Random forest meteorological normalisation models for Swiss PM<sub>10</sub> trend analysis

Abstract

Abstract.Meteorological normalisation is a technique which accounts for changes in meteorology over time in an air quality time series.Controlling for such changes helps support robust trend analysis because there is more certainty that the observed trends are due to changes in emissions or chemistry, not changes in meteorology. Predictive random forest models (RF; a decision tree machine learning technique) were grown for 31 air quality monitoring sites in Switzerland using surface meteorological, 5 synoptic scale, boundary layer height, and time variables to explain daily PM 10 concentrations. The RF models were used to calculate meteorologically normalised trends which were formally tested and evaluated using the Theil-Sen estimator. Between continuing to decrease at similar rates as in the past. The ability for RF models to be interpreted was leveraged using partial dependence plots to explain the observed trends and relevant physical and chemical processes influencing PM 10 concentrations.Notably, two regimes were suggested by the models which cause elevated PM 10 concentrations in Switzerland: one related to poor dispersion conditions and a second resulting from high rates of secondary PM generation in deep, photochemically active boundary layers. The RF meteorological normalisation process was found to be robust, user friendly and simple to implement, 15 and readily interpretable which suggests the technique could be useful in many air quality exploratory data analysis situations.

Abrupt but smaller than expected changes in surface air quality attributable to COVID-19 lockdowns

“…In this case, emission reductions and the unfavorable meteorological conditions drove changes of approximately −19.3 and +38.6% in the observed levels, respectively, leading to an overall +19.2% increase in PM 2.5 . Our results demonstrate that meteorological variations, rather than emission changes on the scale of those occurring during the COVID-19 lockdowns, dominate short-term variability in air pollutant concentrations, which is consistent with previous studies ( 12 , 14 , 20 , 24 ).…”

“…To do this, we first build an RF model for each pollutant and for each year (December to May). The RF model–based weather normalization technique was introduced in Grange et al ( 12 ). Briefly, the RF model was built independently for each period (December 2015 to May 2016, December 2016 to May 2017, December 2017 to May 2018, December 2018 to May 2019, and December 2019 to May 2020), each pollutant, and each site type within a city.…”

“…A major caveat in a number of these studies is that meteorology moderates the link between emissions and pollutant concentrations, and so, weather changes can mask the changes in emissions on air quality ( 12 – 14 ). Such methods cannot explain the observed severe pollution events during the lockdowns in some cities ( 15 – 17 ).…”

“…Machine learning offers an alternative and reliable method in quantifying changes in air quality due to emissions and meteorological factors ( 12 – 14 ). Myllyvirta and Thieriot ( 22 ) used a random forest (RF) method ( 13 ), which was developed for assessing long-term air quality changes, to estimate the short-term changes in NO 2 and PM 10 in Europe due to the COVID-19 lockdowns (see Materials and Methods).…”

Machine Learning Approaches for Outdoor Air Quality Modelling: A Systematic Review

“…Among the most recent studies we find Grange et al [22]. This study aims to (i) build a predictive model of PM 10 based on meteorological, atmospheric, and temporal factors and (ii) analyze PM 10 trend during the last ten years.…”