Assessing the Need for Site-Specific Calibration of ATP Bioluminescence Sensors for Airborne Bacteria Monitoring in Indoor Environments

Continuous monitoring of airborne bacteria is critical for early warning and timely intervention (e.g., emergency ventilation or air cleaning) in indoor environments. However, it remains challenging as standard culture methods are slow and laborious, lacking real-time capability. Adenosine triphosphate (ATP) bioluminescence sensors offer rapid on-site detection, but they suffer from the major limitations that other airborne microorganisms (e.g., fungi) also contain ATP and that different bacterial species may have different ATP levels. To overcome these limitations, a key step for the deployment of ATP bioluminescence sensors in real-world environments is to establish site-specific calibration curves to convert the bioluminescence signal (RLU/m³) to the airborne bacteria concentration (CFU/m³). This study aims to explore two questions: (1) Can a calibration curve established in a subset of rooms be reliably applied to a separate room within the same indoor environments? (2) If a single calibration curve is used across different indoor environments, what level of error would such an approach introduce? We first developed an ATP bioluminescence sensor equipped with a custom-built wet-cyclone sampler and an automation system. Laboratory tests using E. coli and B. subtilis were first conducted to confirm the correlation between the ATP signal and airborne bacterial concentration. Field measurements were then carried out in a shared office and two elderly homes. The results show that the calibration curve established in a subset of rooms could be reliably applied to a separate room within the same indoor environment for airborne bacteria monitoring, with a normalized root mean square deviation of around 20%. Therefore, automated ATP bioluminescence sensors are suited for continuous, site-specific relative risk monitoring of airborne bacteria, and the monitoring can serve as an indicator for timely engineering controls. However, based on the analysis of data from this study and the literature, when applying a single calibration curve to different indoor environments, the level of error could reach around 150%. Therefore, if a site-specific calibration curve is unavailable, extra care is needed to account for the significant errors when interpreting the data.