Indoor air quality in primary schools: real-time monitoring and predictive modeling of PM10 in Kenitra, Morocco

Abstract

The health of children may be adversely influenced by the air quality in schools because they are more sensitive to indoor air pollutants. PM10, which consists of tiny particles that are 10 µm in size or smaller, carries notable dangers associated with breathing issues and the escalation of asthma. This research marks the inaugural continuous monitoring of indoor air quality (IAQ) in Moroccan primary schools, which was conducted in Kenitra from May 27 to September 22, 2023. Indoor monitoring occurred during unoccupied times (May–July), while outdoor data collection was continued until late September. We investigated the PM10 concentrations while considering temperature and humidity using IoT sensor technology. Results from the investigation indicated that the PM10 levels found indoors were moderate, with study-long average concentrations of 15.3 µg/m3 at the urban residential school (Site 1) and 12.3 µg/m3 at the school on a heavy vehicle road near the Sebou River (Site 2). The levels of PM10 outdoors fluctuated significantly, as they were shaped by traffic flow and weather variations. Diurnal patterns revealed morning peaks and afternoon decreases due to natural processes. An exhaustive study was executed, concentrating on three machine learning frameworks—Random Forest, CatBoost, and XGBoost—to reveal the undisclosed indoor PM10 levels. The XGBoost framework demonstrated significant predictive accuracy for hourly PM10 (R2 = 0.77; MAE = 2.08 µg/m3; RMSE = 2.82 µg/m3), highlighting the efficacy of ensemble algorithms in environmental forecasting. The machine learning framework enhances confidence in the accurate representation of indoor air quality and provides a robust basis for sustaining pollution oversight. The investigation supplies a fundamental PM10 dataset for Morocco, aiding future epidemiological examinations and specialized indoor air quality actions in academic settings. It emphasizes the necessity of combining cost-effective sensor networks with sophisticated machine learning to address indoor air quality challenges in developing urban environments.