BIOLOGICAL

CONSERVATION

Biological Conservation 121 (2005) 117–126

www.elsevier.com/locate/biocon

Landscape patterns influencing bird assemblages in a

fragmented neotropical cloud forest
Miguel Angel Mart
ınez-Morales *

Department of Zoology, University of Cambridge Downing Street, Cambridge CB2 3EJ, UK

Received 8 June 2003; received in revised form 18 March 2004; accepted 3 April 2004

Abstract

The cloud forest is one of the rarest and most threatened ecosystems in Mexico, although it contributes highly to the country’s
biological diversity and provides important ecological services. It is a naturally fragmented ecosystem, but anthropogenic defor-
estation and fragmentation has been severe. Consequently, it is essential to identify landscape patterns critical for the conservation
of cloud forest. In order to understand how landscape patterns affect diversity in this ecosystem, this study explores the conse-
quences of cloud forest fragmentation on bird diversity in eastern Mexico. I analysed the response of bird species richness and
abundance as a function of forest fragment size, shape, topographical complexity, altitudinal range, connectivity, and proportion of
landscape forested area in a system of 13 cloud forest fragments. Fragment shape was the main characteristic positively related to
species richness in the bird community, but a differential response to landscape patterns was also detected. Fragment size was the
main characteristic influencing the segment of the bird community depending mostly on forest, that is to say, forest interior and
generalist species. In contrast, the extent of forest edge, expressed as fragment shape, produced a positive response of forest border
species. Both, forest dependent and border dependent species positively responded to the extent of their suitable habitat. The
immediate and most effective ecologically oriented conservation strategy for the region is the conservation of the largest cloud forest
fragments.
Ó 2004 Elsevier Ltd. All rights reserved.

Keywords: Landscape patterns; Forest fragmentation; Tropical montane cloud forest; Bird communities; Mexico

1. Introduction Cloud forest is one of Mexico’s rarest and most threa-

The tropical montane cloud forest in Middle America
is a naturally fragmented ecosystem owing to the geo-
logical history of the region. As a result of this historical
fragmentation, cloud forests hold disproportionate high
levels of endemicity and offer the possibility of studying
the important biogeographic and ecological processes
which underpin its spatial and temporal dynamics (e.g.,
Hern
andez-Ba~ nos et al., 1995). Fragmentation of the
cloud forest in the region has increased considerably due
to human activities and settlements, threatening its ex-
istence (Escalona et al., 1995; V
azquez-Garc
ıa, 1995).
*
Present address: Centro de Investigaciones Biol

ogicas, Universidad
Aut

onoma del Estado de Hidalgo Apartado Postal 69, Plaza Ju
arez,
Pachuca, Hidalgo, C.P. 42001 M
exico.
E-mail address: migmarti@uaeh.reduaeh.mx (M. Angel Mart
ınez-
Morales).
tened ecosystems, but at the same time, contributes
highly to the country’s biological diversity (Flores-Vill-
ela and Gerez, 1994; G
omez-Pompa et al., 1995). Based
on analysis of patterns of distribution, diversity and
endemism of bird faunas, Hern
andez-Ba~ nos et al. (1995)
defined biogeographic regions of cloud forest in Middle
America. This allowed them to identify regions of cloud
forest that require conservation action. In Mexico, such
regions included eastern Mexico north of the Isthmus of
Tehuantepec, Southern Sierra Madre, and interior
Oaxaca, all of which remain practically unprotected.
Once priority regions for conservation of cloud forest
have been identified, it is paramount to define areas for
conservation at a finer scale. The definition of such areas
has to be guided, among other aspects, by studies at
landscape level. Birds are an important ecological tool
for such studies because their taxonomy and distribution
are well known, and because inventory methods are well

0006-3207/$ - see front matter Ó 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.biocon.2004.04.015
118 M. Angel Mart
ınez-Morales / Biological Conservation 121 (2005) 117–126

developed. Furthermore, a remarkable coincidence in

patterns of diversity and endemicity has been found
among different taxonomic groups in Middle America
(Ramammoorthy et al., 1998); thus, optimal conserva-
tion strategies should generally coincide. Analyses of
patterns of avian diversity can be a good indicative of
the effects of cloud forest fragmentation on biodiversity.
The characteristics of forest fragments that have been
consistently reported to have an important relationship
with diversity include fragment size, the extent and na-
ture of fragment edge, fragment connectivity, and the
extent of forest outside the fragment (Ambuel and
Temple, 1983; Howe, 1984; Lovejoy et al., 1986; Blake
and Karr, 1987; Bierregaard et al., 1992; Laurance,
1994; Willson et al., 1994; Murcia, 1995; Robinson
et al., 1995; Bellamy et al., 1996; Kattan and Alvarez-
L
opez, 1996; Arango-V
elez and Katan, 1997; Laurance
and Bierregaard, 1997; Price et al., 1999; Stratford and
Stouffer, 1999; Cornelius et al., 2000). The aim of this
study was to identify fragment or landscape patterns
having a relationship with the structure and composition
of the bird community of a fragmented cloud forest.

2. Methods

2.1. Study area

This study was carried out in 13 fragments of tropical

montane cloud forest in northeastern Hidalgo, eastern
Mexico (Fig. 1). This area practically represents the
northernmost distribution of the cloud forest of the Si-
erra Madre Oriental. It is located on the windward slope
of the Sierra Madre and is usually found at altitudes
between 750 and 2400 m (Luna et al., 2000). Within this
altitudinal band, pine-oak forest and oak forest occur in
areas where humidity is not high enough for the devel-
opment of cloud forest. Below this altitudinal band, it is
replaced by tropical rain forest, and above this band,
pine forest and pine-oak forest become the main vege-
tation types. The forest fragments surveyed were within
an altitudinal band ranging from 1000 to 2000 m. The
aggregate area of all fragments surveyed was approxi-
mately 26,600 ha. Most of this forest was located in
northeastern Hidalgo, except for about 7700 ha, which
extended to the neighbouring state of Veracruz, east of
Hidalgo. The cloud forest in this region is included
within the ‘‘Southern Sierra Madre Oriental’’ Endemic
Bird Area, which has a critical priority ranking for Fig. 1. Location of the 13 cloud forest fragments surveyed in north-
conservation action (Stattersfield et al., 1998). eastern Hidalgo, Mexico. (1) Tlahuelompa, (2) Coatempa, (3)
Tlanchinol, (4) Lontla, (5) Mojonera, (6) Malila, (7) Tenango, (8)
2.2. Analyses of the geographic information Ahuacatl
an, (9) Xochicoatl
an, (10) Soyatla-1, (11) Soyatla-2, (12)
Molocotl
an-1, (13) Molocotl
an-2.

The characteristics of the forest fragments surveyed in which the fragment was embedded, as well as the
such as size, shape, topographical complexity, altitudi- characteristics of the vegetation matrix surrounding the
nal range, connectivity, and proportion of forested area cloud forest fragments were analysed from a Landsat
 M. Angel Mart
ınez-Morales / Biological Conservation 121 (2005) 117–126
TM multispectral satellite image (recorded 15 October
1991, spatial resolution of pixels 25
25 m), from aerial
photographs (May 1994, scale 1:20,000 and January
1978, scale 1:35,000), and from topographical and veg-
etation cover maps (scales 1:50,000 and 1:250,000).
There was no detailed information regarding when the
fragments became isolated, but according to the infor-
mation gathered, all fragments became isolated before
1978. The satellite image was acquired after being geo-
metrically corrected and projected on to a universal
transverse mercator projection grid at the Instituto de
Geograf
ıa of the Universidad Nacional Aut
onoma de
M
exico (UNAM).
The satellite image was processed with GRASS 4.2
(Geographic Resources Analysis Support System) at the
Direcci
on General de Servicios de C
omputo Acad
emico,
UNAM. The image was classified into land cover cate-
gories using a supervised classification method. For the
classification, the generation of signatures was an iter-
ative process using principal components analysis on
bands 1, 2, 3, 4, 5 and 7 of the satellite image. These
bands were found to provide adequate discrimination
among the land cover classes. Band 6 was eliminated.
Additionally, a digital terrain model was incorporated
to facilitate the classification of the satellite image using
elevation information. Filters or smoothing techniques
which eliminate small patches (smaller than about 2 ha)
and make more regular all other patches, were not ap-
plied after classification of the satellite image. This was
because of the need to detect small forest fragments to
carry out bird surveys. A patch was defined as pixels of
the same land cover category that were horizontally or
vertically adjacent; diagonal pixels were not considered
to be contiguous (Turner, 1990).
The 13 cloud forest fragments identified and selected
for bird surveys were isolated from the image to analyse
their area and perimeter. Shape was measured as the
irregularity of the fragment compared to a circle of
the same area. To assess topographical complexity of
the forest fragments (as a surrogate of habitat hetero-
geneity), the slope of each pixel in the forest fragments
was obtained with the aid of the digital terrain model.
The area (square meters) covered by different slope
values (degrees) was calculated in each forest fragment,
then topographical complexity in a forest fragment was
assessed by the coefficient of variation of the terrain
slope weighted by the area covered by the different slope
values. Thus, higher values of the coefficient of variation
represented higher values of topographical complexity.
Two estimates of connectivity were calculated based on
the mean distance between the focal cloud forest frag-
ment (the fragment surveyed) and the forest fragments
larger than 100 ha surrounding it. The mean distance
between the focal fragment and cloud forest fragments
was defined as connectivity – cf, and the mean distance
between the focal fragment and all forest fragments
119
(cloud forest, pine-oak forest, tropical rain forest) was
defined as connectivity – af. The rationale behind the
100 ha criteria is based on the detection of thresholds in
the bird community structure and composition as cloud
forest fragmentation intensifies (Mart
ınez-Morales,
2001). Additionally, a 500 m buffer was created around
each forest fragment to analyse proportion of forested
area in which each forest fragment was embedded. The
proportion of forested area in which each cloud forest
was embedded was defined as the percentage of total
forested area (any forest type) within the area encom-
passed by the cloud forest fragment and a band of 500 m
width surrounding the fragment. Characteristics of the
cloud forest fragments surveyed are summarized in
Table 1.
2.3. Survey of bird communities
Fieldwork was carried out from June 1997 to August
1999 to survey the bird communities in the 13 cloud
forest fragments. Surveys were performed monthly. In
1997, fragments were surveyed from June to September,
and in 1998 and 1999, from January to August.
Point counts were used to assess species richness and
abundance in each forest fragment. Every forest frag-
ment was surveyed once a month. In total, between 30
and 371 census points (depending on fragment size) were
sampled in the 13 forest fragments, giving a total of 2057
census points sampled in the study area at the end of
fieldwork. On average, eight birds were detected per
point count. The bird community in this cloud forest is
dominated (83%) by rare species (Mart
ınez-Morales,
2001); therefore, sampling effort was concentrated in a
relatively small number of forest fragments to ensure
that most of the species using the fragments were de-
tected. If a less intensive sampling effort had been in-
vested in a larger number of forest fragments, patterns
of the bird community using the fragments would have
been obscured because of the detection of common
species only.
The census period ran for 3 h after dawn, and for 3 h
before dusk. Census points were surveyed for 10 min
and there was a travelling distance of about 5 min (ap-
proximately 200 m) between points, to assure indepen-
dence between neighbouring points (Ralph et al., 1995).
At each census point, all bird species detected visually or
acoustically were recorded, and whenever possible, de-
tection distance from the observer was measured with an
optical rangefinder. Notes on habitat use by bird species
or any other relevant information were also recorded.
The location of the surveyed points was determined in
the field using a GPS unit and an altimeter. To reduce
the variability in bird species composition and abun-
dance among forest fragments due to altitude, bird
censuses were restricted to an altitudinal band from
1300 to 1900 m, which represents a mid altitudinal range
120 M. Angel Mart
ınez-Morales / Biological Conservation 121 (2005) 117–126

Table 1
Characteristics of the 13 cloud forest fragments surveyed in northeastern Hidalgo, Mexico
Forest fragment Area (ha) Shape index Topographical Altitudinal range Connectivity Connectivity Land-
complexity of fragment through cloud through any scape
(coefficient of (altitudinal range forest (m) forest type (m) forested
variation) surveyed) area (%)
1. Tlahuelompa 16,289.0 26.7 92.1 1000–2000 m 603 431 80.0
(1400–1850 m)
2. Coatempa 5619.7 22.0 90.3 1000–2000 m 664 531 79.1
(1600–1900 m)
3. Tlanchinol 1663.9 11.7 117.9 1000–1600 m 106 53 73.3
(1400–1600 m)
4. Lontla 1457.0 14.3 96.1 1000–1650 m 463 231 71.2
(1300–1500 m)
5. Mojonera 974.5 12.7 93.8 1550–2000 m 1102 441 75.2
(1700–1900 m)
6. Malila 266.6 4.6 82.5 1450–2000 m 219 165 77.1
(1450–1700 m)
7. Tenango 231.7 7.2 85.6 1050–1500 m 1454 1892 54.1
(1300–1500 m)
8. Ahuacatl

an 45.9 5.2 91.4 1300–1500 m 754 754 37.9
(1300–1500 m)
9. Xochicoatl

an 43.3 3.7 97.4 1700–1800 m 1496 1908 34.6
(1700–1800 m)
10. Soyatla-1 11.7 3.6 92.4 1500–1700 m 939 939 27.6
(1500–1700 m)
11. Soyatla-2 5.0 2.3 75.2 1650–1700 m 770 770 15.4
(1650–1700 m)
12. Molocotl

an-1 1.6 2.0 77.3 1700 m (1700 m) 1738 1738 13.9
13. Molocotl

an-2 0.6 1.3 86.6 1650 m (1650 m) 1457 1618 8.9

for cloud forest. To reduce biases, all bird censuses were
carried out by myself and only in good weather.
To detect bird species occurring in the forest interior,
point counts were carried out at no less than 200 m from
the forest edge. This ensured that species detected were
effectively at no less than 100 m from the forest edge. On
the other hand, to detect bird species in the forest bor-
der, point counts were done at no more than 50 m from
the forest edge. Apart from exceptional circumstances,
most visual detections occurred at no more than 30–50
m from the point count. In the case of acoustic detec-
tions, however, some species could be heard for dis-
tances of 100 m or more. Therefore, corrections on the
location of the bird species detected were done when
necessary. Forest interior species were those occurring
mostly at more than 100 m from the forest edge. On the
other hand, border species were those occurring mainly
within 100 m of the forest edge. Bird species were as-
signed to either interior, or border categories, if the in-
terior abundance was twice the border abundance or
vice versa. Otherwise the species was considered a forest
generalist. For this classification, only the data from
forest fragments larger than 200 ha were included, be-
cause the smaller fragments (less than 50 ha) were not
large enough to have an interior forest area. Those bird
species that were only recorded in forest fragments
smaller than 50 ha were considered as species from the
vegetation matrix surrounding the fragments. The ra-
tionale behind the 100 m threshold is based on empirical
observations and on the detected effects of edge pene-
tration into the forest on birds (Lovejoy et al., 1986;
Paton, 1994). This was also supported by studies on
abiotic and biotic edge effects in fragmented forests
(Williams-Linera, 1993; Murcia, 1995; Didham, 1997;
Kapos et al., 1997; Turton and Freiburger, 1997;
Mesquita et al., 1999).
2.4. Bird species richness and abundance
Before any analyses were performed to assess the
response of the bird community to cloud forest frag-
mentation, the observed species richness in the forest
fragments surveyed was standardized. Species-richness/
sampling-effort curves for each forest fragment surveyed
were constructed to estimate the expected species rich-
ness in each forest fragment (Table 2). Two species ac-
cumulation functions were used, the exponential and
Clench’s functions (Sober
on and Llorente, 1993). These
two functions were selected because their assumptions
fairly well described the situations in which data were
collected.
Bird species abundance in forest fragments was also
standardized as the mean number of individuals de-
tected in 100 surveyed points. This abundance estimate
took account of the seasonality of the species (e.g., mi-
gratory species). Thus, the total number of point counts
 M. Angel Mart
ınez-Morales / Biological Conservation 121 (2005) 117–126 121

Table 2
Summary of the results obtained from the non-linear regression functions fitted to the field data on bird species accumulation, in each forest fragment
and in the whole study area
Forest fragment Non-linear Regression results Bird species Point counts
function 2 a b c
R a b Detected Expected Surveyed Requiredd
1. Tlahuelompa Exponential 0.99 0.7291 0.0058 103 125 304 792
2. Coatempa Exponential 0.99 0.6074 0.0046 107 132 371 1002
3. Tlanchinol Exponential 0.99 0.5729 0.0047 80 122 243 982
4. Lontla Clench’s 0.98 2.8900 0.0391 60 74 87 2533
5. Mojonera Clench’s 0.99 2.6630 0.0282 79 94 149 3512
6. Malila Clench’s 0.95 2.3988 0.0345 58 69 127 2868
7. Tenango Exponential 0.98 1.3191 0.0144 87 92 215 320
8. Ahuacatl
an Exponential 0.99 2.3051 0.0386 59 60 86 119
9. Xochicoatl
an Exponential 0.97 2.1727 0.0291 74 75 118 158
10. Soyatla-1 Exponential 0.99 2.0622 0.0243 85 85 173 189
11. Soyatla-2 Exponential 0.98 1.6493 0.0232 66 71 110 198
12. Molocotl
an-1 Exponential 0.99 3.1686 0.0483 56 66 44 95
13. Molocotl
an-2 Exponential 0.99 4.9998 0.0797 59 63 30 58
Study area Exponential 0.99 0.3851 0.0021 181 184 2057 2196
a 1
Increase rate of the species list at the beginning of the collection with units of species
point counts .
b
Increase rate with units of point counts1 .
c
Predicted asymptote or predicted species richness (a=b).
d
Number of point counts required to register 99% of the total bird species richness.

surveyed only considered the point counts for the
months the species occurred in the study area. Since
detection probability for a certain species was the same
in the forest fragments surveyed, this allowed compari-
sons of the species abundance among forest fragments.
2.5. Analyses of the relationships between landscape
patterns and birds
The response of the standardized bird species richness
in cloud forest fragments as a function of forest frag-
ment size, shape, topographical complexity, altitudinal
range, connectivity, and proportion of landscape for-
ested area was analysed by multiple regression analyses.
The stepwise regression procedure was used to choose
the forest fragment characteristics that have the stron-
gest relationship with bird species richness in this cloud
forest landscape. An F -to-enter probability value of 0.05
and a F -to-remove probability value of 0.10 were used in
all cases. Additionally, forest interior, generalist, and
border species were analysed separately by multiple re-
gression analysis to assess if there was any differential
response to forest fragment characteristics, based on the
species level of dependence on forest.
Detrended correspondence analyses (DCA) were used
to generate scatter diagrams of cloud forest fragments
based on the structure of their bird community. In the
analyses the number of axes extracted in the ordination
analysis was based on the Kaiser’s rule, where the
minimum eigenvalue of an axis should be the average of
all eigenvalues (Legendre and Legendre, 1983). The
scores for axes were calculated by the Hill algorithm,
based on reciprocal averaging (Hill, 1973). Bird abun-
dance data were logðeÞx þ 1 transformed previous to the
analyses to reduce the skewness of the data, resulting in
a more interpretable analysis.
3. Results
The response of the bird species richness to all the
characteristics of the forest fragments analysed was not
statistically significant (R2adj ¼ 0:30, df ¼ 5, P ¼ 0:28), to
a certain extent probably due to the over-parametriza-
tion of the multiple regression. The standardized partial
regression coefficients (b) obtained, however, indicated
the relative importance of the forest fragment charac-
teristics on bird species richness (Table 3). A further
analysis was performed, but characteristics of forest
fragments were selected by the stepwise regression pro-
cedure. A better fit was obtained in the multiple re-
gression, where fragment shape was identified as the
main fragment feature influencing the response of bird
species richness (Table 3).
In the case of bird species of the forest interior and
forest generalist, the best fit of the multiple regression
analysis was obtained when fragment size was selected,
either when analysed separately or when pooled together
(Table 4). Bird species of the forest border did not sig-
nificantly respond to any forest fragment characteristic
(R2adj ¼ 0:06, df ¼ 4, P ¼ 0:57). The stepwise regression
procedure did not select any variable to build a regres-
sion model; thus, the backward elimination procedure
was used to explore the potential variables that might be
relevant in the response of forest border birds. This
procedure produced a marginally significant regression
122 M. Angel Mart
ınez-Morales / Biological Conservation 121 (2005) 117–126

Table 3
Summary of the results obtained from multiple regression analysis, where the response of bird species richness in cloud forest fragments is a function
of the characteristics of forest fragments
Characteristics of forest fragments Partial regression coefficient (b) Significance (P )
All variables included (R2adj
¼ 0:30, df ¼ 5, P ¼ 0:28)
(Constant) 0.985
Log10 size 0.175 0.956
Log10 shape 1.440 0.562
p
Arcsine topographical complexity 0.041 0.918
Log10 altitudinal range )0.568 0.654
1/Log10 connectivity – cf 1.100 0.616
1/Log10 connectivity – af )0.916 0.689
p
Arcsine proportion of landscape forested area )0.332 0.858
Stepwise regression applied (R2adj ¼ 0:60, df ¼ 11, P ¼ 0:001)
(Constant) <0.001
Log10 shape 0.794 0.001

Table 4
Summary of the results obtained from multiple regression analysis, where the response of bird species richness of forest interior, generalist, and
border birds in cloud forest fragments is a function of the characteristics of forest fragments. In the case of forest interior and generalist birds the
stepwise regression procedure was used, whereas the backward elimination procedure was applied for forest border species
Characteristics of forest fragments Partial regression coefficient (b) Significance (P )
Forest interior birds (R2adj ¼ 0:43, df ¼ 11, P ¼ 0:009)
(Constant) 0.445
Log10 size 0.694 0.009
Forest generalist birds (R2adj ¼ 0:87, df ¼ 11, P < 0:001)
(Constant) <0.001
Log10 size 0.938 <0.001
Forest interior and generalist birds (R2adj ¼ 0:66, df ¼ 11, P < 0:001)
(Constant) 0.001
Log10 size 0.827 <0.001
Forest border birds (R2adj ¼ 0:23, df ¼ 9, P ¼ 0:124)
(Constant) 0.064
Log10 size )2.637 0.062
Log10 shape 2.344 0.091

model (R2adj ¼ 0:23, df ¼ 9, P ¼ 0:12) using size and
shape as the most important variables. A negative re-
sponse of forest border birds to fragment size was de-
tected, as was a positive response to shape (Table 4).
Through the arrangement of forest fragments based
on their similarities in bird species abundance and
composition (Figs. 2 and 3), the DCA showed that bird
species assemblages in forest fragments are correlated
with the intensification of cloud forest fragmentation.
Axis 1 in both figures showed a fairly clear gradient in
the intensification of forest fragmentation from low to
high. In addition, axis 2 in both figures seemed to rep-
resent differences in bird species assemblages among
forest fragments caused by topographical and geo-
graphical features. On axis 2 of Fig. 2, fragment 6
(Malila) was isolated from all other fragments because it
had prominent cliffs that were heavily used as roosting
sites by birds such as Tachycineta thalassina (violet-
green swallow) and Strectoprocne zonaris (white-col-
lared swift). The two extreme forest fragments on axis 2
of Fig. 3 (Tenango [7] and Mojonera [5]) do not overlap
altitudinally. The Tenango fragment, located in the
lower range of the cloud forest in Hidalgo, holds species
occurring typically below 1500 m in altitude, such as
Crypturellus cinnamomeus (thicket tinamou), Columba
flavirostris (red-billed pigeon), Attila spadiceus (bright-
rumped attila), Pitangus sulphuratus (great kiskadee),
and Cyanocorax morio (brown jay). On the other hand,
the Mojonera fragment, located in the upper range of
the cloud forest in Hidalgo, was separated from the
other fragments because of the presence of Certhia
americana (brown creeper), which was only recorded in
the upper half of the altitudinal range of the cloud for-
est. Furthermore, since the Mojonera fragment is lo-
cated on the leeward slope of the mountain chain
(contrary to all other forest fragments), it is drier than
other sites. Therefore, bird species such as Cynanthus
latirostris (broad-billed hummingbird), Psaltriparus
minimus (bushtit), and Bombycilla cedrorum (cedar
waxwing), which are associated with drier climates, were
only present in this fragment, contributing to the dis-
tinctiveness of the Mojonera fragment. Fig. 3 also
 M. Angel Mart
ınez-Morales / Biological Conservation 121 (2005) 117–126 123

1.5

(eigenvalue = 0.08, explained variance = 7%)

'7

'3
1.0
'4 '11

'2 '10
Axis 2 '5
'13
'12
'1

0.5 '9
'8

0.0 '6

0.0 0.5 1.0 1.5 2.0

Axis 1
(eigenvalue = 0.25, explained variance = 23%)

Fig. 2. DCA ordination diagram of cloud forest fragments based on the logðeÞ transformed abundance data of bird species. The scale marks in both
axes are multiples of standard deviation. Labels of forest fragments follow Table 1.

5
(eigenvalue = 0.13, explained variance = 13%)

'7
4

3 '4
'8
Axis 2

'6 '9 '10 '13

'3 '12
'1 '11
2

'2
1

0 '5

0 1 2 3 4

Axis 1
(eigenvalue = 0.19, explained variance = 19%)

Fig. 3. DCA ordination diagram of cloud forest fragments based on the presence-absence data of forest interior and generalist bird species. The scale
marks in both axes are multiples of standard deviation. Labels of forest fragments follow Table 1.

showed that forest fragments smaller than 50 ha were documented both in temperate and tropical forests at
quite homogeneous in their bird species assemblages, in several spatial scales (Ambuel and Temple, 1983; Howe,
contrast to fragments larger than 200 ha. 1984; Blake and Karr, 1987; Willson et al., 1994; Bell-
amy et al., 1996; Kattan and Alvarez-L
opez, 1996;
Stratford and Stouffer, 1999; Cornelius et al., 2000). In
4. Discussion this study, however, fragment shape was identified as the
most important fragment characteristic influencing the
4.1. Response of bird species to forest fragment charac- response of species richness in the bird community.
teristics Since fragment shape represents an index of the pro-
portional extent of forest edge in a fragment, its im-
The positive response of bird species richness to forest portance may represent the influence of bird species that
fragment size has been extensively and consistently benefit from the increase in edge extent due to the ir-
124 M. Angel Mart
ınez-Morales / Biological Conservation 121 (2005) 117–126
regularity of forest fragments. Such is the case of forest
border species and birds from the vegetation matrix.
The relevance of fragment shape in this landscape may
be the consequence of the fragmented nature of the
cloud forest in a topographically complex region, and
more recently, because of the deforestation patterns that
have taken place. Deforestation has not followed a
regular pattern mainly because of the topographic
complexity of the region; therefore, most of the forest
fragments are highly irregular. This fact, together with
the intensity of deforestation may have had important
repercussions on the composition and structure of the
cloud forest bird community to such an extent that
area has become less relevant as a predictor of species
richness.
The characteristics of the cloud forest fragments that
produced a differential response of bird species de-
pended on their level of dependence on forest. Among
the forest fragment characteristics analysed, size was
identified as the most important feature having a posi-
tive relationship with species richness of forest depen-
dent birds, that is to say, forest interior and generalist
species. In contrast, forest border species responded
negatively to size of forest fragments, but responded
positively to fragment shape. Other works suggest that
such differential response of bird species to forest frag-
mentation seems to remain irrespective of the scale of
analysis, even in extremely fragmented landscapes
(Bellamy et al., 1996; Graham and Blake, 2001). In
south-east England, Bellamy et al. (1996) found the
same pattern of differential response to forest fragmen-
tation by woodland and edge species, in wood fragments
smaller than 30 ha, where most, if not all fragment area
could be considered as border. Although a differential
response of forest dependent and border dependent
species is evident, both segments of the bird community
responded positively to the extent of their suitable
habitat, that is to say, the extent of forest or the extent
of forest border, respectively.
Connectivity of forest fragments, either by cloud
forest or by any forest type, did not seem to influence
bird species richness at the spatial scale of hundreds of
meters between fragments. This fact has also been re-
ported in other studies in temperate forest fragments
(Ambuel and Temple, 1983). Thus, the persistence of a
given species in a fragment depends upon its regional
population dynamics far more than its population in
that specific forest fragment. However, connectivity of
forest fragments might become important at greater
spatial scales. Price et al. (1999) have demonstrated that
populations of some bird species are regulated in some
manner by the extent of forest outside the patch in
which they occur. They suggested that the spatial scale
of this effect is within a 50 km radius. Furthermore,
connectivity of forest fragments might also become im-
portant when connectivity along an altitudinal gradient
is involved. In this study, an effect of the altitudinal
location of forest fragments on their species composition
was suggested. In species composition, forest fragments
located at the upper altitudinal range of the cloud forest
differed to some extent from the fragments located at the
lower altitudinal range. Therefore, connectivity of forest
fragments with other vegetation types within an altitu-
dinal context seemed to be a relevant factor affecting
bird diversity and ecological processes, such as altitu-
dinal migration, as shown by studies on distribution and
turn over of bird species in altitudinal gradients (Ter-
borgh, 1971; Terborgh and Weske, 1975; Navarro, 1992;
Kattan et al., 1994).
An additional characteristic of forest fragments af-
fecting the bird community was the windward or lee-
ward position of the fragment on the mountain range.
Cloud forest fragments located on the windward slope
are more humid than the fragments located on the lee-
ward slope. Consequently, forest fragments showed
differences in species composition due to the presence of
birds with affinities either for more humid or drier
habitats.
Particular features of forest fragments that are nee-
ded to fulfil the specific requirements of certain bird
species, such as roosting and shelter sites, feeding and
nesting grounds, and protection from predators will
affect the presence and abundance of bird species. In this
study, prominent cliffs were identified as key features of
forest fragments affecting the relative abundance of
some aerial bird species, mainly swifts and swallows.
Some other studies have also identified cliffs as impor-
tant landscape features that represent a significant ad-
ditive influence on avian diversity (Ward and Anderson,
1988; Matheson and Larson, 1998), and also, as im-
portant landscape features for conservation of biodi-
versity in general (Kelly and Larson, 1997; Larson et al.,
1999a; Larson et al., 1999b). Cliffs lend topographical
diversity, and thus, habitat diversity to forest fragments.
Although they may only occupy a small percentage of
the land area, they seem to contribute disproportion-
ately to habitat richness.
4.2. Conservation implications
Forest fragment size was the most important char-
acteristic influencing the response of bird species that
depend on cloud forest; thus, this study suggests that the
immediate and most effective ecologically oriented con-
servation strategy for this region has to be directed to
the conservation of the largest cloud forest fragments.
This will ensure that bird species of particular conser-
vation value, such as species restricted to the cloud
forest, endemic species, and restricted-range species,
among others, will be preserved in landscapes with the
lowest fragmentation intensity. This strategy should
reduce the extinction probability of forest species of
 M. Angel Mart
ınez-Morales / Biological Conservation 121 (2005) 117–126
conservation concern, although the assessment of pop-
ulation viability should also incorporate information on
reproductive success under different fragmentation
scenarios.
Connectivity seemed to be irrelevant for birds moving
between proximate forest fragments. This might be the
result of the historical exposure of birds to the frag-
mented nature of cloud forest. However, the disruption
of forest connectivity involving an altitudinal gradient
has the potential of inducing important adverse effects
on populations of species located at the limits of their
altitudinal range, and also on altitudinal migrants, by
disrupting migratory routes. Therefore, forest continuity
through an altitudinal gradient has to be favoured in the
conservation strategies of the region.
Acknowledgements
Thanks are due to the people of the local communi-
ties that I visited during fieldwork in Hidalgo; to J.L.
Villarreal and L. Heras for help in the analysis of the
satellite image, to the Instituto de Geograf
ıa, UNAM
for the satellite image; to G. Williams-Linera for field-
work advice; to P. Callow and P. Altham for statistical
support; and to M. de L. Brooke, S. Luque, A. Balm-
ford, and N. Davis for comments. M. Mart
ınez-Ramos,
J. Ben
ıtez, D. Gernandt, M.W. Schwartz, and three
anonymous reviewers provided comments to improve
the manuscript. Institutional support was provided by P.
Escalante and V. N
equiz from the Instituto de Biolog
ıa,
UNAM, and D. Wege from BirdLife International. Fi-
nancial support was obtained from CONACYT,
CONABIO, British Airways, and the Cambridge
Philosophical Society.
References
Ambuel, B., Temple, S.A., 1983. Area-dependent changes in the bird
communities and vegetation of southern Wisconsin forests. Ecol-
ogy 64, 1057–1068.
Arango-V
elez, N., Katan, G.H., 1997. Effects of forest fragmentation
on experimental nest predation in Andean cloud forest. Biological
Conservation 81, 137–143.
Bellamy, P.E., Hinsley, S.A., Newton, I., 1996. Factors influencing
bird species numbers in small woods in south-east England. Journal
of Applied Ecology 33, 249–262.
Bierregaard Jr., R.O., Lovejoy, T.E., Kapos, V., Agusto dos Santos,
A., Hutchings, R.W., 1992. The biological dynamics of tropical
rainforest fragments. A prospective comparison of fragments and
continuous forest. BioScience 42, 859–866.
Blake, J.G., Karr, J.R., 1987. Breeding birds of isolated woodlots: area
and habitat relationships. Ecology 68, 1724–1734.
Cornelius, C., Cofr
e, H., Marquet, P.A., 2000. Effects of habitat
fragmentation on bird species in a relict temperate forest in
semiarid Chile. Conservation Biology 14, 534–543.
Didham, R.K., 1997. The influence of edge effects and forest fragmen-
tation on leaf litter invertebrates in central Amazonia. In: Laurance,
125
W.F., Bierregaard, R.O. (Eds.), Tropical Forest Remnants: Ecol-
ogy, Management, and Conservation of Fragmented Communities.
The University of Chicago Press, Chicago, pp. 55–70.
Escalona, G., Torres, M., Navarro, A.G., Villal
on, R., Hern
andez, B.,
Ben
ıtez, H., 1995. Migratory birds of the cloud forest of Mexico.
In: Wilson, M.H., Sader, S.A. (Eds.), Conservation of Neotropical
Migratory Birds in Mexico. Maine Agricultural and Forest
Experiment Station, Miscellaneous Publication 727, USA, pp.
15–33.
Flores-Villela, O., Gerez, P., 1994. Biodiversidad y conservaci
on en
M
exico: vertebrados, vegetaci
on y uso del suelo. CONABIO-
UNAM, Mexico.
G
omez-Pompa, A., Dirzo, R., Kaus, A., Noguer
on-Chang, C.R.,
Ord
~ez, M., de, J., 1995. Reservas de la bi

on osfera y otras
areas
naturales protegidas de M
exico. SEMARNARP-INE-CONABIO,
Mexico.
Graham, C.H., Blake, J.G., 2001. Influence of patch- and landscape-
level factors on bird assemblages in a fragmented tropical
landscape. Ecological Applications 11, 1709–1721.
Hern
andez-Ba~ nos, B.E., Peterson, A.T., Navarro-Sig€ uenza, A.G.,
Escalante-Pliego, B.P., 1995. Bird faunas of the humid montane
forests of Mesoamerica: biogeographic patterns and priorities for
conservation. Bird Conservation International 5, 251–277.
Hill, M.O., 1973. Reciprocal averaging: an eigenvector method of
ordination. Journal of Ecology 61, 237–249.
Howe, R.W., 1984. Local dynamics of bird assemblages in small forest
habitat islands in Australia and North America. Ecology 65, 1585–
1601.
Kapos, V., Wandelli, E., Camargo, J.L., Ganade, G., 1997. Edge-
related changes in environment and plant responses due to forest
fragmentation in central Amazonia. In: Laurance, W.F., Bierreg-
aard Jr., R.O. (Eds.), Tropical Forest Remnants: Ecology, Man-
agement, and Conservation of Fragmented Communities. The
University of Chicago Press, Chicago, pp. 33–44.
Kattan, G.H., Alvarez-L
opez, H., 1996. Preservation and management
of biodiversity in fragmented landscapes in the Colombian Andes.
In: Schelhas, J., Greenberg, R. (Eds.), Forest Patches in Tropical
Landscapes. Island Press, Washington DC, pp. 3–18.
Kattan, G.H., Alvarez-L
opez, H., Giraldo, M., 1994. Forest fragmen-
tation and bird extinctions: San Antonio eighty years later.
Conservation Biology 8, 138–146.
Kelly, P.E., Larson, D.W., 1997. Effects of rock climbing on
populations of presettlement eastern white cedar (Thuja occiden-
talis) on cliffs of the Niagara escarpment, Canada. Conservation
Biology 11, 1125–1132.
Larson, D.W., Matthes, U., Kelly, P.E., 1999a. Cliffs as natural
refuges. American Scientist 87, 411–417.
Larson, D.W., Matthes, U., Gerrath, J.A., Gerrath, J.M., Nekola,
J.C., Walker, G.L., Porembski, S., Charlton, A., Larson, N.W.K.,
1999b. Ancient stunted trees on cliffs. Nature 398, 382–383.
Laurance, W.F., 1994. Rainforest fragmentation and the structure of
small mammal communities in tropical Queensland. Biological
Conservation 69, 23–32.
Laurance, W.F., Bierregaard, R.O. (Eds.), 1997. Tropical Forest
Remnants: Ecology, Management, and Conservation of Frag-
mented Communities. The University of Chicago Press, Chicago.
Legendre, L., Legendre Jr., P., 1983. Numerical Ecology. Elsevier
Scientific Publishing Company, New York.
Lovejoy, T.E., Bierregaard Jr., R.O., Rylands, A.B., Malcolm, J.R.,
Quintela, C.E., Harper, L.H., Brown Jr., K.S., Powell, A.H.,
Schubart, H.O.R., Hays, M.B., 1986. Edge and other effects of
isolation on Amazon forest fragments. In: Soul
e, M.E. (Ed.),
Conservation Biology: The Science of Scarcity and Diversity.
Sinauer Associates, Sunderland, MA, pp. 257–285.
Luna, I., Alc
antara, O., Morrone, J.J., Espinosa, D., 2000. Track
analysis and conservation priorities in the cloud forests of Hidalgo,
Mexico. Diversity and Distributions 6, 137–143.
126 M. Angel Mart
ınez-Morales / Biological Conservation 121 (2005) 117–126
Mart
ınez-Morales, M.A., 2001. Forest fragmentation effects on bird
communities of tropical montane cloud forest in eastern Mexico.
PhD thesis, University of Cambridge, Cambridge.
Matheson, J.D., Larson, D.W., 1998. Influence of cliffs on bird
community diversity. Canadian Journal of Zoology 76, 278–
287.
Mesquita, R.C.G., Delam^ onica, P., Laurance, W.F., 1999. Effects of
the surrounding vegetation on edge-related tree mortality in
Amazonian forest fragments. Biological Conservation 91, 129–
134.
Murcia, C., 1995. Edge effects in fragmented forests: implications for
conservation. Trends in Ecology and Evolution 10, 58–62.
Navarro, A.G., 1992. Altitudinal distribution of birds in the Sierra
Madre del Sur, Guerrero, Mexico. The Condor 94, 29–39.
Paton, P.W.C., 1994. The effects of edge on avian nest success: how
strong is the evidence? Conservation Biology 8, 17–26.
Price, O.F., Woinarski, J.C.Z., Robinson, D., 1999. Very large area
requirements for frugivorous birds in monsoon rainforest of the
Northern Territory, Australia. Biological Conservation 91, 169–
180.
Ralph, C.J., Sauer, J.R., Droege, S., technical editors, 1995. Monitor-
ing bird populations by point counts. Gen. Tech. Rep. PSW-GTR-
149. Albany, CA: Pacific Southwest Research Station, Forest
Service, US Department of Agriculture.
Ramammoorthy, T.P., Bye, R., Lot, A., Fa, J. (Eds.), 1998. Diversidad
biol

ogica de M
exico: or
ıgenes y distribuci

on. Instituto de Biolog
ıa,
UNAM, Mexico.
Robinson, S.K., Thompson III, F.R., Donovan, T.M., Whitehead,
D.R., Faaborg, J., 1995. Regional forest fragmentation and the
nesting success of migratory birds. Science 267, 1987–1990.
Sober
on, J., Llorente, J., 1993. The use of species accumulation
functions for the prediction of species richness. Conservation
Biology 7, 480–488.
Stattersfield, A.J., Crosby, M.J., Long, A.J., Wege, s.C., 1998.
Endemic bird areas of the world. Priorities for biodiversity
conservation. BirdLife Conservation Series No. 7. Birdlife Inter-
national, Great Britain.
Stratford, J.A., Stouffer, P.C., 1999. Local extinctions of terrestrial
insectivorous birds in a fragmented landscape near Manaus, Brazil.
Conservation Biology 13, 1416–1423.
Terborgh, J., 1971. Distribution on environmental gradients: theory
and a preliminary interpretation of distributional patterns in the
avifauna of the Cordillera Vilcabamba, Peru. Ecology 51, 23–40.
Terborgh, J., Weske, J.S., 1975. The role of competition in the
distribution of Andean birds. Ecology 56, 562–576.
Turner, M.G., 1990. Spatial and temporal analysis of landscape
patterns. Landscape Ecology 4, 21–30.
Turton, S.M., Freiburger, H.J., 1997. Edge and aspect effects on the
microclimate of a small tropical forest remnant on the Atherton
Tableland, northeastern Australia. In: Laurance, W.F., Bierregaard
Jr., R.O. (Eds.), Tropical Forest Remnants: Ecology, Management,
and Conservation of Fragmented Communities. The University of
Chicago Press, Chicago, pp. 45–54.
V
azquez-Garc
ıa, J.A., 1995. Cloud forest archipelagos: preservation of
fragmented montane ecosystems in tropical America. In: Hamilton,
L.S., Jivik, J.O. (Eds.), Tropical Montane Cloud Forests. Ecolog-
ical Studies, vol. 110.
Ward, J.P., Anderson, S.H., 1988. Influences of cliffs on wildlife
communities in southcentral Wyoming. Journal of Wildlife Man-
agement 52, 673–678.
Williams-Linera, G., 1993. Vegetaci
on de bordes de un bosque
nublado en el Parque Ecol
ogico Clavijero, Xalapa, Veracruz,
M
exico. Revista de Biolog
ıa Tropical 41, 443–453.
Willson, M.F., de Santo, T.L., Sabag, C., Armesto, J.J., 1994. Avian
communities of fragmented south-temperate rainforest in Chile.
Conservation Biology 8, 508–520.