Unveiling the Complex Drivers of Particulate Matter in Saudi Arabia: Employing Machine Learning and Source Apportionment

Saudi Arabia's arid and semi-arid climatic conditions, characterized by limited rainfall and frequent sand and dust storms, lead to high background levels of atmospheric particulate matter (PM). As a result, PM is the primary pollutant of concern in Saudi Arabia affecting human health, ecosystems, and infrastructure. This study offers an innovative approach by combining advanced machine learning techniques with source apportionment methods to reveal underlying drivers of PM ₁₀ and PM _2.5 levels across diverse sites in the Kingdom. The analysis revealed that PM ₁₀ and PM _2.5 levels frequently exceeded air quality standards set by national and international organizations. Using PM _2.5 /PM ₁₀ ratios, PM sources were categorized as anthropogenic, dust storms, and mixed. On average, dust storms, anthropogenic sources, and mixed sources contributed 4%, 6%, and 90%, respectively. Extreme Gradient Boosting (XGBoost) models were developed for several sites, where PM ₁₀ was used as a modelled variable and PM _2.5 , WS, WD, NO, NO ₂ , temperature, relative humidity, and atmospheric pressure as predictors. The contribution of each predictor was evaluated, among meteorological parameters wind speed demonstrated the highest contribution followed by relative humidity. Wind speed was found to positively correlate with particulate levels, which highlighted its vital role in resuspending and transporting dust particles from deserts, and elevating dust pollution in urban areas. Overall, NO and NO ₂ had a weak impact on PM ₁₀ . Considering NO and NO ₂ as a surrogate for anthropogenic activities, especially for road traffic, we can conclude that the contribution of anthropogenic activities was minimal in controlling the levels of PM ₁₀ . Considering the dominant role of meteorology, it is concluded that overall, the contribution of anthropogenic sources to PM is limited in Saudi Arabia and that a major fraction of PM comes from natural sources dominated by meteorological parameters, which should be excluded while assessing compliance with health threshold.