Modelling of weather variability effect on fitophenology

Phenology models are useful tools for various sectors of human activity, particularly for all environmental studies. An historical, 46-year phenological record, and meteorological data set for period 1955–2000 were used to analyse the abilities of the statistical models to predict flowering and leaf unfolding time for a set of wild and cultivated plants in Slovenia. With single phenological model, we have predicted timing of phenophase for particular plant on the base of previous phenological data of the same plant or on the base of previous phenological data of the other plants. The most frequently included independent variables in such models were common silver birch, dandelion and horse-chestnut. It was stated that these plants may be used as phenological indicators in given conditions. Correlation analysis and linear multiple regression were applied to establish the relationship among phenological development and climatic variables (temperature, rainfall, North Atlantic Oscillation, day length). Different thresholds temperatures have been selected for eight different locations with the smallest standard deviation in growing degree day's method to calculate thermal time. Various plant species responded differently to the same climatic factors and were best fitted to certain geographic region. The timings of spring phenophases strongly correlated with temperature of the precedent months, on the other hand rainfall and North Atlantic Oscillation explained smaller part of phenological variability. Photothermal time significantly improved results of phenoclimatic models when taken into account instead of thermal time. The validity of the results was tested with cross validation method and using independent data set for the year 2000, respectively. Considering the high year-to-year variability of phenological events, the models presented provide satisfactory estimations of the leaf unfolding and flowering dates. Formal equations presented in this study could be powerfully extended and applied to other sites and plants, provided that a sufficiently long time series of phenological and meteorological data were available.