SHORT COMMUNICATIONS 
155 
FIG. 1. Relationship between Julian dates of the 50th percentile of the seasonal total of migrating Sharp-shinned Hawks 
and year of migration at Hawk Ridge Bird Observatory, Duluth, Minnesota, 1974-2009. 
shinned Hawk as a focal species because its 
migration is likely linked to spatial and temporal 
movement of its neotropical songbird prey 
(Rosenfield and Evans 1980, Viverette et al. 
1996, Goodrich and Smith 2008). 
METHODS 
Study Site and Data Collection .—The Hawk 
Ridge Bird Observatory (HRBO) is in boreal 
forest at the western end of Lake Superior in 
Duluth, Minnesota (49° N, 92° W), and is a well 
known concentration point for migrant raptors 
(Goodrich and Smith 2008). Raptors are counted 
hourly at HRBO from about 15 August through 30 
November using standardized techniques estab¬ 
lished by the Hawk Migration Association of 
North America (HMANA) (Ruelas Inzunza 2005, 
Farmer et al. 2008). Detailed descriptions of daily 
and seasonal coverage of counts at HRBO are 
provided by Fanner et al. (2008). Migration of 
Sharp-shinned Hawks primarily occurs from mid- 
August through October at HRBO with peak 
monthly totals of migrating birds typically 
occurring in September (Rosenfield and Evans 
1980, Goodrich and Smith 2008). We obtained 
HRBO count data for Sharp-shinned Hawks 
during 1974-2009 from HMANA’s Hawkcount. 
org/month web site (www.hawkcount.org/month_ 
summary, php). 
Data Analyses.—We used simple linear regres¬ 
sion to assess how Julian date for the 50th 
percentile of the total number of autumnal 
migrating hawks in each of 36 study years 
changed across time (i.e., to ascertain if a shift 
in migration had occurred). We also used Julian 
dates for four percentiles (25, 50, 75, and 99.5) of 
total migration counts in each year to describe the 
extent of the shift (in days) of the migration. We 
truncated the 100th percentile at 99.5 to minimize 
the effect of late-migrating “stragglers” that 
might skew results. We calculated the difference 
in number of Julian dates (days) for each 
percentile in each year for 1975-2009 relative to 
Julian dates for the respective percentiles in 1974, 
the first year of the study. Julian dates earlier and 
later than those in respective percentiles for 1974 
were assigned negative and positive values, 
respectively. We used the average (± SE) of 
those differences to enumerate the approximate 
shift in days in migration for each respective 
percentile since 1974. We chose 1974 as the 
comparative year to demonstrate the extent of the 
shift in timing of migration because it was 
representative of the earlier timing of migration 
at the outset of the study (Fig. 1). We also report 
the shift for all combined percentiles, 1975-2009. 
Further, we separate the percentiles in the first 
19 years of study (1975-1993) from those during 
