Modelling of plant phenological development

Abstract

Possibilities of predicting phenological development of wild vegetation and cultivated plants were studied, based on meteorological parameters in Slovenia for the period 1955-2000. Correlation analysis and linear multiple regression were used to establish the relationship among phenological development and meteorological parameters. Correlation coefficients between the starting dates of phenological phases with the air temperature of the previous 2-3 months were relatively high (0.6 to 0.85). North Atlantic Oscillation Index was also statistically significant correlated with average air temperatures from December to March and spring phenophases appearance; however the amount of monthly precipitation in actual conditions was not significantly correlated to mostly of discussing phenophases. Lower threshold temperatures were determined with the least standard deviation of growing degree-days. Variability of calculated heat sums among years was relatively high (10% to 30%). The inclusion of photothermal time for location Ljubljana significant improved the models. With linear multiple regression we formed phenological and phenoclimatic models. The most frequently included independent variables of phenological models were birch, dandelion and horse chestnut, we can state that they were phenological indicators in given conditions. We could explain from 50% to 79% of variability with phenological models, respectively 86% on average with phenoclimatic models. Thermal time, mean monthly temperatures and photothermal time for Ljubljana were most frequently included variables in phenoclimatic models. We tested models with cross validation method; predictions were quite accurate with at least 5 days foresight period.