Effect of geographical parameters on PM10 pollution in European landscapes: a machine learning algorithm-based analysis

Abstract

BackgroundPM10, comprising particles with diameters of 10 µm or less, has been identified as a significant environmental pollutant associated with adverse health outcomes in European cities. Understanding the temporal variation of the relationship between PM10 and geographical parameters is crucial for sustainable land use planning and air quality management in European landscapes. This study utilizes Conditional Inference Forest modeling and partial correlation to examine the impact of geographical factors on monthly average concentrations of PM10 in European suburban and urban landscapes during heating and cooling periods. The investigation focuses on two buffer zones (1000 m and 3000 m circle radiuses) surrounding 1216 European air quality monitoring stations.ResultsResults reveal importance and significant correlations between various geographical variables (soil texture, land use, transportation network, and meteorological) and PM10 quality on a continental scale. In suburban landscapes, soil texture, temperature, roads, and rail density play pivotal roles, while meteorological variables, particularly monthly average temperature and wind speed, dominate in urban landscapes. Urban sites exhibit higher R-squared values during both cooling (0.41) and heating periods (0.61) compared to suburban sites (cooling period R-squared: 0.39; heating period: R-squared: 0.51), indicating better predictive performance likely attributed to the less heterogeneous land use patterns surrounding urban PM10 monitoring sites.ConclusionThe study underscores the importance of investigating spatial and temporal dynamics of geographical factors for accurate PM10 air quality prediction models in European urban and suburban landscapes. These findings provide valuable insights for policymakers, urban planners, and environmental scientists, guiding efforts toward sustainable and healthier urban environments.