The American Kestrel (Falco sparverius) is one of the most widespread raptors in the Americas, breeding in eastern and western North America, north to the tree line and south into most of Central and South America (Smallwood and Bird 2002). Although it is a partial migrant (sensu Bildstein 2006), large proportions of Canadian and U.S. populations migrate south in autumn, with northern breeders more migratory than those breeding farther south. A small proportion migrates as far south as northern South America. Larger numbers are recorded at coastal hawkwatch sites in eastern North America than at inland eastern sites, but this pattern does not hold in western North America, where migrating kestrels typically are at least as common at Intermountain (i.e., the area between the Rocky Mountains to the east and the Sierra and Cascade mountains to the west) and Rocky Mountain hawkwatch sites as they are in coastal California (Goodrich and Smith 2008).

Previous analyses of migration counts reported generally stable to increasing trends for kestrels in northeastern North America from the 1970s to 1990s (Bednarz et al. 1990, Titus and Fuller 1990, Hussell and Brown 1992, Mueller et al. 2001). Bednarz et al. (1990), however, noted a decline in counts of kestrels at Hawk Mountain Sanctuary, Pennsylvania, from 1973 to 1986, and Mueller et al. (2001) reported a significant decline from 1989 to 1999 at Cedar Grove, Wisconsin. In western North America, mixed trends were reported in the Rocky Mountains, with generally stable populations elsewhere until 2001 (Hoffman and Smith 2003). No continent-wide assessment of kestrel population trends has been attempted until now, nor has any previous study of migration counts attempted to estimate population rates of change across a large area of the continent.

The research reported here was conducted as part of a long-term population monitoring program, the Raptor Population Index. Analyses were initially conducted for population surveillance, with further investigation of the American Kestrel populations prompted by the large number of negative population trends observed. We calculated annual population indexes based on autumn kestrel counts at 20 hawkwatch sites throughout North America (see also Farmer et al. 2008a and Smith et al. 2008a, 2008b). We then estimated trends in these indexes for periods-of-record (variable lengths, 6–39 yr) to test whether kestrel populations declined in North America over the last several decades.

Methods

We analyzed autumn counts of visible migrating kestrels at 20 hawk migration hawkwatch sites in three regions: eastern North America from Minnesota eastward (hereafter “East”), western North America from New Mexico westward (“West”), and along the shores of the Gulf of Mexico (“Gulf of Mexico”; Table 1). Standard methods described in Farmer et al. (2007, 2008a) and Smith et al. (2008a, 2008b) were used to conduct the migration counts.

Table 1

Locations of migration hawkwatch sites in North America.

We used the counts to calculate effort-weighted, annual indexes of abundance for American Kestrels at each hawkwatch site using methods described by Farmer et al. (2007) and Farmer and Hussell (2008). This analytical approach was similar to that used previously in analyses for both diurnal and nocturnal migrants (Hussell 1981, 1985, Hussell and Brown 1992, Francis and Hussell 1998). Analyses for the Grand Canyon, Arizona (1997–2005 only) and Veracruz, Mexico involved combining counts from two sites following methods outlined in Hussell (1981) and Francis and Hussell (1998). As noted by Farmer et al. (2007), these methods are more appropriate than those previously used to analyze raptor migration counts because they account for variations in effort and seasonal passage, as well as the skewed frequency distributions of such counts.

To estimate population trajectories (patterns of change over time), we fitted a polynomial regression model to the indexes (Francis and Hussell 1998, Farmer et al. 2007). We then estimated trends (geometric mean rate of change; sensu Link and Sauer 1997) in annual indexes for each hawkwatch site for full periods-of-record, and calculated average annual indexes for successive 5-yr periods for use in within-site comparisons as recommended by Farmer and Hussell (2008). This method allowed for the fitting of complex curves, which more accurately reflect changes in the population index values than a more traditional linear regression (Farmer et al. 2007).

Farmer and Hussell (2008) suggested three criteria to determine whether population trends are of conservation significance: (1) the rate of change in the most recent 10 yr and its statistical significance, (2) the comparison of recent (last decade) population indexes to the long-term average, and (3) the comparison of current (5-yr average) population indexes to those for successive 5-yr periods in the historical record. As noted by Farmer and Hussell (2008), “significant recent declines to population levels below the long-term average and especially 5-yr averages lower than ever recorded previously would be cause for concern and action.” Accordingly, in addition to calculating trends, we constructed paired t-tests (with Bonferroni adjustment, where appropriate) to make the comparisons (2) and (3) above.

We estimated and tested the significance (α  =  0.05) of a trend between two preselected years on the polynomial curves using a two-tailed t-test as described by Francis and Hussell (1998) and Farmer et al. (2007). The null hypothesis tested in each case was that the actual trend was zero. Our trend analysis transformed year terms in the regression equation so that the first-order term estimated the geometric rate of change between the two sets of years (Francis and Hussell 1998, Farmer et al. 2007). We constructed 95% confidence intervals (CIs) around the estimated trend for the longest available time series for each hawkwatch site as described in Farmer et al. (2008a).

For comparison to independent population indexes, we obtained Breeding Bird survey (BBS; Sauer et al. 2007 and J. Sauer pers. comm.) and Christmas Bird Count (CBC; National Audubon Society 2002) data for areas likely to be either the origin or destination of migrants passing hawkwatch sites within each region (Table 2). We used BBS trends calculated by the estimating-equations algorithm (Sauer et al. 2007) and calculated trends in CBC indexes using log-linear regression.

Table 2

Regions used in the analysis of Breeding Bird Survey (BBS) and Christmas Bird Count (CBC) annual indexes.

Results

East

Migration counts decreased significantly from 1974 to 2004 at rates of 1.6–4.5% per yr from the Atlantic Coast to the Appalachian Mountains (Table 3). Migration counts decreased nonsignificantly over this period between the Appalachian Mountains (Hawk Mountain, Waggoner's Gap) and the eastern Great Lakes (Holiday Beach), and increased significantly in the western Great Lakes (Hawk Ridge). From 1994 to 2004, statistically significant decreases of 3.3–8.5% per yr occurred at hawkwatch sites from the Atlantic Coast to the eastern Great Lakes, with a nonsignificant decrease in the western Great Lakes (Table 3, Fig. 1).

Figure 1

Recent population trends for American Kestrels at eight eastern and midwestern (1994–2004), and 12 western and Gulf Coast (1995–2005) autumn migration hawkwatch sites in North America. Trend magnitudes (arrow height) are expressed in percent change per yr. Flight magnitudes (arrow width) are depicted in three ranges (<100, 100–1000, and >1000) of birds per yr. A bidirectional arrow indicates that the estimated trend is 0.0% per yr.

Table 3

Estimated population trends (% change per yr ± 95% confidence interval) and test statistics (2-tailed t-test) for periods-of-record and the most recent decade at 20 hawk migration watch sites in North America.

At Atlantic Coast and coastal plain hawkwatch sites, mean count indexes from 2000 to 2004 were significantly lower than means from all previous 5-yr periods and the long-term (1974–2004) means. Total declines over 30 yr at these hawkwatch sites ranged from 82.7% to 110.3%. The differences weakened in the Appalachian Mountains and eastern Great Lakes, and total 30-yr changes at hawkwatch sites in this region ranged from −35.3% to +23.3%. At Hawk Ridge, Minnesota, the mean index from 2000 to 2004 was significantly greater than those from all preceding 5-yr periods except 1995–99, with a total increase of 108.7% over 30 yr (Table 4, Appendix).

Table 4

Percentage differences between mean annual count indexes for the most recent 5 yr (2000–04) and mean indexes for previous 5-yr periods and long-term (period-of- record) mean index at 10 autumn migration hawkwatch sites having ≥15 yr of American Kestrel counts.

BBS abundance indexes for kestrels decreased significantly (1.4% per yr) in northeastern North America from 1976 to 2003 (J. Sauer pers. comm.), as did CBCs in northeastern (4.6% per yr) and southeastern (1.4% per yr) North America.

Discussion

Migration counts, BBS data, and CBCs indicate that populations of the American Kestrel have (1) declined in much of northeastern North America since 1974, but increased or remained stable around the western Great Lakes; (2) declined in portions of western North America since the mid-1980s, with increasingly widespread declines coinciding with the onset of regional drought in the late 1990s (Hoffman and Smith 2003, Smith et al. 2008a); and (3) declined in the eastern Gulf of Mexico, but increased or remained stable elsewhere in the Gulf of Mexico over the last decade (Fig. 1). Declines in migration counts can be indicative of changes in migration geography (e.g., migratory short-stopping). Migratory short-stopping, however, should produce a pattern of declining migration counts, stable to increasing BBSs, and increasing CBCs in the north but stable or decreasing CBCs in the south. Instead, we see declining BBSs and CBCs that coincide with the changes in migration counts.

In the East, migration counts, BBS data, and CBCs indicate that kestrel populations are declining, but that these declining trends weaken west of the Appalachian Mountains. Results from Hawk Ridge Bird Observatory and Holiday Beach Migration Observatory suggest that breeding populations of American Kestrels north of the Great Lakes were increasing or stable until recently, when they began to decline. Comparison of recent indexes to previous 5-yr periods also reflects these patterns and shows that population indexes at most hawkwatch sites were much higher in the 1970s than at present.

In the East, the three easternmost hawkwatch sites recorded declines of >50% over 30 yr, a decline that places kestrels in the most vulnerable category defined by Partners in Flight (Rich et al. 2004). The IUCN rates any population undergoing a decline of 30% over 10 yr as vulnerable (IUCN 2001), and five eastern hawkwatch sites recorded declines of this magnitude (Table 4). With respect to the conservation status criteria of Farmer and Hussell (2008), trends in this region clearly indicate cause for conservation concern, with five hawkwatch sites showing recent indexes below long-term averages and currently lower than in any previous 5-yr period on record.

In the West, kestrel counts showed moderate to strong declining patterns during the past decade at all migration sites except the Manzano Mountains, suggesting that populations in the central and southern Rocky Mountains may be more secure at present than in other regions of the West extending from the northern Rocky Mountains to the Pacific Northwest. Whether the recent declines are entirely due to recent widespread drought or a combination of other factors is unknown, but further investigation of potential factors appears warranted. Average rates of change currently range from roughly −4% to −12% per yr where declines are occurring. Prudent management should take these declines into account regardless of cause before undertaking any actions that may further depress kestrel populations. Only the trends in the western Rocky Mountains, however, indicate cause for concern according to the criteria of the IUCN, Partners in Flight, or Farmer and Hussell (2008).

Declines in kestrel populations migrating past raptor migration watch sites may be due to several factors. Kestrels continued to be exposed to high levels of DDT well into the late 1970s, even after the pesticide was banned in the United States in 1972 (Smallwood and Bird 2002), and DDT interferes with successful reproduction in the species (Porter and Wiemeyer 1969, Lincer 1975). Unknown numbers of kestrels die due to poisoning by other pesticides in agricultural areas, and we have recently recovered two kestrels that were killed by organophosphates or carbamates (one bird), and rodenticides (one bird; C. Farmer unpubl. data).

Populations of Cooper's Hawks (Accipiter cooperii) increased throughout northeastern North America from 1974 to 2004 (Farmer et al. 2008a). Research at Hawk Mountain Sanctuary suggests that this species sometimes preys upon American Kestrels in settings where open fields are interspersed with woodlots (Farmer et al. 2006). Other recent research suggests that intraguild predation can be a significant factor determining the population density and dispersion of raptors (e.g., Petty et al. 2003, Sergio et al. 2007).

At the same time, much of this region has been reforested or developed, replacing foraging habitat for kestrels with forests or suburban land cover that provide fewer feeding, nesting, and migratory stopover opportunities.

Since 1999, West Nile virus apparently has affected numerous bird species throughout the continent (CDC 2009). Although the effect of the virus on kestrel populations is unknown, approximately 95% of the adults using nest boxes in 2004 in southeastern Pennsylvania near Hawk Mountain Sanctuary had been exposed to the virus, although this exposure rate has since declined (Medica et al. 2007, Medica and Bildstein 2009). Productivity in this nest-box program declined 57% between 2000 and 2004, stemming primarily from a 44% decline in nesting attempts. West Nile virus seems unlikely as a causal agent for the long-term declines reported here, as they predate the arrival of West Nile virus by more than 20 yr in some cases (e.g., at Cape May, New Jersey).

Overall, migration count data suggest substantial recent declines in populations of American Kestrels across much of North America, and consequently strong cause for conservation concern. Whether these declines represent a crisis for the species or simply a return to numbers present prior to agricultural development is unknown due to the lack of information regarding historical population levels. Patterns of decline over the last decade suggest that the declines are not merely a return to historical levels, however, because the declines have occurred consistently across large regions with very different ecologies and recent histories of development.

A gradient from east to west is apparent in trend estimates for northeastern North America, with stronger declines occurring at coastal hawkwatch sites (Fig. 1). This pattern may indicate that negative influences on populations are strongest in areas from which kestrels migrate near dense human populations along the Atlantic Coast. More recent declines in western North America, however, appear to have occurred in concert with a prolonged drought.

The species' North American population is estimated to exceed one million birds (Ferguson-Lees and Christie 2001), with some previous trend estimates suggesting stable or increasing populations in many areas (e.g., Titus and Fuller 1990, Kirk and Hyslop 1998, Hoffman and Smith 2003). The widespread, significant declines we found using new, more powerful methods of analysis (Farmer et al. 2007) clearly warrant further investigation to identify causal factors and determine appropriate conservation measures. The duration and magnitude of the declines we have documented suggest that this common raptor may not remain common in the future.