The impact of data assimilation into the meteorological WRF model on birch pollen modelling

Abstract

We analyse the impact of ground-based data assimilation to theWeather Research and Forecasting (WRF) meteorological model on parameters relevant for birch pollen emission calculations. Then, we use two different emission databases (BASE – no data assimilation, OBSNUD – data assimilation for the meteorological model) in the chemical transport model and evaluate birch pollen concentrations. Finally, we apply a scaling factor for the emissions (BASE and OBSNUD), based on the ratio between simulated and observed seasonal pollen integral (SPIn) to analyse its impact on birch concentrations over Central Europe. Assimilation of observational data significantly reducesmodel overestimation of air temperature,which is themain parameter responsible for the start of pollen emission and amount of released pollen. The results also show that a relatively small bias in air temperature from the model can lead to significant differences in heating degree days (HDD) value. This may cause the HDD threshold to be attained several days earlier/later than indicated from observational data which has further impact on the start of pollen emission. Even though the bias for air temperature was reduced for OBSNUD, the model indicates a start for the birch pollen season that is too early compared to observations. The start date of the seasonwas improved at two of the 11 stations in Poland. Data assimilation does not have a significant impact on the season's end or SPIn value. The application of the SPIn factor for the emissions results in a much closer birch pollen concentration level to observations even though the factor does not improve the start or end of the pollen season. The post-processing of modelled meteorological fields, such as the application of bias correction, can be considered as a way to further improve the pollen emission modelling.