Fig. 1

Variations over a calendar year of day length [h:min(m)], radiation (J/m²), temperature (1984–2007; °C) and rainfall (1980–2008; mm). Box plots are presented for radiation, temperature and rainfall (1980–2008; mm); white dots minimum and maximum values, black dots mean, box 25th and 75th percentiles, line inside the box: 50th percentile (median). The year can be subdivided into four climatic units (T1, T2, T3 and T4). The statistical significance of this subdivision was tested using a principal component analysis (PCA) based on the four climatic variables collected monthly over the previous 18 years, followed by a between-class analysis (Dolédec and Chessel 1987). This quantified the proportion of the PCA’s inertia explained by a variation across the months gathered by a qualitative variable, in this case, climatic typology (Lebreton et al. 1991). In our case, 53.25 % of PCA inertia is explained by the qualitative variables (climatic units: T1, T2, T3 and T4). We then tested the significant differences between climatic units using a permutation test (999 permutations for estimating the P-value with significance set at 1%). A preliminary test pointed to a significant variation in climate across the climatic units (P < 0.001***), underscoring the reliability of this climatic typology