Reiter and Andersen • ARCTIC FOXES, LEMMINGS, AND CANADA GEESE 
269 
TABLE 1. Variables considered in models of nest success of Eastern Prairie Population Canada Geese in northern 
Manitoba, Canada and predicted associations under the "bird-lemming" hypothesis. The predicted association for hatch 
date (HATCH) and nest density (D) with nest success was based on previous studies of factors intluencing nesting Canada 
Geese (positive = good goose condition; negative = more predators). 
VraUe 
Description 
Predicted association 
with nest success 
LP Peak lemming year (I = peak; 0 = other) Positive 
LT Trough lemming year. 1-year following peak (1 = trough. 0 = other) Negative 
FD Fox den occupancy Positive 
Density of nests/100 ha of wetland Positive or Negative 
HATCH Median hatch date (1 Jun = l) Negative 
T Trapping of arctic fox (I = trap, 0 = no trap) Positive 
trough years. 1 year following a peak. Reiter and 
Andersen (2008b) adapted methods described by 
Danell et al. (1999) to generate a relative measure 
of lemming abundance. This index was calculated 
by collecting a random sample of scars on willow 
'Salix spp.) plants caused by lemming feeding, 
quantifying the distribution of scar ages using 
dendrochronological techniques, and modeling the 
distribution of scar ages using non-linear Poisson 
regression. The model-based correction was neces- 
sar y because distribution of scar ages collected in a 
single year was biased as older scars were harder to 
detect because of plant growth and there were 
relatively fewer older scars due to natural plant 
death. The units of this index were the predicted 
number of scars of each scar age that would be 
counted if all scars were equally detected and no 
plants died. This method estimated relative Jem- 
wing abundance front the end of the previous plant 
growing season to the following spring in each year 
°n the Nestor One study area for a period of years 
extending through the age of the oldest scar. 
Previous studies found that small mammal popula¬ 
tion indices for winter to spring using scar analysis 
Were proportional to small mammal captures in the 
subsequent summer months (Erlinge el al. 1999. 
Predavee et al. 2001). We defined a year of peak 
doming abundance (LP) as a year when the 
relative abundance of lemmings was one standard 
Aviation (SD) larger than the long-term mean 
1 1992—2004). We classified the year following a 
as a lemming trough (LT) when there was a 2 
SD decline from the abundance in the previous peak 
• vear LP and LT were included separately in models 
35 factors (Table 1). 
Fox Den Occupancy .—We visited arctic fox 
dens (FD) on and near the study area as part of 
annual Epp Canada Goose research activities. We 
ev aluated dens for evidence of recent (i.e., since 
snow melt) activity based on the presence of foxes 
or fresh fox sign (e.g., digging, prey remains, or 
scat; Macpherson 1969) indicative of an occupied 
den. We calculated the proportion of occupied 
dens each year for 1993 to 2004. and the 95% Cl 
for the proportion of occupied dens in each year 
using the percentile method and the distribution of 
1.000 bootstrapped estimates of fox den activity 
(Efron and Tibshirani 1993). 
Dens were surveyed only once during nest¬ 
searching activities in 1993—2004; thus, it was 
unclear whether fresh activity actually indicated an 
occupied den (i.c.. the den was occupied by foxes 
for the entire goose nesting season and not just 
visited by foxes prior to our visit). Our definition of 
this covariate was important for interpreting our 
data as related to the bird-lemming hypothesis. This 
hypothesis predicts the proportion of dens with 
fresh activity should be negatively associated with 
annual nest success if den activity was an index to 
relative fox abundance. However, if den activity 
was considered a measure of den occupancy, which 
should be high when lemmings were abundant and 
if lemmings buffer bird nests against depredation, 
our fox den covariale would be predicted to be 
positively associated with nest success (Table 1). 
We used data collected during repeated visits to 
arctic fox dens in 2004-2007 to test the ability of 
fresh fox sign during the first visit to successfully 
predict occupancy. These results suggested the 
proportion of dens with fresh activity at first visit 
was positively correlated (Pearson’s correlation 
coefficient, r = 0.82, n = 4 years) with den 
occupancy, and w'e considered our fox den data to 
represent den occupancy. 
Statistical Analyses and Hypotheses .—We cal¬ 
culated summary statistics for all variables in our 
analysis and used autocorrelation functions (Box 
and Jenkins 1976) as preliminary assessments of 
