The problem of developing a 2-week-on ahead forecast of atmospheric cypress pollen levels is tackled in this paper by developing a principal component multiple regression model involving several climatic variables. The efficacy of the proposed model is validated by means of an application to real data of Cupressaceae pollen concentration in the city of Granada (southeast of Spain). The model was applied to data from 11 consecutive years (1995-2005), with 2006 being used to validate the forecasts. Based on the work of different authors, factors as temperature, humidity, hours of sun and wind speed were incorporated in the model. This methodology explains approximately 75-80% of the variability in the airborne Cupressaceae pollen concentration.