Records of pollen concentrations and meteorological data from five years of sampling in North London are used to develop models to predict daily Poaceae pollen counts. These models are developed using a variety of techniques including the use of a third-order polynomial curve to forecast potential pollen concentration and the use of accumulated meteorological variables. They are then tested for accuracy in forecasting the 1991 Poaceae pollen season. A model developed using accumulated average temperature, daily rainfall, relative humidity and maximum temperature as predictor variables attains an accuracy of 71 for the 1991 season. These models are designed specifically for their ease of use in practice and their reliance on meteorological variables that may be forecast in advance. They are also novel in attempting a forecast of daily pollen concentration based upon an estimate of the potential severity of the season instead of using a retrospective record of pollen abundance as previous authors have done.
展开▼