Meta-analytic means and the relationships between intrinsic and
extrinsic moderators
The global meta-analytic mean calculated over 1007 effect sizes and
representing 681 species and extracted from 249 publications, was 0.68
(95% CI = 0.48 to 0.83; Table S4). This is the back-transformed mean
survival rate of all birds included in the analysis. Overall, the joint
extrinsic / intrinsic model explained variation in survival well (Fig.
S4, adjusted r 2 = 0.43). When we estimated
separate meta-analytical means for the six data subsets, we found
similar values with overlapping 95% confidence intervals between the
global mean and mean effect sizes calculated for passerines vs.
nonpasserines, Old World vs. New World biogeographic realms, and
estimates from mainland vs. island birds (Fig. 2; Table S4). In
addition, all models had values of P <0.0001 for
QE and I2 >90%,
which indicated that substantial heterogeneity remained unexplained
among studies and warranted our subsequent step of evaluating moderator
variables.
We found evidence supporting the hypothesis of a latitudinal gradient in
survival, and this effect was most apparent in the northern hemisphere.
When we examined model predictions from a single-predictor model of
latitude over the entire dataset, the odds of survival decreasing were
1.35 times greater for every 1° increase in latitude in the northern
hemisphere compared to the southern hemisphere (Fig. 3a). Similarly, the
global joint model showed a negative, albeit nonsignificant, effect of
latitude on survival for northern hemisphere species (β = -0.121,
95% CI = -0.293 to 0.050), while estimates for those inhabiting the
southern hemisphere were close to zero (β = -0.016, 95% CI =
-0.130 to 0.097; Table 1, Fig. 4). Driving this global trend at northern
latitudes was a significant negative effect size for passerine birds
(β = -0.252, 95% CI = -0.448 to -0.056; Fig. 4b) and a
marginally significant effect of species / populations from the mainland
(β = -0.142, 95% CI = -0.315 to 0.031; Fig. 4d). In contrast,
effect sizes calculated for southern latitudes were generally smaller,
and the overall slope of the meta-regression line of the global model
was shallower compared to the northern hemisphere (Figs 3a and 4). Only
New World species (i.e., birds from South America) showed a significant
negative association with latitude (β = -0.211, 95% CI = -0.378
to -0.045; Fig. 4c). Of the extrinsic climate moderators we considered,
temperature seasonality was the most competitive within our AIC model
selection framework (Table S2 and S3), although only marginally so
compared to minimum winter temperature. Regardless of which climate
moderator was used in the joint model, the effect calculated over the
global dataset and for all data subsets was nonsignificant (Fig. 4).
In general, the relationship between survival and intrinsic life history
traits was stronger than those of either climate or latitude (Fig. 4;
Table 1). Effect size estimated from the global model was positive for
mass (β = 0.236, 95% CI = 0.189 to 0.284; Fig 3c) and negative
for clutch size (β = -0.507, 95% CI = -0.641 to -0.373; Fig 3d),
which means that avian survival was higher for larger birds and for
those with smaller clutch sizes. With the exception of mass for island
species, similar results for both moderators were found for all data
subsets (Fig. 4). When we included nonmigrant as a moderator in the
global model, the effect size was small and positive, with confidence
intervals marginally overlapping zero (β = 0.016, 95% CI =
-0.005 to 0.039; Fig. 4), suggesting higher survival for year-round
residents. This effect was strongest for passerines (β = 0.147,
95% CI = 0.011 to 0.254; Fig. 4b) and Old World birds (β =
0.237, 95% CI = 0.062 to 0.412; Fig. 4c).