1254394718

131

wing and tarsus; Rising & Somers, 1989) to obtain an index (PCI) of structural body size (hereafter named “size”). Then, we added size (F 1,595= 333.7; p < 0.0001) in a GLM testing for the effects of datę (F|^₉₅= 14.5; p < 0.001) and the variable ^łłrelative time of capture” (time sińce sunrise / day length, hereafter “time of capture”) (F1.595 = 79.5; p < 0.0001) on M_b and used the residuals as our values of M,.

BMR was dependent on M_b (Fi, 222 = 46.0; p < 0.0001) but not on datę (p = 0.3) while Msum was influenced by both M_b (Fi, m = 31.7; p < 0.0001) and datę (F|_t n& = 13.9; p< 0.001). We therefore conducted two sets of survival analyses, one including raw BMR and Msum and the other including residual BMR and residual Msum, controlling for the effect of M_b and datę.

We used raw Het values in survival analyses sińce Het was not related to datę (p = 0.7), time of capture (p = 0.5), M_b (p = 0.3) or size (p = 0.6).

When a bird was measured several times within a winter (for within winter analyses) or in morę than one winter (for the among year analyses), we calculated the mean of each physiological parameter for a given bird and used these means in the survival analyses.

Our analyses were therefore conducted using both raw and residuals of BMR and Msum. However, when one of the residual parameters was found to significantly affect apparent survival, we reported corresponding uncorrected values in the text and figures to allow for comparison with other studies. These analysis were performed with R version 3.0.3 (R Core Team, 2014).

5.4.4.2 Selection of the basie model

M„ Het, BMR and Msum were available for 81 individuals for the within winter survival analysis and for 56 individuals for the among year analysis, we therefore restricted our databases to these birds for further analyses.

Survival analyses were performed using the software MARK version 7.2 (White & Bumham, 1999). To analyse within winter survival (n = 81), we used data recorded during the first winter for cohort 1 and data from the second winter for cohort 2. For both winters, we grouped encounter occasions in 3 periods (period 1 = Sept to Nov, period 2 = Dec to Jan;