The impact of the spatial resolution of vegetation cover on the prediction of airborne pollen concentrations over northern Italy

Accurate pollen forecasting models can help the self-management of allergic respiratory diseases. Our study introduces and validates, for the first time, a pollen modelling system covering the Veneto Region (Italy) at the 3 km spatial resolution for 2019. The model simulated the pollen dispersion, diffusion and deposition processes, using vegetation cover (VC) maps, phenological pollen emission algorithms, and meteorological forecasting. We have specifically analysed the influence of the spatial resolution of VC maps on predicted airborne pollen concentrations for alder, birch, olive, grass, and ragweed. Two VC datasets were used: CAMS VC: the European CAMS dataset at ca. 10 km horizontal resolution; detailed VC: high-resolution datasets (from 250 m to 1 km spatial resolution). Predicted daily averaged concentrations obtained with CAMS and detailed VC were compared to the observations collected at 15 monitoring stations using model performance indicators and pollen seasonal-derived parameters. A stratified analysis assessed performance variations in lowland versus mountain environments. The results showed a reduction of the root mean square error (RMSE) for alder and birch pollen using the detailed VC (detailed VC vs. CAMS VC: 15.7 vs. 133.6; 17.8 vs. 52.5 p/m3, respectively), while higher RMSE resulted for grass (24.5 vs. 20.7 p/m3). Similar RMSEs were obtained for olive and ragweed pollen (3.8 vs. 4.0; 3.9 vs. 3.9 p/m3, respectively). Results from the differences in Seasonal Pollen Integrals (SPIn) were consistent with the RMSE patterns. The onset of pollen seasons was more accurately predicted than their end. The general improvement of pollen predictions obtained with the detailed VC was particularly evident in the mountains. Incorporating data from detailed vegetation maps into atmospheric dispersion models has significantly improved predictions for arboreal pollen (alder, birch, olive), especially in complex surfaces where high-resolution input data is crucial.