Zhou el al. • NEST-SITE SELECTION AND NESTING SUCCESS OF THRUSHES 497 
FIG. 2. Trees and shrubs available at nest sites and random plots, and percentage of vegetation used for nesting by 
Grey-backed Thrushes in Dagang Forestry Farm, China, in 2008. 
Significance tests indicated the average density of 
shrubs at nest sites (1.21 ± 0.46 indi viduals/m 2 ) was 
significantly higher than at random plots (0.95 ± 
0.42 individualsVm 2 ), and basal area of small trees at 
nest sites (2.24 ± 0.78 m 2 /ha) was also greater than at 
random plots (1.64 ± 0.88 nr/ha) (Table 3). The 
average height of large and small trees was 8.97 ± 
1.53 m and 4.33 ± 0.56 m. respectively. The 
correlation matrix indicated density of shrubs was 
negatively correlated with ground cover {r s = 
-0.332, P = 0.001) and height of ground cover (r s 
= -0.237, P = 0.003) (Table 2). 
A set of four candidate models whose AAIC,- 
were <2 was selected from the potential models 
(Table 4), and we compared them with the global 
model (Model 1. Table 4: all variables included) 
and null model (Model 2, Table 4; no variables). 
We used P < 0.05 to indicate a good model when 
testing for the fitness of a model with log- 
likelihood ratio tests. All models except null 
model were potentially good (P < 0.001). 
including the global model. The model building 
procedures indicated Model 3 with the highest 
of 0.325 was the best model. The global model 
only had a IV, of 0.004 (Table 4). 
Height of ground cover, density of shrubs, and 
basal area of small trees were included in the best 
model with the relative importance of 0.599,1.000, 
and 1.000, respectively. Density of shrubs (P = 
0.027) and basal area of small trees (P = 0.012) 
were also significant in the models based on Wald 
statistics, but height of ground cover was not (P = 
0.134). The relative importance of other variables 
excluded from the best model was 0.004, except 
that distance to edge accounted for the importance 
of 0.444. Grey-backed Thrushes built nests in areas 
with a lower height of ground cover, and a higher 
density of small trees and shrubs. 
Successful and Depredated Nests .—The dis¬ 
tance from nest to main stem and horizontal 
exposure at the nest-location scale, and density of 
shrubs at the nest-patch scale met the criteria of 
variable selection (P ^ 0.25), respectively, and 
correlation analyses did not indicate a strong 
correlation among them. Distance from the nest to 
the main stem of the nest tree was the only 
significant variable associated with nesting suc¬ 
cess at the 0.05 level. 
Model 3 met the needs of model fit and 
parameter parsimony, and was the best with the 
highest Wi of 0.202 (Table 5). Distance from nest 
to main stem and horizontal exposure were 
included in the best model with relative impor¬ 
tance of 0.647 and 0.509, respectively. Density of 
shrubs was not included in the best model, 
although its relative importance reached 0.644. 
Nests with closer proximity to the main stem of 
the nest tree and higher horizontal concealment 
were more likely to be successful. 
There were some indications that nest exposure 
was associated with predation for Grey-backed 
Thrushes. Forty-two nests had high levels of 
exposure (s 4) with at least one of the three 
exposure indices (upper, ground-level, or hori- 
