Modelling hourly spatio-temporal PM2.5 concentration in wildfire scenarios using dynamic linear models

Particulate matter with aerodynamic diameter < 2.5 μm (PM₂.₅) is one of the main pollutants generated in wildfire events with negative impacts on human health. In research involving wildfires and air quality, it is common to use emission models. However, the commonly used emission approach can generate errors and contradict the empirical data. This paper adopted a statistical approach based in evidence of ground level monitoring and satellite data. An hourly PM₂.₅ spatio-temporal model based on a dynamic linear modelling framework with Bayesian approach was proposed in a territorial context with a reduced number of monitoring stations for particulate matter. The model validation is complicated by the fact that all monitoring stations are used in the model calibration. The novel validation method proposed considered both the particulate matter with aerodynamic diameter < 10 μm (PM₁₀) recorded as daily value from 24-h mean every six days as well as the PM₂.₅/PM₁₀ ratio. Modelling was carried out to provide satisfactorily the exposure level of PM₂.₅ in a case study of wildfire event.