Grasses characteristically produce a huge amount of small pollen grains, which pose a risk to allergy sufferers. In many aerobiological studies, great variations in the behaviour of the grass pollen season are stressed. We state that in Rzeszów and Ostrowiec Św. there is some regularity in the pattern of the main grass pollen seasons, which is clearly double-peaked. The aim of our work was to elaborate the algorithm which defines the main grass pollen season. Next, the null hypothesis was tested about the lack of difference between daily pollen concentrations and meteorological parameters. Grass pollen seasons were defined using the method of fitting two bell curves. The estimated grass pollen season is characterised by two periods of high or relatively high concentrations, separated by a period of low concentration. In order to investigate the time dependence of the correlation between pollen concentration and the weather parameters, the Gaussian-weighted correlation coefficient has been calculated. Maximum temperature, mean temperature and sunshine positively correlated with pollen concentrations, but relative air humidity and rainfall on the previous day had a negative effect. The temperatures of the second and third ten-day periods of April were the best independent variables for forecasting the beginning and peak dates of the main pollen seasons. An analysis of the results shows that the pattern of successive flowering in grass species and meadow cutting dates appear to be the factors which cause the characteristic bimodal behaviour of the grass pollen season.