Algorithm for forecasting the total amount of airborne birch pollen from meteorological conditions of previous years

Abstract

The birch tree (genus Betula L.) disperses airborne pollen annually from April to June, causing severe symptoms in pollinosis sufferers. Because of interannual variations in pollen levels, there is an urgent need to develop a forecasting model with greater precision in order to provide accurate information to patients and medical personnel regarding airborne pollen levels. We developed an algorithm for forecasting the total amount of airborne birch pollen. This equation suggested that the total amount of airborne pollen in a given season could be estimated using only the meteorological data from previous years. In order to discover potential predictive relationships, a data set including airborne pollen data from 1996 to 2015 and meteorological data from 1990 to 2014 was used to construct forecasting models. Statistical evaluation results were examined to select the optimal model, showing that forecasting models obtained the highest accuracy when using meteorological data from June and the best model performance was achieved using the average daily maximum air temperature and solar radiation of the previous five years. We also developed an extended model that included relative humidity, which demonstrated better predictive capability. These findings clarify that a model with greater predictive power can be constructed using the meteorological conditions from the previous five years. In order to assess this conclusion, the algorithm was tested by forecasting the total amount of airborne birch pollen in Japan with good results.