Five ways to define a pollen season: exploring congruence and disparity in its attributes and their long-term trends

We applied five sets of criteria for pollen season (PS) definition, which are associated with the ‘clinical’, ‘grains’, ‘logistic’, ‘moving average’ and ‘percentage’ methods, on a 30-year time series of daily airborne pollen concentration values of Cupressaceae (cypress family), Oleaceae (olive family) and Poaceae (grass family), from Thessaloniki, Greece. These methods could identify a pollen season for more than 90% of the study period for all three taxa, except for the clinical that identified less than 40% for Oleaceae. The estimated values of the PS start, end, and duration varied largely, in a method-specific way. Even significant reverse patterns of change were recorded for the same attribute and taxon, as for the Poaceae PS end date that shows a significant advance earlier in the year with the moving average method and delay with the percentage. As the season peak date is method-independent and, hence, directly comparable, we recommend this attribute to be examined in airborne pollen studies. Results taken with the percentage method could be compared with those of the past for Thessaloniki, with a more than 10-year shorter time series. No climate-related change was detected for Cupressaceae PS attributes, but an earlier peak was now detected for Oleaceae and Poaceae, and a later end and a longer duration for Poaceae, suggesting an increased allergy risk for this taxon. The different criteria concurrently applied for PS definition that led to even greatly diverging results for the same PS attribute is an issue for the science and calls for efforts at standardization.