Study Areas

We have been directing nest-box programs for American Kestrels in eight locations in Canada and the eastern U.S. In general, nest boxes were placed in habitats apparently suitable for kestrels, open areas such as meadows, hayfields, early oldfield successional communities, agricultural fields, and open parkland (Smallwood and Bird 2002). The nest boxes in the Yukon Territory (managed by DM) extended from near the southern border of the territory (approximately 60°N) almost to tree line (approximately 66°N), traversing open patches within the boreal forest. The study area in Saskatchewan (RDD and GRB) was in the vicinity of Besnard Lake, and was approximately centered on 55°20′N, 106°05′W; a detailed description has been published previously (Bortolotti 1994). The study area in Massachusetts (JM and MJM) was approximately centered on 41°45′N, 70°40′W, and habitats included commercial cranberry bogs. The New Jersey study area (JAS; centered on 41°00′36″N, 74°50′41″W) has been described previously (Smallwood and Wargo 1997). The Pennsylvania study area (JRK, BR, and SR) was in the vicinity of Hawk Mountain Sanctuary (40°38′N, 75°59′W; see Klucsarits et al. 1997). The nest boxes in northern Virginia and central Maryland (MFC) were centered on 39°01′N, 77°25′W. The Georgia study area (JWP, TFB, and KB) was in the vicinity of Ft. Gordon military base (32°23′N, 82°14′W; see Breen and Parrish 1997). The southernmost study area was in north-central Florida (RJM and JAS), approximately centered on 30°01′N, 82°52′W, and was described by Miller and Smallwood (1997).

Nest-box Data

The number of nest boxes in each program tended to vary among years, and all increased during the first few years of each program. The median and maximum number of nest boxes available for each program was 15 and 64 in the Yukon, 308 and 388 in Saskatchewan, 64 and 69 in Massachusetts, 109 and 129 in New Jersey, 158 and 213 in Pennsylvania, 75 and 86 in Virginia and Maryland, 100 and 100 in Georgia, and 34 and 60 in Florida. The Florida nest boxes reported here were monitored each year since 1990, and are part of a larger program (>600 nest boxes monitored intermittently since 1995; J. Smallwood unpubl. data). For all eight nest-box programs, the number of years of study range from 11 (Georgia) to 24 (Virginia/Maryland).

Each nest-box program operated under its own protocol, so the number of monitoring visits to each box per season varied somewhat; e.g., each nest box in the Pennsylvania study area was visited 2–5 times each season, depending on occupancy (Katzner et al. 2005) while nest boxes in New Jersey were visited at intervals of 21–28 d (Smallwood et al. 2003). However, we believe that the monitoring effort was sufficient that the calculated occupancy rates (number of nest boxes in which at least one kestrel egg was observed/number of nest boxes available × 100%) are comparable among years and among nest-box programs. Except for the Pennsylvania program, which is a continuation of one that began in the 1960s, all occupancy rates presented here begin with the year of program establishment, i.e., the first year that nest boxes were made available to the study population.

The introduction of nest boxes to a population that is nest-site-limited is expected to result in an initial increase in the occupancy rate for those nest boxes. This initial increase might mask a longer term underlying population trend. Therefore, to test for the presence of an underlying trend, we omitted the first years of data from each nest-box program if those years represented an initial increase. Trends were tested with regression analyses. Six of the eight data sets met the requirement of normality. The two that did not (Yukon and Georgia) were successfully normalized with the Box-Cox transformation method (λ-values of 0.1892 and −1.4789, respectively; NIST/SEMATECH 2006). Although transformed values were used in tests of significance, the untransformed slopes are presented in the results.

Survey Data

We obtained data on trends in kestrel sightings from the USGS Breeding Bird Survey (BBS), available online ( http://www.mbr-pwrc.usgs.gov/bbs/trend/tf07.html) for the years corresponding to our nest-box data, 1984–2007 (Sauer et al. 2008). Locations included the U.S. Fish and Wildlife Service (USFWS) administrative regions and Canada. The analysis available online is linear regression on the selected dataset.

If increasing Cooper's Hawk density were responsible for the decline in kestrels, we would expect trends in Cooper's Hawk populations to be generally inversely related to trends in kestrel populations. To test this hypothesis, we obtained BBS data for both species, by physiographic region, for 1966–2007, the longest period available in the BBS dataset, and 1980–2007, the available period most closely concurrent with our nest-box data (Sauer et al. 2008). Trends for both species co-occurred in 42 physiographic regions for the period 1966–2007, and in 41 regions for the period 1980–2007. Not all of these trends were significant, but their use in a meta-analysis should be unbiased with respect to the relationship between the two species. Because these data were not normal, we used nonparametric statistical treatments. Using region as the sampling unit, we tested the association between kestrel and Cooper's Hawk population trends in two ways. First, we compared the slopes of the population trends with Spearman's correlation coefficients. Second, we compared just the directions (increase or decrease) of the trends with Fisher's exact tests.

We also obtained population trends for both species based on National Audubon Society Christmas Bird Count data, 1959–88, which were categorized by U.S. state and Canadian province (Sauer et al. 1996). Trends for kestrels and Cooper's Hawks co-occurred in 40 states and provinces. Using state or province as the sampling unit, we followed the same procedure as above to compare the slopes and directions of the population trends for the two species.

All statistical procedures, except those received from the USGS Breeding Bird Survey website, were performed with JMP software (SAS Institute 2004).

Population Trends

The initial increase in occupancy rates that followed the introduction of nest boxes was expected. Local kestrel populations commonly are nest-site-limited (Cade 1982, Smallwood and Bird 2002); indeed, an apparent lack of nest cavities often provides the rationale for establishing a nest-box program. Compared with most larger falconiform species, kestrels have relatively high reproductive potentials (they mature quickly and lay large clutches; Brown and Amadon 1968) and once released from nest-site limitation, populations can respond quickly (e.g., Toland and Elder 1987, Smallwood and Collopy 2009). Most of the nest-box programs experienced an increase in occupancy rates for 4–6 yr after establishment. The Georgia program reached its peak occupancy rate in its second year, suggesting that there may have been a relatively large floater population in that study area.

All eight nest-box programs exhibited significant declines following the initial response to the local increase in nest-site availability. The similarity in the slopes of those declines is remarkable, considering the wide geographic range the various nest-box programs represent. The results from these nest-box programs strongly support the conclusion that a significant, widespread decline in kestrel populations has been occurring in recent years.

Results from the Breeding Bird Surveys, 1984–2007, also support the conclusion that there has been a widespread, although not uniform, decline in North American kestrel populations. Population trends were negative for all Canadian provinces for which data were available and for four of the seven USFWS regions. However, all significant population changes were negative, resulting in a highly significant negative trend survey-wide. Most of the nest-box programs were in Region 5 (northeastern U.S.) and Canada, where overall declines were highly significant (Fig. 2). The mean annual decline estimate from Canadian BBS data, 3.20%, was even greater than the annual decline experienced by nest-box programs in the Yukon (2.7%) and in Saskatchewan (1.9%). The increase detected in USFWS Region 4, where the Georgia and Florida nest-box programs are located, was not significant. BBS trends for kestrels from 1984 to 2007 in just Georgia and Florida also were not significant, but were based on small sample sizes, only 5 and 15 survey routes, respectively (Sauer et al. 2008).

Possible Causes

West Nile virus (WNV) is transmitted between bird reservoir hosts by mosquito vectors (CDC 2009). Since its first appearance in North America in 1999, WNV has been detected in >250 species of wild birds, including American Kestrels (CDC 2009). Thus, there was a possibility that the decline in kestrel populations could be related to the spread of WNV. However, data from our nest-box programs demonstrated that declines began before WNV arrived on the continent. Occupancy rates for the nest boxes in Pennsylvania underwent a net decline for the entire period covered, 1986–2007, although there was an increase during the most recent two breeding seasons. Of the seven nest-box programs for which we have occupancy data since establishment, six began their declines prior to the arrival of WNV. The only exception was New Jersey, which is the most recently established program. In general, the early-established programs have experienced the longest declines.

Kestrels appear to have had substantial exposure to WNV. In southern Quebec, WNV antibodies were detected in blood samples from 17 of 28 (61%) adult kestrels captured during the breeding season, 2003–05 (D. Bird unpubl. data). In eastern Pennsylvania, the exposure rate was even higher; during 2004, 21 of 22 (95%) wild-caught adults tested positive for WNV antibodies (Medica et al. 2007). Although most birds infected with WNV survive and acquire life-long immunity (CDC 2009), corvids are particularly prone to become ill or die (Eidson et al. 2001). Some raptors also may have heightened vulnerability. Nemeth et al. (2006) suggested that wild kestrels would be at greater risk of mortality, even though the kestrels they experimentally infected with WNV survived. Nevertheless, the WNV-exposed kestrels in Quebec bred normally (D. Bird unpubl. data), and Medica et al. (2007) reported both normal reproduction and body weights for exposed kestrels in Pennsylvania.

If WNV or another pathogen were the principal agent of the decline, we would expect rapid selection, as the most vulnerable genotypes would be excluded from the population, leaving the most resistant genotypes. The subsequent recovery would likely be rapid; kestrel populations demonstrate this capability when released from nest-site limitation (Smallwood and Collopy 2009). Thus, the expected population trend due to a serious pathogen would be a brief, marked decline (e.g., Crosbie et al. 2008) followed by a rapid recovery. In contrast, we have been observing prolonged, steady declines.

Another possible cause of the decline in kestrels is an increase in predation by Cooper's Hawks. Cooper's Hawks are known to prey upon kestrels (Farmer et al. 2008b), and there is evidence that Cooper's Hawks may learn to “trapline” (i.e., forage along an established route; Stiles 1995) nest boxes for recently fledged kestrels (B. Millsap unpubl. data). The BBS data suggest that Cooper's Hawk populations increased significantly in the United States from 1984 to 2007 (5.30% annually, N  =  568 routes, P < 0.001; Sauer et al. 2008). However, no significant increases were found concurrently in Canada, perhaps due to small sample sizes (N  =  39 routes, P  =  0.42). Predation by Cooper's Hawks cannot explain the declines we observed in the two kestrel nest-box programs in Canada; Cooper's Hawks do not occur in the Yukon Territory (Rosenfield and Bielefeldt 1993), and they are very uncommon in the Saskatchewan study area (Gerrard et al. 1996).

We expected that if an increase in Cooper's Hawk predation were an important factor in the decline of kestrels, then the population trends of the two species generally would have an inverse relationship. Our analysis of both BBS data and Christmas Bird Count data gave no evidence of such a relationship. It is possible, however, that kestrel populations could be negatively affected by dietary shifts in Cooper's Hawks, independent of the density of Cooper's Hawks.

Habitat loss or degradation has been considered the most important factor in the decline of avian populations (e.g., Fitzpatrick 2004), and may be involved in the observed decline in kestrels. However, in our study areas, the relationship between declining occupancy of our nest boxes and possible changes to the surrounding habitat is not clear. In New Jersey, for example, there was no obvious change in the land use immediately surrounding the nest boxes; 82% of the nest boxes that were vacant in 2007 had been occupied in one or more previous years (J. Smallwood unpubl. data). There has been no evidence of the effects of reduced habitat quality, such as low prey availability, or problems with disease or toxic contamination; all kestrels appeared healthy and the mean nesting success (attempts resulting in at least one chick surviving to banding age, generally 18–22 d) was 84.0% during the decline (J. Smallwood unpubl. data), which compares favorably with other kestrel populations (Smallwood and Bird 2002). Although the habitat in the study area appears suitable and the kestrels appear healthy and have high reproductive success, there simply are fewer kestrels using the nest boxes. Although these analyses are subjective, we nonetheless believe they indicate that the main cause of decline likely is operating somewhere beyond the immediate vicinity of these nest boxes.

Declines in breeding populations may be associated with an increase in mortality during the nonbreeding season, perhaps related to habitat change on the wintering grounds or along migration routes. However, not all the kestrels we studied in our nest-box programs are migratory. In North America, the tendency to migrate decreases from north to south, and the northernmost breeding populations tend to winter in the southernmost wintering grounds (Smallwood and Bird 2002). Populations in the middle latitudes, approximately 44–36° N, are partially migratory and appear to respond to local conditions, migrating short distances during relatively harsh winters (Bird and Palmer 1988). The populations in our study areas in Georgia and Florida are resident Southeastern American Kestrels (F. s. paulus; Howell 1932). Thus, our study populations span the range of entirely migratory to entirely nonmigratory, and all are declining.

Migratory kestrels are vulnerable to many mortality factors, including those related to habitat quality, along the routes and on the wintering grounds. Resident populations, however, also may wander to some extent outside the breeding season (Bird and Palmer 1988), and thus may be affected by habitat issues beyond the immediate vicinity of the nest boxes. Further, the decline in our nest-box breeding populations could reflect the respective regional declines. Although the numbers of marked adults in the Florida and New Jersey programs are small, there appears to be substantial turnover of adults from year to year (J. Smallwood unpubl. data). Thus, regional declines might reduce the number of individuals from outside the study areas that replace nest-box-breeding individuals that have either dispersed or died between breeding seasons.

Conclusions

The widespread decline in kestrel populations detected in the USGS Breeding Bird Surveys during recent decades was corroborated by patterns of nest-box occupancy by both migratory and resident populations of kestrels. Because the decline in kestrels breeding in nest boxes began before the arrival of WNV in North America, the virus clearly is not the primary cause of the decline. Further, we found no evidence that increasing trends in Cooper's Hawk populations were associated with the decreasing trends in kestrels. Although habitat loss and degradation were not evident in the vicinity of the nest boxes, these factors may nevertheless be important, particularly in reducing the number of kestrels available for occupying nest boxes.