Mind the weather signal in the Seasonal birch Pollen Integral

Operational pollen forecast models are potentially powerful tools for patients with allergic rhinitis symptoms caused by airborne pollen. Such a warning system can inform people in a timely manner so preventive measures and adapted medication doses can be taken. In Belgium a birch pollen forecast framework has been established based on the pollen emission and transport model SILAM (System for Integrated modeLling of Atmospheric composition) using a bottom-up approach. This implies, however, that spatially distributed birch pollen emission sources should be assessed before the start of the pollen release season.

We hypothesize that pre-seasonal meteorological-based proxies can be used in combination with the observed Seasonal Pollen Integral (SPIn) for updating the birch pollen emission source map into SILAM prior to the start of the birch pollen season.

Here we analyze the correlations between these pre-seasonal proxies and SPIn observations of birch pollen at the aerobiological surveillance network of Belgium for the period 1987 to 2019. Based on the correlations, temporal scaling factors are derived for updating the main birch pollen emission source map for Belgium (with 2018 as reference year). We evaluate the updated SILAM runs driven by ECMWF ERA5 meteorology by comparing multi-seasonal (2013-2019) modelled levels with daily observed pollen data from the surveillance network.

Preliminary analysis indicates that implementing updated pollen emission source maps in SILAM runs increase the model performance indicator R² (correlation coefficient) by 24% for daily airborne birch pollen levels, and by 90% for the SPIn values at all measurement sites of the network. However, by identifying and adjusting the impact of the weather effect on the observed SPIn values during the pollen season, the correlations with the pre-seasonal meteorological-based proxies as well as the SILAM model performance increase drastically. The R² between modelled and observed SPIn increases from 0.37 without any scaling to 0.71 including scaling and to 0.83 including weather adjusted scaling.

This shows the high potential for improving the modelling and forecasting of the birch pollen levels if pre-seasonal environmental data are included to assess the state of the spatial distributed birch pollen emission sources prior to the start of the pollen release season.