Statistical models to predict daily airborne pollen concentration from meteorological conditions have been evaluated using several years' data from Stockholm. ForBetulapollen values for the multiple correlation coefficient (R) were up to 0.8 using a model describing the seasonal course as well as the day-to-day variation. Models of this kind cannot interpret causally pollen variation. However, the daily temperature values from early spring have a large influence on the time of the beginning of the pollen season and also on the subsequent day-to-day variation. Based on this study, pollen forecasts have been initiated in Stockholm in the spring of 1981.