Beckett and Proudfoot • NORTHERN SAW-WHET OWL MIGRATION 
527 
TABLE 1. Migration route deviation in Northern Saw-whct Owls recaptured within 0.5' latitude from banding location 
>| year after banding. ’Distance' represents straight-line distance between banding and recapture locations. A large 
proportion of same-site recaptures occurred at two sites west of Lake Michigan. We present a truncated data column that 
does not include information from these two sites to examine their effect on the overall trend. Further information is 
presented for owls banded within specific regions. The two sites west of Lake Michigan are excluded from the ’Great 
Lakes' group. 
Dioancc (km) 
< 
All owls 
n = 512) 
Truncated data 
<» = 312) 
Great Lakes 
(n = 161) 
Appalachian Mountains 
In = 41) 
Atlantic 
(ft 
: seaboard 
= 50) 
n 
H, 
« 
% 
ft 
% 
» 
% 
» 
% 
<20 
174 
33.9 
161 
51.6 
66 
40.1 
12 
29.3 
34 
68.0 
£50 
334 
65.5 
202 
64.7 
95 
59.0 
15 
36.5 
40 
80.0 
<100 
368 
72.2 
230 
73.7 
107 
66.5 
25 
61.0 
44 
88.0 
<300 
487 
95.0 
291 
93.2 
47 
91.3 
38 
92.7 
48 
96.0 
>301 
25 
4.9 
21 
6.7 
14 
8.7 
3 
7.3 
2 
4.0 
There was no significant relationship between 
adult: juvenile ratio and latitude for all years 
combined (n = 81,184. r = 0.008. P = 0.77. 
Fig. 6). The relationship was significant in irrup¬ 
tion year 2003 and non-irruption years 2006 and 
2008, when tested separately by year. The 
relationship was insignificant and in inconsistent 
directions in 7 of 10 years (Table 3 ). There was no 
significant difference in adult versus juvenile mean 
banding latitude for all years combined (P — 0.46, 
Table 2). There was a significant difference in 9 of 
10 years when tested separately, but the differences 
were not in a consistent direction. 
Surface interpolation of age ratios showed areas 
of predicted high and low values (Fig. 7), but 
these areas were patchy and localized. Highest 
adult: juvenile ratios were predicted in Wisconsin, 
Virginia, northern New England, and New York. 
Lowest adult: juvenile ratios were predicted in 
eastern Ontario north of Lake Huron, eastern 
Quebec, around Lake Erie, and at coastal stations 
in Massachusetts, Rhode Island. New' Jersey, and 
around Chesapeake Bay (Fig. 7A). Results were 
similar in irruption years (Fig. 7B). but with lower 
overall ratios across the surface. Results for non¬ 
irruption years were similar to the interpolation 
•or all owls (Fig. 7C), but with highest adult: 
juvenile ratios also occurring across Virginia, 
West Virginia, Kentucky, and Indiana. 
DISCUSSION 
Migration Timing .—Peak migration activity 
occurred progressively southward over the course 
of the season, suggesting that Northern Saw-whet 
Owls migrate in distinct fronts. This trend is 
consistent using multiple analyses, indicating the 
strength of this trend and the reproducibility of the 
results (Figs. 2-3). This supports the southbound 
trend that researchers have supposed for decades 
based on the accretion of regional observations 
(Mueller and Berger 1967, Weir et al. 1980, 
Erdman et al. 1997, Brittain et al. 2009). We 
could expect a less-striking latitudinal gradient 
and more irregular distributions of the owls 
banded at each latitude if Northern Saw-whet 
Owls were moving southward haphazardly over 
the entire migration season. The observed trend 
implies that fall migration is uniform and not a 
random seasonal dispersal in search of better 
resources. 
Our results closely match those of Holroyd and 
Woods (1975), the only other study that compares 
mean Northern Saw-whet Owl capture dates 
among multiple banding stations in eastern North 
America. The distribution of banding dates for 
most of their study regions were within our 
predicted means (Fig. 3). Our predicted means for 
Massachusetts, Maryland, and New Jersey were 
1-2 weeks later than indicated in Holroyd and 
Woods (1975). Mean banding day in Ontario was 
predicted ~1 week earlier than they indicated. 
This inconsistency may be due to varying weather 
patterns, shifts in population centers that change 
migration distances or the small sample size (n = 
4,802) available for their study. The overall 
similarity in results despite the >30 year differ¬ 
ence in sampled populations (1955-1969 in theirs 
vs. 1999-2008 in ours) shows the long-term 
temporal consistency of this species' fall migra¬ 
tion. This similarity also shows these results may 
be reproducible using other methodologies. The 
abrupt late mean banding day predicted in central 
Illinois may be explained by low banding station 
density in that area. This area may reveal 
