ecologist at WSU, said the traditional process typically takes six to seven months, sometimes a year, before analysis can begin. A team of undergraduate assistants and a graduate student verify identifications by hand.
Early AI tools filtered out blank images-often 60-70% of the total-but still required humans to review tens of thousands of photos containing animals. The new study tested whether that final human step could be eliminated.
Using SpeciesNet, a general AI model developed by Google, researchers ran images through a fully automated pipeline with no human review. Fully automated processing now takes just a few days, cutting a months-long bottleneck to roughly a week.
Why the speed matters for conservation
Faster processing means wildlife managers and researchers can move from data collection to decision-making more quickly. This could enable near real-time monitoring of species such as jaguars, wolves, and grizzly bears.
"We're not trying to replace people," Thornton said. "The goal is to help researchers get to answers faster so they can make better decisions about managing and conserving wildlife."
The efficiency gains could be particularly valuable for smaller or underfunded conservation groups. Researchers may also expand monitoring efforts without being limited by data processing capacity.
Where AI still falls short
The system works well for common species and standard ecological models. Very rare and easily confused species remain problematic for AI detection.
Dan Morris, a senior staff research scientist at Google who helped create SpeciesNet, said the key question wasn't whether the AI got every image right. "It was whether the ecological conclusions you care about would end up being basically the same," Morris said.
Even when the AI made mistakes-misidentifying animals or missing detections-the overall models remained robust. Occupancy models rely on repeated observations over time, so individual errors don't necessarily compromise results.
Human review is still needed for many other applications of camera trapping data beyond species occupancy analysis.
Broader implications
The researchers made part of their dataset publicly available to support tools like SpeciesNet, which rely on shared data to improve. The study took a practical approach: rather than developing new algorithms, the team asked whether current tools could handle the analyses researchers already perform.
The answer, for many common species, is yes.
Learn more about how AI is applied in research and data analysis through AI Research Courses and AI Data Analysis Courses.
Your membership also unlocks:
