Advancements in non-invasive bat surveillance — ESA 2025 Presentation
By Sarah Lagattuta, Field Research Fellow
How many bats are there?
This seemingly simple question has been the central force driving my research efforts for the past four years. Bat species play critical roles in global disease dynamics, and understanding their populations is key to developing predictive models for emerging infectious disease risk. But how do we accurately count the bats ourselves, and better yet, how do we leverage machine learning to count them for us?
This project has been an incredible journey, through leading multiple years of field work using thermal cameras at diverse field sites, coordinating a team to manually count bats in hours of footage, generating a library of thousands of manually annotated images to train a computer model, and collaborating with the UC Davis DataLab to adapt a multi-object tracking framework to this highly complex task.
I recently had the pleasure of presenting this research at the Ecological Society of America 2025 Annual Meeting. It was a joy to share my work with fellow wildlife enthusiasts (and some fellow bat counters!) and to learn from leading experts in ecology and One Health. Now, as a Field Research Fellow at the NSF Center for Pandemic Insights, I'm continuing this work by fine-tuning the bat counting model and using it to understand bat population dynamics at our field sites across California. This project is just a small part of NSF-CPI's broader mission, advancing technology to build scalable wildlife surveillance tools that improve our understanding of disease emergence and spread.
Lagattuta, S., Gardner, N.R., Freeman, J., Brooks, W., Sun, E., Johnson, C.K. (August 2025). Advancements in non-invasive bat surveillance: Integrating thermal imagery and multi-object tracking machine learning models for censusing bats in diverse, human-made roost environments. Talk presented at the Ecological Society of America Annual Meeting, Baltimore, MD.