Online City Air Toxicity Monitoring Using Rats

CHENYU ZHU, Maosheng Yao, Peking University

     Abstract Number: 641
     Working Group: Health-Related Aerosols

Abstract
Conventional approaches to studying the health effects of air pollution such as particulate matter (PM) are generally offline, which involve air sampling, lab-controlled exposure, then followed by tissue sectioning and blood examinations. Here, we have used rats for real-time sensing air toxicity through analyzing breath-borne biomarkers from rats by an integrated sensor array. The eight gaseous biomarkers in the exhaled breath of rats include total volatile organic compounds (TVOC), CO₂, CO, NO, H₂S, H₂O₂, O₂, and NH₃. The results showed that PM_2.5 significantly affected the relative levels of multiple breath-borne biomarkers from rats, especially NO, H₂S, H₂O₂, and O₂. Further, we have developed an automated, low-cost, and time-resolved system for noninvasive monitoring of eight breath-borne biomarkers from rats. We deployed the system to 13 cities in China to monitor the health effects of real-world air pollutants on a 24/7 basis during the 2023 winter and 2024 spring. Using the results of eight biomarkers, we have derived an air toxicity index (ATI) for assessing the overall health effects of city air. Using the EconML model, we quantified the causal effects of the same magnitude of change in the mass concentrations of PM_2.5–10 and PM_2.5 on the levels of breath-borne VOCs, CO, NO, H₂O₂, and H₂S from rats in each city. Results showed that PM_2.5–10 and PM_2.5 had different health effects in different cities and seasons. Using Deep Neural Networks (DNN) coupled with SHapley Additive exPlanations (SHAP), the contribution of PM_2.5 to the overall air toxicity was quantified, which exhibited complex nonlinear patterns in real-world environments. Overall, this work pioneers a new research paradigm for non-invasive, real-time monitoring of PM health effects and has yielded significant novel findings in field applications.