Abstract / Summary
Prevention of pollen exposure is a way to assist allergic individuals to better cope with their disease. To enable this prevention, a prediction of next day(s) pollen flight is needed. The most accurate predictions are from numerical models which calculate the expected airborne pollen concentrations using simulation of emission combined with atmospheric dispersal processes (as in the SILAM model). However, current numerical models are based on historical averages and make predictions for the whole of Europe, which is not always optimal for a specific location. In this study we adapted a continental numerical model forecast to regional conditions in two areas in Germany: Bavaria in the South, and Leipzig and Berlin in eastern Germany. Eleven automatic pollen monitors were continuously running in these areas, allowing predicted pollen concentrations to be compared with observations. Adaptation of the modelling process to a regional scale increased the correlation (r) between observations and predicted pollen flight by about 38% for birch (up to 0.71 ± 0.07) and 18% for grasses (up to 0.57 ± 0.09). The most effective optimised parameters for birch pollen were; (i) adjusting the grid from a rotated to a non-rotated Latitude–Longitude coordinate system; (ii) updating the distribution and abundance map of potential pollen sources; (iii) adjusting the rate of pollen released in a year; and (iv) the timing based on temperature sum for birch. For grass pollen, the same parameters were optimized, except for the temperature sum. Here, for adjusting the timing, a phenological calendar method was used.