Prediction of annual variations in atmospheric concentrations of grass pollen. A method based on meteorological factors and grain crop estimates.

We performed an aerobiologic observation of the grasses present in Madrid for 14 years (1978-1991), using volumetric air samplers. The counts obtained show that the major grass pollen release period (average daily grass pollen counts greater than 50 grains/m3 of air) occurs in the months of May and June, although lower counts can occur some days from the end of January onward. There are wide year-to-year variations in total atmospheric grass pollen counts, expressed as the total sum of the mean daily concentrations from April 1st to July 30th (ranging from 2568 to 6624). A strong, statistically significant correlation, based on Spearman's rank test and/or simple and multiple linear regressions, was found between the total grass seasonal count and preseasonal rainfall from October to March (R2 = 0.64; P = 0.0429). The meteorological variable which gave the correlation with greatest statistical significance (R2 = 0.97; P = 0.0016) was the average monthly preseasonal humidity from October to March. A good correlation was also found between March estimates of wheat, rye and barley crops and the total grass count (R2 = 0.73; P = 0.006). A model was designed from the above mentioned humidity variable through a multilinear regression analysis, and it was possible to predict, at the beginning of April, total seasonal counts for 1989 (predicted = 5468; actual = 4410; average error = 24%), 1990 (5033; 6090; -17%) and 1991 (3930; 2568; 53%). These data may help clinicians to predict and prepare themselves for the intensity of the grass pollen season and to explain yearly variations in the severity of symptoms.