Machine Learning Model Predicts Indoor Ozone Exposure with Unprecedented Accuracy

Researchers have developed the first large-scale machine learning model capable of predicting hourly indoor ozone concentrations using easily accessible predictors, marking a significant advancement in air pollution exposure assessment. This development addresses a critical gap in understanding human health risks, as people spend 70%–90% of their time indoors where ozone levels differ substantially from outdoor measurements.

The study, published in Eco-Environment & Health on July 9, 2025, used random forest algorithms trained on over 8,200 hours of indoor ozone data collected from 23 households across 18 Chinese cities. Researchers from Fudan University and the Chinese Academy of Sciences demonstrated that including window-opening behavior significantly improved prediction accuracy, raising cross-validation R² from 0.80 to 0.83 and lowering RMSE from 7.89 to 7.21 parts per billion.

This research matters because ozone exposure represents a substantial global health burden. In 2021 alone, long-term ozone exposure contributed to nearly 490,000 deaths worldwide. Traditional exposure assessments have relied primarily on outdoor data, but indoor environments create complex dynamics where ventilation, building materials, and human behavior all influence actual exposure levels. The new model overcomes limitations of traditional mechanistic models that require detailed indoor parameters difficult to obtain at scale.

The model's predictor variables include outdoor ozone levels from high-resolution datasets, meteorological parameters such as temperature, humidity, wind, solar radiation, boundary-layer height, and surface pressure, along with window-opening status recorded manually by volunteers. Predictor-importance analysis revealed surface pressure, temperature, and ambient ozone as dominant factors, with ventilation emerging as a crucial behavioral determinant. The study found indoor ozone concentrations were 40% lower than outdoor levels during daytime hours, highlighting the buffering effect of indoor environments.

"Most exposure studies still rely on outdoor ozone data, but that's only half the story," said Prof. Xia Meng, senior author of the study. "Our findings show that ventilation behavior—something as simple as whether a window is open or closed—can change exposure dramatically." The research demonstrates that integrating behavioral data with meteorological information through machine learning enables more precise estimation of indoor ozone at large scales, which will strengthen epidemiological studies and help guide public-health interventions.

The implications extend to multiple sectors. This practical, low-cost strategy can be integrated into health-risk assessments, smart-home monitoring systems, and public-health surveillance platforms, enabling policymakers and scientists to better understand indoor-outdoor exposure differences. The model's ability to capture hourly ozone fluctuations and regional differences—performing better in southern than northern China and in cold versus warm seasons—provides nuanced insights for targeted interventions.

Future applications could extend this framework to other pollutants such as fine particulate matter or nitrogen dioxide, incorporate smart sensors for automated window tracking, and expand monitoring to diverse climatic zones. The research, detailed in the journal article available at https://doi.org/10.1016/j.eehl.2025.100170, represents a bridge between environmental modeling and daily life, promoting healthier indoor environments in rapidly urbanizing regions where air pollution remains a persistent challenge.