State-of-art in modelling particulate matter (PM) concentration: a scoping review of aims and methods

Air pollution is the one of the most significant environmental risks to health worldwide. An accurate assessment of population exposure would require a continuous distribution of measuring ground-stations, which is not feasible. Therefore, significant efforts are spent in implementing air-quality models. However, a complex scenario emerges, with the spread of many different solutions, and a consequent struggle in comparison, evaluation and replication, hindering the definition of the state-of-art. Accordingly, aim of this scoping review was to analyze the latest scientific research on air-quality modelling, focusing on particulate matter, identifying the most widespread solutions and trying to compare them. The review was mainly focused, but not limited to, machine learning applications. An initial set of 940 results published in 2022 were returned by search engines, 142 of which resulted significant and were analyzed. Three main modelling scopes were identified: correlation analysis, interpolation and forecast. Most of the studies were relevant to east and southeast Asia. The majority of models were multivariate, including (besides ground stations) meteorological information, satellite data, land use and/or topography, and more. 232 different algorithms were tested across studies (either as single-blocks or within ensemble architectures), of which only 60 were tested more than once. A performance comparison showed stronger evidence towards the use of Random Forest modelling, in particular when included in ensemble architectures. However, it must be noticed that results varied significantly according to the experimental set-up, indicating that no overall best solution can be identified, and a case-specific assessment is necessary.