大分大学医学部 (大分県由布市挟間町) におけるスギ花粉飛散予測の検証

Abstract

We investigated the accuracy of the prediction of Japanese cedar-pollen dispersion at the Oita University Faculty of Medicine Complex for use in future prediction models. We compared the predicted and actual total pollen counts per year, the initial day of pollen scattering, and the daily pollen counts from 2005 to 2014. Total pollen counts per year were predicted by the amount of sunshine in the Hita area, which was considered to be the source of pollen in our previous study, during the previous July. The initial day of pollen scattering was predicted using the variation pattern in air temperature. Daily pollen counts were predicted by performing a multivariate analysis using the type-II quantification method. The differences between the actual and predicted total pollen counts per year increased after 2011. The correlation coefficient between the total pollen count and the amount of sunshine duration in the Hita area during the previous July greatly decreased in the same period, and was surpassed by the correlation coefficient for the Nakatsu area. We speculated that the primary source of pollen had moved from the Hita refion to the Nakatsu region. The average difference between the predicted and actual initial day of pollen scattering was 3.2 days. This difference was decreased when we reanalyzed the data using the Nakatsu area as the source of pollen. With regard the prediction of the daily pollen counts, the overall accuracy was 60%-70%, but the accuracy for the full-scale scattering period was 40%-60%. Our results suggest that the prediction of atmospheric Japanese cedar-pollen dispersion must be routinely reviewed. In addition, when estimating total pollen counts and the initial day of pollen scattering, it is necessary to consider the pollen source area. This method must be examined further for its accuracy in predicting the daily pollen counts.