气候变化领域集成服务门户

Aerobiological studies mostly apply linear models when analysing long-term changes in pollen concentrations. There is some evidence that linear models are not fully suitable for describing these changes in recent decades since pollen concentrations of some species have increased until 1990 and decreased subsequently. In order to describe the behaviour of annual pollen concentrations (a factor contributing to the incidence of allergies) in a more flexible way, we made use of Bayesian statistics that describe discontinuities (i.e. change points) and quantify the direction and speed of changes. We examined long-term aerobiological pollen data of Betula spp. (birch), Corylus spp. (hazel) and Poaceae (grasses) from six stations in Switzerland (Basel, Buchs, Davos, Munsterlingen, Neuchatel and Zurich) for the period 1985-2014. For most of the analysed stations and species, the one change point model was considered as the best model. The only exceptions were Corylus pollen time series recorded at Neuchatel and Zurich that could be better described with the linear model. The results indicated that all pollen records of Corylus were characterized by increased pollen concentrations in recent years. For Betula pollen concentrations, four sites were associated with recent decreases; increases were only found for Buchs and Munsterlingen. Regarding Poaceae pollen, half of the stations (N=3) were linked to decreases. Trends of linear regressions differed considerably in magnitude or even differed in sign compared to the Bayesian results. Our results indicate that the choice of the statistical method is of major importance when interpreting aerobiological data. Further studies should focus on the reasons (climate, land use changes, etc.) responsible for changes in atmospheric pollen loads in detail.