Ecology of the Northern Goshawk in Fennoscandia

Abstract

We reviewed studies on the Northern Goshawk (Accipiter gentilis) carried out in northern Europe
(Fennoscandia) since the 1950s concerning the following: diet composition, breeding performance, movements,
home range, survival, and population trends. Goshawks feed mainly on forest grouse throughout the
year in boreal forests but rely more on Ring-necked Pheasants (Phasianus colchicus) and hares (Lepus spp.) in
mixed deciduous-coniferous forests in southern Fennoscandia. Breeding density of the goshawks varies from
one–fi ve pairs/100 km2, on average three pairs/100 km2. Mean clutch size (3.5), brood size (2.8), and productivity
of fl edglings (2) per occupied territory have remained stable over the decades irrespective of the decline of
the forest grouse. Proportion of grouse in the diet as well as breeding output closely followed the density of
grouse during the1950s–1970s with relatively dense grouse populations but this close connection has recently
disappeared, probably due to a decline of grouse and disappearance of their multi-annual cycles. Goshawks are
the most important cause of mortality among forest grouse, and grouse density, in turn, affects the dispersal
distances of juvenile goshawks. Because of the narrower diet width of males compared to that of females, males
tend to move over longer distances than females. Among adults, females move more than males, like in other
raptors. Median distances moved by juveniles range from 50–100 km but some individuals can travel up to
>1,000 km. After the dispersal phase, juveniles tend to establish more or less stable ranges before moving to
the fi nal breeding range. Not much is known about the site tenacity of breeders but in good conditions males,
at least, likely remain on their territories throughout their life. Winter range size varies from 2,000–10,000 ha
depending on sex, age, and the quality of the habitat or of the prey size. Juvenile males suffer from higher mortality
than juvenile females but this difference disappears by the third year of life. Based on fi eld studies and
museum data, roughly one-third of juvenile hawks succumb because of starvation, one-third of trauma or trauma
and starvation-disease, and one-fi fth to one-third are killed by hunters. Productivity of goshawk populations has
not changed during the years of declining trends found in many local studies, which may indicate an increased
adult mortality. Annual mortality among the adults may likely not exceed 30% without a decline of the breeding
population. The ultimate reason behind declining goshawk populations is likely the change in the forest bird
community due to intensifi ed forestry which has negatively affected the populations of main prey, forest grouse.
Problems in nourishment of goshawks occur during the winter after migratory birds have moved to south.

Full text

141
ECOLOGY OF THE NORTHERN GOSHAWK IN FENNOSCANDIA
RISTO TORNBERG, ERKKI KORPIMÄKI, AND PATRIK BYHOLM
Abstract. We reviewed studies on the Northern Goshawk (Accipiter gentilis) carried out in northern Europe
(Fennoscandia) since the 1950s concerning the following: diet composition, breeding performance, move-
ments, home range, survival, and population trends. Goshawks feed mainly on forest grouse throughout the
year in boreal forests but rely more on Ring-necked Pheasants (Phasianus colchicus) and hares (Lepus spp.) in
mixed deciduous-coniferous forests in southern Fennoscandia. Breeding density of the goshawks varies from
one–fi ve pairs/100 km2, on average three pairs/100 km2. Mean clutch size (3.5), brood size (2.8), and productiv-
ity of fl edglings (2) per occupied territory have remained stable over the decades irrespective of the decline of
the forest grouse. Proportion of grouse in the diet as well as breeding output closely followed the density of
grouse during the1950s–1970s with relatively dense grouse populations but this close connection has recently
disappeared, probably due to a decline of grouse and disappearance of their multi-annual cycles. Goshawks are
the most important cause of mortality among forest grouse, and grouse density, in turn, affects the dispersal
distances of juvenile goshawks. Because of the narrower diet width of males compared to that of females, males
tend to move over longer distances than females. Among adults, females move more than males, like in other
raptors. Median distances moved by juveniles range from 50–100 km but some individuals can travel up to
>1,000 km. After the dispersal phase, juveniles tend to establish more or less stable ranges before moving to
the fi nal breeding range. Not much is known about the site tenacity of breeders but in good conditions males,
at least, likely remain on their territories throughout their life. Winter range size varies from 2,000–10,000 ha
depending on sex, age, and the quality of the habitat or of the prey size. Juvenile males suffer from higher mor-
tality than juvenile females but this difference disappears by the third year of life. Based on fi eld studies and
museum data, roughly one-third of juvenile hawks succumb because of starvation, one-third of trauma or trauma
and starvation-disease, and one-fi fth to one-third are killed by hunters. Productivity of goshawk populations has
not changed during the years of declining trends found in many local studies, which may indicate an increased
adult mortality. Annual mortality among the adults may likely not exceed 30% without a decline of the breeding
population. The ultimate reason behind declining goshawk populations is likely the change in the forest bird
community due to intensifi ed forestry which has negatively affected the populations of main prey, forest grouse.
Problems in nourishment of goshawks occur during the winter after migratory birds have moved to south.
Key Words: breeding, cause of death, diet, Fennoscandia, habitat choice, movements, Northern Goshawk, preda-
tion, survival.
ECOLOGÍA DEL GAVILÁN AZOR EN FENNOSCANDIA
Resumen. Revisamos estudios sobre el Gavilán Azor (Accipiter gentilis) llevados a cabo en el norte de Europa
(Fennoscandia) desde 1950, relacionados a lo siguiente: dieta, composición, desempeño de reproducción,
movimientos, rango del hogar, sobrevivencia, y tendencias de población. Los gavilanes se alimentaron
principalmente de gallo del bosque (Tetraonidae) en bosques boreales, durante todo el año, pero dependían más
en el Faisán de collar (Phasianus colchicus) y liebres (Lepus spp.) en bosques deciduos mixtos de coníferas, en
el sur de Fennoscandia. La densidad de reproducción del azor varía de uno a cinco pares/100 km2, en promedio
tres pares /100 km2. La media del tamaño de la puesta (3.5), el tamaño de la pollada (2.8) y la productividad
de los volantones (2) por territorio ocupado, ha permanecido estable sobre los años, independientemente al
decaimiento del gallo del bosque. La proporción del gallo del bosque en la dieta, así como la producción-
rendimiento reproductivo, siguieron muy de cerca la densidad del gallo del bosque durante 1950s–1970s,
con relativamente poblaciones densas de gallo del bosque, pero esta cercana conexión ha desaparecido
recientemente, probablemente debido al decaimiento del gallo del bosque y a la desaparición de sus ciclos
multi-anuales. Los Gavilanes son la causa más importante de la mortandad entre los gallos del bosque y de la
densidad de los mismos, por lo tanto, infl uye en las distancias de dispersión de los gavilanes juveniles. Debido
a la estrechez en la dieta de los machos, comparada con la de las hembras, los machos tienden a moverse sobre
distancias más largas que las hembras. Entre los adultos, las hembras se mueven más que los machos, como en
otros raptores. Las distancias medias en las que se mueven los juveniles van desde 50–100 km, pero algunos
individuos pueden viajar por arriba de >1,000 km. Después de la fase de dispersión, los juveniles tienden
a establecer rangos más o menos estables, antes de pasar al rango fi nal reproductivo. No se conoce mucho
acerca de la tenacidad de sitio de los reproductores, pero en buenas condiciones los machos al menos pueden
permanecer en sus territorios por toda su vida. El tamaño del área de ocupación durante el invierno varía de
Studies in Avian Biology No. 31:141–157

STUDIES IN AVIAN BIOLOGY
142 NO. 31
The Northern Goshawk (Accipiter gentilis) is
one of the most numerous raptor species in northern
Europe (hereafter Fennoscandia; Fig. 1). Due to its
relatively high density and dietary preferences for
small game species, especially forest grouse which
are favored objects for sport hunting, the Northern
Goshawk is probably the most hated species of bird
of prey in much of Europe. It has been estimated
that 5,000–6,000 goshawks were killed annually in
Finland in the 1970s (Moilanen 1976) and 2,000 in
the 1960s in Norway (Nygård et al. 1998). In spite
that it has been now protected in all countries of
North Europe—not until 1989 in Finland—it is still
persecuted by humans. Research on Fennoscandian
goshawks was initiated from diet investigations
carried out in the 1950s in Finland and Sweden
(Höglund 1964b, Sulkava 1964) and also in Norway
(Hagen 1952). Since then, several studies on food
habits during the breeding season have been car-
ried out in Finland (Huhtala 1976, Wikman and
Tarsa 1980, Lindén and Wikman 1983, Tornberg
and Sulkava 1991, Tornberg 1997), Sweden (Widén
1987), and Norway (Selås 1989). Winter diet has
been studied by stomach contents (Höglund 1964b)
2,000–10,000 ha dependiendo del sexo, la edad y la calidad del hábitat, o del tamaño de la presa Los machos
juveniles sufren de una mayor mortandad que las hembras juveniles, pero esta diferencia desaparece al tercer
año de vida. Basado en estudios de campo y datos de museos, aproximadamente un tercio de halcones juveniles
sucumben debido a inanición, un tercio por trauma o enfermedad de trauma e inanición, y de un quinto a un
tercio son matados por cazadores. La productividad de las poblaciones de gavilán no ha cambiado durante los
años de tendencias de declinación, encontradas en varios estudios locales, lo cual probablemente indique una
incrementada mortandad adulta. La mortandad anual entre los adultos probablemente no exceda de 30%, sin un
decaimiento en la población reproductiva. La última razón detrás del decaimiento de las poblaciones de gavilán,
es probablemente el cambio en la comunidad de aves de bosque, debido a la intensa actividad forestal, la cual ha
afectado negativamente a las poblaciones de la presa principal, gallo del bosque. Problemas en la alimentación
del gavilán, ocurren durante el invierno, después de que las aves migratorias se han movido hacia el sur.
FIGURE 1. Map of Fennoscandia showing main study sites of Northern Goshawks. 1. Sulkava (1964), 2. Höglund (1964a),
3. Huhtala (1976), 4. Lindén and Wikman (1983), 5. Kenward et al. (1981b), 6 Widén (1987), 7. Selås (1997a), 8. Kenward
et al. (1999), 9. Tornberg (1997), 10. Nygård et al. (1998), 11. Byholm et al. 2003), and 12. R. Tornberg, E. Korpimäki,
V. Reif, S. Jungell and S. Mykrä (unpubl. data).

NORTHERN GOSHAWK IN FENNSCANDIA—Tornberg et al. 143
and by radio tracking since the late 1970s in Sweden
(Kenward et al. 1981, Widén 1987) and in Finland
(Tornberg and Colpaert 2001). Breeding performance
of goshawks is also well documented in all North
European countries; most long-term studies have
been carried out in Finland (Sulkava 1964, Lindén
and Wikman 1980, Huhtala and Sulkava 1981,
Lindén and Wikman 1983, Tornberg and Sulkava
1991, Sulkava et al. 1994, Byholm et al. 2002a) but
also in Sweden (Widén 1985b, Kenward et al. 1999)
and Norway (Selås 1997b). A countrywide survey of
grouse was started in Finland in 1964, which enables
a more accurate estimation of goshawk impact on
grouse (Lindén and Wikman 1983, Tornberg 2001).
In Sweden, an evaluation was done by Widén (1987).
In farmland areas of Sweden, goshawks hunt pheas-
ants more than grouse; Kenward et al. (1981b) esti-
mated the impact of goshawk predation on released
and wild pheasant stocks in central Sweden in the
late 1970s.
Because goshawks use the same nesting territories
year after year, they have become a popular species
with bird banders. Around 2,000 goshawk nestlings
are currently banded annually in Finland, mostly by
volunteers. As a result, recovery rates of goshawks
have been one of the highest among the banded birds
(nearly 50,000 being banded since 1913 when bird
banding was started in Finland; Valkama and Haapala
2002, Byholm et al. 2003). When shooting of gos-
hawks was allowed, around 20% of banded goshawks
were later recovered. These days recovery rates are
around 10%. Total number of recoveries in Finland
now exceeds 8,000 birds (Valkama and Haapala
2002) and similar situations prevail in Sweden and
Norway. These large databases have enabled sev-
eral analyses of movements, mortality, and causes
of death of goshawks in all Fennoscandian coun-
tries (Haukioja and Haukioja 1971, Saurola 1976,
Marcström and Kenward 1981a, Widén 1985b,
Halley 1996, Byholm et al. 2003), as well as more
specifi c studies on, e.g., sex allocation of goshawks
in relation to varying environmental conditions
(Byholm et al. 2002a, 2002b). As an easily trappable
species, banded goshawks are often captured alive
which has given more insight to their movements
(Marcström and Kenward 1981b, Neideman and
Schönebeck 1990). Large radio-tracking projects in
central Sweden in 1970–1980 were also based on
extensive live trapping that gave light to patterns of
age- and condition-related movements (Kenward et
al. 1981a). Pooling data from breeding performance,
survival, and movements of an animal population
facilitates building a population model. On the large
Baltic Sea island of Gotland, Sweden, this was done
using productivity data of breeding goshawks com-
bined with extensive radio-tagging of juvenile and
adult goshawks (Kenward et al. 1991, 1999).
Goshawks have also been an ideal species for
museum work due to large collections of specimens
in zoological museums. Earliest studies were on
taxonomic aspects (Voipio 1946) and later killed
and naturally dying birds were studied in relation
to changes in morphology (Tornberg et al. 1999),
causes of death (Tornberg and Virtanen 1997), or
body condition (Marcström and Kenward 1981a,
Sunde 2002).
In this paper we summarize all noteworthy pub-
lished papers on the ecology of Northern Goshawks
in Finland, Sweden, and Norway. We attempt to doc-
ument the goshawk’s position in those areas based on
past and current studies and to conclude and predict
the future development of goshawk populations,
as well as to outline future needs in research. We
add also some previously unpublished data on diet,
breeding, and home range size collected near Oulu in
northern Finland during 1987–2003. For a descrip-
tion of this study area and the methods, see Tornberg
(1997) and Tornberg and Colpaert (2001).
STUDY AREA
Fennoscandia is composed of three north
European countries, Norway, Sweden, Finland, and
parts of western Russia (Kola peninsula and Russian
Karelia). Although situated between latitudes 55–
70º N this area is mainly characterized by boreal
forests (between latitudes 60–70º N) and mixed
coniferous-deciduous forests in southern Sweden
and Norway (between latitudes 55–60º N). The
northernmost parts of Finland and the Scandinavian
mountain range, Köli, belong to the arctic zone. All
important goshawk studies carried out in the area are
shown in Fig. 1.
CHARACTERS OF THE FENNOSCANDIAN
GOSHAWK
Scandinavian goshawks belong to the nominate
race Accipiter gentilis gentilis. Finland is a transi-
tion zone between the nominate race and the east-
ern paler and larger A. g. buteoides (Voipio 1946).
Finnish goshawks are larger than Swedish ones
based on both body mass and wing length indicating
that Finnish goshawks belong to the larger buteoides
race (Table 1). Winter weights in Sweden are derived
from extensive trapping of goshawks in central
and southern Sweden (Marcström and Kenward
1981b). Weights of Finnish hawks were obtained

STUDIES IN AVIAN BIOLOGY
144 NO. 31
from trapped birds in Oulu during 1990–1999. Wing
lengths were measured from the fl exed wrist to the
end of longest primary with feathers fl attened and
straightened.
DIET
BREEDING SEASON
A major proportion of the diet of the goshawk
was woodland grouse (Tetraonidae) in all food
habit investigations in Fennoscandia (Höglund
1964b, Sulkava 1964, Huhtala 1976, Lindén and
Wikman 1983, Widén 1987, Selås 1989, Tornberg
1997). Four grouse species are preyed upon by
goshawks—Capercaillie (Tetrao urogallus), Black
Grouse (Tetrao terix), Hazel Grouse (Bonasa
bonasia), and Willow Grouse (Lagopus lagopus).
Grouse proportions in the goshawk diet are highest
in western Finland declining to the west and south
(Table 2). It must be remembered that proportions
of grouse in diet studies based on the collection of
prey remains may depend whether remains were col-
lected only in the nest or also in the vicinity of the
nest and whether the two groups are pooled (Sulkava
1964). Proportions of grouse in the diet at the begin-
ning of the nesting season may be up to 80% but
tend to decline later in the breeding season (Table
2). However, depending on the collection method,
the proportion of soft and digestible grouse chicks
might easily be underestimated in the diet (Höglund
1964b, Sulkava 1964, Grønnesby and Nygård 2000).
Recently, with grouse numbers lower than in the
1950s, grouse proportions actually declined during
the breeding season when more vulnerable prey,
like juvenile corvids and smaller passerine birds,
TABLE 1. AVERAGE MEASUREMENTS TAKEN FROM SWEDISH AND FINNISH GOSHAWKS. DATA FROM SWEDEN BY MARCSTRÖM AND
KENWARD (1981A) AND FROM FINLAND BY TORNBERG ET AL. (1999) AND TORNBERG (UNPUBL. DATA).
Central Sweden Northern Finland
Male N Female N Male N Female N
Winter weight adult 866 52 1,328 60 933 12 1,485 18
Winter weight juvenile 839 289 1,229 215 828 11 1,384 21
Wing length adult 323 37 366 69 330 26 372 29
Wing length juvenile 323 308 363 197 327 79 367 86
TABLE 2. DIET COMPOSITION OF GOSHAWKS DURING THE BREEDING SEASON IN DIFFERENT LOCALITIES IN FENNOSCANDIA. UPPER
ROW FOR EACH PREY SPECIES OR GROUP = DIET COMPOSITION DURING THE NEST-BUILDING AND INCUBATION PERIODS AND LOWER
ROW = DIET DURING THE NESTLING PERIOD. GROUSE CHICKS ARE FOUND ONLY DURING THE NESTLING PERIOD.
Localitya
1 2 3 4 5 6 7
Grouse adult 63.7 20.4 72.6 56.3 29.7
11.1 4.8 5.0 14.9 24.7 (14.5)d 14.0
Grouse juvenile 43.3 14.1 23.4 9.6
Corvids 5.3 8.4 10.0 26.2
10.5 23.9 17.6 7.0 11.4 28.3 15.0
Other birds 9.0 49.5 (13.5)b 22.1 29.9
19.6 47.0 68.5 51.6 38.8 42.8 68.0
European red squirrel 15.2 12.5 4.7
(Sciurus vulgaris) 10.2 7.9 5.1 0.8 6.3
Other mammals 6.8 9.2 (14.0)c 6.9 (14.3)c
2.0 1.8 3.7 2.3 9.2 (14.5) (3.0)b
Unidentifi ed 3.2
N 664 535 2101 557 462
342 772 641 128 649 442 367
a Location and source of data: 1. western Finland 1949–1959 (Sulkava 1964), 2. central Sweden. 1954–1959 (Höglund 1964b), 3. southern Finland 1977–1981
(Wikman and Tarsa 1980), 4. north-western Finland 1963–1976 (Huhtala 1976), 5. northern Finland 1965–1988 (Tornberg and Sulkava 1991), 6. central
Sweden 1977–1981 (Widén 1985a), and 7. southern Norway 1983–1988 (Selås 1989).
b Includes corvids.
c Includes squirrels..
d Includes grouse chicks.

NORTHERN GOSHAWK IN FENNSCANDIA—Tornberg et al. 145
become available (Lindén and Wikman 1983, Selås
1989, Tornberg 1997). As grouse chicks grow, they
become more and more profi table as prey and their
proportion of the diet can increase up to 50% toward
the autumn (Tornberg 1997).
The Black Grouse is generally the most important
grouse species by number and biomass in the diet of
goshawks (Huhtala 1976, Widén 1987, Selås 1989,
Tornberg 1997). In Oulu (Fig. 1), its proportion dur-
ing the breeding season was 25–30%. In more south-
ern parts of the boreal forests, however, Hazel Grouse
may be more important (Sulkava 1964, Lindén and
Wikman 1983). When analyzing dietary proportions
against availability in the fi eld, the small grouse spe-
cies, Willow Grouse and Hazel Grouse weighing
0.3–0.7 kg, may be preferred over the larger Black
Grouse weighing 0.9–1.3 kg (Tornberg 1997). Large
Capercaillies are relatively rare in goshawks’ diet,
limited to females weighing 2 kg during the breed-
ing season. The proportion of mammals in the diet of
goshawks varies between 10–20% in most studies.
The most common mammal species is the European
red squirrel (Sciurus vulgaris) whose proportion can
sometimes reach 30%, particularly in poor grouse
years (Sulkava 1964). Young mountain hares (Lepus
timidus) are numerically the second most important
mammalian prey but by biomass they can exceed
red squirrels (Tornberg 1997). Interestingly, young
mountain hares were very rare prey specimens in the
1950s (Sulkava 1964).
The well-documented decline of forest grouse
in Finland (Lindén and Rajala 1981, Väisänen et
al. 1998) has affected prey choice of goshawks.
Changes of grouse density in the province of Oulu
in northern Finland and the corresponding proportion
of grouse in the diet of goshawks are presented in
Fig. 2. A second order polynomial gave the best fi t
for both the grouse density (r2 = 0.587, F = 24.870,
P <0.001) and proportions of grouse in the diet in
spring (r2 = 0.476, F = 11.353, P = 0.003). It seems
that grouse are slowly recovering from the long-term
decline. Correspondingly, goshawks have quickly
responded to this recovery. During grouse lows, gos-
hawks attempt to switch to preying more on corvids,
thrushes, and pigeons (Tornberg and Sulkava 1991,
Sulkava 1999). Interestingly, these species form the
main diet of the goshawk in central Europe (Opdam
et al. 1977, Toyne 1997); grouse are usually not found
in the diet there but Phasianidae can sometimes form
a considerable proportion in the diet (Manosa 1994).
WINTER DIET
Systematically collected data on goshawk’s win-
ter diet are still scarce. Höglund (1964b) analyzed
stomach contents in the 1950s–1960s in Sweden
FIGURE 2. Density changes of forest grouse in the province of Oulu in northern Finland and corresponding proportions of
grouse in the diet of the Northern Goshawk in spring. Density data for grouse were obtained from grouse censuses by the
Finnish Game Research Institute and data for goshawk diets from the1960s and 1970s are from Huhtala (1976) and for the
1980s and 1990s are from Tornberg and Sulkava (1991) and Tornberg (unpubl. data).

STUDIES IN AVIAN BIOLOGY
146 NO. 31
(N = 130), and found that the proportion of grouse
was only 8%, i.e., less than half of that in the summer
diet whereas the proportion of mammals increased
from 10–35%. Later studies carried out by radio
tracking in Sweden partly confi rmed Höglund’s
fi ndings. In the winters 1977–1981, red squirrels
alone comprised 84% (N = 61) of goshawks’ winter
diet in central Sweden (Widén 1987). In agricultural
areas of central Sweden, goshawks killed mainly red
squirrels (33%), Ring-necked Pheasants (23%) and
European hares (Lepus europaeus) (14%) that were
killed only by females (Kenward et al. 1981b). Due
to the large size of hares (3–3.5 kg), they accounted
for 37% of the food intake by females, whereas males
got 43% of their food from pheasants but females
only 3%. Based on a radio-tracking study in northern
Germany, goshawks killed mostly pheasants (41%)
and rabbits (Oryctolagus cuniculus) (27%, N = 145)
during winter (Ziesemer 1983). In northern Finland,
a radio-tracking study during 1991–1995 revealed
that dietary proportion by number of mountain hares
and red squirrels was 55% (N = 55) and the biomass
of hares alone was 70% (Tornberg and Colpaert
2001). Mountain hares were killed only by females.
Correspondingly, as in farmlands, males hunted red
squirrels and grouse more than females did. We pres-
ent here the combined data of Tornberg and Colpaert
(2001) and new winter diet data from the vicinity of
Oulu during 1999–2002. Excluding predation events
near human settlements and a dump site where
brown rats (Rattus norvegicus) were prey, the pro-
portion of grouse was almost the same as in summer
diet (37.6 % vs. 34.2%; Tornberg and Sulkava 1991;
Table 3). Diet differed between the sexes in spite of
few data being available for analysis. In farmland
areas of central Sweden, an intersexual difference
was found only for hares (Kenward et al. 1981b) but
no difference was found in woodland areas (Widén
1987). During the breeding season, diets of the sexes
were not found to differ substantially (Grønnesby
and Nygård 2000).
FUNCTIONAL RESPONSE
When diet proportion or kill rate of a predator
is plotted against the number of prey individuals,
a functional response curve is obtained. Holling
(1959) described three curve types: increase in the
prey consumption of the predator may be linear
(type I), convex (type II), or concave (type III) as
a function of prey number. A type II curve is found
when consumption in low prey density increases
more rapidly than the number of prey and a type
III curve occurs when consumption in low densities
TABLE 3. WINTER DIET OF GOSHAWKS IN THE OULU AREA, NORTHERN FINLAND. DATA ARE BASED ON PUBLISHED RESULTS BY TORNBERG
AND COLPAERT (2001) DURING 1991–1995 AND TORNBERG (UNPUBL. DATA) DURING 1999–2002.
Weight Male Female Total
classes
a N % N % N %
Mountain hare adult E 17 38.6 17 27.9
Capercaillie male E 2 4.5 2 3.3
(Tetrao urogallus)
Capercaillie female D 1 2.3 1 1.6
Mountain hare juvenile D 1 5.9 1 1.6
Black Grouse male D 6 13.6 6 9.8
(Tetrao tertix)
Black Grouse female C 3 17.6 2 4.5 5 8.2
Willow Grouse C 1 2.3 1 1.6
(Lagopus lagopus)
Hazel Grouse B 4 23.5 4 9.1 8 13.1
(Bonasa bonasia)
European red squirrel B 6 35.3 9 20.5 15 24.6
(Sciurus vulgaris)
Great Spotted Woodpecker A 1 5.9 1 1.6
(Dendrocopos major)
Crossbill A 1 5.9 1 1.6
(Loxia curvirostra)
Small passerine A 1 5.9 1 1.6
Small mammals A 2 4.5 2 3.3
Totals 17 44 61
a Weight classes of prey: A = 0–100 g, B = 100–500 g, C = 500–1,000 g, D = 1,000–2,000 g, E = >2,000 g.

NORTHERN GOSHAWK IN FENNSCANDIA—Tornberg et al. 147
increases slower than number of prey. All curve
types level off at high prey densities because the
predator becomes satiated. Curve types predict
different outcomes for the stability in the predator-
prey interaction. Type II tends to destabilize and
type III to stabilize prey population (Holling 1959,
Begon et al. 1996).
Based on the existing studies in Finland and
Sweden, goshawks’ functional response may be con-
cave (Lindén and Wikman 1983), convex (Wikman
and Tarsa 1980, Tornberg and Sulkava 1991), or only
a weak response (Widén 1985a, Tornberg 2001). It
is likely that goshawks show a type III response for
grouse in southern areas of Fennoscandia where they
are less dependent on grouse as a stable food and
where alternative prey is richly available. Whereas
in the north, where grouse form the major part in the
diet and alternative prey are scarce, a concave or no
response is found.
BREEDING OUTPUT OF GOSHAWKS
BREEDING DENSITY AND QUALITY OF THE BREEDERS
Because goshawks use the same breeding sites
fairly regularly year after year, breeding densities in
intensively studied areas can be reliably estimated.
Reliability is also increased by the fact that breeding
territories are very regularly spaced in a continu-
ous woodland area (Widén 1985b, Selås 1997b). In
southern Norway, mean distances during 1980–1990
varied from 4.5–5.4 km (Selås 1997b). In the vicinity
of Oulu, distance between regularly occupied ter-
ritories was around 4 km (Tornberg 2001). Studies
carried out in western and southern Finland during
the 1950–1970s show that goshawk density was
around fi ve pairs/100 km2 when all nests studied were
active (Huhtala and Sulkava 1981). In more restricted
coastland areas of south Finland a breeding density
of fi ve–eight goshawk pairs/100 km2 was reported
during 1977–1983 (Forsman and Solonen 1984).
Breeding density may have declined since the 1970s
and is probably around three pairs/100 km2 at present
in large parts of Fennoscandia (Widén 1997). In the
vicinity of Oulu, breeding density is, however, still
around fi ve territories/100 km2 (Tornberg 2001), but
due to a yearly average occupancy rate of about 80%,
real breeding density falls to four pairs/100 km2 and
recently even lower (R. Tornberg, unpubl. data). For
comparison, densities in central and southern Europe
tend to be higher but varying considerably depending
on the area, e.g., in northwest Germany from 3.6–7.4
pairs/100 km2 (Krüger and Stefener 1996) and in cen-
tral Poland from 9–13.9 pairs/km2 (Olech 1998).
Physiologically, goshawks are able to breed as
yearlings. In reality this takes place in females but
not in males that likely can not provide enough food
for the females during the courtship phase. On the
island of Gotland, males and females entered the
breeding population in the second year (Kenward
et al. 1991). Their proportion among breeders was
<10%. Females did not breed as yearlings due to a
saturated breeding population but had to wait for
vacancies in their second year of life. In western
Finland and in the Oulu area, percentage of females
breeding as yearlings was about 5–10% annually (P.
Byholm and R. Tornberg, unpubl. data).
CLUTCH AND BROOD SIZES
Goshawks start breeding very early in spring;
nest building can be initiated in mild winters and in
good food conditions by late February (Huhtala and
Sulkava 1981). Initiation of nesting is likely con-
nected with the start of breeding by grouse, which is
stimulated by high temperatures (Nielsen and Cade
1990). Start of egg laying takes place in western
Finland around 20 April (Sulkava 1964, Huhtala
and Sulkava 1981, Tornberg 1997, Byholm et al.
2002a). Yearly average clutch size can vary from
2–4 depending on food conditions, usually the avail-
ability of grouse (Byholm 2005). Based on extensive
data from western Finland during good grouse years
in 1960s–1970s mean clutch size was 3.51 (± 0.06,
N = 164; Huhtala and Sulkava 1981). In the vicin-
ity of Oulu, yearly clutch size during poor grouse
years in 1988–2002 varied from 2.9–4.2, ( = 3.59 ±
0.07, N = 148). Consequently, grouse density seems
not to strongly determine the mean clutch size,
although high peaks or deep lows of grouse usu-
ally are refl ected in the clutch size (Sulkava 1964,
Huhtala and Sulkava 1981, Sulkava et al. 1994).
Clutch size declines signifi cantly with the postpon-
ing of the start of egg laying (Huhtala and Sulkava
1981, Sulkava et al. 1994, Byholm et al. 2002a). In
lowland Britain, clutch size seem to higher than in
Finland 3.96 (± 0.11, N = 47; Anonymous 1990), but
is, on average, the same in central Poland (3.54, N =
143; Olech 1998).
Brood size in large data sets is always about
0.5–0.6 lower than clutch size due to partial brood
loss (Byholm 2005) Hence, average brood size in
western Finland has varied in the 1950–1970s in
data collected in different localities, from 2.78–3.13
(Huhtala and Sulkava 1981). In the vicinity of
Oulu, during 1988–2002, average brood size was
2.89 (± 0.12, N = 163). Mean brood size for whole
Finland during 1989–1998 was 2.79 (± 0.05, N =

STUDIES IN AVIAN BIOLOGY
148 NO. 31
2,822; Byholm et al. 2002a). Hence, it seems that
mean brood size has not declined since the 1950s
in Finland although numbers of main prey, grouse,
have decreased remarkably since then (Lindén and
Rajala 1981). This is not necessarily surprising
because alternative prey (migratory birds) is richly
available during summer. Greatest mortality in gos-
hawks’ broods takes place soon after hatching when
the youngest nestling in the brood usually dies or
one egg does not hatch (Sulkava 1964, Huhtala and
Sulkava 1981, Anonymous 1990, Byholm 2005).
Mortality is higher in nests originally having four
eggs than those having three eggs (Byholm 2005).
Mortality is relatively low during the post-fl edging
dependence period. Interestingly, goshawk brood
size is spatially well synchronized over large area up
to over 300–400 km (Ranta et al. 2003). In England,
brood size based on a small data set collected over
several years was somewhat lower than in Finland
2.76 (± 0.16, N = 45; Anonymous 1990), but higher
in central Poland 2.91 (N = 400; Olech 1998).
Goshawk nestlings leave the nest at the age of
44–46 d (Kenward et al. 1993a) and reach indepen-
dence at the age of 75–82 d (Kenward et al. 1993b).
In the vicinity of Oulu, where mean hatching date is
1 June, young goshawks leave their nesting territory
around mid-August. Reaching independence means
a jump in the mortality of young goshawks, which
continues high during the fi rst winter as illustrated by
the accumulation of dead goshawks to the Zoological
Museum of the University of Oulu (Fig. 3). This has
been verifi ed by a large radio-tracking project on
Gotland (Kenward et al. 1999). Adult mortality
peaked in late winter-early spring (Haukioja and
Haukioja 1971).
OCCUPANCY RATE AND PRODUCTIVITY
In birds of prey using serviceable breeding
sites—old stick-nests, cliffs, or nest-boxes—
occupancy rate counted as breeding sites used per
sites available gives a reasonable estimate of size
of the breeding population (Forsman and Solonen
1984). Populations of birds living in stable and pre-
dictable conditions can also be stable from year to
year (Hunt 1998). Goshawks living in northern areas
and having high winter mortality very seldom fi ll
serviceable breeding sites for long periods. In south-
ern Finland, mean occupancy rate was 68% in an
8-yr study of around 30 territories checked annually
(Lindén and Wikman 1983). In a long-term study
carried out in western Finland during 1979–1996,
mean occupancy rate was 45% (Hakkarainen et al.
2004, Tornberg et al. 2005). In this study, the number
of territories checked annually increased from 16 to
173 during the study. In the vicinity of Oulu, the cor-
responding fi gure was 83% during 1987–2003; num-
ber of territories annually checked increased from
FIGURE 3. Number of Northern Goshawks accumulated monthly by the Zoological Museum of University of Oulu,
1964–2003.

NORTHERN GOSHAWK IN FENNSCANDIA—Tornberg et al. 149
10–32. In the study area at Oulu, occupancy rate
declined strongly but remained stable in study area
of western Finland (Hakkarainen et al. 2004) during
the study years. Declining occupancy rates during a
long study may depend on the improving familiarity
of the research area in the course of investigation
when less used territories are discovered. It is not
surprising that in the western Finland study area the
number of occasionally used territories increased
during the study years (Hakkarainen et al. 2004).
Productivity is measured as young produced per
breeding pair, i.e., per occupied territory (Steenhof
1987). Productivity in the previous studies varied
from 1.8–2.1. Annual variation was substantial, being
highest in southern Finland (C.V = 32.3%) and low-
est in western Finland (C.V = 17.1%). In the vicinity
of Oulu, C.V. was 22.6%. Productivity on Gotland
during 1977–1981 was much lower at 1.36 young/
occupied territory (Kenward et al. 1999). Even farther
south in northwest Germany, Kruger and Stefener
(1996) reported productivity to vary between 0.5–1.8.
In central Poland, in a long-term study, it was fairly
high at 2.25 (Olech 1998). Obviously, goshawks tend
to compensate for higher mortality by man/natural
causes or both in the north and east by higher produc-
tivity (see Kenward et al. 1991)
NUMERICAL RESPONSE
It is not surprising that breeding output as esti-
mated by average clutch and brood sizes follows the
population density of grouse. Breeding attempts of
goshawks failed almost totally after a very cold win-
ter and poor grouse population in western Finland
in 1956 (Linkola 1957, Sulkava 1964). No obvious
differences were found in the mean clutch and brood
sizes between good grouse years in 1950–1970s and
relatively poor grouse years in the 1980–1990s. Yet,
yearly clutch and brood sizes tend to follow grouse
population fl uctuations (Lindén and Wikman 1980),
usually with a 1-yr time lag (Sulkava et al. 1994).
Connection between grouse population density and
goshawks’ breeding output seems to be strongest
in central and zone of the boreal forest (Lindén and
Wikman 1980, Sulkava et al. 1994, Tornberg et al.
2005) while it seems to disappear in southern zone of
boreal forest (Lindén and Wikman 1983). In Norway,
breeding success of goshawks seems not to follow
grouse fl uctuations but may be indirectly linked with
multi-annual vole cycles (Selås and Steel 1998).
Clutch and brood sizes may often poorly represent
the dynamics of the whole goshawk population. We
did not fi nd any obvious correlation between brood
size of goshawks and grouse density in Oulu area
during the 1990s. Better estimates in this sense may
be population productivity and occupancy rate that
also take into account the failed pairs (Steenhof 1987).
In the Oulu area, population productivity closely fol-
lowed the density variation of grouse until 1996 (r =
0.863, N = 10, P <0.001), but thereafter the connection
disappeared (Fig. 4). Yet, the overall correlation dur-
ing the whole study period was signifi cant (r = 0.558,
N = 17, P <0.05). In addition, a positive correlation
(r = 0.549, N = 19, P<0.05) between grouse density
and territory occupancy rate of goshawks with a 2-yr
lag was found in western Finland in a long-term study
during 1979–1996 (Tornberg et al. 2005). Similar
relationship seems to prevail between winter cen-
suses of goshawks and multi-annual fl uctuations of
forest grouse (Tornberg and Väisänen, unpubl. data).
However, we found no correlation between occu-
pancy rate of goshawks and density indices of grouse
in the Oulu area with any time lags. A reason for these
discrepancies in brood size and occupancy rates may
be the decline of grouse populations and disappear-
ance of the multi-annual cycles in grouse population
fl uctuation (see Fig. 2).
GOSHAWK PREDATION ON GROUSE—TOTAL
RESPONSE
Pooling functional and numerical responses
yields a total response or kill rate of the predator
to varying densities of prey. Predation impact is
defi ned as a function of kill rate to density of prey.
Further, predation rate is obtained when predation
impact is plotted against density of prey. (Keith
et al. 1977, Lindén and Wikman 1983; Korpimäki
and Norrdahl 1989, 1991). Three studies of the
goshawk’s predation impact on woodland grouse
(Lindén and Wikman 1983, Widén 1987, Tornberg
2001) and one study on pheasants (Kenward 1977,
Kenward et al. 1981a) have been carried out in
Fennoscandia. Lindén and Wikman (1983) reported
that goshawks took 12% of the adult Hazel Grouse in
southern Finland during the 4-mo breeding season;
on an annual basis predation impact would be 36%.
In central Sweden, territorial goshawks killed 14%
of Black Grouse males and 25% of females during
the breeding season, but during winter, predation on
grouse was negligible (Widén 1987). A grouse study
carried out in the same area by radio-tagged birds
gave almost the same mortality estimate (20%) for
Black Grouse females during the breeding season
(Angelstam 1984). In northern Finland, goshawks
prey on all four available grouse species (Tornberg
2001). Based on a recent predation estimate for the
breeding season, goshawks killed 22% of Willow

STUDIES IN AVIAN BIOLOGY
150 NO. 31
Grouse, 16% of Hazel Grouse, 9% male Black
Grouse, 14% of female Black Grouse, 4% female
Capercaillies, and 7% of grouse chicks. On an
annual basis, numbers for adult grouse were almost
the same (Tornberg 2001). It seems that the goshawk
is the most important predator of adult grouse dur-
ing the breeding season accounting for 30–50% of
adult grouse mortality excluding large Capercaillies
(Widén 1987, Tornberg 2001). Impact of winter pre-
dation by the goshawks on woodland grouse is still
unresolved due to incomplete and small data sets on
winter diet, but in most years it might be as large as
mortality during the breeding season.
Goshawks kill substantial numbers of pheasants
in southern Fennoscandia, their predation impact
being strongly density dependent. Where wild
pheasant stocks prevail, loss by goshawk predation
was 55% for females and 18% for males, but where
captive-born pheasants were released, losses were
substantially higher, goshawks were responsible
for 90% of kills during the winter (Kenward 1977,
Kenward et al. 1981b). Predation studies usually
neglect the impact by non-breeders, which can be
considerable in years of increasing and high predator
populations (Rohner 1996). Healthy raptor popula-
tions should minimally contain around 30–40% non-
breeders (Hunt 1998).
Elsewhere, we (Tornberg 2001, Tornberg et al.
2005) have suggested that goshawk predation may
have a destabilizing effect on grouse population
due to obvious time lags in numerical response of
goshawks to varying grouse densities and a high
proportion of grouse in the diet also during poor
grouse years (Fig. 5). In this sense, the predation
impact of goshawks on forest grouse appears to be
similar to the predation impact of Gyrfalcons (Falco
rusticolus) on ptarmigans (Lagopus spp.) in Iceland
(Nielsen 1999). The lagging numerical response of
goshawks to varying densities of grouse is obviously
different from numerical responses of various vole-
eating owls and raptors to multi-annual vole cycles
in Fennoscandia, because their numerical responses
track varying vole densities without obvious time
lags (Korpimäki 1985, 1994). In conditions more
natural than the present in northern European boreal
forests, goshawks may have had a remarkable role in
driving grouse cycles.
MOVEMENTS OF GOSHAWKS
The goshawk is regarded as a resident raptor but
individuals in their fi rst year of life are mobile and
some of them show directional movement southward
in autumn and northward in spring (Marcström and
Kenward 1981b). These movements can take young
birds >1,000 km from their natal areas (Sulkava
1964, Saurola 1976, Halley 1996). However, most of
the birds do not orient systematically southward but
disperse randomly around their natal area (Sulkava
1964, Saurola 1976, Marcström and Kenward
FIGURE 4. Productivity of the Northern Goshawk population and grouse density of the previous autumn in the Oulu area
from 1987–2003.

NORTHERN GOSHAWK IN FENNSCANDIA—Tornberg et al. 151
1981b, Halley 1996, Byholm et al. 2003). Sulkava
(1964) showed that dispersal distances of the young
goshawks were negatively related to abundance of
grouse in the natal area. Byholm et al. (2003) con-
fi rmed this fi nding recently and also showed that
birds in late broods dispersed farthest, especially
males. Dispersal distances also seem to be related to
sex and age. Juvenile males tend to be most mobile
(Kenward et al. 1981b, Marcström and Kenward
1981, Neideman and Schönbeck 1990, Byholm et
al. 2003, but see Halley 1996). Median distance for
male hawks banded as nestlings and found dead dur-
ing the fi rst winter after reaching the independence
was 80 km but only 34.5 km for females (N = 213;
Byholm et al. 2003). In Norway, however, females
moved more (median 109 km) than males (median
68.5 km, N = 77; Halley 1996). Hawks found dead in
adult plumage had moved less far than those found
as juveniles (Halley 1996, Byholm et al. 2003).
Because birds could not be tracked, this may hint
at return movements to the natal area after matu-
rity (Halley 1996). Distance traveled by adults of
both sexes tends to be the reverse of that found in
juveniles. A similar tendency has been found also in
radio-tracking studies (Kenward et al. 1981b) and
when trapping and banding hawks after the breed-
ing season (Marcström and Kenward 1981). Figure 6
illustrates the spread of juvenile goshawks banded
as nestlings in the Oulu area. Most birds are found
on the coastline of Bothnia Bay, Baltic Sea. Long-
distance travelers seem to have moved in various
directions.
Higher mobility of juvenile males than females
is also apparent in trapping results from southern
Sweden (Neidemen and Schönebeck 1990). A reason
may be that food supply for males is lower than that
for females. Tornberg (2000) estimated that food base
of females is three times larger than that of males,
mainly due to mountain hares (weighing 3–4 kg)
and Capercaillie males (weighing 4 kg), prey that is
nearly out of the males’ hunting capacity. Kenward et
al. (1993b) found that juvenile males moved further
than females on Gotland when young rabbits reached
full size. The food scarcity hypothesis is also sup-
ported by the trapping results in southern Sweden
that showed an increase in proportion of males in
years 1984–1987 when grouse population numbers
were exceptionally low (Fig. 2). Juvenile males also
starve more often than females (Tornberg et al. 1999,
Sunde 2002). Southward migrations of goshawks in
North America are related to food scarcity, especially
during low phases of the 10-yr population cycles of
snowshoe hares (Lepus americana; Keith and Rusch
1989). There, however, differences between the
FIGURE 5. Predation rate by the Northern Goshawk on adult grouse in the Oulu area, 1989–1998 (redrawn from Tornberg
2001).

STUDIES IN AVIAN BIOLOGY
152 NO. 31
sexes in the length of migration, has not been docu-
mented.
In adult goshawks, males seem to be the more
philopatric sex (Kenward et al. 1981b, Widén
1985b, Byholm et al. 2003), a fact common in many
raptors (Newton 1979a, Korpimäki et al. 1987,
Korpimäki 1993). Higher philopatry in males might
be connected to their more active role in territory
defense and brood rearing (Newton 1979a, Byholm
et al. 2003). Also, males trapped as adults are less
reluctant to leave their home ranges than females
(Kenward et al. 1981a, Widén 1985b). In the Oulu
area, one breeding radio-tagged female deserted
her family during the fl edging period of her young
and shifted to nest in a different territory in the next
year. The fl edglings were then successfully reared by
the male. Another female trying to nest near the city
dump of Oulu in 1994 was found 2 yr later 100 km
south eaten by an Eagle Owl (Bubo bubo). Fairly
little is still known about site and mate tenacity in
breeding goshawks in Europe and further study is
badly needed.
One may argue that dispersers moving farther
are in a poorer condition than those moving less.
Investigating movements of trapped and either
banded or radio-tagged hawks in Sweden did not
explain the length of the movement or site tenacity of
the trapped birds (Kenward et al. 1981a, Marcström
and Kenward 1981b, Widén 1985b). In fact, males
that were generally in poorer condition in late win-
ter were more reluctant to leave the study area than
females (Widén 1985b).
HOME RANGE
Juvenile goshawks are very mobile during their
fi rst year of life; post-fl edging dispersal can take
them >1,000 km from their natal areas but most
of the young hawks settle within 100 km. Young
hawks tend to maintain home ranges before settling
in the fi nal breeding territory (Halley et al. 2000).
Those juvenile hawks that were radio-tracked dur-
ing November–December usually stayed near the
trapping site in central Sweden and northern Finland
(Kenward et al. 1981b, Tornberg and Colpaert 2001).
Winter ranges of different goshawk individuals can
overlap extensively. This happens especially in
areas with high food supply like near release pens of
pheasants (Kenward 1977). So, wintering goshawks
seem not to defend their home ranges. This was the
FIGURE 6. Finding sites of the juvenile Northern Goshawks banded as nestlings in the the Oulu area, 1962–2002. Data
obtained from the Ringing Centre of the Natural History Museum of the University of Helsinki.

NORTHERN GOSHAWK IN FENNSCANDIA—Tornberg et al. 153
case also in Oulu (Fig. 7) where breeding birds did
not try to displace visitors. Some observations of
resident breeders hint that they know the core areas
of their neighbors and avoid visits there.
Winter range sizes have been found to be related
to landscape structure. In farmland areas of Sweden,
range size correlated negatively with the amount of
forest edge in the range (Kenward 1982). Because
most of the kills took place near woodland edges,
range size seems to relate negatively to the amount
of good habitat, i.e., forest edge. Correspondingly,
range size correlated negatively with the amount of
mature forest, a preferred hunting habitat, in boreal
forests of northern Finland (Tornberg and Colpaert
2001). Range size seems to respond fl exibly either
to the quantity or the quality of the food resource.
Hawks that kill mostly large prey or live in areas
with high food supply have the smallest ranges
(Kenward 1982, Nygård et al. 1998). It is no wonder
that juveniles being less experienced hunters than
adults have larger ranges (Kenward et al. 1981b).
One might also expect larger winter home ranges for
males that have a narrower food base than females.
However, in boreal forests of central Sweden males’
range size (5,110 ha, maximum polygon) was even
slightly smaller than that of females’ (6,179 ha). In
this study, however, goshawks fed mainly on squir-
rels that might be more suitable prey for smaller
males than larger, less agile females (Widén 1987).
In the Oulu area, average winter range size (maxi-
mum polygon) was 7,091 ha (± 3,935 ha, N = 9) for
males and 5,710 ha (± 664 ha, N = 15) for females,
but the difference was not statistically signifi cant.
HABITAT CHOICE
Goshawks are known to be old-forest special-
ists. This is, however, largely based on studies of
the characteristics of the breeding habitats (Widén
1997, Penteriani 2002). Radio-tracking studies have
shed light over the habitat use of goshawks outside
and during the breeding season. As stated above,
FIGURE 7. Winter ranges of the Northern Goshawks near Oulu in the winter 1992–1993. Ranges marked as follows:
1. adult female (breeding in the area), 2. adult male, 3. adult male (breeding in the area), 4. adult female (breeding 15 km
southwest from the area), 5. adult female (breeding near dump site), and 6. juvenile female.

STUDIES IN AVIAN BIOLOGY
154 NO. 31
goshawks favored forest edges in farmland areas
of central Sweden. Yet, in a more forest-dominated
area radio-tagged birds thrived best in large patches,
avoiding edges (Widén 1989). They preferred mature
forests over younger stands. Correspondingly, gos-
hawks also preferred mature forests in Oulu, but
rather average sized patches that hint at favoring
edges as hunting habitats. Goshawks used young for-
ests proportionately to their availability but avoided
open areas (Tornberg and Colpaert 2001). Because
locating a goshawk is possible only when the bird is
perched, it is impossible to know how much they fl y
over open terrain. Goshawks hunt with a short-stay,
perched technique, perching 3–5 min and then fl ying
200–300m to a new perch (Widén 1984).
SURVIVAL AND CAUSES OF DEATH
A large number of banded hawks and good success
at recapturing them have enabled reliable estimates
of goshawk survival. Haukioja and Haukioja (1971)
estimated the mortality of goshawks to be 63% in the
fi rst year assuming that 60% of the bands found were
returned, 33% in the second year, 20% in the third,
and stabilizing at around 10% in older age classes.
Using a larger data set, Saurola (1976) estimated cor-
responding numbers as 64%, 35%, 18%, and 15%. It
must be remembered that goshawks in Fennoscandia
were under heavy persecution in 1960s–1970s with
5,000–6,000 goshawks, a remarkable proportion
of the annual production, being killed annually by
humans in Finland alone (Moilanen 1976). Analyses
based on band recoveries may be biased, however,
because young age classes are likely to be found eas-
ier than older specimens. Moreover, during the time
when shooting was allowed, hawks killed by humans
were likely to be overrepresented in total recoveries
and young hawks prevailed among those being shot.
Kenward et al. (1991, 1999) found in a large radio-
tracking study on Gotland that 47% of the band recov-
eries were from killed hawks, whereas only 36% from
radio-tagged birds. In addition, radio-tagged hawks
showed an unbalanced mortality in young age classes
in relation to sex—by 1 April, 46% of the males had
died in their fi rst year but only 31% of the females. In
the second year, still more males (41%) than females
(29%) died, but in older age classes mortality was bal-
anced being 21% for both sexes.
Telemetry data collected in the Oulu area during
1991–1995 (N = 26; Tornberg and Colpaert 2001)
were analyzed along with new data on eight tagged
birds from the winters 1999–2003 (four adult males,
one yearling male, two adult females, and one juve-
nile female) to get a survival estimate for winter
months from 10 November to the end of February.
We pooled the data over the years using a staggered
entry method (Pollock et al. 1989). Mortality in
adults (N = 26, males and females together) was 37%
and for juveniles, 81% (N = 8). Because this method
is very sensitive to small sample sizes, our estimate
for juveniles is probably unreliable. The estimate for
adults is very high compared to those obtained from
band recoveries or telemetry data collected in more
southern areas but is not necessarily unrealistic.
Annual mortality may be a bit higher than estimated
for winter months because natural mortality of adult
hawks can still be high in March and April (Fig. 3).
Autopsies of naturally dying hawks on Gotland
revealed that starvation was the most important cause
of death (37%; Kenward et al. 1991), 33% of hawks
died of trauma, and 22% of the combination of dis-
ease and starvation. Based on autopsies of goshawks
brought to the Zoological Museum of the University
of Oulu, 35% of hawks had died of starvation, 25%
from collisions, 15% from a combination of trauma
and starvation, and only 13% from shooting (N =
165; Tornberg and Virtanen 1997). Among banded
hawks, the most important cause of death in the
1960s–1970s was killing by humans (83%; Saurola
1976). Similarly, shooting was the most common
cause of death in Norway; before protection about
50% of birds found had been shot. After protection
this cause of death fell to 5% (Halley 1996). After full
protection of goshawks in 1989 in Finland, killing by
humans declined but starvation may have increased
due to intensifi ed competition for food. Earlier, hawks
prone to starve were often shot when they approached
human settlements (Haukioja and Haukioja 1971).
Hence, the cessation of shooting did not necessarily
increase the number of young hawks because starva-
tion among juveniles may have increased.
POPULATION STATUS AND TRENDS OF THE
GOSHAWK IN FENNOSCANDIA
It is reasonable to argue that decline of a prey
population induces a decline in the population of
its predators. This typically concerns specialized
predators (Begon et al. 1996) because generalists
can switch to another prey if one prey type declines.
The Northern Goshawk could be considered a gen-
eralist predator based on the wide spectrum of prey
species in its diet. Because most diet studies have
been performed during the breeding season when the
greatest variety of suitable prey species, especially
vulnerable juveniles, is available, food niche can be
very wide. More focus should be directed to winter
when availability of prey is more restricted.

NORTHERN GOSHAWK IN FENNSCANDIA—Tornberg et al. 155
Recent estimates show that goshawk still is
one of the most common raptors in Fennoscandia.
Several studies carried out in different localities in
Fennoscandia, however, hint at a decline in breed-
ing densities of goshawks. Widén (1997) reviewed
nine studies and found a decline in eight of them.
Selås (1998a) reported a decline in the breed-
ing density in southern Norway from nine pairs/
100 km2 in the 1950s to three pairs to the 1980s
but a slight increase to four pairs/100 km2 in the
mid-1990s. Recently, density has fallen back to the
previous three pairs/100 km2 (Selås 1998b, Selås,
pers. comm.). In central Norway, breeding density
in the 1990s was very low, only one pair/100 km2
(Nygård et al. 1998). It is still diffi cult to evaluate
whether declines reported in some studies indicate
only local declines or whether they indicate a more
general trend. A Finnish country-wide monitoring
program of breeding populations of birds of prey
which was initiated in1982 does not indicate declin-
ing density until the mid-1990s (Väisänen et al.
1998), even though during the 1990s a slight declin-
ing trend was detected (Björklund et al. 2002). The
Swedish monitoring project from 1975 onward for
winter and summer censuses show a 20% declining
trend for winter but a slight increase for summer
densities (Svensson 2002). In Sweden and Norway,
increasing numbers since the 1980s are, however,
expected and obvious as Selås (1998a) has pointed
out. This is due to a sarcoptic mange epidemic in
red foxes (Vulpes vulpes) that caused fox numbers
to crash and caused a corresponding increase in
grouse numbers (Lindström et al. 1994). Hence,
monitoring initiated in the 1970s–1980s does not
necessarily reveal the long-term development of
the goshawk population. Goshawk populations in
central and south Europe seem to be more or less
stable or even increasing (Kruger and Stefener
1996, Olech 1998)
In the Oulu area, occupancy rate of the goshawk
population showed a strong negative trend during
the 1990s (Fig. 8). We analyzed the population
development by Moffat’s equilibrium model (Hunt
1998) which assumes a fi xed number of service-
able breeding sites. The model further assumes that
juveniles start breeding in their second year. Simply
by altering productivity of breeders and survival of
juveniles, sub-adults, and adults, the model predicts
future structure and development of the population.
We used a series of survival values of 63% for adults
(obtained from telemetry data), adjusting survival
values for sub-adults (51%) and for juveniles
FIGURE 8. Occupancy of Northern Goshawk territories in the Oulu area (thick line) and simulations of the number of
breeders with different survival rates by Moffat equilibrium model (Hunt 1998). Uppermost line denotes a survival value
of 0.7 declining by 0.1 in each step.

STUDIES IN AVIAN BIOLOGY
156 NO. 31
(46%) according to estimates obtained from data
by Kenward et al. (1999). We then modeled survival
estimates by increasing each age category by1%.
We set population productivity at two fl edglings/
breeding pair. By the lowest series of values, the
decline was steeper than observed which hints that
the survival values used obtained from the telemetry
study are too low. Using values 4% higher yielded a
model that matches the observed line (Fig. 8). With
these values, the population does not contain non-
breeders which could explain the poor correlation
between the occupancy rate of goshawk territories
and grouse density because non-breeders are capable
of responding quickly to changes in prey popula-
tion. It seems that productivity is not a problem in
a goshawk population but rather the poor survival of
adults (Hunt 1998). Using values obtained from band
recoveries in Finland (82%, 65%, and 36%) and pro-
ductivity of two fl edglings/pair gives a balanced
breeding population containing 20% non-breed-
ers. The goshawk population on Gotland remained
stable, adjusted by lower proportion of the females
breeding annually (40%) than the males (70%),
which means that proportion of non-breeders of the
breeders was around 40–50%.
CONCLUSIONS
Recently, a lot of debate has centered on rea-
sons for changes in avian fauna of boreal forests in
Fennoscandia (Haila and Järvinen 1990, Mönkkönen
et al. 1999). The general conclusion derived from
ornithological reports has been that old-forest spe-
cies have declined and species living in young suc-
cessional stages have increased or remained stable
(Väisänen et al. 1986). This is considered to be due
not only to the decline of the mature-forest stands but
also to the fragmentation yielding patches too small
to maintain meta-populations of certain old-forest
specialists (Andrén 1994). The goshawk has been
considered as an old-forest bird based on the nest-
site selection (Penteriani 2002). Widén (1997) con-
cluded that the goshawk has suffered from forestry
because of the decrease of its main hunting habi-
tat—old forests. Young successional stages of boreal
forests, although basically maintaining higher grouse
densities, are often too dense for successful hunting
of the goshawk (Beier and Drennan 1997). Hence,
Widén (1997) considers that habitat degradation is a
more important reason for decline of goshawks than
decline in the prey supply as such. It is, however,
quite evident that the supply of the main prey, forest
grouse, has declined.
In Finland, where grouse counts have been made
since the mid-1960s, decline in all forest grouse spe-
cies has been >50% (see Fig. 2). Modern forestry
with extensive clear cuts, draining of the peat land
bogs, and construction of a dense network of forest
roads have had negative impacts on forest grouse
(Kurki et al. 1997). Clear-cuts may have increased
grasslands that maintain voles and their predators.
During crashes of vole populations, small mammal
predators switch to hunting grouse chicks and thus
lower the productivity of grouse (Angelstam et al.
1984, Henttonen 1989). Removal experiments of
mammalian predators have resulted in higher grouse
populations or at least higher reproductive rate
compared to control areas (Marcström et al. 1988,
Kauhala et al. 2000). We conclude that shrinkage
in the area of mature forests does not explain the
observed negative trends in the goshawk population
per se, but rather the availability of suitable sized
prey during the non-breeding season. Goshawks are
able to live in areas where forest cover is <20% of the
area but where enough prey is accessible (Kenward
1982). In the Oulu area, goshawks preferred fairly
small patches of forests. Surprisingly, the compo-
sition of the winter diet is close to that found in
farmland areas of central Sweden with the difference
that grouse replaced the pheasants (Kenward et al.
1981a; Table 2). Habitat of kill sites did not differ
much from that of the habitat composition available
(Tornberg and Colpaert 2001).
Forest fragmentation has caused a decline in
forest grouse and perhaps also in red squirrels,
whereas it may have increased mountain hare
numbers. Comparisons of mountain hare densities
between Finland and Russian Karelia show a three-
fold higher hare population in Finland compared
to Russian Karelia where forest stands are mostly
at mature stage (Lindén et al. 2000). As found in
winter diet studies, females can but male goshawks
unlike cannot kill full-grown mountain hares. This
has led to a curious situation where females may
have benefi ted from forest fragmentation but males
suffered. This appears to result in a higher starvation
risk and poorer winter condition in male goshawks
(Widén 1985b, Tornberg et al. 1999, Sunde 2002).
It may also explain why breeding output expressed
as clutch and brood sizes do not match well with
the density fl uctuations of grouse. Females in good
condition in spring can lay eggs with a minimal aid
from the males. Therefore, recent changes in forest
structure may have even affected their morphology.
Tornberg et al. (1999) found, based on museum
material from the last 40 yr, that adult males have

NORTHERN GOSHAWK IN FENNSCANDIA—Tornberg et al. 157
become smaller and females larger. This change was
more on the outer morphology (body, wing, tail,
and tarsus length) than bone length. This might be
explained by dietary changes caused by a general
decline of grouse populations—females have found
larger alternative prey than males. Another interest-
ing adaptation that probably originates from a tighter
dependence of males on grouse, appears as a chang-
ing sex ratio in goshawk broods as a function of
grouse density (Byholm 2003). Goshawk pairs pro-
duce signifi cantly more males in good grouse years
compared to poor years. This might be a compensa-
tive response for higher juvenile mortality of males
induced by natural selection.
When evaluating the conservation needs for a
declining raptor species, focus should not be on
only one apparently important fact, but on a wider
scale, e.g., how the change in habitat has affected
the food supply. One must also realize when the
food supply is a limiting factor, it is not likely to
be limiting during the breeding season at northern
latitudes. Kenward (1996) presumes that problems
faced by the goshawks in the sub-boreal region of
North America might be due to poor food supply in
winter. Protection of the goshawks has not increased
goshawk numbers. It can be possible that nowadays,
when more young probably are entering the winter
than during the years when many juveniles were
killed by humans, intra-specifi c competition for food
in goshawk populations has intensifi ed. This may
lead to more starving young birds but also a weaker
winter supply for adults and poorer breeding perfor-
mance in the next spring (Haukioja and Haukioja
1971). In a specialist predator-prey interaction, a
decline of the predator may lead to an increase in
prey population. In goshawk-grouse systems, this
does not necessarily happen these days because
increased impact by mammalian predators harvests
grouse populations independently of their density
(Angelstam et al. 1984, Marcström et al. 1988,
Korpimäki and Norrdahl 1997). In fact, mammalian
predators and goshawks are competing for a com-
mon resource, grouse, which is of vital importance
for the goshawks but not necessarily for mammalian
predators (Selås 1998a). Modern forestry improves
the conditions of mammalian predators and at the
same time harms forest grouse and the predators
dependent on them. All in all, habitat restoration is
the ultimate solution for the sustainable populations
of forest grouse and goshawks.
Future research effort should be directed to
winter ecology of goshawks. Topics like: (1) win-
ter food supply, (2) predation rates on the most
important prey species, (3) hunting habitats with
precise data on kill sites, (4) movements, survival,
and causes of death of different age classes, and (5)
relationships to competitors, should be investigated
with modern fi eld techniques. In addition, we badly
need individual-level studies on goshawks during
both the breeding and non-breeding seasons in
boreal forests. For example, it could be important
to know how the reproductive effort of individual
pairs and members of pairs varies in relation to tem-
poral and spatial density fl uctuations of main prey,
and how sexual differences in the main food supply
induced by modern forestry practices (benefi cial
for females, costly for males) affects reproduc-
tive effort, division of duties during the breeding
season, and reproductive success of individual
goshawks.
ACKNOWLEDGMENTS
This paper is dedicated to the pioneer investiga-
tor of goshawks and other Fennoscandian raptors and
owls, S. Sulkava, Department of Zoology, University
of Oulu. We are grateful for valuable comments that
V. Selås, V. Penteriani, and S. Sulkava made on the
manuscript. We further thank the Finnish Game
Research Institute and Ringing Centre of Finnish
Museum of Natural History for data on grouse densi-
ties and ringed goshawks.