Wildfires Impact Assessment on PM Levels Using Generalized Additive Mixed Models

Wildfires are relevant sources of PM emissions and can have an important impact on air pollution and human health. In this study, we examine the impact of wildfire PM emissions on the Piemonte (Italy) air quality regional monitoring network using a Generalized Additive Mixed Model. The model is implemented with daily PM10 and PM2.5 concentrations sampled for 8 consecutive years at each monitoring site as the response variable. Meteorological data retrieved from the ERA5 dataset and the observed burned area data stored in the Carabinieri Forest Service national database are used in the model as explanatory variables. Spline functions for predictive variables and smooths for multiple meteorological variables' interactions improved the model performance and reduced uncertainty levels. The model estimates are in good agreement with the observed PM data: adjusted R-2 range was 0.63-0.80. GAMMs showed rather satisfactory results in order to capture the wildfires contribution: some severe PM pollution episodes in the study area due to wildfire air emissions caused peak daily levels up to 87.3 mu g/m(3) at the Vercelli PM10 site (IT1533A) and up to 67.7 mu g/m(3) at the Settimo Torinese PM2.5 site (IT1130A).