Ecological Monographs, 80(4), 2010, pp. 651?669
C211 2010 by the Ecological Society of America
Landscape matrix and species traits mediate responses of Neotropical
resident birds to forest fragmentation in Jamaica
CHRISTINA M. KENNEDY,
1,4
PETER P. MARRA,
2
WILLIAM F. FAGAN,
1
AND MAILE C. NEEL
3
1
Behavior, Ecology, Evolution and Systematics, Department of Biology, University of Maryland, College Park, Maryland 20742 USA
2
Smithsonian Migratory Bird Center, National Zoological Park, 3001 Connecticut Avenue, Washington, D.C. 20008 USA
3
Department of Plant Science and Landscape Architecture and Department of Entomology, University of Maryland,
College Park, Maryland 20742 USA
Abstract. Land cover and land use surrounding fragmented habitat can greatly impact
species persistence by altering resource availability, edge effects, or the movement of
individuals throughout a landscape. Despite the potential importance of the landscape matrix,
ecologists still have limited understanding of the relative effects of different types of land cover
and land uses on species patterns and processes in natural systems. Here we investigated
whether Neotropical resident bird communities in limestone forest patches differed if they
were embedded in three different human-dominated matrix types (agriculture, peri-urban
development, and bauxite mining) relative to sites in continuous forest in central Jamaica. We
found that species richness, community composition, and abundances were matrix-dependent,
with agricultural landscapes supporting greater avian diversity and more intact community
assemblages than either peri-urban or bauxite landscapes. Abundance of almost 70% of
species differed in forest embedded in the different landscape matrix types. Traits related to
resource use best predicted species responses, including diet guild, nest height, habitat
association, and foraging strata. Insectivores, frugivores, canopy nesters, understory and
canopy foragers, and forest-restricted species rarely observed in matrix habitats had lower
abundances in forest fragments embedded in human-dominated matrix types than in
continuous forest. In contrast, nectarivores, omnivores, granivores, ground and multi-strata
nesters, ground foragers, and species regularly in matrix habitats were least sensitive to forest
fragmentation. Results suggest that structure, composition, and land use disturbance regimes
in matrix areas impact overall habitat quality in landscapes by potentially mediating resource
availability inside as well as outside forest habitat. This study reinforces the importance of
differentiating among land cover and land uses in fragmentation research and lends support to
the hypothesis that resource availability may be a primary factor driving Neotropical bird
responses to fragmentation.
Key words: birds; Caribbean; community ecology; habitat fragmentation; life history traits; matrix
effects; Neotropics; tropical conservation.
INTRODUCTION
In early assessments of forest fragmentation effects,
landscapes were often simply dichotomized between
forest (i.e., habitat) or non-forest (or non-habitat). This
binary view was due in part to strong influences from
island biogeography (MacArthur and Wilson 1967) and
later to metapopulation theories (Hanski 1998). The
drastic contrast in structure and species composition
between natural forest remnants and modified non-
forest habitats made this simplification seem reasonable,
and many insights regarding the effects of habitat loss
and isolation were gained under this landscape model
(Fahrig 2003). In reality though, landscapes are complex
mosaics of heterogeneous land cover types and species
are affected not only by the size, shape, and spatial
location of primary habitat, but also by the structure
and the composition of the land cover surrounding
isolated patches (termed ??the matrix??) (Haila 2002,
Kupfer et al. 2006).
Despite the potential importance of the matrix, its
effects are only beginning to be understood (Kupfer et
al. 2006), and empirical data remain limited in terms of
the range of matrix types, species responses, and
potential mechanisms that have been examined. Most
studies to date have compared species patterns (i.e.,
richness, community composition, or occupancy) in
forest patches surrounded by intensely human-modified
matrix (typically agriculture) to those found in patches
surrounded by matrix types that are more similar in
structure to once-contiguous forest. Specifically for bird
communities, researchers have compared forest patches
embedded in pasture to patches embedded in secondary
forest (Laurance et al. 2002), exotic tree plantations
Manuscript received 22 May 2009; revised 19 January 2010;
accepted 18 February 2010. Corresponding Editor: T. R.
Simons.
4
Present address: Department of Plant Science and
Landscape Architecture, University of Maryland, 2331 Plant
Sciences Building, College Park, Maryland 20742-4452.
E-mail: cmk6@umd.edu
651
(Estades and Temple 1999, Hobson and Bayne 2000,
Renjifo 2001, Lindenmayer et al. 2002, Wethered and
Lawes 2003), or managed forests (i.e., undergoing
silviculture; Andren 1992, Aberg et al. 1995, Bayne
and Hobson 1997, Norton et al. 2000). Results based on
comparisons of agricultural vs. forested landscapes
indicate that matrix effects on species are strongest
when structural contrast to remnant habitat is high (i.e.,
when vegetation structure differs in physiognomy or
cover). Matrix types with high structural contrast are
thus predicted to be less permeable to movement and
more hostile to species than those with low-contrast
boundaries (e.g., Stamps et al. 1987, Forman 1995,
Strayer et al. 2003). Therefore, in forested landscapes,
land uses that create ??open?? deforested habitats, such as
for agriculture or residential development, may be
considered equally hostile to forest-dwelling species.
This prediction, however, remains largely untested given
the paucity of studies that examine the relative impacts
of different types of human-modified land cover on
species patterns and processes.
The most commonly studied mechanism by which
matrix type affects species is its effect on inter-patch
movement (i.e., dispersal hypothesis) (e.g., Ricketts
2001, Gobeil and Villard 2002, Revilla et al. 2004,
Bender and Fahrig 2005). The landscape matrix,
however, can mediate processes that influence species?
area and isolation relationships other than dispersal
(Ewers and Didham 2006, Kupfer et al. 2006). First,
different matrix types may provide alternative or
supplemental resources (e.g., food or nest sites) that
support greater species abundances than would be
expected based on amounts or quality of primary
habitat alone (i.e., habitat compensation hypothesis)
(e.g., Gascon et al. 1999, Norton et al. 2000, Brotons et
al. 2003, Luck and Daily 2003, Cook et al. 2004).
Second, the type and extent of contrast between matrix
habitats and forest remnants can also mediate the
magnitude of edge effects, such as nest predation and
parasitism (Donovan et al. 1997, Chalfoun et al. 2002,
Driscoll and Donovan 2004), or the alteration of within-
patch conditions (e.g., vegetation structure, microcli-
mate) (Saunders et al. 1991, Ries et al. 2004) (i.e., edge
effects hypothesis). Third, anthropogenic land use
disturbance, such as hunting, logging, noise, and
burning in remaining habitat patches, may vary as a
function of matrix type (i.e., disturbance hypothesis)
(Hobbs 2001, Laurance and Cochrane 2001, Peres
2001). Collectively, these hypotheses outline mecha-
nisms by which ??matrix effects?? may drive colonization?
extinction dynamics in fragmented landscapes (e.g.,
Vandermeer and Carvajal 2001), with greater research
attention having been focused on the former (dispersal
and edge effects) rather than the latter (disturbance)
hypotheses.
With few notable exceptions (e.g., Renjifo 2001,
Rodewald and Yahner 2001, Rodewald and Bakermans
2006), previous empirical research has failed to identify
unique impacts of the matrix in part because forest
amount and spatial configuration are commonly con-
founded with one another and because the amount of
forest is further confounded with matrix composition
and associated human land uses (Rodewald 2003,
Laurance 2008). Avoiding such conflation of factors is
essential, because the influence due to habitat elements
likely outweighs the influence of matrix elements
(Goodwin and Fahrig 2002); thus, matrix effects can
easily go undetected. Moreover, multiple human-mod-
ified land cover types with similar structural ??edge
contrast?? (Strayer et al. 2003) are rarely examined in a
single setting; thus, few empirical data exist to evaluate
the relative impact of different human activities on
fauna within forest remnants.
Here, we investigate how forest-dependent Neotrop-
ical resident birds in Jamaica respond to habitat patches
in landscapes that are similar in structural habitat
fragmentation but that are surrounded by different
human-modified land cover types. Specifically, we
examined whether species richness, community compo-
sition, or abundances of birds differed in forest
fragments embedded in three dominant land use types
in the Caribbean: agriculture (i.e., pasture), residential
(peri-urban) development, or mining for bauxite (i.e.,
human-dominated matrices) relative to sites in contin-
uous forest (i.e., natural ??matrix??). We sampled forest
patches in replicate human-dominated landscapes that
were similar in the major components of fragmentation
(i.e., forest amount and configuration) to isolate the
influence of the surrounding matrix. Throughout the
Caribbean (Lugo 2002, Evelyn and Camirand 2003) and
in many other parts of the world (DeFries et al. 2004),
land cover is increasingly being converted from agricul-
ture to more intensive urban and extractive land uses.
Despite their pervasiveness, these intensively developed
matrix types have rarely been assessed to understand
whether they differentially impact species existing in
fragmented forest remnants.
When different landscape matrices have been exam-
ined, species responses to habitat fragmentation often
appear idiosyncratic (e.g., Mac Nally 2007). Even
though researchers have addressed associations between
species traits and species responses to fragmentation for
over 20 years (e.g., Laurance 1991, Hansen and Urban
1992, Pereira et al. 2004, Lampila et al. 2005, Sigel et al.
2006), a clear consensus has yet to emerge on which
traits govern species-specific responses and why (Henle
et al. 2004). Moreover, patterns detected are often
contingent upon which landscape matrix types are
examined (Bender et al. 1998, Debinski and Holt 2000,
Ewers and Didham 2006), a factor that is not often
explicitly addressed. To elucidate potential mechanistic
explanations underlying species responses to forest
conversion, we selected a suite of traits proposed to
influence species persistence in fragmented systems
(Ewers and Didham 2006). We considered phylogenetic
CHRISTINA M. KENNEDY ET AL.652
Ecological Monographs
Vol. 80, No. 4
relatedness, body size, rarity, geographic and altitudinal
ranges, clutch size, nest type, nest height, diet guild,
foraging strata, and habitat association. Each of these
traits can affect dispersal ability, resource acquisition,
and/or population growth potential (Henle et al. 2004);
thus, we predicted these traits would relate to species
responses to human-altered forested landscapes.
METHODS
Study area
We conducted research in Manchester and Clarendon
Parishes in central Jamaica (17856
0
24
00
?18811
0
6
00
N,
77823
0
13
00
?77837
0
5
00
W). This region lies in the premon-
tane moist forest climatic zone (Holdridge 1967), with
mean annual temperature of 268C. Rainfall is bimodal,
peaking in May/June and September/October, with 1000
mm/yr average (Jamaican Meteorological Service, un-
published data). The karst substrate that underlies the
region yields a topographically complex landscape with
numerous white limestone hills and plateaus, ranging
between ;400 and 800 m elevation (Porter 1990) and
separated from one another by small to moderate-sized
valleys.
In pre-Columbian times, this region was covered in
wet limestone forest (Asprey and Robbins 1953)
composed of evergreen and semi-deciduous trees,
referred to as Evergreen Season Forest formation
(Beard 1944, 1955). Dominant canopy species included
broadleaf (Terminalia latifolia), Jamaican cedar (Cedrela
odorata), sweetwoods (Nectandra spp.), and bulletwoods
(Daphnopsis spp.) (Asprey and Robbins 1953). Today,
,30% of native forest remains (Evelyn and Camirand
2003), which is within the range at which the effects of
fragmentation are postulated to occur (Fahrig 2003).
Forest is now restricted to small hilltop remnants on
limestone outcrops, surrounded by valleys cleared for
three dominant land uses: agriculture (i.e., primarily
cattle pasture), residential development, and mining for
bauxite. The vast majority of forest fragments in this
region are C20100 ha, with only a few large forest tracts
remaining along hilltop ridges. Large-scale deforestation
occurred by the 18th century, largely for agricultural
development (Eyre 1987a, b). Since the 1950s, land cover
change in Jamaica has resulted largely from conversion
of agriculture to residential development and mining for
bauxite. The exact rate of ongoing deforestation in the
region is unknown but is estimated to be as low as 0.1%
annually for the country, with current land use change
largely occurring among human-modified matrices
(Evelyn and Camirand 2003). We are thus investigating
the role of the matrix in a region that has undergone
historic rather than contemporary forest loss and
fragmentation. Locations and extent of forest fragments
in this region have remained fairly stationary in recent
time but are embedded within a changing matrix. This
setting provided a unique opportunity to investigate the
influence of matrix land cover on Neotropical birds in
fragmented forests.
Site selection
We surveyed 20 1-km
2
landscapes that were typical of
land cover patterns in central Jamaica: six landscapes
comprising continuous forest and 14 landscapes in
which forest has been fragmented by agriculture (N ?
5), by residential (peri-urban) development (N ? 4), or
by bauxite mining (N ? 5; Fig. 1). The 1-km
2
scale is
biologically relevant given known territory sizes of forest
songbirds (Robbins et al. 1989, Terborgh et al. 1990)
and movement patterns for Jamaican birds (Cruz 1981,
Kennedy 2009). To increase the probability that birds
were independently sampled, replicate landscapes were
separated by .1?26 km, with the exception of peri-
urban landscapes due to logistical constraints. This
distance range should be sufficient to prevent overlap in
territories or daily movement of the majority of
Jamaican forest birds within our sampling periods
(A. M. Haynes-Sutton, personal communication;S.E.
Koenig, unpublished data). We identified landscapes
based on 2001?2002 IKONOS imagery, land cover
maps (Forestry Department 1999), and field verification.
To assess the spatial characteristics of forest habitat, we
digitized forest cover from IKONOS multispectral pan-
sharpened imagery (1-m resolution; Space Imaging
2002) and ground-truthing surveys using ArcGIS 9.3
(ESRI 2008). Forest habitat was categorized by a closed
canopy and visual dominance of native broadleaf trees;
producer?s and user?s accuracy for this cover type were
estimated at 92%and 84%, respectively (Kennedy 2009).
We selected fragmented landscapes that contained a
similar proportion and spatial configuration of remnant
forest, but that were dominated by only one of the three
target matrix types. Human-dominated landscapes
contained ;36% of forest cover and ;20 fragments
that were an average of 4?6 ha. Shape complexity of
forest fragments (1.3?1.4 perimeter-area fractal dimen-
sion), inter-patch distances (20?30 m), and patch
connectivity (;34% of patches interconnected) were
also similar among the different landscape matrix types.
(Comparisons of patch- and landscape-level patterns of
forest fragmentation in agricultural, peri-urban, and
bauxite landscapes are detailed in Appendix A.) To
serve as reference sites, we selected landscapes in the
largest intact forested areas in the region (Fig. 1).
Matrix land cover composition and vegetation struc-
ture differed substantially among the three types of
human-dominated landscapes. We conducted .700
vegetation surveys documenting both land cover com-
position and foliage structure in matrix areas (see
Appendix B for full details). Agricultural landscapes
were dominated by introduced pasture and herbaceous
gardens (;60%), followed by tree-lined fencerows
(18%), paddock trees (9%), and secondary growth of
Acacia stands (10%) that were interspersed in valleys
between forested hilltops. Peri-urban landscapes con-
sisted mainly of low-density residential housing and
roads that were surrounded by lawns (;10%), herba-
ceous gardens (9%), fruiting tree gardens (26%),
November 2010 653AVIAN RESPONSE TO LANDSCAPE MATRIX
ornamental shrubbery (19%), ornamental trees (12%),
and mixed woodlands (11%). Bauxite landscapes were
former agricultural lands that had been converted to
mining within the past 10 years; relictual forests were
surrounded by exposed bauxitic soils with vegetation
cover dominated by planted grassland or ferns (;78%)
and recent growth of Acacia trees (19%). Peri-urban and
agricultural matrices had greater foliage cover and
vertical complexity than bauxite lands, largely due to
the presence of scattered trees in peri-urban areas (i.e.,
ornamental tree gardens, vacant woodlots) and in
agricultural pasture (i.e., paddock trees, live fences,
fencerows; Appendix B).
We surveyed an average of five forest fragments per
replicate landscape using stratified random sampling to
represent the patch size distribution. Twenty-two of
these fragments were sampled in an agricultural matrix,
19 in a peri-urban matrix, and 27 in a bauxite-mining
matrix. Qualitative assessment based on aerial photo-
graphs taken in 1968 indicated that sampled fragments
have been in existence for at least 40 years (Evelyn
1997), although the nature of matrices may have
changed within this time period. Within forested
landscapes, we selected 31 ??pseudo-patches?? by ran-
domly accumulating consecutive samples along transects
that were located in continuous forest, such that
sampled areas were approximately equal in size to
patches in fragmented landscapes. We refer to both
pseudo-patches sampled in a natural ??matrix?? and forest
fragments sampled in human-modified matrices as
??patches.?? In total, 99 forest patches were sampled
across 20 landscapes.
Sampled patches also had similar forest area (3.89 6
0.45 ha; mean 6 SE) and isolation (33.58 6 3.48 m,
160.80 6 19.79 m, and 2381.75 6 147.07 m to the
nearest fragment.0.5 ha, 5 ha, and 100 ha, respectively;
Appendix A). All patches across all landscapes were
located between 400 and 800 m elevation. Even though
surveyed landscapes had similar environmental condi-
tions (e.g., elevation, climate, soil substrate), vegetation
structure in forest patches differed among the four
matrix types based on vegetation surveys conducted at
point count stations (Appendix C). Patches in agricul-
tural landscapes, and to a lesser extent in continuous
forest, had greater stand basal area, leaf area index, tree
diameter, tree canopy height, and tree cover than
patches in peri-urban and bauxite landscapes. Forest
fragments embedded in bauxite and peri-urban matrices
had lower and relatively more open canopies and a
greater proportion of herbaceous cover and low shrubs
(Appendix C). We did not exclude patches where
selective logging or moderate human use has occurred,
because (1) all forest in the region is unprotected
secondary forest that is subject to ongoing human
FIG. 1. Locations of the 20 1-km
2
landscapes surveyed in Manchester and Clarendon Parishes on the island of Jamaica as
shown in the context of the West Indies.
CHRISTINA M. KENNEDY ET AL.654
Ecological Monographs
Vol. 80, No. 4
encroachment and (2) we wanted to document bird
richness and abundance patterns in landscapes under-
going disturbance that was representative of each matrix
type.
Sampling of resident bird communities
We conducted point counts at 286 locations on two or
three separate occasions from early February to mid-
June during the height of breeding activity (Raffaele et
al. 1998) each year for three consecutive breeding
seasons (2005?2007). Point counts were conducted along
a centrally placed transect in each of the 68 forest
patches in human-dominated landscapes and along 1?3
randomly placed transects (averaging 1500 m in length)
in each forested landscape. We surveyed an average of
12?15 stations per replicate landscape per occasion. We
conducted sampling proportional to forest area to
ensure representative coverage, typically adding one
station for each additional hectare (conditional on the
terrain). To minimize double-counting, stations were
located 100 m apart and .25 m from a matrix?forest
boundary, and during each census, we mapped locations
and flight paths of all birds seen or heard. Surveying
intensity was uniform among landscapes and sufficient
in representing resident bird diversity in the region based
on species accumulation curves that reached definitive
asymptotes in all four landscape types during the survey
period (Appendix D; Gotelli and Colwell 2001).
At each point count station, we recorded the number
of individuals per species seen or heard during a 10-min
count (recorded in three sequential intervals: 2 min, 3
min, and 5 min) and in a 25-m fixed-radius area (Hutto
et al. 1986). We selected a 25-m radius because it was
found to provide the most reliable detection across sites
based on field tests in our system and in the Caribbean
(Wunderle and Waide 1993). We conducted surveys
between 06:00 and 10:00 on clear days without rain.
Each site was visited by one of three trained observers
that were rotated for repeat counts within a season, and
the order of site visitation was rotated throughout each
field season to diminish detection bias due to observer
and time-of-day effects. To investigate potential differ-
ences in species detectability among matrix types due to
potential variation in noise, vegetation structure, or bird
abundances in different landscape contexts (e.g., Simons
et al. 2007), we modeled species detection probabilities
with and without detection heterogeneity among the
four matrix types based on removal method analyses
(after Farnsworth et al. 2002). We found no evidence
that species detection varied by matrix type as deter-
mined by model comparisons (Appendix E). Based on
this finding, we had no a priori reason to suspect
systematic biases in species detectability in our study;
thus, we used count data as indices of bird abundance in
our analyses. Detected trends based on observed point
counts are expected to reflect those based on true
abundances, because we implemented standardized
survey protocols to control for confounding detection
biases due to landscape matrix effects and those due to
observer, time, or sampling effort (Johnson 2008,
Etterson et al. 2009).
To determine species occurrence in and potential use
of matrix habitats, we conducted 241 point counts in
2005 in matrix land cover surrounding forest patches in
fragmented landscapes. Two to four point count stations
were located midway between adjacent forest fragments
and .100 m from a matrix?forest boundary in at least
two cardinal directions. We conducted matrix point
counts using similar protocol as forest point counts (i.e.,
10-min count, 25-m fixed radius) during a similar time
span (early March to mid-June). Matrix surveys were
used to determine categorical bird?habitat associations
for trait analyses.
Species traits
We considered the influence of 11 traits on bird
responses to landscape matrix type. Trait values for each
species were determined from published information in
field guides and primary literature, consultation with
ornithological experts, and personal field observations
(Appendix F). We recorded species taxonomic order and
derived body mass and clutch size based on averaged,
published estimates. Species were classified into diet
guilds based on dominant food sources consumed (i.e.,
frugivore, nectarivore, insectivore, omnivore, granivore,
and carnivore). We categorized species foraging and
nesting locations into four height zones: ground (,0.5
m), understory (shrub layer to midstory canopy, up to 5
m), canopy (upper forest layer,.5 m), or multiple strata
(commonly using more than one height zone). Nest type
was classified as open or closed, with the former
containing large openings (i.e., cup, saucer, and
platform nests) and the latter being partially enclosed
(i.e., cavity, burrow, sphere, and pendant nests).
Geographic range was based on whether species
distributions were restricted to Jamaica, the Caribbean,
the New World Tropics, or spanned both Nearctic and
Neotropical regions. Altitudinal range was based on
presences or absences of species distributions in three
elevation classes in Jamaica: lowland, mid-elevation,
and montane.
We determined species habitat associations and rarity
based on our own field data. We classified species as
forest-restricted, generalist, or open-associated by com-
paring their average densities in continuous forest to
their average densities in matrix habitats. Species were
classified as restricted to forest habitat if their average
densities were at least three times greater in forest than
in matrix areas. In contrast, species were classified as
associated with open habitats if their average densities
were at least three times greater in matrix areas than in
continuous forest. Species with densities within a factor
of three between forest and matrix habitats were
classified as generalists. We chose this cutoff to ensure
that species categorized as forest- and open-associated
exhibited a strong affinity for their respective cover
November 2010 655AVIAN RESPONSE TO LANDSCAPE MATRIX
types. This admittedly arbitrary cut-off was necessary,
because no standard exists in the literature to delineate
habitat specificity. Our resulting classifications, howev-
er, closely matched habitat associations published in
field guides (e.g., Downer and Sutton 1995, Raffaele et
al. 1998) and with the opinion of ornithologists with
expertise in Jamaican avifauna (P. P. Marra, A. M.
Haynes-Sutton, and H. A. Davis, personal communica-
tions). Based on those convergences, we are confident
our habitat classifications reflect potential differential
forest vs. matrix usage. Rarity was based on the average
density of each species in its primary habitat over the
three-year sampling period (e.g., open-associated relied
on average densities in matrix habitats and forest-
restricted and generalist species relied on average
densities in forest patches).
Statistical analyses
Community composition.?Resident native birds de-
tected in our study region are included in analyses, with
the exception of nocturnal and aquatic birds, vultures,
swifts, and swallows. We used nonmetric multidimen-
sional scaling (NMDS) to describe variation in the
composition of resident bird communities in forest
patches embedded in the four matrix types. Nonmetric
multidimensional scaling is a nonparametric ordination
technique effective for graphically depicting multivariate
relationships in ecological data, via maximizing the rank
correlation between calculated distances in an original
matrix and distances in reduced ordination space
(Clarke 1993). The NMDS was performed using a
Bray-Curtis dissimilarity matrix derived from species
relative abundances at the patch level, where patches
were standardized by dividing species abundances by the
total number of point counts conducted within each
patch across all years. To avoid spurious effects of rare
species, we excluded species that occurred in ,5% of
patches and standardized community matrices by
species maximum. Overall statistical significance of the
ordination was determined based on a Monte Carlo
unrestricted permutation test with 100 randomizations.
Standardized avian species composition was compared
among matrix types using multiresponse permutation
procedure (MRPP; based on Bray-Curtis distance
matrix), which compares distances among samples
within a priori groups from those derived from a
randomization procedure (Mielke and Berry 2001,
McCune and Grace 2002). To determine the importance
of spatial dependence among sampled patches, we tested
for the overall correlation between the standardized
species community matrix with a corresponding matrix
of geographic distance between patches using a Mantel
test (Bray-Curtis distance matrix, N ?9999 randomiza-
tions; Legendre and Legendre 1998).
Species richness and individual species abundances.?
The influence of the landscape matrix on species richness
and individual abundances was analyzed using linear
and generalized linear mixed models, respectively. Both
richness and abundances were based on total point
counts in each forest patch for each sampling occasion.
Species richness was estimated using Chao1; this
estimator adjusts for bias due to missed species as
determined by the number of rare species observed
within samples based on abundance data (Colwell and
Coddington 1994) and has been found to be more
robust than other richness estimators (Walther and
Moore 2005). Based on model diagnostics (e.g., histo-
grams and plots of residuals vs. fitted values) and
dispersion scores, we modeled richness via a normal
distribution and abundances via a Poisson probability
distribution using a log link function.
Because sampling was conducted hierarchically (i.e.,
patches sampled within landscapes) and over multiple
time intervals (i.e., within and across three years), we
examined eight models that captured three distinct error
structures that were logical based on our sample design:
(1) correlations among repeat observations of a patch
(temporal correlation); (2) correlations among patches
within a landscape (spatial correlation); and (3) corre-
lations among patches within a landscape across
different sampling occasions (temporal and spatial
correlations; Pinheiro and Bates 2002). Based on the
lowest Akaike information criterion, corrected for small
sample sizes (AIC
c
), and using maximum likelihood
estimation (Burnham and Anderson 2002), the most
supported model for species richness and for ;60% of
the bird species accounted for within-patch correlation
due to repeated measures but lacked explicit correlation
among years and due to patch location. Our final models
therefore included landscape matrix type as the fixed
effect and patch identity as a random effect, using a
Laplace likelihood approximation, which is considered
more accurate for count data with small means (Bolker
et al. 2009).
Significance of differences in species richness and
relative bird abundances among forest patches embed-
ded in the four matrix types was determined based on
ANOVA F tests and Wald v
2
tests, respectively. Post
hoc pairwise comparisons were conducted using Tukey?s
multiple comparison procedure to separate treatment
means (Westfall and Young 1993). We used a family-
wise a?0.10 to indicate a biologically relevant response.
Failing to detect an effect when species were in fact
differing in abundance by landscape matrix type
(making a Type II error) had as severe a consequence
as falsely detecting an effect (making a Type I error).
Therefore, using a family-wise a , 0.10 would have led
us to substantially under-predict species that were
responding to forest fragmentation given stronger
evidence to the contrary, which would have impeded
our subsequent analyses.
Matrix response and species traits.?We categorized
species into one of four response classes based on
differences in relative abundance in forest patches
among the human-dominated matrix types relative to
a continuous forest matrix, as detected by mixed models
CHRISTINA M. KENNEDY ET AL.656
Ecological Monographs
Vol. 80, No. 4
and post hoc tests: (1) lower abundance(s) in patches
embedded in one or more of the human-dominated
matrix types (type L); (2) higher abundance(s) in patches
embedded in one or more of the human-dominated
matrix types (type H); (3) mixed-matrix response, with
higher abundance in patches in at least one human-
dominated matrix type and lower abundance in patches
in a different human-dominated matrix type (type M);
and (4) no significant difference in abundance in patches
among any of the three human-dominated matrix types
relative to intact forest (type N).
We used a classification decision tree analysis to assess
whether species responses shared biological traits. This
nonparametric method determines membership in pre-
defined groups based on a suite of characteristics using
recursive data partitioning. It is well suited for non-
normal, intercorrelated, multivariate data characteristic
of life history traits (De?ath and Fabricius 2000). We
selected the final model based on a series of 1000 10-fold
cross-validations, using the 1-SE rule and Gini index of
impurity and prior probabilities proportional to sample
sizes. Overall statistical significance of the final tree was
determined based on Monte Carlo resampling (N?1000
randomizations; Breiman et al. 1993).
We tested the relative importance of each trait with
regard to species sensitivity to forest fragmentation and
landscape matrix type in two ways. First, we calculated
the ability of each trait to distinguish among matrix
response types as determined by the decrease in impurity
attributable to the best surrogate split for each variable
on the final classification tree (Breiman et al. 1993). We
also conducted goodness-of-fit tests to determine the
statistical significance of each trait in relation to matrix
response. These univariate tests complement the tree-
based approach by detecting important variable(s) that
may be masked in a tree-based framework (McCune and
Grace 2002, Maindonald and Braun 2003). For both the
classification decision tree analysis and the goodness-of-
fit tests, we excluded the one species with a mixed-matrix
response, because this unique response was insufficient
to analyze independently. To increase the reliability of
the contingency table analyses, we increased cell counts
by combining the seven species with higher abundances
in patches in fragmented landscapes with species that
lacked an abundance response. This contrast provided a
distinct comparison between birds likely to be negatively
affected by forest fragmentation (i.e., exhibit lower
abundances, type L) vs. those that may not (i.e., exhibit
higher or constant abundances in human-dominated
landscapes relative to continuous forest, types H and N).
Due to multicollinearity among traits, separate tests
were conducted on each trait and based on randomized
v
2
tests using Monte Carlo simulations (N ? 1000000
randomizations), which are more accurate for sparse cell
counts (Sokal and Rohlf 1995). We examined adjusted
residuals from v
2
tests to determine classes that were
driving significant overall differences (Everitt 1992).
Statistical packages.?Statistical analyses were per-
formed using the R statistical system (version 2.8.1; R
Development Core Team 2008) and SAS (version 9.2;
SAS Institute, Cary, North Carolina, USA). Richness
(Chao1) estimation, NMDS, MRPP, and Mantel tests
were performed using the R ??vegan?? package (version
1.13.1; Oksanen et al. 2008). Model selection for linear
and generalized linear mixed models was conducted
using the ??nlme?? (Pinheiro et al. 2008) and ??lme4??
(Bates et al. 2008) packages, respectively. The final linear
mixed model was implemented using ??nlme?? and
multiple comparisons performed via ??multcomp?? pack-
age (Hothorn et al. 2008) in R; final generalized linear
mixed models were implemented using PROC GLIM-
MIX in SAS. The v
2
tests were performed using ??stats??
package and classification regression trees using the
??rpart?? package in R (Therneau and Atkinson 2009).
RESULTS
Species richness and community composition
In total we detected 16996 resident birds in forest
patches among all landscapes and observed 44 resident
species, of which 23 were endemic to Jamaica. Estimated
richness of resident birds significantly differed in patches
embedded in the four different matrix types (mixed-
model ANOVA, F
3,95
? 6.29 P ? 0.0006). Based on
pairwise comparisons, forest patches in an agricultural
matrix had greater richness than did patches in peri-
urban and bauxite matrices (P?0.0225 and P?0.0027,
respectively). Patches embedded in a continuous-forest
matrix exhibited richness greater than patches in a
bauxite matrix (P?0.0075) and marginally greater than
patches in a peri-urban matrix (P ? 0.0595). Avian
richness did not differ in patches embedded in forest vs.
agricultural matrices (P ? 0.9341), nor did it differ
between patches in peri-urban and bauxite matrices (P?
0.9665; Table 1).
We retained 41 of the 44 resident species detected in
the community analyses based on the .5% detection
rule. The NMDS ordination resulted in a three-axis
solution, with a final stress of 17.702, which is within the
range reliable for community data and unlikely to have
been obtained by chance (Monte Carlo test, P , 0.001;
McCune and Grace 2002). The three axes together
represented 96.9% of the variance in resident bird
communities, using a fit-based R
2
measure (Oksanen et
al. 2008). Patches within the same matrix type tended to
group together in ordination space, indicating a
similarity in bird community composition (Fig. 2).
Substantial overlap in patches in forest and agricultural
matrices in ordination space indicated shared species
assemblages. Despite these similarities, the MRPP
analyses confirmed that forest bird communities differed
significantly among all matrix types (A ? 0.0587, P ,
0.0001; Fig. 2). Differences in community assemblage
among matrix types cannot be attributed to mere spatial
correlation among patches (Mantel test, r?0.0869, P?
0.102).
November 2010 657AVIAN RESPONSE TO LANDSCAPE MATRIX
TABLE 1. Estimated means (and SE) of linear and generalized linear mixed models for richness and abundances of native resident
birds in forest patches in agricultural (N?22), peri-urban development (N?19), or bauxite mining (N?25) landscapes, or sites
in intact forest (N ?31) in central Jamaica during the study period (2005?2007).
Species and response type
Forest Agriculture Peri-urban Bauxite
Mean SE Mean SE Mean SE Mean SE
Species richness 26.25
a
1.76 27.30
a
1.35 21.79
b
1.92 20.94
b
1.81
Lower abundance in peri-urban and bauxite landscapes
Arrow-headed Warbler (Dendroica pharetra) (e) 0.48
a
0.10 0.76
a
0.18 0.06
b
0.02 0.14
b
0.04
Jamaican Becard (Pachyramphus niger) (e) 0.23
a
0.06 0.29
a
0.09 0.02
b
0.01 0.01
b
0.01
Jamaican Elaenia (Myiopagis cotta) (e) 0.28
a
0.07 0.39
a
0.11 0.02
b
0.01 0.11
c
0.04
Jamaican Pewee (Contopus pallidus) (e) 0.36
a
0.10 0.25
a
0.09 0.00
b
0.00 0.01
b
0.01
Jamaican Woodpecker (Melanerpes radiolatus) (e) 0.67
a
0.11 0.68
a
0.14 0.31
b
0.07 0.22
b
0.05
Rufous-tailed Flycatcher (Myiarchus validus) (e) 0.28
a
0.07 0.36
a
0.10 0.05
b
0.02 0.04
b
0.02
Lower abundance in peri-urban landscapes
Greater Antillean Bullfinch (Loxigilla violacea) 1.04
a
0.15 1.12
a
0.19 0.61
b
0.12 0.95
ab
0.15
Jamaican Lizard-Cuckoo (Saurothera vetula) (e) 0.06
a
0.03 0.04
ab
0.02 0.00
b
0.00 0.03
ab
0.02
Jamaican Vireo (Vireo modestus) (e) 1.41
a
0.22 1.49
a
0.28 0.45
b
0.10 0.84
ab
0.16
Olive-throated Parakeet (Aratinga nana) 0.15
a
0.05 0.13
a
0.06 0.02
b
0.01 0.14
a
0.06
White-crowned Pigeon (Columba leucocephala) 0.68
a
0.12 0.83
a
0.17 0.28
b
0.07 0.60
a
0.12
Yellow-shouldered Grassquit (Loxipasser anoxanthus) (e) 0.34
a
0.06 0.26
a
0.06 0.09
b
0.03 0.42
a
0.09
Lower abundance in bauxite landscapes
Black-faced Grassquit (Tiaris bicolor) 0.13
ab
0.04 0.06
bc
0.02 0.25
a
0.06 0.019
c
0.01
Higher abundance in agricultural and bauxite landscapes
Loggerhead Kingbird (Tyrannus caudifasciatus) 0.19
b
0.04 0.46
a
0.10 0.30
ab
0.07 0.44
a
0.09
Higher abundance in peri-urban and bauxite landscapes
Smooth-billed Ani (Crotophaga ani) 0.00
b
0.00 0.03
ab
0.02 0.05
a
0.03 0.05
a
0.03
Higher abundance in agricultural landscapes
Jamaican Spindalis (Spindalis nigricephala)C160 (e) 0.68
b
0.13 1.33
a
0.27 0.64
b
0.14 0.54
b
0.11
Jamaican Oriole (Icterus leucopteryx) 0.57
b
0.08 0.99
a
0.14 0.55
b
0.09 0.75
ab
0.11
White-bellied Dove (Leptotila jamaicensis) 0.13
b
0.03 0.36
a
0.08 0.14
b
0.04 0.26
ab
0.06
White-winged Dove (Zenaida asiatica) 0.03
b
0.01 0.22
a
0.07 0.04
b
0.02 0.06
b
0.03
Higher abundance in peri-urban landscapes
Vervain Hummingbird (Mellisuga minima) 0.15
b
0.04 0.21
ab
0.06 0.37
a
0.10 0.15
b
0.04
Higher abundance in agricultural and lower abundance in bauxite landscapes
White-eyed Thrush (Turdus jamaicensis) (e) 0.29
b
0.07 0.72
a
0.19 0.12
bc
0.04 0.06
c
0.03
No difference in abundance in human-dominated landscapes than intact forest
Bananaquit (Coereba flaveola) 2.28 0.28 2.39 0.35 2.19 0.34 2.12 0.29
Chestnut-bellied Cuckoo (Hyetornis pluvialis) (e) 0.12 0.03 0.09 0.03 0.13 0.04 0.06 0.02
Common Ground-Dove (Columbina passerina) 0.08 0.03 0.06 0.03 0.06 0.03 0.05 0.02
Jamaican Euphonia (Euphonia jamaica) (e) 0.54 0.10 0.75 0.16 0.47 0.11 0.52 0.11
Jamaican Mango (Anthracothorax mango) (e) 0.02 0.01 0.06 0.03 0.01 0.01 0.04 0.02
Jamaican Tody (Todus todus) (e) 0.92 0.13 1.25 0.21 1.08 0.19 0.84 0.14
Mangrove Cuckoo (Coccyzus minor) 0.04
ab
0.02 0.03
ab
0.02 0.00
b
0.00 0.07
a
0.03
Northern Mockingbird (Mimus polyglottos) 0.01 0.01 0.03 0.02 0.06 0.03 0.05 0.02
Orangequit (Euneornis campestris) (e) 2.63
ab
0.31 3.33
a
0.46 2.27
ab
0.34 2.08
b
0.28
Red-billed Streamertail (Trochilus polytmus) (e) 1.65 0.23 2.18 0.35 1.82 0.31 1.74 0.26
Ruddy Quail Dove (Geotrygon montana) 0.15
ab
0.05 0.32
a
0.10 0.07
b
0.03 0.12
ab
0.04
Sad Flycatcher (Myiarchus barbirostris) (e) 0.26
ab
0.05 0.48
a
0.10 0.17
b
0.04 0.14
b
0.04
White-chinned Thrush (Turdus aurantius) (e) 0.99 0.12 1.26 0.18 1.34 0.19 1.41 0.18
Yellow-faced Grassquit (Tiaris olivacea) 0.11 0.03 0.08 0.03 0.09 0.03 0.15 0.05
Zenaida Dove (Zenaida aurita) 0.08 0.02 0.19 0.06 0.12 0.04 0.17 0.05
Notes: P values are based on ANOVA F tests and Wald v
2
tests (values in boldface are significant at P , 0.05). Superscripted
lowercase letters indicate results of pairwise comparisons among matrix types based on post hoc Tukey?s honestly significant
difference (hsd) tests. Species are categorized based on abundance differences among human-dominated matrix types relative to
intact forest. Species included in community analyses but with insufficient detections for Poisson mixed models were: American
Kestrel (Falco sparverius), Jamaican Crow (Corvus jamaicensis) (e), Red-tailed Hawk (Buteo jamaicensis), Rufous-throated Solitaire
(Myadestes genibarbis), and Stolid Flycatcher (Myiarchus stolidus). Species detected in the study region, but with insufficient
detections to include in any analyses were: Crested Quail Dove (Geotrygon versicolor) (e); Yellow-billed Parrot (Amazona collaria)
(e); and Greater Antillean Grackle (Quiscalus niger). ??(e)?? Indicates species endemic to Jamaica. Primary diet guild abbreviations
are: C, carnivore; F, frugivore; G, granivore; I, insectivore; N, nectarivore; O, omnivore. Primary habitat association abbreviations
are: FR, forest-restricted; G, generalist; OA, open-associated. Primary nesting height or foraging strata abbreviations are: G,
ground; U, understory; C, canopy; M, multiple strata.
C160 Previously named Jamaican Stripe-headed Tanager (Spindalis nigricephalus) (Banks et al. 2000).
CHRISTINA M. KENNEDY ET AL.658
Ecological Monographs
Vol. 80, No. 4
Individual species responses
Thirty-six species that were detected in .15% of
patches had sufficient occurrences to model via Poisson
regressions (i.e., likelihood functions converged with
reliable model fit and parameter estimates). Relative
abundances of 69.4% of these species differed signifi-
cantly in forest patches among the four matrix types,
which is greater than would be expected by chance
(binomial test, P ? 0.014). Thirteen species had lower
abundances in one or two of the human-dominated
matrix types (agriculture, peri-urban, and/or bauxite;
Table 1). Six of these species had abundances that were
lower in forest fragments in peri-urban and bauxite
matrices relative to a forest matrix, but their abun-
dances in fragments in an agricultural matrix were
similar to that in an intact forest. Seven species had
reduced abundances in only one human-dominated
matrix type; for all but one of these species, abundances
were lowest in peri-urban forest fragments. The Black-
faced Grassquit was the exception, being least abun-
dant in fragments in a bauxite matrix, but equally as
abundant in fragments embedded in a peri-urban
matrix as in a forest matrix (Table 1). No species was
consistently lower in abundance in fragments in all
three human-dominated matrix types. Two species,
however, showed patterns suggesting this response, but
were too scarce for statistical analyses: 87%of 15 Stolid
Flycatcher detections were in continuous forest and all
eight Rufous-throated Solitaire detections were in
continuous forest.
No species exhibited abundances that were consis-
tently higher in forest fragments in all three human-
dominated matrix types, but seven species were higher in
abundance in one or two of the human-dominated
matrix types relative to continuous forest. Of these
species, five were higher in abundance in fragments in an
agricultural matrix but lower in abundance in fragments
in peri-urban and bauxite matrices. Two species were
more abundant in two types of human-dominated
landscapes relative to continuous forest: the Loggerhead
Kingbird was higher in abundance in forest embedded in
both agriculture and bauxite and the Smooth-billed Ani
was higher in abundance in forest embedded in both
bauxite and peri-urban development. Only one species,
the Vervain Hummingbird, was more abundant in forest
patches in a peri-urban matrix than in all other matrix
types.
The White-eyed Thrush was the only species that
exhibited a mixed matrix response, with abundances
higher in agricultural patches but lower in bauxite
patches. The remaining 15 species did not differ in
abundance in forest patches in human-dominated
matrices relative to a forest matrix. Four of these
species, however, had abundances that varied among the
three types of human-dominated landscapes.
Overall, peri-urban and bauxite landscapes had the
highest frequency of resident species with reduced
abundances relative to continuous forest: eight species
were lower in abundance in bauxite patches and 12
species were lower in abundance in peri-urban patches
(Fig. 3). No species was less abundant in patches in
agricultural landscapes relative to forested landscapes.
Six species were more abundant in forest patches in an
agricultural matrix whereas only three species were more
abundant in patches in bauxite and/or peri-urban
matrices relative to a forest matrix.
TABLE 1. Extended.
P Diet
Habitat
association
Nest
height
Foraging
strata
0.0006
,0.0001 IFRUM
,0.0001 C C
,0.0001 IFR
0.0001 C U
0.0001 IG M
,0.0001 IFRC C
0.0834 F FR M U
0.0482 CF U
0.0001 IFR U
0.0294 FG C M
0.0074 FFRM C
0.0008 FG U U
0.0001 GOAU G
0.0100 OOAM M
0.0455 OOAM G
0.0125 FG U C
0.0188 IG M
0.0093 GG G G
0.0004 GOAU
0.0616 N G U U
,0.0001 OFRM M
0.9382 N G M M
0.2974 C G M M
0.7754 G G U G
0.4265 F G M C
0.1648 N OA C U
0.3179 I FR G M
0.0408 C FR U U
0.2325 O OA M G
0.0857 N FR NA M
0.6130 N G U M
0.0275 GFRG G
0.0007 IG UU
0.1974 O G M G
0.5329 G OA G G
0.1446 G OA M G
November 2010 659AVIAN RESPONSE TO LANDSCAPE MATRIX
Role of species traits
The most parsimonious classification tree model (with
the greatest prediction accuracy) included only 1 of 11
traits (diet guild) and predicted two matrix response
types. Of observed bird responses to landscape matrix,
66% were correctly classified based on diet guild alone,
which is greater than expected by chance (Monte Carlo
simulation, P ? 0.0079; Fig. 4). The model correctly
classified 84.6% of species with observed lower abun-
dances (type L), and 80.0% with observed similar
abundances in patches in human-dominated matrices
compared to a forest matrix (type N). Misclassifications
largely stemmed from the erroneous categorization of
the seven species with higher abundances in fragmented
landscapes (type H). Five of these species were
incorrectly predicted to have no difference in abundance
(i.e., misclassified as type N) and two of these species
were incorrectly predicted to have lower abundances in
fragmented forest relative to continuous forest (i.e.,
misclassified as type L).
Based on both variable importance from the decision
tree analysis and statistical significance from v
2
tests,
diet guild, nest height, habitat association, and, to a
lesser extent, foraging strata were strongly associated
with bird responses to landscape matrix (Table 2).
Taxonomic order, geographic and altitudinal range,
body size, rarity, clutch size, and nest type had weak
prediction power and lacked any statistical association.
Diet guild.?A total of 70%of insectivores and 67%of
frugivores had lower abundances in patches in human-
dominated matrices than a forest matrix. In contrast,
100% of nectarivores and omnivores and 86% of
granivores had higher or similar abundances in frag-
ments (Figs. 4 and 5a). Carnivores lacked a consistent
response, with the Jamaican Lizard-Cuckoo exhibiting
lower abundances and the Chestnut-bellied Cuckoo and
the Mangrove Cuckoo exhibiting similar abundances
between human-dominated and forested landscapes.
Habitat association.?The extent to which a species
was known to be a forest specialist rather than to use
matrix habitats impacted its response to fragmentation.
Forest-restricted species exhibited the greatest reduc-
tions in abundance, with 70% lower, 30% similar, and
none higher in abundance in forest patches in human-
dominated matrices than in a forest matrix (Fig. 5b). In
FIG. 2. Nonmetric multidimensional scaling (NMDS) ordination (stress ? 17.0170) of resident bird communities in 99 forest
patches in four landscape matrix types over the study period (2005?2007). For illustration purposes NMDS axis 1 and axis 2 (of a
three-dimensional solution) are presented, which capture most of the variation in community structure and depict the overall
pattern. Dimensions represent the relative position among sampled patches based on species assemblages, with patches with similar
avian composition containing similar scores in multidimensional space. Community composition among the four matrix types
significantly differed (based on familywise a ? 0.05) based on overall and pairwise comparisons based on multiresponse
permutation procedure (MRPP) results (inset).
CHRISTINA M. KENNEDY ET AL.660
Ecological Monographs
Vol. 80, No. 4
contrast, ;90% of birds associated with open habitats
and;80%of generalist birds had greater (38%and 29%,
respectively) or equal abundances (50% each) in
fragments as compared to continuous forest.
Nest height.?The dominant height at which a species
nested also impacted its fragmentation response. Eighty-
six percent of canopy-nesting species were less abundant
and 14% were equally abundant in forest patches in
human-dominated matrices relative to a forest matrix
(Fig. 5c). In contrast, 100% of ground-nesting species
either had similar abundances (75%) or had higher
abundances (25%) and 80% of multi-strata nesters had
similar or higher abundances in patches in human-
dominated matrices (60% and 20%, respectively) than in
a forest matrix. Understory nesters failed to exhibit a
consistent response, with roughly one-third of species
with higher (31%), lower (38%), or equal abundances
(31%) in patches in human-dominated landscapes
relative to intact forest.
Foraging strata.?Canopy foragers and, to a lesser
extent, understory foragers had lower abundances in
fragmented landscapes than did ground foragers (Fig.
5d). Two-thirds of canopy foragers were lower in
abundance in fragments, whereas 17% were higher and
another 17%were equal in abundance in fragments as in
intact forest. Abundances were lower for 56% of
understory foragers, equal for 33%, and higher for only
11% (one species) in patches in human-dominated
TABLE 2. The strength of association among the 11 traits with
landscape matrix responses by resident birds in central
Jamaica.
Trait Variable importance P
Diet guild 100.00 0.0105
Nest height 72.93 0.0113
Habitat association 67.57 0.0092
Foraging strata 45.12
Taxonomic order 30.45 0.2135
Geographic range 28.93 0.3944
RarityC160 25.54 0.4888
Nest typeC224 19.29 0.1566
Clutch size? 18.26 0.1621
Altitudinal range 8.46 0.2639
Body mass} 2.61 0.7461
Notes: Variable importance was determined by calculating
the change in impurity (i.e., Gini index) when a trait was
substituted for the original variable on the final decision tree
(i.e., diet) and is expressed as the relative magnitude of the total
decrease in impurity (based on normalized quantiles). The
variable with the greatest prediction accuracy is attributed the
highest value (100), and the variable with the lowest prediction
accuracy is attributed the lowest value (0). P values from
randomized v
2
tests are provided; values in boldface are
significant at P , 0.05.
C160 Variable importance was based on continuous distribution;
v
2
test was based on classified groups (,0.25 and .0.25
density).
C224 One bird with unknown nest height was excluded from the
v
2
test.
? Variable importance was based on continuous distribution;
v
2
test was based on classified groups (,3 and .3 eggs).
}Variable importance was based on continuous distribution;
v
2
test was based on classified groups (,15, 15?50, and .50 g).
FIG. 3. Percentages of resident bird species in agricultural,
peri-urban, or bauxite mining landscapes that exhibited
significantly lower abundance, higher abundance, or no
abundance difference relative to forested landscapes over the
study period (N ?36 species).
FIG. 4. Predicted response type, percentage of observations
correctly classified per response type, the number of species (in
parentheses), and the distribution of the observed bird
responses per group based on classification tree analysis of
bird responses to the landscape matrix in relation to 11 traits.
Species responses were categorized as: (1) type L, lower in
abundance in any of the human-dominated matrix type(s)
relative to intact forest; (2) type H, higher in abundance in any
of the human-dominated matrix type(s) relative to intact forest;
and (3) type N, no difference in abundance in fragmented
landscapes relative to intact forest. Diet guild was the only trait
retained in the final model, with two matrix types predicted
(type L and type N) (with 34% misclassification rate).
November 2010 661AVIAN RESPONSE TO LANDSCAPE MATRIX
matrices than in a forest matrix. In contrast, all but one
ground-foraging species had greater (30%) or similar
(60%) abundances in human-dominated landscapes as
compared to forested landscapes. Species that foraged
among multiple strata failed to exhibit a strong
response, with 20% higher, 30% lower, and 50%
exhibiting no change in abundance among fragmented
vs. forested landscapes.
DISCUSSION
Ostensibly, landscapes in Jamaica would seem to
adhere to the classic binary habitat vs. non-habitat
model in that historically forested areas have been
converted to similar matrix types comprising novel
habitats with relatively little or no forest cover. This
study demonstrates, however, that simplistic models of
habitat fragmentation do not adequately reflect respons-
es of Neotropical bird communities in this region. Suites
of species were profoundly affected by whether once-
continuous forest was converted to agricultural, peri-
urban, or bauxite mining development. Almost 70% of
bird species differed in abundance in patches among the
four landscape matrix types. More than 36% of species
had lower abundances in forest fragments embedded in
peri-urban and bauxite mining matrices relative to a
forested matrix. Another 20% had higher abundances in
fragments embedded in an agricultural matrix than in a
forest matrix. Overall, no species was consistently more
or less abundant in all human-dominated landscape
types. Species richness and community composition
FIG. 5. Adjusted residuals of the v
2
analyses relating the number of resident bird species exhibiting two matrix responses types
to (a) diet guild, (b) habitat association, (c) nest height, and (d) foraging strata. The two response types were classified as (1) type L,
lower in abundance in any of the human-dominated matrix type(s) relative to intact forest, and (2) type H?N, higher in abundance
or similar in abundance among human-dominated landscapes relative to intact forest. The number above each bar represents the
number of species observed to exhibit the respective response type and trait class combination. Adjusted residuals with the largest
absolute values indicate the class driving overall significant differences among matrix responses and the trait in question.
CHRISTINA M. KENNEDY ET AL.662
Ecological Monographs
Vol. 80, No. 4
behaved similarly to species abundances, with an
agricultural matrix having little effect on richness (Table
1) and community composition (Fig. 2) relative to a
forest matrix; but peri-urban and bauxite matrix types
had substantial impacts on both multispecies response
metrics. The agricultural matrix had seemingly little
effect even on forest-restricted species and some
generalist species, and species associated with open
habitats actually increased in this landscape type. In
contrast, peri-urban and mined landscapes had lower
native resident bird diversity and had communities that
were strikingly different from those in forested land-
scapes, in that many insectivores and frugivores were
absent or less abundant. Different measures of species
responses (i.e., richness, community composition, and
abundances) thus provided convergent results, which
strengthens the reliability of these patterns. Moreover,
matrix-specific responses were detected despite similar
climate, geology, elevation, and forest type and despite
controlling for the amount and configuration of forest
cover among anthropogenically fragmented landscapes.
Role of species traits and possible mechanisms mediating
responses to forest fragmentation
Four proximate mechanisms have been proposed to
explain divergent species responses to forest fragmenta-
tion as a function of the composition and the
configuration of matrix land cover: (1) differential
impedance or facilitation of inter-patch movement
(e.g., Renjifo 2001, Ricketts 2001, Gobeil and Villard
2002, Revilla et al. 2004, Bender and Fahrig 2005); (2)
differential alteration of interspecific interactions, in
particular predation (Rodewald and Yahner 2001,
Chalfoun et al. 2002); and differential mediation of
resources either (3) by habitat compensation via the
addition of alternative or supplemental resources (e.g.,
food or nesting sites) in certain matrix areas (e.g.,
Gascon et al. 1999, Norton et al. 2000, Brotons et al.
2003, Cook et al. 2004) or (4) by the disparate reduction
of within-patch habitat quality (i.e., microclimate and
vegetation structure) due to dissimilar edge effects
(Saunders et al. 1991, Ries et al. 2004) or different
human disturbances (Friesen et al. 1995, Rodewald and
Bakermans 2006).
Collectively, our results are consistent with the
hypothesis that resource availability may be the
important driver of bird community changes in Jamai-
ca?s fragmented landscapes. Although we cannot tie any
single trait to a particular mechanistic explanation, in
part because of the correlational nature of our data,
several traits emerged as significant and in combination
support resource availability as an important mecha-
nism. Diet guild, habitat association, nest height, and
foraging stratum best predicted the variation in bird
responses to forest fragmentation across matrix types.
These four traits relate to a species ability to acquire and
use resources, including food and nest sites, in human-
dominated landscapes. In particular, Jamaica?s insectiv-
orous birds had lower abundances in fragmented
landscapes relative to other guilds, which is consistent
with growing evidence that insectivores are declining
disproportionately in tropical forest remnants (e.g.,
Castelletta et al. 2000, Ribon et al. 2003, Sodhi et al.
2004, Sigel et al. 2006, Stouffer et al. 2009). Such
declines are thought to be potentially due to the
interdependent effects of the loss of microhabitats and
the decline of prey availability (e.g., Burke and Nol
1998, Zanette et al. 2000). Agricultural landscapes
supported a greater number of insectivores, which could
be a consequence of forest fragments in these landscapes
providing greater foraging substrates due to their greater
leaf area indices (Appendix C). Frugivores were also less
abundant in human-dominated landscapes in the region.
As with insectivores, survival of frugivorous birds has
been found to be lower in other fragmented tropical
forests (Kattan et al. 1994, Ribon et al. 2003, Ruiz-
Gutierrez et al. 2008), potentially due to lack of year-
round fruit in deforested tropical areas (Sodhi et al.
2004). Understory and canopy foragers and canopy
nesters also had lower abundances, which may be due to
the reduction in the vertical complexity of forest
structure and the loss of canopy and emergent trees in
peri-urban and bauxite mining landscapes (Appendix
C), an effect that has also been shown to accompany
tropical forest fragmentation (Tabarelli et al. 2004). In
contrast, Jamaican nectarivores, omnivores, and grani-
vores were insensitive to fragmentation, particularly in
agricultural matrices. Edge- and matrix-foraging birds
are increasingly shown to persist and even thrive in
fragmented tropical systems (e.g., Stouffer and Bierre-
gaard 1995, Renjifo 1999, Sigel et al. 2006) due to
potential cross-boundary subsidies (Fagan et al. 1999,
Cantrell et al. 2001). A species tolerance of forest
conversion is increasingly linked to its ability to utilize
resources in matrix habitats (e.g., Laurance 1991,
Gascon et al. 1999, Henle et al. 2004).
In contrast to many studies that indicate matrix
effects are best explained by dispersal or movement
limitation in temperate (e.g., Ricketts 2001, Be?lisle and
Clair 2002, Gobeil and Villard 2002) and tropical
regions (e.g., Robinson 1999, Renjifo 2001, Laurance
et al. 2002, Sekercioglu et al. 2002, Stratford and
Robinson 2005), we found that traits linked to dispersal
abilities were not associated with species?matrix re-
sponses. Body size and taxonomic order, which tend to
correlate significantly with potential dispersal power of
birds (Sutherland et al. 2000), did not emerge as
important predictors. Geographic and altitudinal range
sizes, which relate in part to species dispersal and
establishment abilities (Gaston 1996), also failed to
predict landscape matrix responses. Moreover, frugi-
vores had disproportionately lower abundances in forest
remnants than in intact forest; but we would not predict
that these species would be dispersal limited due to their
adaptations to search for resources that are patchily
distributed in both space and time (Bowler and Benton
November 2010 663AVIAN RESPONSE TO LANDSCAPE MATRIX
2005). The karst countryside in central Jamaica is
characterized by small forested hilltops, often ,10 ha
in size, which are separated by other land uses but
remain in close proximity (e.g., hundreds of meters
apart) relative to potential bird dispersal. Such small
distances between forest patches may not prohibit
frequent movement of many bird species. Moreover,
the evolutionary history of Caribbean avifauna has
likely promoted selection of species with stronger
dispersal abilities and fewer physiological or morpho-
logical limitations than mainland counterparts; these
characteristics have allowed them to (re)colonize and
(re)establish in island habitats and to withstand large-
scale natural disturbance events such as hurricanes
(Lack 1976, Ricklefs and Bermingham 2008).
Other frequently cited causes of forest bird declines
are increased nest parasitism and predation (e.g.,
Robinson et al. 1995, Lampila et al. 2005), the impact
of which can vary by landscape context (Donovan et al.
1997, Rodewald and Yahner 2001, Chalfoun et al. 2002,
Driscoll and Donovan 2004). Traits such as nest type
and nest height may affect a species susceptibility to nest
predation, with open- and ground-nesting species found
to be at greater risk (Ford et al. 2001, Chalfoun et al.
2002, Lampila et al. 2005, but see Martin 1993). In
Jamaica, however, abundance differences between frag-
mented and intact forest were not related to nest type.
Moreover, species predicted to be most sensitive to nest
predation (ground nesters) were least likely to exhibit
lower abundances and species predicted to be least
sensitive to nest predation (canopy nesters) were more
likely to exhibit lower abundances in human-dominated
landscapes. Traits related to reproductive potential (i.e.,
rarity and clutch size) also failed to predict bird
abundance patterns. Thus, differential functional con-
nectivity (i.e., via altering species dispersal and recolo-
nization events) and differential population growth
potential (i.e., via population sinks due to higher
predation) do not appear to be dominant mechanisms
underlying bird responses to forest fragmentation in
Jamaica.
Impacts of matrix land cover and land use interact
with forest fragmentation effects
We propose that birds in Jamaica may be influenced
more by the extent to which the landscape matrix
mediates the availability of critical resources, either via
resource supplementation in matrix habitats or differ-
ential reduction of within-patch forest resources or some
combination of both. In many cases, matrix habitats
may be hospitable for native species and may provide
supplemental or additional resources that allow for
population maintenance or growth in fragmented
systems (Norton et al. 2000, Brotons et al. 2003, Cook
et al. 2004). In comparison to bauxite lands, peri-urban
and agricultural matrices contained greater vegetation
cover and complexity (Appendix B). Vegetation in
agricultural matrices (e.g., pasture, paddock trees, and
live fences) may have provided additional resources for
omnivores and granivores, which were more abundant
in these landscapes than in other matrix types. Similarly,
nectarivores may have benefited from residential gar-
dens, given these birds were equally or more abundant in
peri-urban landscapes than in the other human-modified
landscapes.
Beyond these obvious modifications to external
matrix habitats, landscape matrix type can also impact
internal forest conditions. For example, surface-mining
activities in bauxite areas involve large-scale removal of
vegetation and topsoil and the creation of open pits of
exposed earth, all of which could alter soil water
retention, create dust pollution, and lead to biogeo-
chemical and hydrologic changes (Bell and Donnelly
2006). All of these factors likely impact within-forest
microclimate and structure and can alter the composi-
tion of fauna, but the degree and direction of these
changes are largely unknown (Simmons et al. 2008).
Even after post-mining restoration, ecological commu-
nities may not fully recover to their original state
(Parrotta and Knowles 1999, 2001). Bird species in peri-
urban landscapes may be disproportionately affected by
forest disturbance from human activities (e.g., hunting,
selective logging, noise, spread of fire or invasives; e.g.,
Theobald et al. 1997, Marzluff et al. 2001, Miller et al.
2003, Rodewald and Bakermans 2006). Forest remnants
in bauxite and peri-urban matrices have low stature and
more open canopies, less structural complexity, and
greater percentage of shrub layer (Appendix C); these
factors indicate that these forests may have undergone
greater disturbance and/or be in earlier successional
stages than forests in agricultural matrices (Asprey and
Robbins 1953). Agricultural areas tend to be in large
private land holdings, which afford a greater level of
forest protection than commonly found in bauxite and
peri-urban areas.
Even though the landscapes we surveyed had similar
environmental conditions, vegetation structure differed
in remnant forests as a function of matrix land cover,
potentially due to associated land use practices in matrix
areas. Rarely will alterations to the surrounding
landscape proceed without associated changes to the
internal habitat (Laurance et al. 2002, Laurance 2008).
Thus, effects on forest conditions must be considered
part of the collective matrix effects in Jamaica. Avian
community structure could be driven as much by the
degree of disturbance to the internal properties of
remnant forest as by the external properties of the
matrix.
Data limitations, research needs, and other caveats
Several factors must be considered when interpreting
our results. First, several species were detected in low
abundance and with low probability (Appendix E),
which increases uncertainty about whether species
patterns are robust or will remain consistent over the
long term. Second, we may have failed to detect
CHRISTINA M. KENNEDY ET AL.664
Ecological Monographs
Vol. 80, No. 4
relationships between species traits and their responses
to matrix type due to the limited ways in which we were
able to categorize their life history attributes. There is a
general lack of biological data for many Jamaican bird
species; additional research on foraging, nesting, and
movement patterns must be completed before more
refined considerations of species traits are possible. We
relied in part on information provided in field guides
(Appendix F), because this level of detail is often all that
is available to researchers when determining trait
specificities. Even under the best circumstances, traits
are surrogates for actual mechanistic responses that are
best tested experimentally. Measures of reproductive
output, survival, and movement for species in a range of
habitats in different landscape contexts are ideally
needed to isolate mechanistic explanations. Such de-
tailed investigations, however, are rarely possible for
even the best-studied species, much less entire commu-
nities. Thus, examining surrogates such as traits that
help inform mechanistic explanations behind why
groups of species respond similarly to landscape change
can be informative first steps (Ewers and Didham 2006),
particularly in poorly studied regions such as the tropics.
Historic declines and temporal differences among
land uses must also be considered. Examination of
current-day abundance patterns may underestimate the
effects of forest fragmentation on species persistence in
regions such as the West Indies, where flora and fauna
have been altered by a long history of habitat
conversion, human disturbance, and introductions of
novel predators (e.g., rats, mongoose; Ricklefs and
Bermingham 2008). Abundance patterns detected in
Jamaica?s fragmented landscapes may be underesti-
mates, given that the reference baseline is itself
unprotected secondary forest that has been altered by
decades of selective logging, fuelwood collection, and
human-caused fire (Forestry Department 2001, Tole
2001). Moreover, the full extent of species responses
may not have been captured uniformly among land-
scapes due to the temporal differences among land use
practices. Bauxite matrices were more recently converted
(,10 years); therefore, their full impacts may not be
fully manifested. Future research in this system, and
elsewhere, should specifically address the temporal
effects of land use practices and the relative importance
of extinction vs. colonization processes on species long-
term persistence.
Conservation implications
Treating all anthropogenically modified lands as a
single cover type (i.e., non-habitat) in fragmentation
research and/or conservation planning has consequences
for biodiversity. All matrix types examined in our study
could be classified as having similar ??edge contrast??
(Strayer et al. 2003) in that they differed dramatically in
structure, composition, and microclimate from once-
contiguous native forest. Consequently, matrix areas
could be considered equally hostile to native species.
Bird communities, however, responded differently to the
three human-modified matrix types. Patches surrounded
by an agricultural matrix maintained greater native bird
diversity and more intact community assemblages than
did patches embedded in a peri-urban or a bauxite
mining matrix. Treating all modified lands equally and
categorizing Jamaica?s landscapes into ??matrix?? vs.
??forest?? components would have hidden these important
patterns. The trajectory of land conversion in many
regions of the world is from subsistence agriculture to
increasing urbanization (DeFries et al. 2004). Since the
1950s, land cover change in Jamaica (Evelyn and
Camirand 2003) and on other Caribbean islands (Lugo
2002) results largely from conversion of agriculture to
residential development and mining for bauxite. Our
study strongly suggests that such land conversion would
cause the loss of a large proportion of the Neotropical
bird community, even without additional forest loss or
isolation. Although we have not experimentally con-
firmed the mechanism, evidence suggests that human
land use practices in the matrix may be driving
differential abundance patterns via mediating resource
availability both within the forest and in external matrix
habitats.
ACKNOWLEDGMENTS
We thank A. Hayes-Sutton and O. Evelyn (Jamaica
Forestry Department) for sharing their knowledge of the
study region and for providing logistical field support; R.
Defries (Columbia University) and S. Schill (The Nature
Conservancy) for GIS and remote sensing data and consul-
tation; S. Koenig (Windsor Research Centre) for bird life
history data; J. Hines and J. Nichols (USGS Patuxent
Wildlife Research Center) for statistical guidance and model
programming; and L. Ries (University of Maryland), T.
Ricketts (World Wildlife Fund), A. Rodewald (Ohio State
University), and two anonymous reviewers for constructive
inputonpreviousdraftsofthismanuscript.Fieldworkwas
directed by C. M. Kennedy with essential field assistance by
H. Davis and C. Samuels. Research was made possible by the
support of many Jamaican researchers, nonprofit and
governmental organizations, and individual private land-
owners. Jamaica Forestry Department, Windalco, Alpart
Mining Venture, Jamalco bauxite mining companies, S&G
Road Surfacing Materials Ltd., and many private farmers
and Mandeville citizens permitted access to their lands.
Funding was provided to C. M. Kennedy by the NASA Earth
System Science Program (Doctoral Fellowship), the Ful-
bright U.S. Scholarship Program, the Washington Explorer?s
Club, the Cosmos Club Foundation, and the University of
Maryland (Behavior, Ecology, Evolution and Systematics
Graduate Program and an Ann G. Wylie Dissertation
Fellowship),andprovidedtoC.M.KennedyandP.P.
Marra by the Smithsonian Institution (James Bond Trust)
andprovidedtoP.P.MarrabytheNationalScience
Foundation.
LITERATURE CITED
Aberg, J., G. Jansson, J. E. Swenson, and P. Angelstam. 1995.
The effect of matrix on the occurrence of Hazel Grouse
(Bonasa bonasia) in isolated habitat fragments. Oecologia
103:265?269.
Andren, H. 1992. Corvid density and nest predation in relation
to forest fragmentation: a landscape perspective. Ecology 73:
794?804.
November 2010 665AVIAN RESPONSE TO LANDSCAPE MATRIX
Asprey, G. F., and R. G. Robbins. 1953. The vegetation of
Jamaica. Ecological Monographs 23:359?412.
Banks, R. C., C. Cicero, J. L. Dunn, A. W. Kratter, H. Ouellet,
P. C. Rasmussen, J. V. J. Remsen, J. A. Rising, and D. F.
Stotz. 2000. Forty-second supplement to the American
Ornithologists? Union Check-list of North American Birds.
Auk 117:847?858.
Bates, D., M. Maechler, and B. Dai. 2008. lme4: linear mixed-
effects models using S4 classes. R package version 0.999375-
28. R Foundation for Statistical Computing, Vienna,
Austria. hhttp://lme4.r-forge.r-project.org/i
Bayne, E. M., and K. A. Hobson. 1997. Comparing the effects
of landscape fragmentation by forestry and agriculture on
predation of artificial nests. Conservation Biology 11:1418?
1429.
Beard, J. S. 1944. Climax vegetation in tropical America.
Ecology 25:127?158.
Beard, J. S. 1955. The classification of tropical American
vegetation-types. Ecology 36:89?100.
Be? lisle, M., and C. C. S. Clair. 2002. Cumulative effects of
barriers on the movements of forest birds. Conservation
Ecology 5:9.
Bell, F. G., and L. J. Donnelly. 2006. Mining and its impact on
the environment. Taylor and Francis, New York, New York,
USA.
Bender, D. J., T. A. Contreras, and L. Fahrig. 1998. Habitat
loss and population decline: a meta-analysis of the patch size
effect. Ecology 79:517?533.
Bender, D. J., and L. Fahrig. 2005. Matrix structure obscures
the relationship between interpatch movement and patch size
and isolation. Ecology 86:1023?1033.
Bolker, B. M., M. E. Brooks, C. J. Clark, S. W. Geange, J. R.
Poulsen, M. H. H. Stevens, and J.-S. S. White. 2009.
Generalized linear mixed models: a practical guide for
ecology and evolution. Trends in Ecology and Evolution
24:127?135.
Bowler, D. E., and T. G. Benton. 2005. Causes and
consequences of animal dispersal strategies: relating individ-
ual behaviour to spatial dynamics. Biological Reviews 80:
205?225.
Breiman, L., J. Friedman, R. Olshen, and C. Stone. 1993.
Classification and regression trees. Chapman and Hall, New
York, New York, USA.
Brotons, L., M. Monkkonen, and J. L. Martin. 2003. Are
fragments islands? Landscape context and density?area
relationships in boreal forest birds. American Naturalist
162:343?357.
Burke, D. M., and E. Nol. 1998. Influence of food abundance,
nest-site habitat, and forest fragmentation on breeding
ovenbirds. Auk 115:96?104.
Burnham, K. P., and D. R. Anderson. 2002. Model selection
and multimodel inference: a practical information-theoretic
approach. Second edition. Springer Science and Business
Media, Fort Collins, Colorado, USA.
Cantrell, R. S., C. Cosner, and W. F. Fagan. 2001. How
predator incursions affect critical patch size: the role
of the functional response. American Naturalist 158:368?
375.
Castelletta, M., N. S. Sodhi, and R. Subaraj. 2000. Heavy
extinctions of forest avifauna in Singapore: lessons for
biodiversity conservation in southeast Asia. Conservation
Biology 14:1870?1880.
Chalfoun, A. D., F. R. Thompson, and M. J. Ratnaswamy.
2002. Nest predators and fragmentation: a review and meta-
analysis. Conservation Biology 16:306?318.
Clarke, K. R. 1993. Nonparametric multivariate analyses of
changes in community structure. Australian Journal of
Ecology 18:117?143.
Colwell, R. K., and J. A. Coddington. 1994. Estimating
terrestrial biodiversity through extrapolation. Philosophical
Transactions of the Royal Society B 345:101?118.
Cook, W. M., R. M. Anderson, and E. W. Schweiger. 2004. Is
the matrix really inhospitable? Vole runway distribution in an
experimentally fragmented landscape. Oikos 104:5?14.
Cruz, A. 1981. Bird activity and seed dispersal of a montane
forest tree (Dunalia arborescens) in Jamaica. Biotropica 13:
34?44.
De?ath, G., and K. E. Fabricius. 2000. Classification and
regression trees: a powerful yet simple technique for
ecological data analysis. Ecology 81:3178?3192.
Debinski, D. M., and R. D. Holt. 2000. A survey and overview
of habitat fragmentation experiments. Conservation Biology
14:342?355.
DeFries, R. S., J. A. Foley, and G. P. Asner. 2004. Land-use
choices: balancing human needs and ecosystem function.
Frontiers in Ecology and the Environment 2:249?257.
Donovan, T. M., P. W. Jones, E. M. Annand, and F. R.
Thompson. 1997. Variation in local-scale edge effects:
mechanisms and landscape context. Ecology 78:2064?2075.
Downer, A., and D. Sutton. 1995. Birds of Jamaica: a
photographic field guide. Cambridge University Press, Cam-
bridge, UK.
Driscoll, M. J. L., and T. M. Donovan. 2004. Landscape
context moderates edge effects: nesting success of wood
thrushes in central New York. Conservation Biology 18:
1330?1338.
ESRI [Environmental Systems Research Institute]. 2008.
ArcGIS. Version 9.3. Environmental Systems Research
Institute, Redlands, California, USA.
Estades, C. F., and S. A. Temple. 1999. Deciduous-forest bird
communities in a fragmented landscape dominated by exotic
pine plantations. Ecological Applications 9:573?585.
Etterson, M. A., G. J. Niemi, and N. P. Danz. 2009. Estimating
the effects of detection heterogeneity and overdispersion on
trends estimated from avian point counts. Ecological
Applications 19:2049?2066.
Evelyn, O. B. 1997. Deforestation in Jamaica: an analysis of the
data. Forestry Department, Kingston, Jamaica.
Evelyn, O. B., and R. Camirand. 2003. Forest cover and
deforestation in Jamaica: an analysis of forest cover estimates
over time. International Forestry Review 5:354?363.
Everitt, B. S. 1992. The analysis of contingency tables. Second
edition. Chapman and Hall, New York, New York, USA.
Ewers, R. M., and R. K. Didham. 2006. Confounding factors in
the detection of species responses to habitat fragmentation.
Biological Reviews 81:117?142.
Eyre, L. A. 1987a. Deforestation in Jamaica. Journal of the
Scientific Research Council of Jamaica 6:17?24.
Eyre, L. A. 1987b. Jamaica: Test case for tropical deforestation?
Ambio 16:338?343.
Fagan, W. F., R. S. Cantrell, and C. Cosner. 1999. How habitat
edges change species interactions. American Naturalist 153:
166?182.
Fahrig, L. 2003. Effects of habitat fragmentation on biodiver-
sity. Annual Review of Ecology, Evolution, and Systematics
34:487?515.
Farnsworth, G. L., K. H. Pollock, J. D. Nichols, T. R. Simons,
J. E. Hines, and J. R. Sauer. 2002. A removal model for
estimating detection probabilities from point-count surveys.
Auk 119:414?425.
Ford, H. A., G. W. Barrett, D. A. Saunders, and H. F. Recher.
2001. Why have birds in the woodlands of southern Australia
declined? Biological Conservation 97:71?88.
Forestry Department. 1999. Land use/cover types areas 1989
and 1998 per watershed management unit and protection
status, Jamaica. Trees for Tomorrow Project, Kingston,
Jamaica.
CHRISTINA M. KENNEDY ET AL.666
Ecological Monographs
Vol. 80, No. 4
Forestry Department. 2001. National forest management and
conservation plan. Jamaica Forestry Department, Kingston,
Jamaica.
Forman, R. T. T. 1995. Land mosaics: the ecology of
landscapes and regions. Cambridge University Press, Cam-
bridge, UK.
Friesen, L. E., P. F. J. Eagles, and R. J. MacKay. 1995. Effects
of residential development on forest dwelling Neotropical
migrant songbirds. Conservation Biology 9:1408?1414.
Gascon, C., T. E. Lovejoy, R. O. Bierregaard, J. R. Malcolm,
P. C. Stouffer, H. L. Vasconcelos, W. F. Laurance, B.
Zimmerman, M. Tocher, and S. Borges. 1999. Matrix habitat
and species richness in tropical forest remnants. Biological
Conservation 91:223?229.
Gaston, K. J. 1996. Species-range-size distributions: patterns,
mechanisms and implications. Trends in Ecology and
Evolution 11:197?201.
Gobeil, J. F., and M. A. Villard. 2002. Permeability of three
boreal forest landscape types to bird movements as
determined from experimental translocations. Oikos 98:
447?458.
Goodwin, B. J., and L. Fahrig. 2002. How does landscape
structure influence landscape connectivity? Oikos 99:552?
570.
Gotelli, N. J., and R. K. Colwell. 2001. Quantifying biodiver-
sity: procedures and pitfalls in the measurement and
comparison of species richness. Ecology Letters 4:379?391.
Haila, Y. 2002. A conceptual genealogy of fragmentation
research: from island biogeography to landscape ecology.
Ecological Applications 12:321?334.
Hansen, A. J., and D. L. Urban. 1992. Avian response to
landscape pattern: the role of species life histories. Landscape
Ecology 7:163?180.
Hanski, I. 1998. Metapopulation dynamics. Nature 396:41?49.
Henle, K., K. F. Davies, M. Kleyer, C. Margules, and J. Settele.
2004. Predictors of species sensitivity to fragmentation.
Biodiversity and Conservation 13:207?251.
Hobbs, R. J. 2001. Synergisms among habitat fragmentation,
livestock grazing, and biotic invasions in southwestern
Australia. Conservation Biology 15:1522?1528.
Hobson, K. A., and E. Bayne. 2000. Effects of forest
fragmentation by agriculture on avian communities in the
southern boreal mixed woods of western Canada. Wilson
Bulletin 112:373?387.
Holdridge, L. R. 1967. Life zone ecology. Tropical Science
Center, San Jose, Costa Rica.
Hothorn, T., F. Bretz, and P. Westfall. 2008. multcomp:
simultaneous inference in general parametric models. R
package version 1.0-0. R Foundation for Statistical Com-
puting, Vienna, Austria.
Hutto, R. L., S. M. Pletschet, and P. Hendricks. 1986. A fixed-
radius point count method for nonbreeding and breeding
season use. Auk 103:593?602.
Johnson, D. H. 2008. In defense of indices: the case of bird
surveys. Journal of Wildlife Management 72:857?868.
Kattan, G. H., H. Alvarezlopez, and M. Giraldo. 1994. Forest
fragmentation and bird extinctions: San Antonio eighty years
later. Conservation Biology 8:138?146.
Kennedy, C. M. 2009. Matrix effects on individual and
community-level responses of birds to forest fragmentation
in Jamaica. Dissertation. University of Maryland, College
Park, Maryland, USA.
Kupfer, J. A., G. P. Malanson, and S. B. Franklin. 2006. Not
seeing the ocean for the islands: the mediating influence of
matrix-based processes on forest fragmentation effects.
Global Ecology and Biogeography 15:8?20.
Lack, D. 1976. Island biogeography illustrated by the land
birds of Jamaica. University of California Press, Berkeley,
California, USA.
Lampila, P., M. Monkkonen, and A. Desrochers. 2005.
Demographic responses by birds to forest fragmentation.
Conservation Biology 19:1537?1546.
Laurance, W. F. 1991. Ecological correlates of extinction
proneness in Australian tropical rain forest mammals.
Conservation Biology 5:79?89.
Laurance, W. F. 2008. Theory meets reality: how habitat
fragmentation research has transcended island biogeographic
theory. Biological Conservation 141:1731?1744.
Laurance, W. F., and M. A. Cochrane. 2001. Special section:
synergistic effects in fragmented landscapes. Conservation
Biology 15:1488?1489.
Laurance, W. F., T. E. Lovejoy, H. L. Vasconcelos, E. M.
Bruna, R. K. Didham, P. C. Stouffer, C. Gascon, R. O.
Bierregaard, S. G. Laurance, and E. Sampaio. 2002.
Ecosystem decay of Amazonian forest fragments: a 22-year
investigation. Conservation Biology 16:605?618.
Legendre, P., and L. Legendre. 1998. Numerical ecology.
Second edition. Elsevier Science, Amsterdam, The Nether-
lands.
Lindenmayer, D. B., R. B. Cunningham, C. F. Donnelly, H.
Nix, and B. D. Lindenmayer. 2002. Effects of forest
fragmentation on bird assemblages in a novel landscape
context. Ecological Monographs 72:1?18.
Luck, G. W., and G. C. Daily. 2003. Tropical countryside bird
assemblages: Richness, composition, and foraging differ by
landscape context. Ecological Applications 13:235?247.
Lugo, A. E. 2002. Can we manage tropical landscapes? An
answer from the Caribbean perspective. Landscape Ecology
17:601?615.
MacArthur, R. H., and E. O. Wilson. 1967. The theory of
island biogeography. Princeton University Press, Princeton,
New Jersey, USA.
Mac Nally, R. 2007. Use of the abundance spectrum and
relative-abundance distributions to analyze assemblage
change in massively altered landscapes. American Naturalist
170:319?330.
Maindonald, J., and J. Braun. 2003. Data analysis and graphics
using R: an example-based approach. Cambridge University
Press, New York, New York, USA.
Martin, T. E. 1993. Nest predation among vegetation layers
and habitat types: revising the dogmas. American Naturalist
141:897?913.
Marzluff, J. M., R. Bowman, and R. Donnelly. 2001. Avian
ecology and conservation in an urbanizing world. Kluwer
Academic, Boston, Massachusetts, USA.
McCune, B., and J. B. Grace. 2002. Analysis of ecological
communities. MjM Software Design, Gleneden Beach,
Oregon, USA.
Mielke, P. W., and K. J. Berry. 2001. Permutation methods: a
distance function approach. Springer-Verlag, New York,
New York, USA.
Miller, J. R., J. A. Wiens, N. T. Hobbs, and D. M. Theobald.
2003. Effects of human settlement on bird communities in
lowland riparian areas of Colorado (USA). Ecological
Applications 13:1041?1059.
Norton, M. R., S. J. Hannon, and F. K. A. Schmiegelow. 2000.
Fragments are not islands: patch vs landscape perspectives on
songbird presence and abundance in a harvested boreal
forest. Ecography 23:209?223.
Oksanen, J., R. Kindt, P. Legendre, B. O?Hara, G. L. Simpson,
M. Henry, H. H. Stevens, and H. Wagner. 2008. Vegan:
community ecology package. R package version 1.13-1. R
Foundation for Statistical Computing, Vienna, Austria.
hhttp://vegan.r-forge.r-project.org/i
Parrotta, J. A., and O. H. Knowles. 1999. Restoration of
tropical moist forests on bauxite-mined lands in the Brazilian
Amazon. Restoration Ecology 7:103?116.
November 2010 667AVIAN RESPONSE TO LANDSCAPE MATRIX
Parrotta, J. A., and O. H. Knowles. 2001. Restoring tropical
forests on lands mined for bauxite: examples from the
Brazilian Amazon. Ecological Engineering 17:219?239.
Pereira, H. M., G. C. Daily, and J. Roughgarden. 2004. A
framework for assessing the relative vulnerability of species
to land-use change. Ecological Applications 14:730?742.
Peres, C. A. 2001. Synergistic effects of subsistence hunting and
habitat fragmentation on Amazonian forest vertebrates.
Conservation Biology 15:1490?1505.
Pinheiro, J. C., and D. M. Bates. 2002. Mixed effects models in
S and S-Plus. Springer, New York, New York, USA.
Pinheiro, J., D. Bates, S. DebRoy, and D. Sarkar, R Core
Team. 2008. nlme: linear and nonlinear mixed effects models.
R package version 3.1-89. R Foundation for Statistical
Computing, Vienna, Austria.
Porter, A. R. D. 1990. Jamaica: a geological portrait. Institute
of Jamaica Publications, Kingston, Jamaica.
R Development Core Team. 2008. R: A language and
environment for statistical computing. Version 2.8.1. R
Foundation for Statistical Computing, Vienna, Austria.
Raffaele, H., J. Wiley, O. Garrido, A. Keith, and J. Raffaele.
1998. A guide to the birds of West Indies. Princeton
University Press, Princeton, New Jersey, USA.
Renjifo, L. M. 1999. Composition changes in a subandean
avifauna after long-term forest fragmentation. Conservation
Biology 13:1124?1139.
Renjifo, L. M. 2001. Effect of natural and anthropogenic
landscape matrices on the abundance of subandean bird
species. Ecological Applications 11:14?31.
Revilla, E., T. Wiegand, F. Palomares, P. Ferreras, and M.
Delibes. 2004. Effects of matrix heterogeneity on animal
dispersal: from individual behavior to metapopulation-level
parameters. American Naturalist 164:E130?E153.
Ribon, R., J. E. Simon, and G. T. De Mattos. 2003. Bird
extinctions in Atlantic forest fragments of the Vicosa region,
southeastern Brazil. Conservation Biology 17:1827?1839.
Ricketts, T. H. 2001. The matrix matters: effective isolation in
fragmented landscapes. American Naturalist 158:87?99.
Ricklefs, R., and E. Bermingham. 2008. The West Indies as a
laboratory of biogeography and evolution. Philosophical
Transactions of the Royal Society B 363:2393?2413.
Ries, L., R. J. Fletcher, J. Battin, and T. D. Sisk. 2004.
Ecological responses to habitat edges: mechanisms, models,
and variability explained. Annual Review of Ecology,
Evolution, and Systematics 35:491?522.
Robbins, C. S., D. K. Dawson, and B. A. Dowell. 1989. Habitat
area requirements of breeding forest birds of the middle
Atlantic states. Wildlife Monographs 103:1?34.
Robinson, S. K., F. R. Thompson, T. M. Donovan, D. R.
Whitehead, and J. Faaborg. 1995. Regional forest fragmen-
tation and the nesting success of migratory birds. Science 267:
1987?1990.
Robinson, W. D. 1999. Long-term changes in the avifauna of
Barro Colorado Island, Panama, a tropical forest isolate.
Conservation Biology 13:85?97.
Rodewald, A. D. 2003. The importance of land uses within the
landscape matrix. Wildlife Society Bulletin 31:586?592.
Rodewald, A. D., and M. H. Bakermans. 2006. What is the
appropriate paradigm for riparian forest conservation?
Biological Conservation 128:193?200.
Rodewald, A. D., and R. H. Yahner. 2001. Influence of
landscape composition on avian community structure and
associated mechanisms. Ecology 82:3493?3504.
Ruiz-Gutierrez, V., T. A. Gavin, and A. A. Dhondt. 2008.
Habitat fragmentation lowers survival of a tropical forest
bird. Ecological Applications 18:838?846.
Saunders, D. A., R. J. Hobbs, and C. R. Margules. 1991.
Biological consequences of ecosystem fragmentation: a
review. Conservation Biology 5:18?32.
Sekercioglu, C. H., P. R. Ehrlich, G. C. Daily, D. Aygen, D.
Goehring, and R. F. Sandi. 2002. Disappearance of
insectivorous birds from tropical forest fragments. Pro-
ceedings of the National Academy of Sciences USA 99:
263?267.
Sigel, B. J., T. W. Sherry, and B. E. Young. 2006. Avian
community response to lowland tropical rainforest isolation:
40 years of change at La Selva Biological Station, Costa Rica.
Conservation Biology 20:111?121.
Simmons, J. A., W. S. Currie, K. N. Eshleman, K. Kuers, S.
Monteleone, T. L. Negley, B. R. Pohlad, and C. L. Thomas.
2008. Forest to reclaimed mine land use change leads to
altered ecosystem structure and function. Ecological Appli-
cations 18:104?118.
Simons, T. R., M. W. Alldredge, K. H. Pollock, and J. M.
Wettroth. 2007. Experimental analysis of the auditory
detection process on avian point counts. Auk 124:986?999.
Sodhi, N. S., L. H. Liow, and F. A. Bazzaz. 2004. Avian
extinctions from tropical and subtropical forests. Annual
Review of Ecology, Evolution, and Systematics 35:323?345.
Sokal, R. R., and F. J. Rohlf. 1995. Biometry: the principles
and practice of statistics in biological research. Third edition.
W. H. Freeman, New York, New York, USA.
Space Imaging. 2002. IKONOS imagery products and product
guide. Space Imaging, Thornton, Colorado, USA.
Stamps, J. A., M. Buechner, and V. V. Krishnan. 1987. The
effects of edge permeability and habitat geometry on
emigration from patches of habitat. American Naturalist
129:533?552.
Stouffer, P. C., and R. O. Bierregaard. 1995. Effects of forest
fragmentation on understory hummingbirds in Amazonian
Brazil. Conservation Biology 9:1085?1094.
Stouffer, P. C., C. Strong, and L. N. Naka. 2009. Twenty years
of understory bird extinctions from Amazonian rain forest
fragments: consistent trends and landscape-mediated dynam-
ics. Diversity and Distributions 15:88?97.
Stratford, J. A., and W. D. Robinson. 2005. Gulliver travels to
the fragmented tropics: geographic variation in mechanisms
of avian extinction. Frontiers in Ecology and the Environ-
ment 3:91?98.
Strayer, D. L., M. E. Power, W. F. Fagan, S. T. A. Pickett, and
J. Belnap. 2003. A classification of ecological boundaries.
BioScience 53:723?729.
Sutherland, G. D., A. S. Harestad, K. Price, and K. P.
Lertzman. 2000. Scaling of natal dispersal distances in
terrestrial birds and mammals. Conservation Ecology 4:16.
Tabarelli, M., M. J. C. Da Silva, and C. Gascon. 2004. Forest
fragmentation, synergisms and the impoverishment of
Neotropical forests. Biodiversity and Conservation 13:
1419?1425.
Terborgh, J., S. K. Robinson, and T. A. Parker III. 1990.
Structure and organization of an Amazonian forest bird
community. Ecological Monographs 60:213?238.
Theobald, D. M., J. R. Miller, and N. T. Hobbs. 1997.
Estimating the cumulative effects of development on wildlife
habitat. Landscape and Urban Planning 39:25?36.
Therneau, T. M., and B. Atkinson. 2009. rpart: recursive
partitioning. R package version 3.1-43. R Foundation for
Statistical Computing, Vienna, Austria. hhttp://mayoresearch.
mayo.edu/mayo/research/biostat/splusfunctions.cfmi
Tole, L. 2001. Jamaica?s disappearing forests: physical and
human aspects. Environmental Management 28:455?467.
Vandermeer, J., and R. Carvajal. 2001. Metapopulation
dynamics and the quality of the matrix. American Naturalist
158:211?220.
Walther, B. A., and J. L. Moore. 2005. The concepts of bias,
precision and accuracy, and their use in testing the
performance of species richness estimators, with a literature
review of estimator performance. Ecography 28:815?829.
CHRISTINA M. KENNEDY ET AL.668
Ecological Monographs
Vol. 80, No. 4
Westfall, P. H., and S. S. Young. 1993. Resampling-based
multiple testing: examples and methods for P-value adjust-
ment. John Wiley and Sons, New York, New York, USA.
Wethered, R., and M. J. Lawes. 2003. Matrix effects on bird
assemblages in fragmented Afromontane forests in South
Africa. Biological Conservation 114:327?340.
Wunderle, J. M., and R. B. Waide. 1993. Distribution of
overwintering Nearctic migrants in the Bahamas and Greater
Antilles. Condor 95:904?933.
Zanette, L., P. Doyle, and S. M. Tremont. 2000. Food shortage
in small fragments: evidence from an area-sensitive passerine.
Ecology 81:1654?1666.
APPENDIX A
Forest fragmentation pattern among human-dominated landscape types (Ecological Archives M080-023-A1).
APPENDIX B
Habitat structure within human-dominated matrices (Ecological Archives M080-023-A2).
APPENDIX C
Vegetation structure within forest patches by matrix type (Ecological Archives M080-023-A3).
APPENDIX D
Sampling effort across landscapes in 2005?2007 (Ecological Archives M080-023-A4).
APPENDIX E
Estimating species detection probabilities by matrix type (Ecological Archives M080-023-A5).
APPENDIX F
Traits of native resident Jamaican birds (Ecological Archives M080-023-A6).
November 2010 669AVIAN RESPONSE TO LANDSCAPE MATRIX