NC State Study Finds AI-Generated Images Can Support Species Conservation, but Not Replace Real Data

Researchers at North Carolina State University’s Center for Geospatial Analytics found that AI-generated images can improve computer models used for species identification and biodiversity monitoring when real-world photos are scarce, but that synthetic images cannot substitute for real data. The study, published September 2, 2026 in the journal “Remote Sensing in Ecology and Conservation” under the title “Synthetic Imagery Improves Ecological Classification When Real Data Are Scarce, Not Direct Substitutes,” was led by research scholar Thomas Lake alongside co-authors Brennen Farrell, Ross Meentemeyer and Chris Jones.

The team generated 40,000 AI-generated, or “synthetic,” images across 20 common North American tree genera using Stable Diffusion v1.5, then combined them with real photographs drawn from the community-science platform iNaturalist and the Auto Arborist Dataset, a street-view tree imagery collection. They trained ResNet-50 image-classification models on controlled mixtures of real and synthetic images to test four scenarios: replacing real images entirely with synthetic ones, supplementing real images when data were scarce, using synthetic images as a conventional data-augmentation technique, and transferring synthetic images across different image domains.

Synthetic images help most when real data on a species is scarce

Replacing real training images with synthetic ones reduced the models’ performance on held-out real test sets, though the size of the decline varied between the iNaturalist and Auto Arborist image domains, the study found. Adding synthetic images to a limited pool of real images did improve classification accuracy, but those gains shrank as more real images became available. In cross-domain tests, synthetic supplementation helped some scarce-data cases but did not close the gap with models trained on transferred real-image data.

“The greatest potential for this technology may be in studying species and places we know relatively little about,” Lake said in the university’s announcement. Co-author Chris Jones noted that photographic coverage is uneven across species: “Some species are photographed thousands of times, while others may be rare, occur in remote places or simply receive much less attention.”

Lake cautioned that the synthetic images, while visually convincing, missed details that matter for accurate identification. “Most of the AI-generated images looked plausible at first glance,” he said. “But really, they didn’t capture the fine details and variations that exist in the real world, telling us that these images can look realistic while still missing subtle information that matters in practice.” The research was supported by the U.S. Department of Agriculture’s National Institute of Food and Agriculture through its Agriculture and Food Research Initiative, Data Science for Food and Agriculture Systems program (award no. 2022-67021-36465).

Community science remains the core data source

Lake said the findings reinforce, rather than reduce, the importance of community science platforms such as iNaturalist, where members of the public contribute photographs of plants, animals and other organisms. “AI-generated imagery does not eliminate the need for field observations or community science; it gives us new ways to use those observations, for recognizing species and noticing how our Earth is changing, potentially sooner,” he said. A related China solar policy study linking renewable-energy siting to bird diversity loss illustrates a separate but connected challenge in ecological monitoring: pairing large-scale environmental change with reliable species data.

The researchers stressed that synthetic images work best as a supplement for under-documented species and places rather than a wholesale replacement for photography. For common, well-photographed species, real images from public contributors continue to provide the fine-grained detail, such as leaf shape, bark texture and growth form, that synthetic images tend to miss.

Background

Generative image models such as Stable Diffusion have been used since the early 2020s to create synthetic training data for computer-vision tasks in fields ranging from medical imaging to autonomous vehicles. Their application to ecological classification is more recent, as conservation researchers have sought ways to fill data gaps for rare or poorly documented species without additional costly fieldwork. The NC State study is among the first to systematically test synthetic imagery across multiple scenarios, including direct replacement of real data, supplementation, conventional augmentation and cross-domain transfer, for a defined set of species. The Center for Geospatial Analytics at NC State, where the study was conducted, focuses on applying geospatial data science, including satellite imagery and machine learning, to conservation and environmental-change questions. The study’s authors say the next step for the field is testing where synthetic imagery can most reliably close data gaps without introducing identification errors that could mislead conservation decisions.

Sources: NC State College of Natural Resources; Remote Sensing in Ecology and Conservation

Featured image: photo by Radosław Krupa on Pexels (free Pexels license).

Become a Sponsor

Our website is the heart of the mission of WINSS – it’s where we share updates, publish research, highlight community impact, and connect with supporters around the world. To keep this essential platform running, updated, and accessible, we rely on the generosity of you, who believe in our work.

We offer the option to sponsor monthly, or just once choosing the amount of your choice. If you run a company, please contact us via info@winssolutions.org.

• donate monthly
• donate once only

Select a Donation Option (USD)

$5.00 (p/m) $10.00 (p/m) $20.00 (p/m) Other (p/m)

Enter Donation Amount (USD)

Figen Sekin

I specialize in sustainability education, curriculum co-creation, and early-stage project strategy. At WINSS, I craft articles on sustainability, transformative AI, and related topics. When I’m not writing, you’ll find me chasing the perfect sushi roll, exploring cities around the globe, or unwinding with my dog Puffy — the world’s most loyal sidekick.