Airborne grass pollen is one of the most important causes of pollen allergy worldwide. Still, the identification of its specific emission sources is a challenging task because of the ubiquitous nature of this pollen type. Previous studies have explored the contribution of the most abundant grass species and grassland habitats to airborne grass pollen at a local spatial scale using in situ phenological samplings and remote sensing data. This background allowed us to hypothesise that wild grasslands are a more important pollen emission source than human-managed areas dominated by grass species. The main aim of this study was to extend this hypothesis by modelling the grass pollen load measured by aerobiological stations in temperate and Mediterranean areas during the last two decades using remotely sensed land cover data at continental scale. The mean Annual Pollen Integral (APIn) for stations in the Iberian Peninsula and Central Europe was calculated for a minimum of 10 years during the last two decades. Years with APIn values below the 10th percentile and above the 90th percentile for each station were removed to get a more centred mean and to avoid anomalous years. The potential sources considered in this study were derived from the continental-scale remotely sensed products provided by the Copernicus Land Monitoring Service. The combination of different products allowed us to characterise the spatial distribution of different grass-dominated habitats, namely: intensively managed grasslands (seasonally cut and tilled pastures and arable lands), wild grasslands (natural or semi-natural grasslands including Iberian "dehesa" systems). The relationship between airborne pollen and potential emission sources was established compared to human-managed grasslands and crops. Wild grassland distribution explained between 60-70% of the variance of the mean grass APIn, depending on the spatial resolution of the source map and the maximum influence distance considered. A detailed analysis of the emission influence and source distance of wild grasslands provided an influence mathematical function with maximum values of correlation of 0.75 (Spearman's test) between airborne pollen and source abundance. Conversely, intensively human-managed pastures and arable lands only explained about 10-20% of the variance of the mean grass APIn, obtaining a non-coherent influence function. These results constitute valuable evidence of the land-use and vegetation types with the major airborne pollen contribution of grasses at a large spatial scale. Our findings are the basis for comprehensively study the factors involved in the long-term pollen trends of grasses by information of the land use and land cover changes for the last decades in Europe. Also, the influence function for the most important pollen sources can be combined with short-term meteorological data to develop a forecasting model of the interannual APIn of grasses.
