prilagođeno pretraživanje po punom tekstu

ŠUMARSKI LIST 3-4/2018 str. 26     <-- 26 -->        PDF

The temperature base was 5 °C, which has been used for other forest tree species as it is presented in formula 1 (Cannel and Smith 1983, Chassagneux and Choisnel 1986, Kasprzyk 2009, Dobrowolska et al. 2011).
Descriptive statistics were calculated for all analysed variables, and the level of statistical significance was set at 5 %. To classify clones by GDD for the study period (2004-2013), we used a k-means clustering that divided all 53 investigated clones into three groups depending on the leaf unfolding (early, intermediate and late flushing groups). By using repeated measures analysis of variance, we tested whether and how other analysed variables (precipitation, day of year, and insolation) differ between the GDD clusters. Our model contains three parameters – effect of GDD groups (between subject) the effect of years, and the interaction years × GDD groups (within subject) – because we wanted to test whether our analysed variables (according to GDD groups) behaved approximately the same throughout the analysed period (Davis 2002). For testing sphericity we use Mauchly`s Sphericity test and if sphericity assumption has be violated we use Greenhouse-Geisser Epsilon (G-G) and Huynh-Feldt-Lecoutre epsilon (H-F-L) correction factors which adjust the univariate test degrees of freedom for within subject testing. To determine which of the analysed variables best describes GDD for each GDD group, we used a regression analysis regardless of the clones and the year. First, an univariate regression was performed for each analysed variable. Then, we performed a multivariable (multiple) regression for variables that were statistically significant in the univariate models. Finally, we choose the most appropriate model that served as the best predictor for GDD using differences in R2 values as the selection criterion (Sokal and Rohlf 1995).
The chilling accumulation calculations were based on thetemperature difference between Tmin and Tmax in the period between 1st November and 28th February (formula 2), with ≤ 5 °C adapted from the Thermal Time model (Cannell and Smith, 1983). All statistical analyses were conducted using Statistica 11.0 (StatSoft Inc. 2013).
•If the Tmin is equal to the Tbase then the chilling value for that day is 0.
•X represents an interval in degrees between the Tmin and the Tbase, or between Tmin and Tmax if Tmax is less than the Tbase.
•If the Tmin is higher than the Tbase value then chilling for that day is 0.
The averaged data are clearly segregated between the three groups of Q. robur phenoforms (early flushing, intermediate and late flushing) for all measured parameters during the duration of the experiment, as shown in Table 1. The start date of leaf unfolding phenoform is clearly defined by the GDD model (Table 1 and Figure 1) .
The mean values of GDD were 118.42 for the early flushing group, 188.34 for the intermediate group and 261.61 for the late group of oak clones. All other measured parameters (precipitation, day of year, and insolation) are similarly separated and are divided into three separate groups (Figure 2). For the early flushing group, the mean day of year is 85 and insolation is 271.04 hours, which achieved the GDD requirements and the clones of that group started with the leaf unfolding. The intermediate group required an average of 98 days and 342.49 hours of insolation, whereas the late flushing group required 109 days and 411.57 hours of insolation. The calculated Coefficient of Variation (C.V. %) values for all parameters are lowest for the late flushing form group, likely because they belong to a group with at least of 9 genotypes. In contrast, the intermediate group contains 13 of genotypes, and the early group contains 31 genotypes.
The results of k-means clustering of the clones according to their values of GDD for the period of 2004-2013 are shown in Fig. 1. The clones are clearly classified into three groups (early, intermediate and late flushing).
To examine how the other analysed variables (precipitation, day of year, and insolation) affected the GDD groups, the variables were tested with repeated measures ANOVA to determine whether they individually participate in the GDD groups, their behaviour over years and their behaviour over years under certain clusters. Because sphericity assumption has been violated we use G-G and the H-F-L correction factors. In Table 2 we gave adjusted p value for both. In all tests the p value before and after using correction factors p values were the same (p <0.0001). Sphericity test, G-G episolon and H-F-L epsilon for Precipitation (Chi2=514.61; df =44; p<0.001; G-G=0.3337; H-F-L=0.3574); Insolation (Chi2=348.74,df =44; p <0.001; G-G=0.4446; H-F-L=0.4881) and day of year (Chi2=313.64 df=44; p <0.001; G-G=0.3994; H-F-L=0.4341).
The results of the repeated measures ANOVA (Table 2) show that there are statistically significant differences between each group for all of the analysed variables. There is a statistically significant difference over the years and significant differences in the interaction between the GDD group and the years, which means that all the analysed variables for individual GDD group over the year do not behave the same (Figure 2).