1254394722

135

Table 5.2 Model selection testing for the effect of mean size-independent body mass (M₅), haematocrit (Het), residual basal metabolic ratę (resBMR) and residual summit metabolic ratę (resMsum) on the return ratę of both cohorts during winter. (W_Aicc: weight of the model, Par : number of parameters, Dev.: Deviance)

	Model	AlCc	AAICc	wAICt	Ukelihood	Par	Dev. 1
i	CD(Hct+resMsum) p(.)	76.0	0.0	0.3	1.00	4	67.4
2	<D(Hct+resBMR+resMsum) p(.)	76.6	0.6	0.2	0.73	5	65.7
3	OtM.+M^+Hct+Hct^resBMR+resBMR^resMsum+resMsum^2) P(.)	78.2	2.2	0.1	0.33	10	54.7
4	<D(M,+Hct+resMsum) p(.)	78.2	2.3	0.1	0.32	5	67.4
5	0(M,+Hct+resBMR+resMsum) p(.)	78.7	2.7	0.1	0.26	6	65.4
6	O(resMsum) p(.)	79.1	3.1	0.1	0.21	3	72.7
7	<D(resBMR+resMsum) p(.)	80.3	4.4	0.0	0.11	4	71.8
8	<D(M,+M, ’) p(.)	80.7	4.7	0.0	0.09	4	72.1
9	<D(M,+resMsum) p(.)	80.7	4.8	0.0	0.09	4	72.2
10	<D(M,+resBMR+resMsum) p(.)	81.0	5.1	0.0	0.08	5	70.2
11	<D{resMsum+resMsum^2) p(.)	81.1	5.2	0.0	0.08	4	72.6
12	(D(Hct+Hct^2) p(.)	87.5	11.6	0.0	0.00	4	78.9
13	0(.) p(.)	87.5	11.6	0.0	0.00	2	83.4
14	O(Hct) p(.)	88.0	12.0	0.0	0.00	3	81.6
15	(D(resBMR) p(.)	89.7	13.7	0.0	0.00	3	83.3
16	<D(M,) p(.)	89.7	13.8	0.0	0.00	3	83.4
17	0(M,+Hct) p(.)	90.0	14.0	0.0	0.00	4	81.4
18	<D{Hct+resBMR) p(.)	90.2	14.2	0.0	0.00	4	81.6
19	0(M,+resBMR) p(.)	91.9	15.9	0.0	0.00	4	83.3
20	0(resBMR+resBMR^2) p(.)	91.9	15.9	0.0	0.00	4	83.3
21	CP(M,+Hct+resBMR) p(.)	92.2	16.2	0.0	0.00	5	81.3

mean t = 1.0

The relationship between Msum and residual Msum was linear and strong (r = 0.89, n= 180 p < 0.0001), allowing us to use the regression eąuation (Msum= 1.535+residual Msum) to back-calculate uncorrected Msum. The relationship between winter survival and uncorrected Msum followed a non-linear curve (figurę 5.2a). Below an Msum of 1.17 W, winter survival probability was less than 10%, between 1.20 W and 1.32 W, survival probability inereased linearly reaching 50% at 1.26 W. When birds expressed an average Msum above 1.35 W, winter survival probability was greater than 90% and individuals with an Msum superior to 1.46 W were expected to have 100% chances of survival.

Analyses with raw BMR and raw Msum showed that the within winter return ratę was best explained by a model where the survival probability was dependent on BMR and Msum