LAUSR

Assimilation of pollen observations for increasing accuracy of the deterministic model forecast and reanalysis faces several roadblocks. The most-evident problem today is the long delay of the data because of the manual character of the observations. Automatic monitors are about to eliminate this issue but more are on the way. This paper shows that the classical assimilation of the model state, i.e. pollen concentrations, has very little effect on the forecasts. Due to short relaxation time of the system, the updates generated by the assimilation are forgotten within a few hours. In a search of approaches with a longer-lasting impact, a numerical experiment is conducted assimilating the total seasonal pollen emission, which controls the overall season severity. It turned out to be a prominent example of parameters affecting the model predictions over the long period—in the retrospective simulations. It remains to be demonstrated that this parameter can substantially benefit from assimilation of the near-real-time data that are becoming available from the automatic pollen monitors.