Search

The research for the thesis is related to the construction of long and short term pollen forecast models for Grass, Olea and Urticaceae in Albania which are the most allergenic taxa in this country. Aerobiology, which is the study of organic particles (bacteria, fungal spores, pollen, small insects) passively transported by the air [Spieksma, 1991], has received very little attention in Albania. The research represents a major advance as it is the first work of this kind in this country. The aims of the research were to investigate the features of the pollen seasons for the mentioned taxa as well as establishing relationships between these features and the main controlling weather variables. The achievement of these aims provided the basis for a further aim of the research which was to construct the long and short term forecasts for Grass, Olea and Urticaceae. Also the research investigated the possibility of constructing short term forecasts for the mentioned taxa on the non-rainy days. The data used for the research were pollen data and meteorological data from Tirana city in the period 1995-2004. It was not possible to have data from the years 1997 and 1999, 2000, 2001 due to practical problems. The meteorological data for the research were obtained from the Meteorological Institute in Tirana. In order to investigate the pollen season features, the pollen season for Grass and Urticaceae were divided into three periods, the pre peak, peak and post-peak since the behavior of the pollen seasonal variation curve differs according to the phases. The Olea pollen season is very short lasting for no more than 40 days so this was divided in two periods namely pre-peak and peak. An important outcome of the research was also the production of a pollen calendar for the main allergenic taxa based on five years of data. The pollen calendar will be useful for allergists and the general public. A lot of meteorological variables were used in the empirical analysis (correlation and regression analysis) in order to investigate which of the weather parameters give most explanation of the features of the pollen season. A number of variables were examined for possible inclusion in the linear regression analysis. The variables were selected after reviewing previous research on the effects of meteorological variables on the production of pollen from the three taxa. Linear regression was used to construct the long term forecasts for Grass, Olea and Urticaceae while multiple regression analyses were used for the construction of the short term forecasts. The forecasts obtained were able to forecast with an accuracy from 50-85% for Grass, Olea and Urticaceae. The models obtained for the non-rainy days were successful for Olea in the pre-peak period. No rainfall was recorded during the peak period. Also the Urticaceae models for the non rainy days were accurate only for the pre and peak period. Neural networks were used as an alternative method to the regression analysis for Grass, Olea and Urticaceae and were very accurate. This method was able to forecast the daily variations of the mentioned taxa as a whole season as well as pre and post peak period. It also increased the accuracy to 96% to forecast the daily variations for the Olea in the prepeak period and 82% in the post-peak period. The accuracy achieved for Urticaceae was 98% in the pre-peak period and 95% in the post-peak and whole pollen season respectively. The skills gained and the forecast models that were constructed through this research will enable Aerobiology to be established in Albania as a scientific discipline. The work will allow the creation of a network pollen system similar to that in other European countries. The results will be for use for the public, for doctors, pharmacists and related bodies. The acquisition of more pollen data through the continuous monitoring sites in Albania will enable the constructed forecast models to be updated.