Big data from population surveys and environmental monitoring-based machine learning predictions of indoor PM2.5 in 22 cities in China

Abstract

Many studies have confirmed that PM_2.5 exposure can cause a variety of diseases. Because people spend most of their time indoors, exposure to PM_2.5 in indoor environments is critical to population health. Large-population, long-term, continuous, and accurate indoor PM_2.5 data are important but scarce because of the difficulties in monitoring the indoor air quality on a large scale. Model simulation provides a new research direction. In this study, an advanced machine learning model was constructed using environmental health big data to predict the daily indoor PM_2.5 concentration data in 22 typical air pollution cities in China from 2013 to 2017. The test R² value of this model reached as high as 0.89, and the RMSE of the model was 9.13. The predicted annual indoor PM_2.5 concentrations of the cities ranged from 54.6 μg/m³ to 82.7 μg/m³, and showed a decreasing trend year by year. The pollution level exceeds the recommended AQG level of PM_2.5 and has potential impact on human health. The results could take a breakthrough in obtaining accurate big data of indoor PM_2.5 and contribute to research on the indoor air quality and human health in China.

Synopsis

This study established a machine learning model and predicted indoor PM_2.5 big data, which could support the research of indoor PM_2.5 and health.