Results

Leaf-out and first flowering day were significantly positively correlated in all years, and correlation coefficients ranged from 0.32 to 0.46 (Pearson’s correlation; 2013: r = 0.46, t196 = 7.29, p < 0.001; 2014: r = 0.41, t201 = 6.44, p < 0.001; 2015: r = 0.32, t205 = 4.77, p < 0.001; Fig. 1).
There was total selection for earlier leaf-out in all years (Table 1a). Selection on leaf-out day in 2015 was nonlinear, fitness being highest in individuals with an early to intermediate leaf-out day (Table 1a; Fig. 2a). There was total selection for earlier flowering in 2013 and 2014, but not in 2015 (Table 1b; Fig. 2b). We found no evidence of nonlinear selection on first flowering day in any of the study years (Table 1b).
There was direct selection for earlier leaf-out in 2013, but in 2015 selection favoured later leaf-out instead (Table 1c; Fig. 2a). There was no direct selection on leaf-out in 2014. Direct phenotypic selection favoured earlier flowering in all three years (Table 1c; Fig. 2b). We found no support for nonlinear selection on leaf-out or first flowering day in any of the study years.
The estimate of direct selection on leaf-out in 2014 was just outside the BCa interval for the corresponding estimate of total selection, suggesting that selection for earlier leaf-out 2014 acted largely via first flowering day (Table 1a,c). The estimates of total and direct selection on leaf-out day in 2013 and 2015, and on first flowering day in all years, did not differ significantly (Table 1b; Fig. 2).
Individuals with shorter development time between leaf-out and first flowering day had higher fitness than individuals with longer development time in 2014 and 2015, but not in 2013 (Appendix S7, Table S1; Fig. 2c). We found no evidence of correlational selection on leaf-out and first flowering day in any of the three study years (Table 1c).