492 
THE WILSON JOURNAL OF ORNITHOLOGY • Vol 124. No. 3. September 2012 
TABLE I. Model selection of unconditional multiple logistic regression using Akaike’s Information Criteria (AIQ. 
Three variables of endorheic lakes (length of lake perimeter, aeolian lunette si/e, and elevation) were modeled to assess 
probability of lake occupation by Magellanic Plovers in Santa Cruz Province. Argentina. Sample size In), number of 
parameters (k), AIC values corrected for small sample si/e (AIQ.), difference in AIC, (AAIC,-). and Akaike's weights (iv.i 
are presented for top ranking models (<4 AAlC r ) and the null model. 
Model 
n 
k 
AIC, 
AAIC,. 
•n 
Lunelle size (ha) +• lake perimeter (m) + elevation (in as!) 
53 
4 
26.23 
0.000 
0.29 
Lunette size (ha) 
53 
2 
26.33 
0.107 
0.28 
Lunette size (ha) +-elevation (m asl) 
53 
3 
26.35 
0.121 
0.27 
Lake perimeter (m) + lunette size (ha) 
53 
3 
27.47 
1.237 
016 
Intercept only 
53 
I 
74.62 
21.83 
0.00 
elevation (all P's > 0.05). Occupancy was 
significantly associated with lake salinity as only 
two of 17 freshwater lakes were occupied 
whereas 12 ol 14 saline lakes were occupied 
(Fisher’s exact test. P < 0.0001). 
1 he size of the aeolian lunette was important in 
all top models at the lake scale and had the highest 
importance variable (Swr = 1.00, Table I). This 
variable was the only one to have a parameter 
estimate that did not include zero (estimate = 
1.515. 95% CJ = 0.67-2.86), indicating a higher 
probability of occupancy by Magellanic Plovers 
on lakes with large than small aeolian lunettes. 
Generally, occupied lakes had perimeters of 950 m 
oi greater and contained aeolian lunettes of at 
least 2 ha (Fig. 2). 
Territory Use.— Lake was included as a random 
effect in GLMM models at this scale. The model that 
resulted from backward stepwise parameter removal 
included two parameters with significant (ot < 0.05) 
P- values: Vegetation at 5 m (intercept ± SE: 2.187 ± 
0.61, parameter estimate: - 1.911 ± 0.567, / = 3.37 
p < 0.0022) and vegetation at 15 m (parameter 
estimate: -3.42 ± 1.155, / = 2.96, P = 0.0049). 
Both parameters indicated increasing the percent 
cover ol vegetation within 15 m of the center point 
decreased the probability of territory use. 
The model for distance to vegetation and 
distance to channel was highest-ranking, alone 
describing 66% of variation in territory occupa¬ 
tion (Table 2). The sum of Akaike’s weights tor 
each variable indicated distance to vegetation and 
distance to channel (both I \vi = 1,00) were 
considerably more important than distance to lake 
(Im 1 / = 0.34). The parameter estimate lor distance 
to vegetation was positive and distance to channel 
was negative, indicating the probability of teni- 
tory use increased wdth increasing distance from 
vegetation and decreasing distance to freshwater 
channels (Fig. 3). 
Nest Site Use .—Nests were directly on the 
substrate, close to the shore, and on the side of (he 
lake with aeolian lunettes. No nests had obstruc- 
tions (e.g., logs, vegetation) other than cobble 
near them. The most parsimonious model describ¬ 
ing nest site characteristics included the additive 
ellects of percent clay and percent cobble (>c/ = 
0.37, Table 3). The model including percent clay 
features ( lakeshore. vegetation, ^nd"freshwater rcgress,on dcsc bbing the effect of proximity to three geographic 
Sample size («), number of parameters (kj AkaikeTl" V, r ma,^ tenilory occu P aiK \V by Magellanic Plovers, 
sample size (QA1C,difference in OA1C fAOAIC ) an > h"™" (A,C >' A1C cor "*ted for dispersion and small 
^ll ^andidate mode, ^ ~ ££* — ^ 
