The search for the best airborne pollen monitoring locations

Abstract

In Europe a quarter of the adult population and a third of all children suffer from allergenic airborne pollen thereby decreasing the quality of life. In order to ease the pollen induced symptoms mitigation measures can be applied. This, however, requires timely information on forthcoming pollen episodes derived from early warning systems. These systems can substantially be improved when pollen observations from strategically well-chosen pollen monitoring stations are assimilated. Here we explore the network quality (i) and network coverage (ii) of the current five pollen monitoring stations in Belgium. As reference dataset we use the spatio-temporal distributions of daily surface airborne birch and grass pollen levels as produced by the operational early warning system for pollen on the website of the Royal Meteorological Institute of Belgium. This system implements the SILAM model (System for Integrated modeLling of Atmospheric coMposition) and ECMWF meteorological data. The ability of the network to reproduce the concentration field over the region of interest is quantified by the RMSE computed from the reference concentration field and the interpolated concentration field for each day of the pollen season. In the first step, time series of the current daily pollen observations in the network are interpolated over space by applying the radial-based function. This results in the daily interpolated concentration fields which we compare with the spatially distributed daily reference data. For evaluating the network coverage of the current five monitoring stations we perform a footprint-based analysis. Footprints relate directly to the fraction of air reaching the monitoring device. By applying pollen emission point sources in the five stations into SILAM that is run in the backward mode (three days back), we can investigate the travelling trajectory of the captured birch and grass pollen in the air observed at the network stations. Nine pollen seasons (2013-2021) were analyzed using ECMWF ERA5 meteorology. First results on the network quality for birch pollen show that over a period of nine pollen seasons more than 60% of the daily RMSE values derived from the interpolated daily concentrations are less than their mean value. This is an indication that the interpolated network performs well compared to the spatio-temporal reference dataset derived from SILAM. For the 2013 birch pollen season more than 80% is reached. In contrast, this is only ~40% for 2020. The applied time scale is of great importance, since at smaller time scales (days, hours) network configurations may degrade faster than on larger time steps (weeks, months, seasons). The footprint-based analysis shows that on average the coverage of the monitoring stations for birch pollen is quite good. There are, however, large differences during the 2013-2022 seasons which might be due to the typical large inter-seasonal variation in birch pollen production. For grass pollen, the average coverage is better, and the inter-seasonal variation much lower.