Modeling and forecasting air pollution for public health protection based on ML and time series models in Gulf Cooperation Council (GCC) countries

BACKGROUND: Air pollution is a serious environmental factor associated with higher rates of illness and death from respiratory, cardiovascular, and other non-communicable diseases. Accurate prediction is vital for assessing health risks, which would guide the public health experts and shape sustainable policies. AIM: This study aims to provide a modelling approach by comparing traditional, non-parametric, and machine learning based hybrid models, the first of its kind to forecast PM2.5 levels across the Gulf Cooperation Council (GCC) nations, to inform data-driven environmental health strategies and public policy development. METHOD: The study utilized the annual PM2.5 dataset from the World Bank database for GCC countries covering the span of 1991–2021. The traditional models, like ARIMA, Naïve, exponential smoothing, non-parametric model, NPAR, and machine learning based hybrid models, were applied. The model accuracy was evaluated by RMSE, MAE, MAPE, nRMSE, and Diebold–Mariano (DM) test. A Rolling Cross-validation procedure was performed to validate the models. KEY RESULTS: The study showed that Qatar consistently showed the highest PM2.5 levels at 87.90 ± 5.78 µg/m³, followed by Bahrain (67.42 ± 4.59 µg/m³) and Kuwait (58.00 ± 5.95 µg/m³). The modelling approach concluded that machine learning based hybrid models performed well across all competing models, with NPAR-NNAR showing the lowest RMSE (KSA = 2.0181, Qatar = 2.2852, Bahrain = 1.3145, and Kuwait = 2.299), while NAIVE-NNAR performed best for UAE = 1.0462 and Oman = 1.6522. The forecasting results showed that GCC countries may experience variations in PM2.5 levels, with Qatar and Bahrain facing the highest concentrations among them in the coming decades. MAJOR IMPLICATION: This is the first multi-country study across the GCC to forecast PM2.5, showing a significant step forward for environmental health planning. The higher accuracy of modelling approaches is important for improving early warning capabilities, anticipating pollution trends that directly affect respiratory and cardiovascular health outcomes.