Analysis of Changes in Pollutant Concentrations Levels Using a Meteorological Normalisation Technique Based on a Machine Learning Algorithm

Abstract

In this study, a methodological procedure combining a technique of meteorological normalisation, based on a random forest algorithm, with trend analysis and the change points detections in air quality time series is developed to analyse changes in pollutant concentrations levels. Data of air pollutants and meteorological parameters, collected over the period 2013–2019 in a rural area affected by anthropic sources of air pollutants, are used to test the procedure. The results appear to be promising in revealing, in a robust way, changes in pollutant levels not clearly observable in the original data.