3.2 The influence of climate change on the size of S. araneus
The weather records from this period revealed several significant trends indicative of a changing climate (scatterplots of six climate parameters are shown in Figure S2). The mean annual temperature increased significantly (F 1,51=28.1,P <0.0001; Figures 4c and S2a) from 6.1 °C to 8.2 °C (as calculated from the regression line). There was also an increase in the mean July temperature (F 1,51=10.5,P =0.0021; Figure S2b), from 17.2 °C to 19.7 °C, but not in the mean January temperature (F 1,51=2.29,P =0.136; Figure S2c). The relationship of the sum of annual precipitation with time was nonlinear (quadratic term:F 1,50=9.30, P=0.0037), with an initial increase leading to a maximum in the mid-1970s, followed by a decrease until 2004 (the decrease in 1971-2004 was statistically significant:F 1,32=11.25, P =0.0021, linear regression) (Figures 4d and S2d). The period of high precipitation was mainly due to increased rainfall in June and July. Continuous increases in mean annual and mean July temperatures coupled with a significant decline in precipitation resulted in an increased nonlinear soil moisture deficit over the years (quadratic term: F 1,50=7.69,P =0.0078; Figures 4e and S2e). This change in the deficit was not significant until 1971 (F 1,18=0.20,P =0.660, linear regression), but there was a significant increase in the soil moisture deficit after 1971 (F 1,32=11.9, P =0.0016). Interestingly, the beginning of the increase in drought corresponded with the inflection points of the smoothed relationships of the skull size vs. years and with the clear decrease in skull size later on (Figure 4). The number of days with snow cover significantly decreased over the study period (F 1,51=4.57, P =0.0374) (Figure S2f).
We found no significant relationships between skull length and the climate parameters in either juvenile or adult shrews in the whole time series, i.e., between 1953 and 2004. However, when we reduced this period to between 1971 and 2004 (the time of the significant decrease in skull length in adults), the relationship between the skull length and the moisture deficit in the soil became highly significant (F 1,20=6.91, P =0.0161,r2 =0.257); an increased deficit resulted in a shorter skull length (Figure 5).
The analysis of the relationships between skull height and climate between 1953 and 2004 revealed significant temperature effects (Table 2). Skull height decreased significantly with increasing mean daily temperatures during the expected lifetime in both juvenile and adult shrews. We also found an impact of winter weather (during the subadult stage) on adult skull height the following summer: the increase in the lowest average daily temperature in February (but not in January) was associated with the decrease in skull height in summer, but this relationship was marginally significant. Moreover, summer skull heights were greater when there were more days with snow cover in the previous January and February. However, the relationship between skull height and the number of days with snow cover >=5 cm in January and February did not reach the significance level (Table 2). Following this pattern in other analyses, there were no significant relationships between skull width and climate parameters. Including more than one weather variable in a model never increased the overall significance.