Can wearable CO₂ and VOC sensors detect defecation and urination events in horses? A new study puts it to the test
Published in Dairy Science and Management
June 2025
In a world increasingly driven by data and precision agriculture, the idea of using wearable sensors to monitor livestock behavior has great appeal—especially for tracking manure emissions, which are both an environmental concern and a management challenge. A recent study by Wright et al. (2025) took a novel approach: placing open-source gas sensors on horses’ tailheads to detect spikes in CO₂ and volatile organic compounds (VOCs) during defecation and urination events.
The Study at a Glance
Researchers monitored four horses in a barn setting over nine days, collecting gas sensor data and validating emission events through video observation. They then applied machine learning models—including Random Forest, Support Vector Machine, and Extreme Gradient Boosting—to determine whether the sensor data could accurately predict these events.
What They Found
-
CO₂ and VOC spikes often coincided with defecation and urination—but also occurred during unrelated behaviors, introducing significant noise.
-
Machine learning algorithms struggled to accurately classify events, especially with such imbalanced data (only a small fraction of timepoints involved actual emissions).
-
Even after testing oversampling, undersampling, and model tuning, predictive performance remained low.
This figure shows CO₂ concentration readings from one horse over a single day. Spikes marked with vertical lines represent confirmed defecation or urination events. While some CO₂ peaks align with these events, many other peaks occur independently—highlighting the challenge of accurately detecting manure events based on gas concentrations alone.
*Figure 3 from Wright et al. (2025), Dairy Science and Management. Shared under CC BY-NC-ND 4.0. https://doi.org/10.1186/s44363-025-00003-z
Why It Matters
This study underscores both the promise and limitations of low-cost, open-source sensing technologies for behavior-based environmental monitoring in livestock. While there’s clear potential, the data show that simply detecting gas spikes is not enough—especially in complex barn environments where other sources of CO₂ may interfere.
Looking Ahead
Wright et al. suggest future research should:
-
Explore multi-sensor approaches (e.g., combining motion, gas, and temperature data)
-
Move experiments into open pastures, where environmental noise may differ
-
Consider edge computing strategies to reduce the data burden and improve real-time detection
Full Paper
Wright, R.K., Ganino, A. & White, R.R. Open-source carbon dioxide and volatile organic compound sensing and associations with defecation and urination events in horses. Dairy Sci. Manag. 2, 2 (2025). https://doi.org/10.1186/s44363-025-00003-z