SAMPLE SIZES

Hoffman and Smith (2003) reported data from 15 species at six sites, collected for 10 to 19 years per site. Overall, the study reports and summarizes well over half a million observations of migrant raptors. Site-specific totals for some species, however, are quite low. For example, although about 1534 Golden Eagles are detected annually in the Bridger Mountains of Montana, only about 28 are sighted each year at Lipan Point, Arizona. In their analyses, however, Hoffman and Smith (2003) do not weight the significance of site-specific results by the volume of migration at those sites; the biological significance of trends at Lipan Point and in the Bridger Mountains are effectively equated.

In addition, Hoffman and Smith (2003) discuss spatial and temporal patterns based on small sample sizes. For example, generalizations about eagle migration in the Intermountain West Flyway were based on two sites (Goshute Mountains, Nevada and Lipan Point, Arizona; according to the authors, the flyway affinities of the Wellsville Mountains, Utah are not clear-cut). The sample size for characterizing the Rocky Mountain Flyway is also effectively two, the Bridger Mountains, Montana and the two New Mexico sites (Manzano and Sandia Mountains). The two sites in New Mexico probably do not represent independent samples from the Rocky Mountain Flyway because of their close proximity, the likelihood that they sample from the same population (Hoffman and Smith 2003), and the fact that they are sampled during different seasons.

We believe that small sample sizes per flyway limit the degree of confidence in regional generalizations. For example, the authors suggest that there is a latitudinal gradient in the tendency of Rocky Mountain migrant eagles to follow a 10-year cycle. This latitudinal gradient hypothesis, however, was based upon just a single northern fall migration site (Bridger Mountains, averaging 1534 eagles per year with no evidence for 10-year cycles) and a single southern fall migration site (Manzano Mountains, averaging only 116 eagles per year with limited evidence for decadal cycling). Elsewhere, conclusions about age-related patterns of population change in the Intermountain Flyway are derived from their sample of two sites in the region. Similarly, their characterization of Golden Eagle ecology in Alaska is based upon data from a single site (Denali National Park, McIntyre and Adams 1999). Although not explicitly cited by Hoffmann and Smith (2003), this Denali study is the one referenced in the paper they do cite (Sherrington 2003), and is the only long-term Alaskan study with published estimates of annual productivity and correlations with prey abundance. These patterns of productivity at one Alaska site may or may not be representative of the species statewide.

In terms of temporal patterns, there is a similar paucity of replication. The suggestion that eagles are tracking cyclic fluctuations in favored prey abundance is intuitively appealing. At some sites described by Hoffman and Smith (2003), there is the appearance of such tracking (see below), but the duration of the study precluded observing even two full cycles. Thus, conclusions about cycling are based on at most one full transition from an early population peak to a later one. We suggest that patterns based on small sample sizes (i.e., sites, population cycles) should be presented explicitly as hypotheses to be tested when more sites or years are added to the data.

PATTERNS OF POPULATION CHANGE

The caption for the cover photo of Condor 105 (Issue 3) states, “In western North America, declines in migration counts of immature Golden Eagles tend to precede peaks in adult abundance, perhaps reflecting declining productivity in response to prey cycles. Hoffman and Smith's comprehensive summary of 24 years of migration data at six count sites (p. 397) identifies this trend.” We suggest that the patterns described here are neither as general nor as clear-cut as both the caption and the article indicate.

Hoffman and Smith (2003) state that at five of the six sites (all except the Bridger Mountains), “adult passage rates followed a similar pattern through at least the mid-1990's: high abundance in the mid-1980s (the Wellsville Mountains surveys may have missed this period), generally sharp declines the next 2–3 years, then mostly steady increases through the mid-1990s (Fig. 5)”. Figure 5, however, does not appear to support these generalizations. In terms of high numbers in the mid-1980s, one of the five sites (Lipan Point) had no data for this period, and, as noted by the authors, a second site (Wellsville Mountains) began too late to verify if it is consistent with the pattern. Thus, the high in the mid-1980s high is demonstrated at only three of the five sites. The subsequent 2–3 year sharp decline is clear only at Goshute and Manzano Mountains, while at Sandia Mountains, the decline is gradual from 1985 through 1991. The third element of the pattern, a steady increase through the mid-1990s, was not found at Wellsville, and the steep drop at Lipan in 1993 precludes describing the increase at that site as steady. The increase by the mid-1990s at Sandia is relatively small when compared to the dramatic spike in 1998; in fact, the eight-year period from 1987 to 1995 comprised the longest interval of generally stable numbers found at any of the sites. Thus, the three-part pattern in adult passage rates only applies to the Goshute Mountains in the Intermountain Flyway, and to the Manzano Mountains in the Rocky Mountain Flyway.

Hoffman and Smith (2003) also state that, “In the Intermountain Flyway, as illustrated by the Goshute Mountains, Lipan Point, and to a lesser degree Wellsville Mountains transition-zone counts, it appeared that the cycle began repeating after a mid-1990s peak. In comparison, except for an unusually high 1998 spike in the Sandia Mountains, adult passage rates remained relatively stable after the mid-1990s at the three other Rocky Mountain sites” (p. 413). These generalizations do not appear to be supported by the data presented in the paper. The cycle suggested for the Intermountain sites is clear only at Goshute Mountains. As Hoffman and Smith (2003) indicate, the dramatic annual fluctuations through the 1990s at Wellsville Mountains obscure the significance of declines at that transitional site, and Lipan Point lacks data for the early phase of the putative cycle and depicts two or three peaks during the 1990s. The outlying peaks could be population responses similar in causality to, but lower in amplitude than, the highest peak in 1996. Conversely, they could simply be noise (either sampling error or rapid responses to short-term environmental phenomena) superimposed on a longer-term, larger-scale pattern of population change. Unfortunately, there are no data presented to clarify this question.

Similarly, we do not see relatively stable passage rates when we inspect the Rocky Mountains data. As the authors note, the 1998 peak at Sandia is inconsistent with the hypothesized pattern. If there were a large number of Rocky Mountain sites, a single exception might not fundamentally alter the picture. There are, however, only three sites being characterized. Sandia clearly doesn't show stability, and based on the “M”-shaped population graph for the Bridger Mountains, an argument could be made that cycles are exhibited at that site, but with a higher frequency and lower amplitude than at Goshute Mountains. Finally, the data from the Manzano Mountains show a very different pattern than the data from either of the other two locations. We conclude that after the mid-1990s, adult passage rates in the Rocky Mountains varied considerably, both among years and among sites.

The authors next suggest “that the migratory abundance of eagles within the Intermountain Flyway cycles with a periodicity of ∼10 years. This tendency is less pronounced in the Rocky Mountain Flyway…” (Hoffman and Smith 2003:413). The Goshute Mountains data show a single interpeak interval of 10 years, which is consistent with hare population cycles of 10– 11 years, and the Lipan Point data suggest an intertrough interval of about 11–12 years (if the two lateral peaks are ignored; otherwise, a three- to four-year cycle is suggested). If the quadratic regression line at Wellsville Mountains roughly traces a cycle, however, the period of this population cycle has a frequency of about 14 years (1987–2001). Potential cycles can be superimposed on the data from the Rocky Mountain sites at Bridger, Manzano, and Sandia as well, but one could interpret the data to suggest that these cycles have frequencies of about four, ≥16, and ≥13 years, respectively. Thus, a cycle of approximately 10 years does not seem to be a general pattern.

Based on Goshute Mountains and Lipan Point data, the authors also suggest that multiyear declines in the numbers of immature eagles tend to precede peaks in adult abundance. Three major peaks in adult numbers are depicted in Figure 5 (two at Goshute and one at Lipan Point). Two minor peaks may also occur at Lipan Point. The first multi-year decline in immatures at Goshute Mountains (1983–1985) precedes the first peak of adults (1986). Similarly at Lipan Point, the major adult peak (1996) is preceded by a dramatic drop in the number of immatures (1992–1995). The second multi-year decline in immatures at Goshute Mountains (1993–1995), however, is of much smaller magnitude than the earlier one, and actually switches to an increase coincident with the peak in adult numbers (1996). Further, this second adult peak at Goshute Mountains was reached after a nearly decade-long increase that markedly predated the small decline in immatures from 1993 to 1995. Thus, a causal link between declines in immatures and subsequent peaks in adults seems unsupported over this interval. At Lipan Point, although the postulated pattern holds for the major peak, the minor peaks in adult numbers do not follow multiyear declines in immatures. Rather, they coincide with a major (1992) and minor (1999) peak of immatures, respectively, followed by immediate declines in both classes. At the other four sites, trends of the two age classes show a variety of patterns, from one which parallels that described by the authors (Bridger Mountains), to those at other sites with no apparent correlation between multiyear declines in immatures and peaks in adults. Therefore, the pattern is neither common to most sites, nor consistent at all times at Goshute Mountains and Lipan Point.

The authors cite data from Alberta (Sherrington 2003) that “also indicate a negative correlation between immature and adult abundance” (Hoffman and Smith 2003:413). We do not believe the authors have demonstrated a negative correlation between immature and adult abundance from their migration data. Instead, they have proposed that multiyear declines in immatures precede adult peaks. That pattern may or may not result in a negative correlation in abundance depending upon the temporal scale of the changes and the relative abundance of the two age classes. We used the authors' data (Fig. 5, p. 408) to compare annual changes in abundance in the two age classes. Considering only Goshute Mountains and Lipan Point, annual changes in adult and immature abundance estimates were in the same direction 69% of the time (18 of 26 changes). Among all six sites combined, changes were in the same direction 68% of the time (51 of 75 changes), significantly more frequently than expected (χ²₁ = 4.7, P = 0.03). Thus, year-to-year population trajectories for the two age classes were much more likely to go in the same direction than in the opposite direction, i.e., the adult and immature components of the population usually responded synchronously.

Hoffman and Smith (2003) then attempt to reconcile paradoxical data, the long-term increase in immatures at Goshute with previous studies that found long-term declines in nesting territories in the Snake River area of Idaho. “Thus counts of immature eagles in the Goshute Mountains appear to accurately reflect major cyclical productivity declines in the northern Great Basin, but overall most likely represent a broad source population that includes birds from the northern Great Basin, which may not be doing particularly well, and birds from more northerly latitudes (i.e., Canada and Alaska) where populations are relatively stable and productive” (p. 414). We find this solution unsatisfying, and will consider the two components of this explanation in turn.

In order to conclude that the Goshute Mountains migration data accurately reflect major cyclical productivity declines in the Great Basin, these data should be compared to, and be correlated with, productivity data collected at multiple breeding sites from throughout the region. In their paper, only productivity data from the Snake River area of Idaho were compared to the Goshute migration site, the very comparison that is problematic because of the contradictory trends. The authors document that there were 10 years between successive population peaks of adults at the Goshute Mountains, which suggests the possibility of a 10-year cycle. From the data presented or cited in their paper, however, we see no basis for generally concluding that there are cyclical productivity declines in the northern Great Basin, or that the Goshute Mountains data reflect such declines.

Hoffman and Smith (2003) explain the increasing number of immatures in the Goshute Mountains by postulating that this growing population is comprised of two components, a relatively unproductive Great Basin population plus birds from more successful northern populations. Elsewhere, however, the authors state that most of the eagles counted at the migration sites are probably relatively short-distance migrants, with the Bridger Mountains being an exception. If the authors are correct, then long-distance migrants from northern populations probably do not make up a significant fraction of the birds detected in the Goshute Mountains. In addition, the most compelling evidence that concern for this species is warranted is the marginally significant negative trend at the Bridger Mountains. Because this site has the most eagles, and is the only site (of the six sites) apparently sampling significant numbers of Canadian or Alaskan birds (Kochert and Steenhof 2002, Kochert et al. 2002), their data suggest that northern populations may be less healthy than those comprised of primarily short-distance migrants moving within the western United States.

THE STATUS OF GOLDEN EAGLES

Hoffman and Smith (2003) concluded that an evaluation of migration data and other sources should generate concern about the status of the Golden Eagle in the western United States. Based on their data, and the data they summarize from other sources, we disagree with this conclusion. For each of the six sites, Hoffman and Smith (2003) performed linear regression on continuous data to evaluate long-term trends in total eagles, adult eagles, immature eagles, and age ratios. Among those 24 analyses, only three produced significant results (i.e., P ≤ 0.05): an increase in the number of immatures at Goshute Mountains, and decreases in immatures and age ratios at Lipan Point. Given the general lack of significant trends, and considering that the annual mean number of eagles observed at Goshute Mountains (the site with a significant increase) is nearly 10 times higher that at Lipan Point, concern about the status of Golden Eagles is unwarranted based on the linear regression analyses.

A cautionary note is appropriate, however, given the results from the Bridger Mountains, the site with the largest flight of migrant eagles among the sites analyzed (i.e., nearly 60% of all eagle detections reported during fall migration). Hoffman and Smith (2003:404) reported a marginally significant (i.e., 0.05 < P ≤ 0.10) decline in total eagles in the Bridger Mountains over the 10 years for which they have data, but no declines for either age class alone at this site (Fig. 5, p. 408). The Bridger Mountains certainly merit continued monitoring due to the large volume of eagle migration, but we would argue that, despite the marginally significant trend in the overall population numbers, the sum total of linear regression data do not generate concern about the species' status in western North America.

Hoffman and Smith (2003) also present several supplemental analyses, including quadratic regressions, rank-trend analysis, and, for Wellsville Mountains, t-tests to compare temporally discontinuous datasets. Significant quadratic trends were found for total birds and adults at Lipan Point, and adults at Wellsville Mountains. Such trends do not necessarily indicate a deterioration of status and in fact, are not unexpected in a species that may be cycling over a period roughly comparable to the study duration. In other words, the quadratic regression may just be tracking the increasing and declining phases of population cycles.

Three other supplemental analyses yielded significant results. A rank-trends analysis of the Goshute Mountains data yielded a marginally significant increase in numbers, and t-tests identified significant long-term declines in both immatures and age-ratios at Wellsville Mountains between the periods 1977–1979 and 1987–2001. As noted by the authors, however, there were no significant linear trends for any of the four variables at Wellsville Mountains over just the last 15 years. Overall, the analyses supplementing the linear regression data produced few significant results, and those results were decidedly mixed relative to the status of the species at various sites. Finally, the Breeding Bird Survey and Christmas Bird Count analyses presented by the authors showed no trend and an increasing trend, respectively, for Golden Eagles in western North America.

Hoffman and Smith (2003) have produced a major contribution to our understanding of raptor migration in North America. We conclude, however, that their analyses do not make a compelling case for concern about the status of Golden Eagles. In coming to that conclusion, we neither disparage the value of raptor migration data generally, nor do we argue that concern for Golden Eagles in the western United States is unwarranted. In the first case, we recognize that raptor migration data can be an important, often critical, component of large-scale raptor population assessments. Titus and Fuller (1990), Lewis and Gould (2000), Kjellen and Roos (2000), and Kochert and Steenhof (2002) all argue convincingly about the value of standardized, long-term datasets for tracking population trends in migrant raptors. Regarding the need for concern about Golden Eagles in the western United States, ongoing habitat change across much of this region (Kochert and Steenhof 2002, Kochert et al. 2002, Knick et al. 2003) provides a convincing rationale for carefully monitoring eagle population trends. We do argue, however, that the datasets analyzed and presented by Hoffman and Smith (2003) do not, in and of themselves, provide evidence of Golden Eagle populations at risk.