ORIGINAL
ARTICLE
Patterns of density, diversity, and
the distribution of migratory strategies
in the Russian boreal forest avifauna
Russell Greenberg*, Anna Kozlenko, Matthew Etterson and
Thomas Dietsch?
Smithsonian Migratory Bird Center, National
Zoological Park, Washington, DC, USA
*Correspondence: Russell Greenberg,
Smithsonian Migratory Bird Center, National
Zoological Park, Washington, DC 20008, USA.
E-mail: greenbergr@nzp.si.edu
Present address: 3760 Eastway Road, South
Euclid, OH 44118, USA.
Present address: US Environmental Protection
Agency, Mid Continent Ecology Division, 6201
Congdon Boulevard, Duluth, MN 55804, USA.
?Present address: Center for Tropical Research,
University of California Los Angeles, Los
Angeles, CA 90095-1469, USA.
ABSTRACT
Aim Comparisons of the biotas in the Palaearctic and Nearctic have focused on
limited portions of the two regions. The purpose of this study was to assess the
geographic pattern in the abundance, species richness, and importance of
different migration patterns of the avifauna boreal forest of Eurasia from Europe
to East Asia as well as their relationship to climate and forest productivity. We
further examine data from two widely separated sites in the New World to see
how these conform to the patterns found in the Eurasian system.
Location Boreal forest sites in Russia and Canada.
Methods Point counts were conducted in two to four boreal forest habitats
at each of 14 sites in the Russian boreal forest from near to the Finnish border
to the Far East, as well as at two sites in boreal Canada. We examined the
abundance and species richness of all birds, and specific migratory classes,
against four gradients (climate, primary productivity, latitude, and longitude).
We tested for spatial autocorrelation in both dependent and independent
variables using Moran?s I to develop spatial correlograms. For each migratory
class we used maximum likelihood to fit models, first assuming uncorrelated
residuals and then assuming spatially autocorrelated residuals. For models
assuming unstructured residuals we again generated correlograms on model
residuals to determine whether model fitting removed spatial autocorrelation.
Models were compared using Akaike?s information criterion, adjusted for small
sample size.
Results Overall abundance was highest at the eastern and western extremes of
the survey region and lowest at the continent centre, whereas the abundance of
tropical and short-distance migrants displayed an east?west gradient, with
tropical migrants increasing in abundance in the east (and south), and short-
distance migrants in the west. Although overall species richness showed no
geographic pattern, richness within migratory classes showed patterns weaker
than, but similar to, their abundance patterns described above. Overall abundance
was correlated with climate variables that relate to continentality. The abundances
of birds within different migration strategies were correlated with a second
climatic gradient ? increasing precipitation from west to east. Models using
descriptors of location generally had greater explanatory value for the abundance
and species-richness response variables than did those based on climate data and
the normalized difference vegetation index (NDVI).
Main conclusions The distribution patterns for migrant types were related to
both climatic and locational variables, and thus the patterns could be explained
by either climatic regime or the accessibility of winter habitats, both historically
and currently. Non-boreal wintering habitat is more accessible from both the
western and eastern ends than from the centre of the boreal forest belt, but the
Journal of Biogeography (J. Biogeogr.) (2008)
? No claim to original US governement works www.blackwellpublishing.com/jbi 1
Journal compilation ? 2008 Blackwell Publishing Ltd doi:10.1111/j.1365-2699.2008.01954.x
INTRODUCTION
Increasingly, a biogeographical perspective has been brought to
the classic ecological question of what determines the distri-
bution and abundance of organisms. Broad comparisons
between regions, or even entire continents, can illuminate
processes and patterns that cannot be discerned in more local
studies. However, such broad comparisons may themselves be
limited because of the restricted geographic scope of data sets.
Such is the problem that faces students of the distribution of
faunas in temperate forests. Although considerable research
has focused on describing patterns and hypothesizing processes
underlying the distribution of temperate-zone faunas in both
the Nearctic and the Palaearctic, these analyses have often
suffered from incomplete coverage, particularly of the Palae-
arctic east of Europe (Haila et al., 1987; Morse, 1989;
Mo?nkko?nen, 1994; Mo?nkko?nen & Viro, 1997; Bo?hning-Gaese,
2005).
For example, a number of studies have been published in
the past few decades that explore patterns of the species
richness and abundance of birds with different migration
patterns (residents, within-Temperate Zone, and tropical
migrants in Palaearctic and Nearctic forests: Herrera, 1978;
Helle & Fuller, 1988; Newton & Dale, 1996; Mo?nkko?nen &
Forsman, 2005). Most of the studies, particularly those
that attempt inter-continental comparisons, are based on
studies of the Western Palaearctic (Europe) and the eastern
Nearctic. Studies that look at avifaunal relationships between
different portions of the boreal forest within each conti-
nent are few and are generally qualitative analyses based
on regional species lists (Wiens, 1975; Erskine, 1977;
Preobrajenskaya, 1982; Mo?nkko?nen & Helle, 1989; Haila &
Ja?rvinen, 1990).
The boreal forest is a compelling biome for comparative
studies of the composition and structure of faunal assemblages
(Haila & Ja?rvinen, 1990; Mo?nkko?nen & Welsh, 1994). The
biome is distributed in an almost continuous band across the
northern continental masses and consists of relatively homo-
geneous, low-diversity stands of conifers, birches, and aspens
throughout. Within each continent, the boreal forest belt
extends north-west to south-east, so that equivalent climatic
zones and associated vegetation tend to be at higher latitudes
in the western portions (Haila & Ja?rvinen, 1990). Although
boreal forests show important spatial variation, they arguably
form the most comparable band of vegetation both across and
between continents. Because of the broad distribution and
similar composition of boreal forests, the study of the
distribution of organisms across the boreal zone provides an
excellent opportunity to evaluate the roles that climate,
physical geography, and historical process have in shaping
local assemblages found at different ends of continents and on
different continents.
Boreal forests are both floristically undiverse and climati-
cally harsh, often experiencing extreme differences in temper-
ature between summer and winter. Because of this, most
animal groups must adapt to the extreme seasonality in climate
and resource availability. Birds are often surprisingly abundant
and diverse in boreal habitats because of their mobility and
their ability to exploit the seasonal summer peak in arthropod
abundance, while avoiding the cold harsh winters. It is not
surprising, therefore, that previous studies have often focused
on the distribution and abundance of species with differing
migratory strategies.
Ecologists have attempted to understand the factors that
control the abundance, diversity, and ecological composition
of bird assemblages through the analysis of divergence and
convergence. Studies of divergence examine how particular
shared taxa respond to variation in climate, geography, and
history. Research in convergence examines how largely unre-
lated faunas respond to habitats that are similar in climate and
geography. The boreal forest affords opportunities to use and
integrate both approaches. Despite this, studies of factors that
determine the abundance, diversity and composition of boreal
forest avifaunas are few.
In this paper, we use a data set of simple quantified surveys
from across the Russian boreal forest to: (1) describe the
geographic patterns in abundance, species richness, and the
composition of the avifauna with respect to overall abundance,
species richness, and the distribution of migratory strategies;
(2) determine whether the distribution of birds and their
different migratory strategies is predictably related to conti-
nental position and gradients of climate and primary
productivity; and (3) test the hypothesis that the variation
abundance, diversity of boreal forest birds and the distribution
tropics are most accessible from the eastern end of the Palaearctic boreal zone, in
terms of distance and the absence of geographical barriers. Based on comparisons
with Canadian sites, we recommend that future comparative studies between
Palaearctic and Nearctic faunas be focused more on Siberia and the Russian Far
East, as well as on central and western Canada.
Keywords
Boreal forest birds, community convergence, migratory strategies, Palaearctic,
Russian birds, Russian Far East, Siberian birds, spatial autocorrelation, taiga,
tropical migrants.
R. Greenberg et al.
2 Journal of Biogeography
? No claim to original US governement works. Journal compilation ? 2008 Blackwell Publishing Ltd
of migration strategies is at least as strongly related to location
within a continent as it is to the differences between the
Nearctic and Palaearctic per se. The study, although limited to
data from only 1 to 2 years from each site, represents the only
attempt known to us to discern patterns across the entire
Eurasian boreal zone, and therefore should provide a working
hypothesis for longer-term studies in the future.
MATERIALS AND METHODS
Study sites
Russian Bird Surveys were conducted primarily at Zapoved-
niks or Scientific Nature Reserves (Colwell et al., 1997). Sites
were selected based on whether there were available local
ornithologists to conduct the surveys. Exceptions to this
consist of fieldwork conducted near the Kostromskaya field
station in Kostroma Province, and Myrnoe, Fomka, and Fir
Island sites along the Yenesei River in Central Siberia. The
locations and approximate elevations of the sites are provided
in Appendix S1, and their locations are plotted in Fig. 1. In
addition, studies using similar methodologies were conducted
in the western and eastern Canadian boreal forest (see
Greenberg et al., 1999a,b), and these data are used in these
analyses as well.
Bird surveys
Bird surveys were conducted from 05:00 to 09:00 h on 1?20
June at each of the study sites (see Fig. 1 and Appendix S1), a
period that corresponds to the peak of singing for most boreal
forest species. The data were all collected during the summer of
1994, with the exception of data gathered in eastern Canada
during the 1996 field season and in Central Siberia and
Kostroma during the 1993 and 1995 field seasons, respectively.
All of the observers, with the exception of those for the Central
Siberian and Kostroma study areas, were experienced Russian
ornithologists. Some of the observers at the two excepted study
areas were North Americans, who went through extensive pre-
field and field training on Palaearctic bird identification before
conducting the surveys. At each study site, the major accessible
forest types were identified, and transects were established
along which point counts (Hutto et al., 1986; Petit et al., 1995)
were conducted at 200-m intervals. In this way, up to fifteen
5-min point counts were conducted per morning. All birds
detected visually or audibly within 50 m of the observer were
recorded. For the analysis, we excluded nocturnal birds and
birds flying over the plot. In addition, we did not include data
from waterfowl, shorebirds, or species that are characteristi-
cally nomadic and gregarious (crossbills). For further analysis,
birds were classified as resident (most of the population
remains resident in the surveyed region throughout the year);
tropical migrant (most of the population winters south of the
Tropic of Cancer); and short-distance migrant (most of the
population migrates from the region of the surveys, but
remains north of the Tropic of Cancer). The classification of
the species is presented in Appendix S2.
Climate data, global vegetation index values, and
habitat classification
Climate data consist of 20-year averages obtained from the
nearest meteorological sites to the study sites that could be
found on the International Research Institute for Climate and
Society C (IRI) data base (IRI/LDEO, 2003). This data base
provides monthly average temperatures, precipitation, and the
number of frost-free days. From these data, we derived six
climatic variables: average maximum temperature in July,
average minimum temperature in January, average difference
between the two previous measures, total precipitation per
Figure 1 Locations of study sites in
Russian and Canadian boreal forest.
1, Kivach; 2, Pinejskyi; 3, Pechoro-Ilichskiy;
4, Kostroma; 5, Visimskiy; 6, Malaya Solosva;
7, Myrnoe; 8, Fir Island; 9, Fomka; 10, Bay
Kalskiy; 11, Barguzinskiy; 12, Komsomolskiy;
13, Sichote-Alin; 14, Amur; 15, Mackenzie;
16, Ontario (see Appendix S1 for further
details).
Russian boreal birds
Journal of Biogeography 3
? No claim to original US governement works. Journal compilation ? 2008 Blackwell Publishing Ltd
150?0'W 165?0'W 180?0' 180?0' 165?0'E 150?0'E 
60?0'N 60?0*N 
Arctic Circle 
66?33'39"N 
D?0'W 45?0,W 30?0'W        15?0'W 0?0' 15?0'E 30?0'E 45?0'E 
year, summer precipitation (May?August), and number of
frost-free days. In order to reduce the redundancy and
multicollinearity among the climate variables we used princi-
pal components analysis (PCA) on the six variables listed
above and used the site-specific loadings of the first two
principal components (PC1 and PC2) as climatic explanatory
variables.
Normalized difference vegetation index (NDVI) values were
obtained from the NOAA National Environmental Satellite,
Data, and Information Service (Kineman & Hastings, 1992).
NDVI values are remotely sensed values based on the overall
photosynthetic activity of a habitat. The NDVI has been used
to represent habitat productivity in a number of studies of bird
distribution (Hurlbert & Haskell, 2003; Sze?p & M?ller, 2005;
Ding et al., 2006). Latitude and longitude (to the nearest 0.1)
were entered into the program, and the interpolated monthly
averages (1985?88) were downloaded. We obtained two
parameters from the NDVI data: the maximum monthly
value (usually June or July, NDVImax); and an index of
seasonality, which was calculated as the difference between the
monthly maximum and minimum divided by the maximum
(NDVIratio, based on Hurlbert & Haskell, 2003). This value
ranges from 0 to 1, so we multiplied it by 100 to obtain a value
between 0 and 100. Finally, we ranked habitats for relative
productivity in the following order (from greatest to least):
1 = riparian; 2 = broad-leafed upland (aspen?birch); 3 =
mixed upland (spruce, fir, aspen, birch etc.); and 4 = pine,
larch, black spruce (Erskine, 1977).
Statistical analysis
We calculated the mean number of birds per individual point
count as an index of relative abundance. For each habitat we
derived the Chao I estimate of total species richness based on a
sample size of 50 point counts (Colwell, 2003). We calculated
these same parameters for tropical migrant, short-distance,
total migrant, and resident species within each habitat
surveyed at each site. We thus had 10 response variables
(species richness and number of individuals in each of five
classes: total birds, tropical migrants, short-distance migrants,
total migrants, and residents) that we wished to test against
the four continental gradients described above, namely
climate (PC1 and PC2), primary productivity (NDVImax and
NDVIratio), longitude, and latitude, which resulted in a total of
40 regression equations. Note that all models included habitat
(HAB) as a fixed effect because we were certain that our
response variables would vary by habitat.
Many previous analyses of continental-scale patterns of the
density and species richness of birds (e.g. Diniz-Filho et al.,
2003; Bini et al., 2004) and of other organisms (Selmi &
Boulinier, 2001) have shown spatial autocorrelation both in
the response variables and in common explanatory variables
(e.g. climate and primary productivity). The existence of
spatial autocorrelation may result in negatively biased esti-
mates of standard errors around estimated regression para-
meters and in inflated Type I error rates (Legendre et al.,
2002). However, when autocorrelation is caused by the
underlying environmental gradient under study, the fitting of
a regression model should remove the autocorrelation, leaving
uncorrelated residuals, and reliable error rates (Diniz-Filho
et al., 2003).
To investigate autocorrelation in our data we generated
spatial correlograms using Moran?s I (Legendre & Fortin,
1989) calculated for 10 distance classes (with upper bounds at
240, 622, 1396, 1707, 2254, 2880, 3490, 4683, 5279, and
6244 km) for all 10 response variables and for PC1, PC2,
NDVImax, and NDVIratio. We interpreted a correlogram as
showing significant autocorrelation when at least one of the
individual estimates of Moran?s I had an associated P-value of
< 0.005 (a = 0.05 with Bonferroni correction for 10 distance
classes tested). Because multiple habitats were sampled within
a site and the transects were not geo-positioned at the time of
the surveys, we arbitrarily set the habitats to be separated by
1 km within each site. Although this introduced some artificial
structure at local scales, we did not think it would affect the
macro-scale patterns in which we were primarily interested
because sites were separated by hundreds to thousands of
kilometres. All correlogram data were generated using the
software Spatial Analysis in Macroecology (sam, Rangel et al.,
2006) and plotted using matlab 7.0 (Mathworks, 2006).
To test the explanatory power of the four continental
gradients of interest, namely climate, primary productivity,
longitude, and latitude, we used simple linear regression (often
referred to as ordinary least squares, or OLS, although we fitted
all models using maximum likelihood) to fit 10 sets of four
models each, corresponding to the 10 response variables tested
against each of the four continental gradients described above.
Because we were certain that our response variables would vary
with habitat, all models included habitat as a main effect, and
each model treated habitats within sites as repeated measures
on the same subject. Upon examination of scatter plots of bird
data, we determined that a continental pattern was often
evident, in that the data in the centre of the continent differed
from those in the western and eastern regions. Therefore, for
location variables (latitude and longitude) we also included a
quadratic term (Table 1).
For each of the 10 sets of four models, we used AICc
(Akaike?s information criterion, adjusted for small sample size;
Burnham & Anderson, 2002) to choose the best regression
model (i.e. the gradient that best explained the variation in the
Table 1 Four fixed-effects models of the continental distribution
of birds by migratory status.
Name Modela
PC Response = HAB + PC1 + PC2
NDVI Response = HAB + NDVImax + NDVIratio
LONG Response = HAB + Longitude + Longitude2
LAT Response = HAB + Latitude + Latitude2
a?Response? refers to one of the 10 categories of species or individuals
by migratory status. LONG, longitude; NDVI, normalized difference
vegetation index; PC, principal component; HAB, habitat.
R. Greenberg et al.
4 Journal of Biogeography
? No claim to original US governement works. Journal compilation ? 2008 Blackwell Publishing Ltd
response variable). For this best model we generated a second
set of correlograms on the residuals to determine whether
the model fitting procedure reduced or removed spatial
autocorrelation in the response variable.
When spatial autocorrelation cannot be removed through
simple regression modelling, it may be possible to incorporate
it directly into the residual error structure during model fitting
(Littell et al., 2006) and thereby avoid biased error rates. These
models are often referred to as generalized least squares (GLS),
although again we fit all models using maximum likelihood to
ensure that our fixed-effects models would be comparable
using AICc. We used GLS to fit all 40 of the previous fixed-
effects models assuming a spherical structure to the error
residuals (Littell et al., 2006). To account for multiple
measurements of the response variable (habitats) within a site
we included a local nugget effect (Littell et al., 2006), which
allows an abrupt discontinuity in the pattern of autocorrela-
tion at very local scales. As above, we ranked all four models
for each of the 10 response variables using AICc. All regression
analyses were performed using Proc Mixed in sas 9.0 (SAS
Institute, 2002).
RESULTS
PCA analysis of climate data
The first two principal components explained 87% of the
variance in climate data from the study sites. The first
component (59% of variance) was correlated primarily with
data indicating temperature extremes (temperature difference,
r = )0.87). The second component (28% of variance) was
correlated primarily with summer rainfall ()0.78). We used
only these components in the regression analysis.
The relationship between the first two PC scores and degrees
longitude from the eastern coast shows that the first principal
component is a good index of continentality (Fig. 2a). Note
that the two Canadian sites fit well within the Eurasian
relationship (e.g. Northwest Territory points fall near Central
Siberia, and Ontario is found near the Russian Far East). The
second PCA shows the east?west gradient (Fig. 2b), primarily
in precipitation, and the Canadian points are, once again,
consistent with the Eurasian pattern.
Spatial autocorrelation and effects on model selection
Analysis of Moran?s I indicated significant autocorrelation
for all variables tested. In general, the four predictor
variables for which we generated correlograms (Fig. 3a?d)
showed a greater degree of autocorrelation than did the 10
response variables (Fig. 3e?n), as judged by the extreme
values of Moran?s I observed at some spatial scales. For each
of the 10 response variables, the best-fit model usually
greatly reduced spatial autocorrelation, in most cases elim-
inating it at large distances (> 2000 km, Fig. 3). After fitting
the linear models, the residuals retained some positive
autocorrelation at very local scales (< 500 km), and, for
some variables, negative autocorrelation at intermediate
scales (500?2000 km). The further imposition of spherical
structure on model residuals had little effect on the AICc
scores or model ranks for any of the 10 response variables
(Tables 2 and 3).
Patterns of abundance
The model containing a quadratic effect of longitude (LONG,
Table 1) was the best description for overall abundance
(Tables 2 and 3), which followed a clear continental pattern
(Fig. 4a) with the lowest abundance towards the centre of the
Eurasian continents and the highest abundances in the
European and Far Eastern sites. Abundance also varied with
the first principal component of climate data (Fig. 4b),
although the PC model was clearly inferior to the LONG
model (Tables 2 and 3) based on AICc scores. The LAT and
NDVI models both performed very poorly (Tables 2 and 3).
Examining Fig. 4, the data points from the Canadian sites fall
within the pattern for the Russian sites.
?20020406080100120
?2.0
?1.5
?1.0
?0.5
0.0
0.5
1.0
1.5
2.0
2.5(a)
(b)
Russia 
Canada 
Europe Siberia Far east
OntarioMackenzie
?20020406080100120
Longitude (? from east coast)
?2.0
?1.5
?1.0
?0.5
0.0
0.5
1.0
1.5
PC
A 
sc
or
e
PC
A 
sc
or
es
Russia 
Canada 
Europe Siberia Far East
OntarioMackenzie
Figure 2 PCA climate scores for study areas plotted against
degrees longitude from the eastern edge of the continent for
(a) Factor 1 (continentality of temperature increasing from
bottom to top of axis) and (b) Factor 2 (precipitation increasing
from top to bottom of axis). The locations of the regions of
Russia are indicated at the bottom of the graphs.
Russian boreal birds
Journal of Biogeography 5
? No claim to original US governement works. Journal compilation ? 2008 Blackwell Publishing Ltd
The relative abundance of tropical and short-distance
migrants
The vast majority of the birds in the boreal forest sites were
migrants, accounting for an average of 81% (? 13 SD) of the
surveyed birds; tropical migrants averaged 43% (? 26); and
short-distance migrants 36% (? 21) of the samples. The
abundance of birds classified by migration types was best
explained by a quadratic model of longitude for residents,
short-distance migrants, and overall migrants (LONG,
(a) (b)
(c) (d)
(e) (f)
(g) (h)
(i) (j)
(k) (l)
(m) (n)
Figure 3 Spatial correlograms for raw predictor variables (a) PC1, (b) PC2, (c) NDVIratio, (d) NDVImax, and for all 10 response variables
both as raw variables (dashed lines) and as residuals after fitting the best model (solid lines). See text for details on the best models.
Response correlograms are species richness of (e) all birds, (g) short-distance migrants, (i) residents, (k) all migrants, and (m) long-distance
migrants, and total individuals of (f) all birds, (h) short-distance migrants, (j) residents, (l) all migrants, and (n) long-distance migrants.
Table 2 AICc differences for simple regres-
sion analyses between the mean abundance
and estimated species number for compo-
nents of the Russian boreal avifauna and
potentially explanatory ecogeographic
variables.
Response TOSP RSP SDSP TRSP MGSP TOIND RIND SDIND TRIND MGIND
Climate PC 1.3 1.7 13.4 0 3.2 5.9 14.7 15 9.6 5.3
NDVI 0 5.9 11.6 13.2 1.3 22 13.3 39.8 23.4 22.4
LONG 2.7 0 0 2.8 3.5 0 0 0 2.8 0
LAT 1.2 3.3 13.2 5.5 0 20.6 8.9 36.8 0 20.6
TO, total; R, residents; SD, short-distance migrants; TR, tropical migrants; MG, migrants; IND,
individuals per survey; SP, estimated number of species per habitat; PC, principal component;
NDVI, normalized difference vegetation index; LONG, longitude; LAT, latitude.
Table 3 AICc differences for regression
analyses between the mean abundance and
estimated species number for components of
the Russian boreal avifauna and potentially
explanatory ecogeographic variables assum-
ing spatially autocorrelated (spherical)
residuals.
Response TOSP RSP SDSP TRSP MGSP TOIND RIND SDIND TRIND MGIND
Climate PC 1.4 1.6 3.5 0 3.2 3.2 11.4 5.2 9.5 1.4
NDVI 0 8.3 5.5 13.6 1.2 6.5 10.8 7.6 8.1 5
LONG 2.8 0 0 2.9 3.4 0 0 0 2.7 0
LAT 1.2 3.3 2.9 5.6 0 10.6 8.8 8 0 7.4
For key to acronyms see Table 2.
R. Greenberg et al.
6 Journal of Biogeography
? No claim to original US governement works. Journal compilation ? 2008 Blackwell Publishing Ltd
1000   2000   3000   4000   5000   6000 0     1000   2000   3000   4000   5000   6000 
Distance (km) 
Tables 2 and 3). None of the other models considered
performed well compared with the LONG model. For tropical
migrants, both the LONG and LAT models performed well,
with the LAT model slightly outperforming the LONG model
(Tables 2 and 3, Figs 5 & 6a). Thus, all of the migratory classes
showed a relationship with longitude, with tropical migrant
and resident abundances increasing to the east (Fig. 6), but
short-distance migrants decreasing to the east.
Species richness
Species richness (Chao I) showed inconsistent patterns over
our set of candidate models. The NDVI model performed best
for overall species richness. Beyond that, each model was
selected as the best model for at least one migratory class
(Tables 2 and 3). However, the differences among top models
were generally small, within a range (? 4 AICc units, Burnham
& Anderson, 2002) that meant they could be considered as
serious competitors. The exception was short-distance migrant
species richness in the simple regression analysis, for which the
quadratic longitudinal model (LONG) was the only reasonable
model (Table 2; see Fig. 6c,d for abundance and species
richness vs. longitude for short-distance migrants). However,
when the residuals were assumed to be spatially autocorrelated,
the relative strength of this model for short-distance migrants
was weakened considerably (Table 3).
DISCUSSION
Geographic patterns
The abundance data from the Russian sites showed two
distinct patterns. First, overall abundance showed a continen-
tal-scale pattern, with the average number of birds per survey
point depressed in Siberia compared with European Russian
and Far Eastern sites. Second, the numerical dominance of
tropical migrants increased along a gradient from west to east,
whereas the numerical dominance of short-distance migrants
increased from east to west. The pattern was less clear for
residents, which accounted for a small portion (< 20%) of the
birds surveyed.
Explanatory power of climate, NDVI and geographic
variables
For the 10 response variables considered (five classes each of
species richness and number of individuals), either the LONG
or the LAT model performed best in eight cases. LONG or LAT
was always the best model for the total number of individuals
detected, either for total birds or for each migratory class.
Thus, our data suggest that patterns in the distribution of
avian abundance and, to a lesser degree, species richness,
across the Russian boreal forest are best explained by simple
geographic position. We offer three possible explanations for
this. First, the data quality for climate and NDVI may have
been too low. We used time periods and spatial scales that were
dictated by the available data sets. More exploration of data
gathered at varying temporal and spatial scales may improve
their fit in these models. Second, with predictable geographic
?200 20 40 60 80 100 120 
Longitude (? from east coast) 
0 
1 
2 
3 
4 
5 
6 
7 (a)
(b)
 Russia  
 Canada  
?2.0 ?1.5 ?1.0 ?0.5 0.0 0.5 1.0 1
PC1 factor score
0 
1 
2 
3 
4 
5 
6 
7 
M
ea
n 
in
di
vi
du
al
s 
pe
r s
ur
ve
y 
po
in
t
M
ea
n 
in
di
vi
du
al
s 
pe
r s
ur
ve
y 
po
in
t
 Russia  
 Canada  
Figure 4 Abundance (mean individuals/point) per study site/
habitat plotted against (a) degrees longitude from the eastern edge
of the continent, and (b) climate-based PC1 factor scores.
44 46 48 50 52 54 56 58 60 62 64 66
Latitude of study sites
?0.5
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
M
ea
n 
nu
m
be
r o
f t
ro
pi
ca
l m
ig
ra
nt
s 
pe
r
su
rv
ey
 p
oi
nt
Russia
 Canada 
Figure 5 Mean individuals of tropical migrants per point for
each habitat/site plotted against degrees latitude.
Russian boreal birds
Journal of Biogeography 7
? No claim to original US governement works. Journal compilation ? 2008 Blackwell Publishing Ltd
patterns in climate, the geographic variables may include both
climate effects and the additional direct influence of geography
as it affects the accessibility of winter habitats. Finally, the
geographic configuration of boreal forest may have an
overarching direct effect on abundance and migration patterns.
In a speculative manner, we explore the possible underlying
causal hypotheses below.
Explanatory hypotheses: climatic regimes in the
breeding grounds or accessibility of wintering
grounds
We present two hypotheses for the explanatory effect of the
variables on the overall distribution of birds and the pattern
for the component migratory classes. First, bird populations
are responding to the two major climatic gradients as indicated
in the PCA analysis: a continental pattern in temperature
regimes resulting in more extreme temperatures and shorter
growing seasons in the mid-continent, and a gradient of
summer precipitation from west to east. Second, the accessi-
bility of tropical and non-tropical wintering areas plays a role
either in patterns of overall abundance or in the relative
importance of tropical vs. within-temperate migration pat-
terns.
The climatic conditions hypothesis would suggest that
climate drives either resource abundance or the amount of
time that habitat is available for reproduction, and that these
in turn affect bird abundance. Arguments have been made for
why climate might affect the proportion of migrants vs.
residents in temperate breeding assemblages (MacArthur,
1959; Herrera, 1978; Newton & Dale, 1996; Hockey, 2005).
However, it is unclear why breeding-location climate should
affect the relative abundance of short-distance vs. long-
distance migrants (Willson, 1976).
The wintering-ground accessibility hypothesis is based on a
growing amount of research that suggests that the distribution
of migratory birds and their flyways can be strongly shaped by
barriers, such as deserts and high mountains, particularly if
they are located close to the potential breeding area (Bensch,
1999; Alerstam, 2001; Henningsson & Alerstam, 2005). In the
Palaearctic, the western and eastern ends of the boreal forest
have no major barriers to non-boreal wintering areas, which
allows for the over-winter survivorship of a large population of
birds (Fig. 7a,b). Siberia has both a greater area of boreal forest
?20020406080100120
?0.5
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0(a)
(b)
(c)
(d)
Russian
Canadian 
?20020406080100120
0
5
10
15
20
25
30
35
40
Es
tim
at
ed
 n
um
be
r o
f t
ro
pi
ca
l
m
ig
ra
nt
 s
pe
ci
es
?0.5
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
M
ea
n 
nu
m
be
r o
f s
ho
rt-
di
st
an
ce
m
ig
ra
nt
s 
pe
r p
oi
nt
2
4
6
8
10
12
14
16
18
20
22
24
26
28
Es
tim
at
ed
 n
um
be
r o
f s
ho
rt-
di
st
an
ce
m
ig
ra
nt
 s
pe
ci
es
M
ea
n 
nu
m
be
r o
f t
ro
pi
ca
l m
ig
ra
nt
s
pe
r s
ur
ve
y 
po
in
t
Longitude (? from east coast)
?20020406080100120
Longitude (? from east coast)
?20020406080100120
Longitude (? from east coast)Longitude (? from east coast)
Figure 6 Mean individuals per point (a) and estimated species richness (b) of tropical migrants, and mean individuals per point (c) and
estimated species richness (d) of short-distance migrants for each habitat/site plotted against degrees longitude from the eastern edge of the
continent.
R. Greenberg et al.
8 Journal of Biogeography
? No claim to original US governement works. Journal compilation ? 2008 Blackwell Publishing Ltd
than Europe and the Russian Far East and a series of profound
barriers to migration south of boreal habitats, including both
extremely arid areas, depicted in Fig. 6, and the Himalayas
(Irwin & Irwin, 2005). The major barrier (the Sahara Desert)
between Europe and tropical Africa will favour boreal species
that winter within the temperate and Mediterranean regions of
Europe. Some of these species, such as the chaffinch, have
adapted to the human-transformed environment, which
provides enhanced winter resources, and they occur at
extremely high abundances (Greenberg et al., 1999a), thus
explaining the high abundance of short-distance migrants in
the Western Palaearctic. In contrast, the Russian Far East
differs from Europe and Siberia in having a historically stable,
barrier-free corridor to the tropics of Southeast Asia, thus
favouring tropical migration.
The pattern of overall abundance could be explained by the
climate or accessibility of wintering grounds hypotheses, as the
abundance is lowest in Siberia where the climate is most
continental and the present-day barriers to both temperate/
Mediterranean and tropical wintering areas are the most
profound. The shift between migration types is probably best
explained by the wintering ground accessibility hypothesis, as
it is unclear why a gradient in summer rain would shift the
migration strategies of breeding species.
Inter-continental comparisons
Most previous work comparing Palaearctic and Nearctic boreal
biota has focused on Europe and Eastern North America. A
few studies have included Siberia and the Russian Far East, but
these are based on regional species lists (Mo?nkko?nen & Helle,
1989; Haila & Ja?rvinen, 1990). This study is the first to be
based on habitat-specific survey data, and can examine
abundance and local patterns of species richness. With only
two regions in North America surveyed in this study, it is
difficult to conduct formal statistical analyses. However, the
values for both climate and metrics characterizing avian
assemblages at the Canadian sites are more similar to those
from the central and eastern portions of the Palaearctic than
they are to those from the European boreal forest, and hence
the Asian boreal forest might be an appropriate region for
future avifaunal comparisons with Nearctic sites.
First, as Fig. 2 shows, both major gradients in climate (PC
factors 1 and 2 vs. longitude) show that the eastern Canadian
site is closest in climate to sites in the Russian Far East, and
that the western Canadian site is similar to those in Central
Siberia. Large-scale patients of habitat distribution, particularly
the connectivity of tropical and temperate forest since the late
Pleistocene (Huntley, 1993), is also similar between the two
continents if the appropriate regions are compared. Patterns of
overall abundance and the abundance and species richness of
tropical and short-distance migrant species (Fig. 6) in the
Canadian sites appear to be consistent with the overall
geographic patterns found in the Eurasian samples. Therefore,
the characterization of the differences in the abundance of
tropical and short-term migrants between the Palaearctic and
Nearctic boreal forest is largely dependent on where in the
continent the data from the comparison are obtained. Based
Closed Forest (> 70% canopy cover)
Present Potential Vegetation
Extreme Desert (< 2% vegetation cover)
Closed Forest
Last Glacial Maximum (18 000 14C years ago)
Extreme Desert
(a)
(b)
Figure 7 Estimated distribution of desert
and closed forest: (a) present-day potential
and (b) Latest Glacial Maximum (from
Adams, 1997, with permission).
Russian boreal birds
Journal of Biogeography 9
? No claim to original US governement works. Journal compilation ? 2008 Blackwell Publishing Ltd
on these preliminary surveys, it would appear that a compar-
ative intercontinental research programme should focus more
on Siberia and the Russian Far East (see also Mo?nkko?nen,
1994), where current information is the most sparse. Finally,
the generality of patterns of bird density and species richness
described in this paper should be tested by conducting a
similar analysis based on sites distributed widely across the
New World boreal forest.
ACKNOWLEDGEMENTS
The following observers conducted surveys in the Russian
Nature Reserves: Kivach ? M. Yacovleva and A.V. Suchov;
Pinejskyi ? A.M. Rykov; Pechora-Ilichski ? V.V. Teplov;
Visim ? E.G. Larin; Malaya Sosva ? V.D. Lichvar; Barguzin ?
A. Ananin; Baykalskiy ? V.V.Baskakov; Visimskiy ? D.
Polushkin; Sikhote-Alin ? S. Yelsukov; and Komsomolskiy ?
V.A. Kolbin. M. L. Cumberpatch of the NOAA Library
helped track down climatic data bases. We thank the Institute
of Evolutionary Morphology and Animal Ecology of the
Russian Academy of Sciences, in particular Katya Preobzaz-
henskaya and Oleg Bursky for providing housing and
logistical support for work in Myrnoe and Kostromomskaya
field stations. Anwar Kerimov hosted us at the Zvenigorod
field station. The following people assisted in gathering data
for this project: Mark Amberle, Sam Droege, Todd Easterla,
Jennifer Gammon, Gjon Hazard, Richard Hoyer, Vitali
Kontorschikov, and Olga Romanenko. Mario Castellanos
assisted with figure preparation. A Pew Fellowship for
Biodiversity and the Environment provided funding to the
senior author.
REFERENCES
Adams, J.M. (1997) Global land environments since the last
interglacial. Oak Ridge National Laboratory, TN. Available
at: http://www.esd.ornl.gov/ern/qen/nerc.html; last accessed
May 2007.
Alerstam, T. (2001) Detours in bird migration. Journal of
Theoretical Biology, 209, 319?331.
Bensch, S. (1999) Is the range size of migratory birds con-
strained by their migratory program? Journal of Biogeogra-
phy, 26, 1225?1235.
Bini, L.M., Diniz-Filho, J.A.F. & Hawkins, B.A. (2004) Mac-
roecological explanations for differences in species richness
gradients: a canonical analysis of South American birds.
Journal of Biogeography, 31, 1819?1827.
Bo?hning-Gaese, C. (2005) Influence of migrants on temperate
bird communities. Birds of two worlds: the ecology and evo-
lution of migration (ed. by R. Greenberg and P.P. Marra), pp.
143?153. Johns Hopkins University Press, Baltimore, MD.
Burnham, K.P. & Anderson, D.R. (2002) Model selection and
multimodel inference: a practical information-theoretic
approach. Springer, New York.
Colwell, R.K. (2003) EstimateS website. Available at: http://
viceroy.eeb.uconn.edu/EstimateS; last accessed May 2007.
Colwell, M.A., Dubyin, A.V., Koroliuk, A.Y., Yu, A. & Sobolev,
N.A. (1997) Russian nature reserves and conservation of
biological diversity. Natural Areas Journal, 17, 56?68.
Ding, T.-S., Yuan, H.W., Geng, S., Koh, C.N. & Lee, P.F.
(2006) Macro-scale bird species richness patterns of the East
Asian mainland and islands: energy, area and isolation.
Journal of Biogeography, 33, 683?693.
Diniz-Filho, J.A.F., Bini, L.M. & Hawkins, B.A. (2003) Spatial
autocorrelation and red herrings in geographical ecology.
Global Ecology and Biogeography, 12, 53?64.
Erskine, A.J. (1977) Birds in boreal Canada: communities,
densities, and adaptations. Canadian Wildlife Service Report
41. Canadian Wildlife Service, Ottawa.
Greenberg, R., Pravosudov, V., Sterling, J., Kozlenko, A. &
Kontorschikov, V. (1999a) Warblers, finches, and tits: the
structure of boreal forest foliage-gleaning guild. Condor,
101, 299?310.
Greenberg, R., Pravosudov, V., Sterling, J., Kozlenko, A. &
Kontorschikov, V. (1999b) Divergence in foraging behavior
of foliage-gleaning birds of Canadian and Russian boreal
forests. Oecologia, 120, 451?462.
Haila, Y. & Ja?rvinen, O. (1990) Northern conifer forests and
their bird species assemblages. Biogeography and ecology of
forest bird communities (ed. by A. Keast), pp. 61?97. SPB
Academic Publishing, The Hague.
Haila, Y., Ja?rvinen, O. & Raivio, S. (1987) Quantitative versus
qualitative distribution patterns of birds in the western
Palearctic taiga. Annales Zoologici Fennici, 24, 179?195.
Helle, P. & Fuller, R.J. (1988) Migrant passerine birds in
European forest successions in relation to vegetation height
and geographical position. Journal of Animal Ecology, 57,
565?579.
Henningsson, S.S. & Alerstam, T. (2005) Barriers and distances
as determinants for the evolution of bird migration links:
the arctic shorebird system. Proceedings of the Royal Society
B: Biological Sciences, 272, 2251?2258.
Herrera, C.M. (1978) On the breeding distribution pattern of
European migrant birds: MacArthur?s theme re-examined.
The Auk, 95, 496?509.
Hockey, P.A.R. (2005) Predicting migratory behavior in
landbirds. Birds of two worlds: the ecology and evolution
of migration (ed. by R. Greenberg and P.P. Marra),
pp. 53?64. Johns Hopkins University Press, Baltimore,
MD.
Huntley, B. (1993) Species richness in north-temperate zone
forests. Journal of Biogeography, 20, 163?180.
Hurlbert, A.H. & Haskell, J.P. (2003) The effect of energy and
seasonality of avian species richness and community com-
position. The American Naturalist, 161, 83?97.
Hutto, R.L., Pletschet, S.M. & Hendricks, P. (1986) A fixed-
radius point count method for nonbreeding and breeding
season use. The Auk, 103, 593?602.
IRI/LDEO (2003) Climate data library. Available at: http://
ingrid.ldgo.columbia.edu/; last accessed May 2007.
Irwin, D.E. & Irwin, J.H. (2005) Siberian migratory divides.
Birds of two worlds: the ecology and evolution of migration
R. Greenberg et al.
10 Journal of Biogeography
? No claim to original US governement works. Journal compilation ? 2008 Blackwell Publishing Ltd
(ed. by R. Greenberg and P.P. Marra), pp. 27?40. Johns
Hopkins University Press, Baltimore, MD.
Kineman, J. & Hastings, D. (1992) Monthly Generalized Global
Vegetation Index from NESDIS NOAA-9 Weekly NDVI
Data (APR 1985-DEC 1988). Available at: http://www.
ngdc.noaa.gov/seg/cdroms/ged_iia/datasets/a01/mgv.htm#top;
last accessed May 2007.
Legendre, P. & Fortin, M.-J. (1989) Spatial pattern and eco-
logical analysis. Vegetatio, 80, 107?138.
Legendre, P., Dale, M.R.T., Fortin, M.-J., Gurevitch, J., Hohn,
M. & Meyers, D. (2002) The consequences of spatial
structure for the design and analysis of ecological field
surveys. Ecography, 25, 601?615.
Littell, R.C., Milliken, G.A., Stroup, W.W., Wolfinger, R.D. &
Schabenberger, O. (2006) SAS for mixed models, 2nd edn.
SAS Institute, Cary, NC.
MacArthur, R.H. (1959) On the breeding distribution pat-
tern of North American migrant birds. The Auk, 76, 318?
325.
Mathworks (2006) MATLAB. Mathworks, Natick, MA.
Mo?nkko?nen, M. (1994) Diversity patterns in Palearctic and
Nearctic forest bird assemblages. Journal of Biogeography, 21,
183?195.
Mo?nkko?nen, M. & Forsman, J.T. (2005) Ecological and bio-
geographical aspects of the distribution of migrants versus
residents in European and North American forest birds.
Birds of two worlds: the ecology and evolution of migration
(ed. by R. Greenberg and P.P. Marra), pp. 131?142. Johns
Hopkins University Press, Baltimore, MD.
Mo?nkko?nen, M. & Helle, P. (1989) Migratory habits of birds
breeding in different stages of forest succession: a compar-
ison between the Palaearctic and the Nearctic. Annales
Zoologici Fennici, 26, 323?330.
Mo?nkko?nen, M. & Viro, P. (1997) Taxonomic diversity in
avian and mammalian faunas of the northern hemisphere.
Journal of Biogeography, 24, 603?612.
Mo?nkko?nen, M. & Welsh, D. (1994) A biogeographic
hypothesis on the effects of human caused landscape
changes on the forest bird communities of Europe and
North America. Annales Zoologici Fennici, 3, 61?70.
Morse, D.H. (1989) American warblers: an ecological and
behavioral perspective. Harvard University Press, Cambridge,
MA.
Newton, I. & Dale, L. (1996) Relationship between migration
and latitude among west European birds. Journal of Animal
Ecology, 65, 137?146.
Petit, D.R., Petit, L.J., Saab, V.A. & Martin, T.E. (1995) Fixed-
radius point counts in forests: factors influencing effective-
ness and efficiency. Monitoring bird population trends by
point counts (ed. by C.J. Ralph, S. Droege and J. Sauer),
pp. 49?56. Gen. Tech. Rep. No. PSW-GTR 149. USDA
Forest Service, Albany, CA.
Preobrajenskaya, E. (1982) Meridianal changes of summer
population in southern taiga. Abundance and distribution of
Siberian vertebrates, pp. 48?67. Novosibirsk, Nauka.
Rangel, T.F.L.V.B., Diniz-Filho, J.A.F. & Bini, L.M. (2006)
Towards an integrated computational tool for spatial ana-
lysis in macroecology and biogeography. Global Ecology and
Biography, 15, 321?327.
SAS Institute (2002) The SAS system for Windows. V. 9.00. SAS
Institute, Cary, NC.
Selmi, S. & Boulinier, T. (2001) Ecological biogeography of
Southern Ocean islands: the importance of considering
spatial issues. The American Naturalist, 158, 426?437.
Sibley, C.G. & Monroe, B.L., Jr (1990) Distribution and tax-
onomy of birds of the world. Yale University Press, New
Haven, CT.
Sze?p, T. & M?ller, A.P. (2005) Using remote sensing data to
identify migration and wintering area and to analyze effects
of environmental conditions on migratory birds. Birds of two
worlds: the ecology and evolution of migration (ed. by
R. Greenberg and P.P. Marra), pp. 390?400. Johns Hopkins
University Press, Baltimore, MD.
Wiens, J.A. (1975) Avian communities, energetics, and func-
tions in coniferous forest habitats. Proceedings of the Sym-
posium on Management of Forest and Range Habitats for
Nongame Birds. Gen. Tech. Rep., WO 1, 226?265.
Willson, M.F. (1976) The breeding distribution of North
American migrant birds: a critique of MacArthur (1959).
Wilson Bulletin, 88, 582?587.
SUPPORTING INFORMATION
Additional Supporting Information may be found in the
online version of this article:
Appendix S1 Location, climate, NDVI, and bird abundance
and species richness data for Russian and Canadian survey
sites.
Appendix S2 Latin names (Sibley & Monroe, 1990) and
migratory bird classes for species of (a) Russian and
(b) Canadian birds included in the analyses.
Please note: Blackwell Publishing is not responsible for the
content or functionality of any Supporting Information
supplied by the authors. Any queries (other than missing
material) should be directed to the corresponding author for
the article.
Russian boreal birds
Journal of Biogeography 11
? No claim to original US governement works. Journal compilation ? 2008 Blackwell Publishing Ltd
BIOSKETCHES
Russell Greenberg received his PhD from the University of California in 1981 and has been the director of the Smithsonian
Migratory Bird Center since 1992.
Anna Kozlenko earned her doctorate from the Russian Academy of Science in 1987, worked as a researcher for the Russian Institute
for Nature Protection for 8 years, and is currently working on information technology for a private firm.
Matthew Etterson received his PhD in conservation biology from the University of Minnesota in 2000 and is currently a senior
research associate at the Mid Continent Ecology Division of the US Environmental Protection Agency in Duluth, MN, and an
adjunct assistant professor of biology at the University of Minnesota Duluth.
Thomas Dietsch received his PhD from the University of Michigan and has gone on to post-doctoral positions at the Smithsonian
Migratory Bird Center and UCLA?s Center for Tropical Research.
Editor: Bradford Hawkins
R. Greenberg et al.
12 Journal of Biogeography
? No claim to original US governement works. Journal compilation ? 2008 Blackwell Publishing Ltd