J. Raptor Res. 39(3):237–246
䉷 2005 The Raptor Research Foundation, Inc.
PATTERNS OF TEMPORAL VARIATION IN GOSHAWK
REPRODUCTION AND PREY RESOURCES
SUSAN R. SALAFSKY1
AND
RICHARD T. REYNOLDS
USDA Forest Service, Rocky Mountain Research Station, 2150 Centre Avenue, Building A, Suite 350,
Fort Collins, CO 80526 U.S.A.
BARRY R. NOON
Department of Fishery and Wildlife Biology, Colorado State University, Fort Collins, CO 80523 U.S.A.
ABSTRACT.—To investigate whether Northern Goshawk (Accipiter gentilis) reproduction is food-limited,
we evaluated the reproductive output from 401 goshawk breeding opportunities on the Kaibab Plateau,
Arizona during 1999–2002. Concurrently, we estimated densities of 10 goshawk prey species (seven birds,
three mammals) using distance sampling. We then assessed the relationship between goshawk productivity (number of fledglings produced) and prey density within and among years by relating the contribution of individual prey species and total prey density to goshawk productivity. We also estimated the
proportion of total diet and biomass for each species that contributed ⱖ3% of all prey items. Total prey
density was highly correlated with variation in goshawk productivity (r2 ⫽ 0.98, P ⫽ 0.012). Red squirrel
(Tamiasciurus hudsonicus) density explained more variation in goshawk productivity than any other species (r2 ⫽ 0.94, P ⫽ 0.031), but density could not be estimated for every predominant prey species in
goshawk diets. However, only red squirrels had a positive and significant relationship to goshawk productivity in terms of frequency (r2 ⫽ 0.97, P ⫽ 0.014) and biomass (r2 ⫽ 0.95, P ⫽ 0.033). Northern
Flickers (Colaptes auratus) and cottontail rabbits (Sylvilagus spp.), which contributed the greatest frequency and biomass, respectively, to goshawk diets, showed no relationship with goshawk productivity.
Even though goshawks on the Kaibab Plateau have a diverse diet and will readily switch to alternate
prey species, goshawk productivity showed significant interannual variation. Our results suggest that the
magnitude of goshawk productivity was determined by total prey density and annual variation was driven
by differences in the densities of critical prey species.
KEY WORDS:
productivity.
Northern Goshawk; Accipiter gentilis; diet; distance sampling; predator-prey dynamics; prey density;
PATRONES DE VARIACIÓN TEMPORAL DE LAS PRESAS Y DE LA REPRODUCCIÓN
DE ACCIPITER GENTILIS
RESUMEN.—Para investigar si la reproducción de Accipiter gentilis se encuentra limitada por la disponibilidad de alimento, evaluamos el rendimiento reproductivo de 401 oportunidades reproductivas de
estos halcones en Kaibab Plateau, Arizona, entre 1999 y 2002. Al mismo tiempo, estimamos las densidades de 10 especies de presas para los halcones (siete aves, tres mamı́feros) utilizando el método de
conteo con distancias variables. Luego determinamos la relación entre la productividad de los halcones
(número de volantones producidos) y la densidad de presas dentro y entre años, relacionando la contribución de cada especie de presa y la densidad total de presas con la productividad de los halcones.
También estimamos la proporción de la dieta total para cada especie que contribuyó más del 3% de
todas las presas en la dieta. La densidad total de las presas se correlacionó fuertemente con la variación
en la productividad de los halcones (r2 ⫽ 0.98, P ⫽ 0.012). La densidad de la ardilla Tamiasciurus
hudsonicus explicó la mayor parte de la variación en la productividad de los halcones con relación a las
otras especies (r2 ⫽ 0.94, P ⫽ 0.031), pero no se pudo estimar la densidad de cada especie de presa
predominante en la dieta de los halcones. Sin embargo, sólo la ardilla T. hudsonicus presentó una
relación positiva y significativa con la productividad de los halcones en términos de frecuencia (r2 ⫽
0.97, P ⫽ 0.014) y biomasa (r2 ⫽ 0.95, P ⫽ 0.033). Las aves del género Colaptes y los conejos, que
contribuyeron la mayor frecuencia y biomasa de la dieta de los halcones, respectivamente, no se cor1
Corresponding author’s email address: salafsky@lamar.colostate.edu
237
238
SALAFSKY
ET AL.
VOL. 39, NO. 3
relacionaron con la productividad de los halcones. A pesar de que los halcones en Kaibab Plateau tienen
una dieta diversa y pueden cambiar fácilmente a especies de presas alternativas, su productividad mostró
una variación interanual significativa. Nuestros resultados sugieren que la magnitud de la productividad
de los halcones fue determinada por la densidad total de presas y que la variación anual fue producida
por las diferencias en la densidad de especies de presas crı́ticas.
[Traducción del equipo editorial]
To understand temporal variation in population
size, it is necessary to focus on the factors that limit
demographic processes, such as reproduction and
survival. Ultimately, the availability of essential resources within a habitat regulates population
growth. Resource availability, specifically food, is
hypothesized to be an important limiting factor of
many raptor populations (Newton 1979). Variations in food supply often result in extensive fluctuations in population demographic parameters
(Gotelli 1998, Newton 1998), but the mechanisms
of food-limitation are difficult to quantify, especially in complex systems. Consequently, most information on the influence of food resources on
population dynamics comes from correlations between reproduction and food abundance (Martin
1987).
The magnitude of the effects of food-limitation
on reproduction is poorly understood, especially
for predators with broad diets, such as Northern
Goshawks (Accipiter gentilis). Goshawks regularly
consume a variety of prey including ground and
tree squirrels, rabbits, medium to large passerines,
woodpeckers, and gallinaceous birds (Squires and
Reynolds 1997). The diversity of prey in their diets
ultimately depends on the abundance and availability of the local bird and mammal fauna, which
varies geographically. In Canada, although goshawks regularly consumed several prey species
(ⱖ5), goshawk reproduction showed a strong functional response to only one species—snowshoe
hare (Lepus americanus; Doyle and Smith 2001). In
contrast, 14 species of birds and mammals regularly contributed to goshawk diets in the southwestern
United States (Reynolds et al. 1992). This diet diversity may stabilize their breeding rates. When
prey populations vary asynchronously, the ability of
goshawks to switch between alternative prey species
may result in less annual variation in reproduction
than in areas where goshawks rely primarily on cyclic populations of a single prey species (Newton
1979).
Our objectives were to: (1) determine if prey resources limit the reproductive rates of goshawks
with relatively diverse diets and (2) describe how
changes in prey populations may influence goshawk productivity (number of fledglings produced). If food is a limiting factor of goshawk productivity, then variation in the number of
fledglings produced should be associated with fluctuations in prey resources. However, if there is a
difference in the contribution of individual prey
species, then goshawk productivity should respond
to fluctuations in the densities of individual prey
species. Finally, if the densities of important prey
species vary in synchrony, then goshawk productivity should exhibit greater temporal variation. To
explore these relationships we studied goshawk
productivity and prey resources on the Kaibab Plateau, Arizona during 1999–2002.
STUDY AREA
The Kaibab Plateau is a large (95 ⫻ 55 km) forested
island, surrounded by shrub-steppe desert, in northern
Arizona. Steep slopes and escarpments form the eastern,
southern, and western edges of the Kaibab Plateau and
create a distinct boundary between the shrub-steppe desert at 1750 m elevation above sea level and the plateau
(maximum elevation 2800 m). The northern edge of the
plateau gradually descends to sagebrush desert, forming
an indistinct boundary between the two landforms.
The study area (1285 km2) on the Kaibab Plateau included forests above 2182 m elevation on the North Kaibab Ranger District of the Kaibab National Forest. Four
forest types dominated the study area: Pinyon-juniper (Pinus edulis-Juniperus spp.) woodlands occupied 106 km2 at
lower elevations, ponderosa pine (Pinus ponderosa) forests
occupied 714 km2 at mid-elevation zones, mixed conifer
(Abies concolor, Pinus ponderosa, Pseudotsuga menziesii, Picea
engelmannii) forests occupied 275 km2 at the highest elevations, and quaking aspen (Populus tremuloides) forests
occupied 112 km2 interspersed among the other forest
types ( Joy 2002).
METHODS
Goshawk Productivity. We estimated annual goshawk
productivity per territory in 1999–2002. A territory was
defined as the area (approximately 11 km2) defended by
a pair of goshawks during the breeding season (Reynolds
et al. 2005). Because goshawks may use more than one
nest within a territory among breeding years (Reynolds
et al. 2005), all nest structures were visited annually in
spring to determine the territory occupancy status. If an
active nest (nest containing eggs or young) was not located within an existing territory, we conducted systematic surveys until we found an active nest or thoroughly
SEPTEMBER 2005
BIOLOGY
searched the entire territory, which required a minimum
effort of 10 person-days (Reynolds et al. 2004). Each year,
we also conducted surveys throughout the study area to
locate territories not detected in previous years (Reynolds and Joy 2005). To determine nest status and fledgling production, all active nests were visited weekly
throughout the breeding season. Goshawk offspring were
counted in the nest 7–10 d prior to fledging or from the
ground after fledging. Goshawk productivity was estimated annually as the mean number of fledglings produced
per territory under study.
Prey Density. To obtain estimates of prey density, we
conducted distance sampling (Buckland et al. 1993)
along line transects from 1999–2002. Sixty 500-m transects were placed randomly throughout the study area
within two strata defined by the forest types (mixed conifer, ponderosa pine) that occupied most of the study
area. We established 30 transects per stratum and characterized each transect by its elevation, tree species composition, and tree density. Within a given year, we sampled each transect during three time periods that
corresponded with specific goshawk reproductive stages:
spring (28 May–24 June)–incubation/hatching stage,
summer (25 June–22 July)–nestling stage, and late summer (23 July–14 August)–fledging stage. To reduce travel
time between transects and to increase sampling efficiency, transects were grouped by location. Transects were
sampled in groups of four per day, and the sampling order of groups was determined using a random number
table. Daily sampling began 0.5 hr after sunrise and was
completed within 3 hr. All transects were sampled by one
observer (Salafsky) during the 4 yr of the study. Sampling
was not conducted during inclement weather (rain,
winds ⬎20 kph) due to reduced probability of prey detection. Prey seen or heard during sampling were identified to species, and the perpendicular distance from the
detected animal to the transect line was measured with a
laser rangefinder (accurate to ⫾1 m). Data were collected on 15 prey species common in goshawk diets on the
Kaibab Plateau (S. Salafsky unpubl. data) and considered
important components of goshawk diets in the southwestern United States (Reynolds et al. 1992).
Goshawk Diet. The species composition of goshawk diets was determined from prey remains (pelage, plumage,
skeletal parts) that were collected from active goshawk
nest sites during weekly visits throughout the breeding
season. Prey remains were pooled by territory and date
collected, identified to species, and paired to assess the
minimum number of individuals consumed (Reynolds
and Meslow 1984). The biomass contribution of individual prey was based on the published mass of each avian
(Dunning 1993) and mammal (Hoffmeister 1986) species. All methods for quantifying raptor diets have inherent biases (Marti 1987). However, Kennedy (1991) reported that estimates of prey use were similar for prey
remain, pellet, and direct observation methods of diet
analysis for goshawks in New Mexico.
Data Analysis. We based goshawk productivity on the
number of fledglings produced per territory under study.
We classified territories based on ⱖ1 attempt to breed on
the territory, the identity of the adult birds, and the average inter-territory distance (Reynolds et al. 2005). A
high density of territories, a tendency of individuals to
239
retain the same territory for life, and a delayed age at
first breeding (Wiens and Reynolds 2005) suggests that
the breeding habitat on the Kaibab Plateau was saturated.
This evidence combined with the observed patterns of
territory occupancy for individual adults over a 14-yr period (S. Salafsky unpubl. data) indicated that goshawks
occupied the territories, even when we found little evidence of birds present. By including all territories rather
than only those that were confirmed ‘‘active’’ or ‘‘occupied,’’ we accounted for all potential breeding opportunities and the full range of variability in the reproductive
quality of territories.
Variable distance sampling data were analyzed with
program DISTANCE, Version 3.5 (Thomas et al. 1998).
Reliable estimates of density from distance sampling depend on several critical assumptions: all individuals on
the transect line were detected, all individuals were detected at their initial location, and all distances were measured accurately (Buckland et al. 1993). Data collection
methods were designed to meet these assumptions. Because variable distance sampling uses a detection function that compensates for differences in detection probabilities among species, habitats, and distances from
transects (Emlen 1971, Buckland et al. 1993), density estimates based on distance data are not confounded by
factors affecting detectability and thus are representative
of the true population size. Prey densities were estimated
separately for mixed conifer and ponderosa pine to account for differences in detection probabilities among
forest types. These estimates were then multiplied by the
proportion of each forest type within the study area and
added together to calculate prey densities for the entire
study area. Annual density estimates were computed only
for species with sufficient sample sizes. Total prey density
was calculated as the sum of the individual prey densities
for species with a sufficient number of detections. We
stratified total prey density by sampling period within
each year to estimate prey densities associated with goshawk breeding phenology.
We used the Tukey-Kramer adjustment for multiple
comparisons of means to test for differences in goshawk
productivity among years (PROC GLM, SAS Institute
1999). Z-statistics were used to test for differences in
mean prey densities among years and sampling periods
(Buckland et al. 1993). To control for Type I error, we
only tested for differences in density between specific
pairwise comparisons (e.g., years of highest and lowest
density). To assess the relationship between goshawk productivity and prey density, we used linear regression
(PROC REG, SAS Institute 1999), where annual goshawk
productivity was the dependent variable, and estimates
for individual prey species and summed over prey species
were used as explanatory variables. Linear regression was
also used to assess the relationship between goshawk productivity and prey species in the diet. In these regressions, annual goshawk productivity was the dependent
variable and percent of total diet or biomass contribution
for individual prey species were assessed as explanatory
variables. We used an information-theoretic approach
(Burnham and Anderson 2002) to identify the prey variables that explained the most annual variation in goshawk productivity per territory. A priori candidate models
were developed to represent the potential effects of prey
240
SALAFSKY
ET AL.
VOL. 39, NO. 3
Figure 1. Mean number of Northern Goshawk fledglings produced per territory (⫾SE) on the Kaibab Plateau,
Arizona, 1999–2002.
density on goshawk productivity. We hypothesized that
goshawk productivity would be most strongly related to
prey densities that contributed the most to goshawk reproduction. Competing models were ranked by their adequacy in explaining the variation in goshawk productivity using Akaike Information Criterion (PROC MIXED,
SAS Institute 1999). To compare the relative importance
of each prey species, we also used cumulative Akaike
weights, which were calculated by summing the weights
across all models that included the variable of interest
(Burnham and Anderson 2002).
RESULTS
Variation in Goshawk Productivity. The number
of goshawk territories used to estimate productivity
was 97 in 1999, 98 in 2000, and 103 in 2001 and
2002 (N ⫽ 401). The proportion of territories with
active nests was 54% in 1999, 58% in 2000, 28% in
2001, and 18% in 2002. Goshawk productivity (x̄ ⫾
SE) varied among years (F3,397 ⫽ 26.78, P ⬍ 0.001)
and ranged from 0.14 ⫾ 0.04 fledglings produced
per territory in 2002 to 1.23 ⫾ 0.14 fledglings produced per territory in 2000 (Fig. 1). There was a
significant decline (F1,199 ⫽ 37.15, P ⬍ 0.001) in
goshawk productivity between 2000 and 2001 (Fig.
1).
Variation in Prey Density. Ten prey species had
sufficient detections to estimate density: American
Robin (Turdus migratorius), Clark’s Nutcracker (Nucifraga columbiana), Downy Woodpecker (Picoides
pubescens), golden-mantled ground squirrel (Spermophilus lateralis), Hairy Woodpecker (Picoides vil-
losus), Kaibab squirrel (Sciurus aberti kaibabensis),
Northern Flicker (Colaptes auratus), red squirrel
(Tamiasciurus hudsonicus), Steller’s Jay (Cyanocitta
stelleri), and Williamson’s Sapsucker (Sphyrapicus
thyroideus). We were unable to estimate densities
for black-tailed jackrabbit (Lepus californicus), Blue
Grouse (Dendragapus obscurus), chipmunk (Eutamias spp.), cottontail rabbit (Sylvilagus spp.), and
rock squirrel (Spermophilus variegatus) due to low
numbers of detections. Detection probability plots
showed little evidence of heaping, measurement
errors, and evasive movement prior to detection.
Total prey density (⫾SE) varied annually and
ranged from 2.22 ⫾ 0.08 individuals ha⫺1 in 2001
to 3.96 ⫾ 0.14 individuals ha⫺1 in 2000 (z ⫽ 10.39,
P ⬍ 0.001). Density also varied significantly among
years for most individual prey species (Table 1) including golden-mantled ground squirrel (z ⫽ 2.18,
P ⫽ 0.015), Hairy Woodpecker (z ⫽ ⫺2.88, P ⫽
0.002), Kaibab squirrel (z ⫽ 2.47, P ⫽ 0.007),
Northern Flicker (z ⫽ 5.70, P ⬍ 0.001), red squirrel (z ⫽ 8.32, P ⬍ 0.001), Steller’s Jay (z ⫽ 3.25, P
⬍ 0.001), and Williamson’s Sapsucker (z ⫽ ⫺2.78,
P ⫽ 0.003). Significant declines in prey densities
were also observed between 2000 and 2001 for
golden-mantled ground squirrel (z ⫽ 2.18, P ⫽
0.015), Kaibab squirrel (z ⫽ 2.47, P ⫽ 0.007),
Northern Flicker (z ⫽ 2.62, P ⫽ 0.005), and red
squirrel (z ⫽ 8.32, P ⬍ 0.001), but only red squirrel
SEPTEMBER 2005
BIOLOGY
241
Table 1. Annual estimates of Northern Goshawk prey density ha⫺1 for American Robin (AMRO), Clark’s Nutcracker
(CLNU), Downy Woodpecker (DOWO), golden-mantled ground squirrel (GMSQ), Hairy Woodpecker (HAWO),
Kaibab squirrel (KASQ), Northern Flicker (NOFL), red squirrel (RESQ), Steller’s Jay (STJA), Williamson’s Sapsucker
(WISA), and all 10 prey species’ densities combined (Total) on the Kaibab Plateau, Arizona, 1999–2002.
1999
2000
2001
2002
SPECIES
x̄
SE
x̄
SE
x̄
SE
x̄
SE
AMRO
CLNU
DOWO
GMSQ
HAWO
KASQ
NOFL
RESQ
STJA
WISA
Total
0.23
0.05
0.20
0.28
0.09
0.11
0.58
1.16
0.41
0.18
3.29
0.05
0.02
0.05
0.30
0.03
0.04
0.08
0.17
0.07
0.04
0.19
0.22
0.05
0.11
0.64
0.05
0.26
0.77
1.38
0.12
0.36
3.96
0.07
0.02
0.03
0.18
0.02
0.07
0.09
0.15
0.05
0.08
0.14
0.27
0.04
0.20
0.22
0.23
0.07
0.48
0.12
0.30
0.28
2.22
0.06
0.01
0.06
0.06
0.06
0.03
0.06
0.04
0.07
0.06
0.08
0.25
0.11
0.11
0.32
0.17
0.08
0.20
0.23
0.33
0.45
2.24
0.06
0.04
0.03
0.10
0.04
0.02
0.04
0.05
0.05
0.09
0.09
density decreased by an order of magnitude (Table
1).
Prey density also varied by sampling period (Fig.
2). However, there were too few observations to
accurately estimate density by sampling period for
most individual prey species, so we report only total prey density by sampling period. Total prey density in the spring sampling period was highest in
2000, followed by 1999, 2002, and 2001 (Fig. 2).
However, the decrease in density was only statisti-
cally significant between 1999 and 2002 (z ⫽ 1.74,
P ⫽ 0.041), and 2002 and 2001 (z ⫽ 6.58, P ⫽
0.005). The only significant decrease in total prey
density between the late-summer sampling period
of one year and the spring sampling period of the
next occurred between 2000 and 2001 (z ⫽ 6.58,
P ⬍ 0.001; Fig. 2).
Goshawk Diets. Goshawks on the Kaibab Plateau
captured and consumed a wide diversity of prey. A
total of 710 individual prey items consisting of 30
Figure 2. Total Northern Goshawk prey density estimates ha⫺1 (⫾SE) by sampling period on the Kaibab Plateau,
Arizona, 1999–2002.
242
SALAFSKY
ET AL.
VOL. 39, NO. 3
Table 2. Prey species each contributing ⱖ3% of all items (N ⫽ 710) to Northern Goshawk diets in terms of percent
frequency and biomass (kg), and their relationship to the number of fledglings produced per goshawk territory on
the Kaibab Plateau, Arizona, during 1999–2002.
SPECIES
Black-tailed jackrabbit
Clark’s Nutcracker
Cottontail rabbit
Kaibab squirrel
Northern Flicker
Red squirrel
Steller’s Jay
NO.
PERCENT
FREQUENCY
R2
P
PERCENT
BIOMASS
R2
P
23
34
125
40
141
87
88
3
5
18
6
20
12
12
⫺0.32
0.03
⫺0.13
⫺0.05
0.87
0.97
⫺0.09
0.43
0.82
0.64
0.78
0.07
0.01
0.70
24
2
42
13
8
7
4
⫺0.36
0.01
⫺0.03
0.02
0.87
0.95
0.04
0.40
0.88
0.84
0.87
0.07
0.02
0.81
species were collected from nest areas during
1999–2002. Seven species each contributed ⱖ3%
of all prey items collected in terms of percent frequency (Table 2). In descending order of percent
of total diet, the most common prey items were
Northern Flickers, cottontail rabbits, red squirrels,
Steller’s Jays, Kaibab squirrels, Clark’s Nutcrackers,
and black-tailed jackrabbits. The descending order
of species biomass contribution to goshawk diets
was: cottontail rabbits, black-tailed jackrabbits, Kaibab squirrels, Northern Flickers, red squirrels,
Steller’s Jays, and Clark’s Nutcrackers (Table 2).
The mean number of prey items per fledgling was
1.8 in 1999, 2.3 in 2000, 5.1 in 2001, and 7.6 in
2002. In contrast the mean biomass of prey items
per fledgling was 0.8 kg in 1999, 0.6 kg in 2000,
2.3 kg in 2001, and 2.8 kg in 2002.
Goshawk Productivity and Prey Resources. We
found a strong positive relationship (r2 ⫽ 0.98, P
⫽ 0.012) between total prey density and goshawk
productivity from 1999–2002 (Fig. 3). Although annual goshawk productivity was highly correlated
with prey density in the spring sampling period (r2
⫽ 0.70, P ⫽ 0.163), summer sampling period (r2
⫽ 0.75, P ⫽ 0.131), and late-summer sampling period (r2 ⫽ 0.79, P ⫽ 0.112), annual prey density
accounted for more of the variation in goshawk
productivity. Based on regression models for each
prey species, only red squirrel density had a significant and positive relationship to goshawk produc-
Figure 3. The relationship between total prey density ha⫺1 and the mean number of Northern Goshawk fledglings
produced per territory on the Kaibab Plateau, Arizona, 1999–2002.
SEPTEMBER 2005
BIOLOGY
243
Table 3. Top 10 models for mean number of Northern Goshawk fledglings produced per territory on the Kaibab
Plateau, Arizona, 1999–2002. Models are ranked based on Akaike’s Information Criteria (AIC) and include model
covariates, number of parameters (K), AIC differences (⌬AIC) and Akaike weights (wi).
MODEL
Total prey species
Red squirrel
Mammal prey species
Northern Flicker
Total prey in late-summer
Hairy Woodpecker
Total prey in summer
Kaibab squirrel
American Robin
Total prey in spring
AIC
K
⌬AIC
Wi
1134.80
1137.70
1138.10
1144.20
1149.50
1150.80
1152.20
1154.20
1155.40
1156.90
3
3
3
3
3
3
3
3
3
3
0.00
2.90
3.30
9.40
14.70
16.00
17.40
19.40
20.60
22.10
0.70
0.16
0.13
0.01
0.00
0.00
0.00
0.00
0.00
0.00
tivity (r2 ⫽ 0.94, P ⫽ 0.031). Red squirrel was also
the only species that had a significant and positive
relationship to goshawk productivity for percent of
diet (r2 ⫽ 0.97, P ⫽ 0.014) and biomass (r2 ⫽ 0.95,
P ⫽ 0.024; Table 2). The densities of mammal prey
species (r2 ⫽ 0.94, P ⫽ 0.033) explained more of
the variation in goshawk productivity than avian
prey species (r2 ⬍ 0.01, P ⫽ 0.949).
Our model selection results showed that total
prey density was clearly the top model (Table 3).
This model, which included an annual summation
of all prey species’ densities, received ⬎70% of the
Akaike weight across the model set (Table 3) and
was more than four times as likely as the next best
model. The only single species models with some
weight of evidence included those for red squirrel
and Northern Flicker (Table 3). However, the red
squirrel density covariate had a higher cumulative
Akaike weight (99%) than Northern Flicker
(71%). All other models based on individual prey
species, avian density, and models of total prey density by sampling period had minimal support and
failed to explain variation in goshawk productivity
(Table 3). When we compared only the models
with total prey density by sampling period in a separate analysis, total prey density summed over all
sampling periods was selected as the best model
(AIC ⫽ 811.10, K ⫽ 3, ⌬AIC ⫽ 0.00, wi ⫽ 0.93).
All other models, including the model with the difference in prey density between late-summer and
the successive spring (AIC ⫽ 873.00, K ⫽ 3, ⌬AIC
⫽ 61.90, wi ⫽ 0.00) and the lowest ranked model
with late-summer prey density from the prior year
(AIC ⫽ 883.60, K ⫽ 3, ⌬AIC ⫽ 72.50, wi ⫽ 0.00),
were not supported by the data.
DISCUSSION
A short-term observational study cannot provide
a strong basis for estimating the causal relationship
between prey resources and annual goshawk productivity. Thus, our study only established a strong
association between variation in prey resources
within the study area and goshawk productivity. Because fluctuations in other limiting factors (e.g.,
climate) may have coincided with changes in prey
resources, we cannot identify the factors ultimately
responsible for variation in goshawk productivity.
However, if the patterns we observed between prey
resources and goshawk productivity were supported by experimental studies that established a relationship between food-supply and goshawk reproduction, then it would be reasonable to infer that
prey resources may be an important limiting factor
of goshawk reproduction on the Kaibab Plateau.
During 1999–2002 we observed high temporal
correlations between goshawk productivity and annual prey density; changes in goshawk productivity
paralleled changes in prey density. Total prey density, in addition to the proportion of active goshawk nests and mean number of fledglings produced, was high in 1999 and 2000 and low in 2001
and 2002. Therefore, it appears that goshawk reproduction on the Kaibab Plateau responded to
inter-annual increases in prey density. Several other studies have also found close ties between measures of goshawk reproduction and the relative
abundances of prey (Huhtala and Sulkava 1981,
Doyle and Smith 1994, Keane 1999). Further, goshawk studies that experimentally manipulated
food-supply found supplemental food may have in-
244
SALAFSKY
fluenced goshawk productivity by increasing nestling survival when background prey-levels were low
(Ward and Kennedy 1996, Dewey and Kennedy
2001). Thus, we suggest that the number of goshawk fledglings produced may be influenced by
fluctuations in prey density.
On an annual basis, the reproductive responses
of goshawks depend on the abundance of prey during critical time periods. Low food resources may
manifest through failure to lay eggs, smaller clutches, and reduced survival of young (Newton 1998).
The abundance of prey may be an important determinant of the ‘‘decision’’ to breed. Goshawks
initiate breeding before most prey species reproduce, so the density of prey during the incubation
period is likely similar to prey levels prior to egglaying. On the Kaibab Plateau, prey densities during the incubation stage were similar in 1999 and
2002, yet goshawk productivity was six times higher
in 1999. In addition, although there was a significant increase in prey density during the incubation
period between 2001 and 2002, goshawk productivity changed little between these years, suggesting
that below a density of ca. 0.8 prey ha⫺1, fewer
fledglings are produced. However, prey density levels prior to egg-laying may alter the threshold effects of prey density on goshawk productivity
through physiological constraints. Assuming our
density estimates represented true population size,
the difference in prey density between late-summer
and the next spring should reflect prey density levels prior to egg-laying. The large decline we observed in prey density between August 2000 and
May 2001 indicated that there was substantial overwinter mortality for prey species. The lower prey
numbers prior to egg-laying may have affected the
ability of females to accumulate sufficient reserves
to produce eggs in 2001.
Our results suggest there is a difference in the
contribution of individual prey species to goshawk
reproduction. Red squirrel density and their percent frequency and biomass contribution to goshawk diet accounted for more variation in goshawk
productivity than any other species. Although rabbits contributed the majority of biomass to goshawk diets (⬎66%), goshawk reproduction was
lower in most years when rabbits contributed the
greatest proportion of biomass to the diets. Further, in ‘‘poor’’ goshawk reproductive years (2001,
2002), the number of prey items and total biomass
per fledgling was twice as high as in ‘‘good’’ reproductive years (1999, 2000). The difference in the
ET AL.
VOL. 39, NO. 3
apparent influence of individual prey species is
likely a result of encounter rates with goshawks.
Goshawks are opportunists and will presumably attempt to capture whatever prey species are readily
available. However, the limited distributions or different activity patterns of some prey species decreases the probability that diurnal goshawks will
encounter them while foraging. Jackrabbits are less
common in upper elevation forests, and although
cottontails are widely distributed across the study
area, they are crepuscular (Hoffmeister 1986). In
contrast, red squirrels are among the heaviest of
the diurnal prey species, with a wide distribution
across the study area (Salafsky 2004). Red squirrels
do not hibernate, which likely increases their importance to goshawks, particularly prior to egg-laying. However, the importance of other prey species
may vary with the spatial distribution of goshawk
territories relative to the spatial distribution of prey
habitats. For example, goshawks with territories located primarily within lower elevation forests may
rely more heavily upon jackrabbits.
In our study, goshawk productivity on the Kaibab
Plateau was more closely associated with variation
in mammal density than in avian density. Goshawks
may consume more mammals than birds in some
areas due to the availability and sizes of local prey
species (Zachel 1985, Widén 1987, Doyle and
Smith 1994). Similar to our study, Boal and Mannan (1994) and Reynolds et al. (1994) found that
goshawks on the Kaibab Plateau consumed a higher proportion of mammalian prey. Other goshawk
studies conducted in northern latitudes identified
a strong link between goshawk reproductive rates
and cyclical variation in hare abundance (McGowan 1975, Doyle and Smith 2001). Because annual variations in predator reproductive rates are
greatest among species with limited diets that are
dominated by cyclic prey (Newton 1979), goshawks
on the Kaibab Plateau may be subject to more
marked variations in productivity due to their reliance on prey species with fluctuating densities.
Goshawks have the ability to switch to alternate
prey when the densities of essential prey species
are reduced (Doyle and Smith 1994). However, if
different prey species’ populations decline simultaneously, then the opportunities for goshawks to
switch to alternative prey species are limited. The
densities of golden-mantled ground squirrels, Kaibab squirrels, Northern Flickers, and red squirrels
declined significantly between 2000 and 2001. Further, these species contributed ⬎39% of all prey
SEPTEMBER 2005
BIOLOGY
items to goshawk diets. Parallel fluctuations in annual densities of important prey species may result
in potentially ‘‘poor’’ and ‘‘good’’ years of prey resources. Thus, it may be that the collective density
of the entire prey community influences the magnitude of variation in goshawk productivity on the
Kaibab Plateau.
In summary, our results indicate that prey density is an important limiting factor of goshawk productivity. Although the temporal correlations between goshawk productivity and prey resources
were consistent over time, other factors may have
varied with prey density and limited goshawk reproduction in our study. Synchronous declines in
prey species’ densities suggests that landscape-level
factors acting at broad spatial scales, such as climate, may interact with prey abundance to limit
goshawk productivity. Because unfavorable weather
conditions may have a greater effect on goshawk
productivity when prey resources are already low,
it is important to study the relationship between
goshawk productivity and prey density over long
time periods and variable environmental conditions.
ACKNOWLEDGMENTS
Many thanks go to the goshawk technicians that collected data for this project. We also thank Clint W. Boal,
Curtis H. Flather, John J. Keane, Rudy M. King, Marc A.
Snyder, J. David Wiens, and three anonymous referees for
reviewing this manuscript. This project was made possible
with funding provided by both the Rocky Mountain Research Station and the Southwestern Region of the U.S.
Department of Agriculture Forest Service.
LITERATURE CITED
BOAL, C.W. AND R.W. MANNAN. 1994. Northern Goshawk
diets in ponderosa pine forests on the Kaibab Plateau.
Stud. Avian Biol. 16:97–102.
BUCKLAND, S.T., D.R. ANDERSON, K.P. BURNHAM, AND J.L.
LAAKE. 1993. Distance sampling: estimating abundance of biological populations. Chapman & Hall,
London, U.K.
BURNHAM, K.P. AND D.R. ANDERSON. 2002. Model selection and inference: a practical information-theoretic
approach, 2nd Ed. Springer-Verlag, New York, NY
U.S.A.
DEWEY, S.R. AND P.L. KENNEDY. 2001. Effects of supplemental food on parental-care strategies and juvenile
survival of Northern Goshawks. Auk 118:352–365.
DOYLE, F.I. AND J.M.N. SMITH. 1994. Population responses
of Northern Goshawks to the 10-year cycle in numbers of snowshoe hares. Stud. Avian Biol. 16:122–129.
AND J.M.N. SMITH. 2001. Raptors and scavengers.
Pages 378–404 in C.J. Krebs, S. Boutin, and R. Boon-
245
stra (EDS.), Ecosystem dynamics of the boreal forest.
Oxford University Press, New York, NY U.S.A.
DUNNING, J.B., JR. 1993. CRC handbook of avian body
masses. CRC Press, Inc., Boca Raton, FL U.S.A.
EMLEN, J.T. 1971. Population densities of birds derived
from transect counts. Auk 88:332–334.
GOTELLI, N.J. 1998. A primer of ecology. Second edition.
Sinauer Associates, Inc., Sunderland, MA U.S.A.
HOFFMEISTER, D.F. 1986. Mammals of Arizona. The University of Arizona Press, Tucson, AZ U.S.A.
HUHTALA, K. AND S. SULKAVA. 1981. Environmental influences on goshawk breeding in Finland. Pages 89–104
in R.E. Kenward and I.M. Lindsay (EDS.), Understanding the goshawk. The International Association for
Falconry and Conservation of Birds of Prey, Oxford,
England.
JOY, S.M. 2002. Northern Goshawk habitat on the Kaibab
National Forest in Arizona: factors affecting nest locations and territory quality. Ph.D. dissertation, Colorado State Univ., Fort Collins, CO U.S.A.
KEANE, J.J. 1999. Ecology of Northern Goshawk in the
Sierra Nevada, California. Ph.D. dissertation, Univ. of
California, Davis, CA U.S.A.
KENNEDY, P.L. 1991. Reproductive strategies of Northern
Goshawks and Cooper’s Hawks in north-central New
Mexico. Ph.D. dissertation, Utah State Univ., Logan,
UT U.S.A.
MARTI, C.D. 1987. Raptor food habitat studies. Pages 67–
80 in B.G. Pendleton, B.A. Millsap, K.W. Cline, and
D.M. Bird, [EDS.], Raptor management techniques
manual. National Wildlife Federation, Sci. Tech. Series No. 10, Washington, DC U.S.A.
MARTIN, T.E. 1987. Food as a limit on breeding birds: a
life-history perspective. Ann. Rev. Ecol. Syst. 18:453–
487.
MCGOWAN, J.D. 1975. Distribution, density and productivity of goshawks in interior Alaska. Final Report to
Federal Aid in Wildlife Restoration, Project Report
W17-3, 4, 5, 6, Alaska Department of Fish and Game,
Juneau, AK U.S.A.
NEWTON, I. 1979. Population ecology of raptors. Buteo
Books, Vermillion, SD U.S.A.
. 1998. Population limitation in birds. Academic
Press, Inc., San Diego, CA U.S.A.
REYNOLDS, R.T., R.T. GRAHAM, M.H. REISER, R.L. BASSETT,
P.L. KENNEDY, D.A. BOYCE, JR., G. GOODWIN, R. SMITH,
AND E.L. FISHER. 1992. Management recommendations for the Northern Goshawk in the southwestern
United States. Gen. Tech. Report RM-217. USDA Forest Service, Rocky Mountain Research Station, Fort
Collins, CO U.S.A.
AND S.M. JOY. 2005. Demography of Northern
Goshawks in northern Arizona, 1991–1996. Stud. Avian Biol.: in press.
, S.M. JOY, AND D.G. LESLIE. 1994. Nest productivity, fidelity, and spacing of Northern Goshawks in
northern Arizona. Stud. Avian Biol. 16:106–113.
246
SALAFSKY
AND E.C. MESLOW. 1984. Partitioning of food and
niche characteristics of coexisting Accipiter during
breeding. Auk 101:761–779.
, G.C. WHITE, S.M. JOY, AND R.W. MANNEN. 2004.
Effects of radiotransmitters on Northern Goshawks:
do tailmounts lower survival of breeding males? J.
Wildl. Manage. 68:25–32.
, J.D. WIENS, S.M. JOY, AND S.R. SALAFSKY. 2005.
Sampling considerations for demographic and habitat
studies of Northern Goshawks. J. Raptor Res. 39:274–
285.
, J.D. WIENS, AND S.R. SALAFSKY. 2005. A review and
evaluation of factors limiting Northern Goshawk populations. Stud. Avian Biol.: in press.
SALAFSKY, S.R. 2004. Covariation between prey abundance
and Northern Goshawk fecundity on the Kaibab Plateau, Arizona. M.S. thesis, Colorado State Univ., Fort
Collins, CO U.S.A.
SAS INSTITUTE. 1999. SAS/STAT User’s Guide, Version 8.
SAS Institute, Inc., Cary, NC U.S.A.
SQUIRES, J.R. AND R.T. REYNOLDS. 1997. Northern Goshawk (Accipiter gentilis). Pages 1–32 in A. Poole and F.
ET AL.
VOL. 39, NO. 3
Gill [EDS.], The Birds of North America, No. 298. The
Academy of Natural Sciences, Philadelphia, PA U.S.A.
THOMAS, L., J.L. LAAKE, J.F. DERRY, S.T. BUCKLAND, D.L.
BORCHERS, D.R. ANDERSON, K.P. BURNHAM, S. STRINDBERG, S.L. HEDLEY, M.L. BURT, F. MARQUES, J.H. POLLARD, AND R.M. FEWSTER. 1998. Distance 3.5. Research
Unit for Wildlife Population Assessment, Univ. of St.
Andrews, U.K.
WARD, J.M. AND P.L. KENNEDY. 1996. Effects of supplemental food on growth and survival of juvenile Northern Goshawks. Auk 113:200–208.
WIDÉN, P. 1987. Goshawk predation during winter, spring,
and summer in a boreal forest in central Sweden. Holarctic Ecol. 10:104–109.
WIENS, J.D. AND R.T. REYNOLDS. 2005. Is fledgling success
a reliable index of fitness in Northern Goshawks? J.
Raptor Res. 39:210–221.
ZACHEL, C.R. 1985. Food habits, hunting activity, and
post-fledgling behavior of Northern Goshawks (Accipiter gentilis) in interior Alaska. M.S. thesis, Univ. Alaska, Fairbanks, AK U.S.A.
Received 3 March 2004; accepted 15 May 2005
Associate Editor: Clint Boal
Richard Reynolds