The 72hour airborne pollen concentration (PMPC) prediction model is devised by using the observation data of pollen from the meteorological observing tower in the Hexi district of Tianjin from 1999 to 2004 and the conventional meteorological data. The different PMPC models are established by means of the multiple linear/nonlinear regression techniques with three schemes for the whole pollen season, spring, summer, and autumn, as well as for various stages, respectively. The comparisons are also conducted between various schemes and models. The results show that the multiple nonlinear regression method is better than that of linear one, and the variseason scheme is better than others, except that for summer.