INVESTIGATION OF AERIAL POLLEN DIVERSITY IN SANTINIKETAN, WEST BENGAL AND PREDICTION OF POLLEN CONCENTRATION: A NEW STATISTICAL APPROACH FOR FORECASTING OF POLLEN SEASON

The present study deals with the investigation of diversity of airborne pollen in the atmosphere of Santiniketan, West Bengal (Eastern part of India) using a Burkard personal volumetric sampler for the two-years period 2014–2015. A total of 66 pollen taxa belonging to 37 angiosperm families and one gymnosperm have been identified. Grass pollen alone dominated the airborn pollen assemblage followed by Cassia sp., Acacia sp., Solanaceae, Asteraceae and Cyperaceae. Pollen grains of Cycas sp., Lagerstroemia sp., Spathodea campanulata, Lantana camara, Eucalyptus sp., Malvaceae, Liliaceae, Parthenium hysterophorus, Carica papaya, Peltophorum pterocarpum, Areca catechu, and Catharanthus roseus were also predominant in the air of Santiniketan. The annual pollen index (API) was 77,272 (pollen/m3 of air) in Santiniketan which reflects the rich pollen diversity of this famous sub-urban tourist spot of West Bengal. The place has the highest annual concentrations of pollen in April (13024/m3 in 2014 and 12160/m3 in 2015) in both of the study years. The studies carried out established the association between aero pollen concentration, meteorological factors and air quality data (pollutants). Yearly variations of pollen seasons could be related to the influence of meteorological factors such as temperature, rainfall, relative humidity and wind speed which have been proven statistically by correlation analysis. PM10 and PM 2.5 were found to be statistically significant with total pollen count. We have made a statistical approach using the multivariate regression analysis and Generalized Linear Model to predict onset, duration and peak pollen season to evaluate the threat imposed by the presence of pollen allergens in the air. For this statistical approach, we have considered the influencing meteorological factors and air pollution data as predictive variables and pollen counts as response variable for 2014 and 2015. The fitted regression equation showed goodness of fit of the proposed model with such an adjusted R 2 value (0.8281) which explaining almost 82.81% of the variability for prediction of weekly pollen counts for future years. This aeropalynological survey may serve as guide for allergologiests to predict and manage the source and the incidence of allergic diseases among local inhabitants. Key words: Pollen, Santiniketan, annual pollen index, statistically significant, prediction, pollen season, Regression analysis, Generalized Linear Model.