Habitat selection and population trends in terrestrial bird species of Robinson Crusoe … · 2017-03-10 · Bird sampling We recorded abundance of terrestrial birds in Robinson Crusoe

Page 1

ORI GIN AL PA PER

Habitat selection and population trends in terrestrialbird species of Robinson Crusoe Island: habitatgeneralists versus forest specialists

Ingo J. Hahn • Pablo M. Vergara • Uwe Romer

Received: 15 October 2010 / Accepted: 2 July 2011� Springer Science+Business Media B.V. 2011

Abstract Habitat loss and degradation on oceanic islands are key processes leading to

population decline of endemic birds and facilitating the establishment of invasive bird

species. In this study, carried out in the Robinson Crusoe Island, we assessed density and

habitat selection of terrestrial bird species, including juan fernandez firecrown and juan

fernandez tit-tyrant, two endemics, as well as green-backed firecrown and austral thrush,

which apparently originate from the mainland. Results show that perturbed habitats con-

tained a low density of the endemic species whereas the mainland species were signifi-

cantly more abundant in perturbed scrub habitats. Bird species show different habitat

selection patterns, with endemics selecting for native forest and mainland species selecting

for perturbed habitats, or using them at random. Bird species experienced temporal trends

in their overall population sizes, with the endemic tit-tyrant suffering a significant decline

in its population size of about 63% between 1994 and 2009. Only mainland species

exhibited temporal changes in habitat use, significantly reducing their densities in the

preferred scrub habitats, possibly as a response to decreased habitat quality. Thrushes

apparently were able to compensate the population decrease in one non native habitat type

by using native forests, a habitat giving them the opportunity of preying on nests of

endemic species. We conclude that endemic bird species behave as specialists whereas the

mainland species must be treated as invasive generalists on Robinson Crusoe Island.

I. J. Hahn (&)Institute of Landscape Ecology, University of Munster, Robert-Koch-Str. 28, 48149 Munster, Germanye-mail: [email protected]

P. M. VergaraDepartment of Engineer Geography, Universidad de Santiago de Chile, Av. Lib. B. O’Higgins 3363,7254758 Santiago, Chile

P. M. VergaraCEUS Llanquihue, Universidad de Santiago de Chile, Santiago de Chile, Chile

U. RomerInstitute of Biogeography, University of Trier, Am Wissenschaftspark 25–27, 54296 Trier, Germany

123

Biodivers ConservDOI 10.1007/s10531-011-0109-x

Page 2

Keywords Anairetes fernandezianus � Habitat selection � Invasive species �Population declining � Sephanoides fernandensis � Juan Fernandez Islands

Introduction

Habitat loss and degradation on oceanic islands are key processes leading to population

decline of endemic birds and facilitating the spread of invasive bird species (Catry et al.

2000; Byers 2002; Soh et al. 2006). These detrimental effects lead to progressive

replacement of endemics by exotic species resulting in increased biotic homogenization

(Castro and Jaksic 2008). Since oceanic islands are biodiversity hot-spots supporting a

large proportion of critically endangered endemic animals (Johnson and Stattersfield 1990;

IUCN 2011), it is critical to understand how species select habitats on islands which are

exposed to increased habitat disturbances, and how these animal decisions determine

population dynamics and persistence.

Human induced disturbances on oceanic islands have resulted in the degradation and

replacement of the original vegetation by new habitats and land uses, including urban,

agricultural and livestock grazing. Island-endemic bird species which are ecologically

specialized tend to be disproportionately vulnerable to those habitat changes since their

population sizes are typically small due to their naturally restricted distributions (Owens

and Bennett 2000; Fordham and Brook 2010). However, endemic and native species may

display broad habitat use patterns on islands, resulting in the spread of their populations

into the new human-created habitats (Blondel et al. 1988; Palomino and Carrascal 2005;

Trainor 2007; Carrascal et al. 2008). On the contrary, the successful establishment of an

exotic (i.e. introduced) species usually requires that the species is able to spread into non-

perturbed habitats after its introduction (Crawley 1986; Sakai et al. 2001). Therefore, both

native and exotic species may become invasive on an island depending on their habitat

selection attributes (Colautti and MacIsaac 2004; Vergara et al. 2011).

Habitat generalism is an important attribute in determining the ability to persist in

heterogeneous and human-perturbed landscapes (Sol et al. 1997; Marvier et al. 2004;

Vergara and Armesto 2009). Habitat generalist species have broad habitat amplitudes

(occupying several habitat types) and usually they behave as opportunistic species, pre-

ferring the habitats offering more resources (Medel and Jaksic 1988; Magura et al. 2003).

Unlike specialist species, which are constrained to use a small habitat spectrum, habitat

generalists may switch their habitat selection pattern over time as expected from habitat

selection theory (Fretwell and Lucas 1970; Rosenzweig 1985; Latta and Faaborg 2002;

Chen et al. 2008). Consequently, changes in population size and in the spatial distribution

of their resources over time may result in a wide temporal variability in habitat use by

generalist species (Diamond 1975; Holt 1993; Mobæk et al. 2009).

Robinson Crusoe Island supports two endemic terrestrial bird species, the juan fer-

nandez firecrown (Sephanoides fernandensis) and juan fernandez tit-tyrant (Anairetesfernandezianus), which have been reported using remnant native forests (Hahn et al. 2005,

2006). The native forests have experienced a rapid spatial reduction and degradation in the

last centuries after human colonization as a result of the land-use change and introduction

of exotic herbivores (Cuevas and van Leersum 2001; Ricci 2006). As a consequence of

habitat loss and degradation, in addition to the potentially high risk from exotic predators

such as rats, cats, and coatis, juan fernandez firecrown and tit-tyrant have population sizes

Biodivers Conserv

123

Page 3

small enough to be considered as critically endangered and near threatened, respectively

(comp. IUCN 2011).

At present two terrestrial bird species native to the mainland, the green-backed fire-

crown (Sephanoides sephaniodes) and the austral thrush (Turdus falcklandii magellanicus)

coexist with the endemic firecrown and tit-tyrant in native forest habitats of the Robinson

Crusoe Island (Hahn et al. 2006, 2009, 2011). Although there is no conclusive evidence,

previous studies suggest that both green-backed firecrown and austral thrush are exotic

species in Robinson Crusoe Island (Audouin 1830; Brooke 1987; Stiles 1987; Colwell

1989; Roy et al. 1998, 1999; Hahn et al. 2006, 2009; C. Smith-Ramirez, pers. comm.).

Irrespective of their origin, green-backed firecrown and austral thrush in southern South

America have a wide geographical distribution, high local abundances and use a wide

variety of habitats, including mature forest, degraded forest, forest plantations as well as

agricultural and urban areas (e.g. Willson et al. 1994; Estades 1997; Anderson and Rozzi

2000; Jaramillo 2003; Vergara and Simonetti 2004; Grigera and Pavic 2007; Vergara and

Armesto 2009; Vergara et al. 2010). The ability to use both native forest and human-

modified habitats in the mainland, coupled with the fact that these species pollinate and

disperse exotic plant species on the island, may make green-backed firecrowns and austral

thrushes invasive species on Robinson Crusoe Island (Duncan et al. 2003; Colautti and

MacIsaac 2004). Although previous studies have reported both species using a variety of

human-created habitats in Robinson Crusoe Island (e.g. Hahn et al. 2005), their habitat

selection patterns and population developments are still unknown.

In this study, we assessed the habitat selection pattern and the temporal variation in

population size of terrestrial bird species of Robinson Crusoe Island. Specifically, we tested

the following hypotheses: (i) The abundance of both green-backed firecrowns and austral

thrushes is equal or greater in disturbed habitat than in native forest. These species ran-

domly use habitats or select for disturbed habitats (since human created habitats are used

by both species in the mainland); (ii) The abundance of the endemic firecrown and tit-

tyrant is greater in native forest. These endemic species behave as habitat specialists

selecting for native forest or low perturbed habitats; (iii) Endemic species exhibit declining

population sizes; (iv) Green-backed firecrowns and austral thrushes exhibit habitat-

dependent temporal variation in abundance due to changes in population size and in the

spatial distribution of habitat resources over time. In particular, the relative abundance

pattern can change when the variability in climatic conditions, the human disturbance

regimes and the population size affect differently to the suitability of each habitat. These

temporal changes in abundance should result in significant interactions between year and

habitat type.

Methods

Study area

Robinson Crusoe Island is the easternmost of the three islands of the Juan Fernandez

Archipelago, located ca. 600 km off the Chilean coast in the South-east Pacific Ocean

(33�3303800S and 78�5604400W). The island area is 47.11 km2, reaching an altitude of 915 m

above sea-level. Juan Fernandez Archipelago belongs to the few places in the Pacific

Ocean which remained untouched by humans until the European colonization. The island

was named after the sailor Alexander Selkirk, who became known as the model of Daniel

Defoe’s novel in the early eighteenth century. The entire island, with exception of the

Biodivers Conserv

123

Page 4

settlement area, is a Chilean National Park since 1935 and UNESCO Biosphere Reserve

since 1977.

Early reports indicate that the island was forested before the arrival of Europeans (e.g.

Skottsberg 1953). The Juan Fernandez Archipelago is characterised by a high number and

proportion of endemic plant (Stuessy and Ono 1998) and animal species (Kuschel 1963).

Amongst the 310 breeding bird species of Chile (Araya et al. 1998) only 15 are endemic,

but five of them, and another three endemic subspecies, are limited to this archipelago.

Therefore, it is not only of major importance for the endemic avifauna of Chile but also of

international interest as an Endemic Bird Area (EBA), being categorised as one of 12

Natural Areas of high conservation importance worldwide (Allen 1984; Stattersfield et al.

1998).

Habitat classification

We used the criteria of Hahn et al. (2005, 2010) as a guide for grouping vegetation units

(hereafter called ‘‘habitat types’’). This habitat classification is focused on habitat features

important for terrestrial birds, such as vegetation structure and dominant plant species, as

well as geographical factors like altitude, exposition and incline. Habitat types were

mapped using recent aerial photographs and geo-referenced QuickBird Satellite Images

available from Google Earth Pro program (Fig. 1). Vegetation from the resulting map was

checked in the field by comparing it with geo-referenced vegetation plots (Hahn et al.

2005, 2010). From these habitat types, grassland is the spatially most extended habitat type

on Robinson Crusoe (Greimler et al. 2002), covering more than half of the island’s surface,

whereas native forests cover one fifth. The remaining three habitats are predominantly

characterised by non-native plants (Table 1).

Fig. 1 Habitat types and their distribution on Robinson Crusoe Island (see Table 1). Habitat types codesare: 1 native forest, 2 exotic scrub, 3 ridge scrub, 4 settlement, 5 grassland

Biodivers Conserv

123

Page 5

Bird sampling

We recorded abundance of terrestrial birds in Robinson Crusoe during their 1994, 2001 and

2009 breeding seasons (see below). Visual identification of breeding birds was straight

forward, using Araya et al. (1998), Jaramillo (2003), and the original species descriptions.

Acoustic identification was possible after learning the bird vocalisations from observation

and taping with a DAT-Recorder (Sony, HD-S100) (e.g. Hahn 2006).

We estimated density of our focal species using fixed-distance line transects (i.e. strip

transects), a sampling method which assumes that all animals within the fixed width of the

transect have a detection probability close to 1.0 (e.g. Skalski et al. 2005). Consequently, in

order to ensure the detection of all birds within the strip transect, we used a 25 m transect

half-width (e.g. Skalski et al. 2005). Such transect half-width distance was selected

because our previous data suggest that the detection probability of all study species in

forest and shrub habitats drops significantly beyond 30 m from the observer due to dense

vegetation structure and the steep topography of the study sites (Hahn et al. 2006; Hahn,

unpubl.). By using such a relatively short detection distance we also minimized the

probability that the same bird was counted twice during a single strip transect (Jimenez

2000).

Table 1 Differences in vegetation structure, dominant plant species, and topography among the mainhabitat types of Robinson Crusoe Island

Habitatcharacteristics

Habitat typea

Native forest Exotic scrub Ridge scrubb Settlement Grassland

Nativity status Mainly native Mainly non-native

Mainly non-native

Mainly non-native Mainly non-native

Altitudinalrange(a.s.l.), m

220–650 100–360 150–915 0–170 0–400

Surface area,ha

990 290 599 109 2723

Dominantplant genera

Myrceugenia,Drimys,Fagara

Aristotelia,Rubus, Ugni

Ugni,Pernettya

Crop/flowers,Eucalyptus,Cupressus

Poa, Stipa,Acaena

Number ofstoreys (tree,shrub, herb)

2–3 1–2 1–2 1–3 1

Vegetationheight, mean(range) m

14 (5–22) 2.5 (1.5–6) 1.5 (0.8–3) Diverse (0.1–26) 0.2 (0–0.5)

Upper storeycover, mean(range) %

75 (50–95) 90 (60–95) 95 (80–98) Diverse (0–95) Diverse (0–95)

Topographicclassification

Forest of themontaneregion

Scrubvegetation ofvalley floorand valleyslope

Scrubvegetationofmountainridges

Cultivated land,gardens and treeplantations nearsea-level

Grasslands,rocky anderosive terrainof basal region

a More detailed information on vegetation structure and plant species dominance may be taken from Hahnet al. (2005, 2010)b Sites with steeper slopes ([40%) and thinner soils than exotic scrub

Biodivers Conserv

123

Page 6

During 1994 a total of 23 strip transects was established, covering all habitat types. The

number of strip transects was relatively proportional to the area of each habitat (Tables 1,

2), with the exception of grassland, where only three strip transects were made, due to few

previous records of individual bird in such habitat type (in fact, no bird species was

observed feeding or nesting in grassland). Although the number of strip transects increased

to 39 and 46 in 2001 and 2009, respectively; the same transects were revisited during later

study period. Despite the sample size differences between years, we did not note important

yearly differences in the standard errors of bird density estimates.

Each strip transect was visited once during the breeding season of the birds, thus

reducing the likelihood of double counting. Strip transects were surveyed from November

5th through January 11th. Within this period of time we noticed no differences in the

conspicuity and activity of these birds, hence we did not expect differences in bird

detectability (e.g. Hahn et al. 2011). Transects were on average 950 m long,[200 m apart

to avoid double counting of birds and[100 m distant from the habitat boundary to reduce

the effect on the abundance of neighbour habitats. Transect length was measured directly

in the field by using a pedometer and correcting for the slope angle. Abundance records in

each year were obtained by the same observer. Line transects were conducted mostly

during morning hours (06:45–09:30 local time) on days without rain and little wind

(\15 km/h). The observer walked slowly (mean speed = 0.5 km/h) along transects and

counted all bird individuals seen or heard either side of the census trail. The perpendicular

distances from the transect to the bird were estimated visually and checked using a laser

rangefinder.

Data analysis

Habitat selection, density and population size of bird species were assessed using Hier-

archical Bayesian Models which provide a powerful tool for modelling spatio-temporal

data like that of this study (Link and Sauer 2002; Clark 2005). Bayesian inference treats

model parameters as random variables and hence provides an approach for fitting hierar-

chical models (Clark 2005). We used Gaussian regression with bird density of each species

as dependent variable. Habitat type, year and the interaction between year and habitat were

specified as fixed effects. In addition, we included altitude in order to control for the

possible effects of within-habitat altitudinal gradients in environmental factors (Table 1).

The following nuisance (random) effects were included: (1) Non-structured spatial terms

representing the locality (geographic area) effect; (2) Sample-level terms, representing

variability accounted for differences between transects within localities; and (3) Obser-

vation-level terms, accounting for temporal random variation in density within transects.

Table 2 Number of strip transects made in the five main habitat types of Robinson Crusoe Island

Year Habitat type

Native forest Exotic scrub Ridge scrub Settlement Grassland

1994 6 4 4 6 3

2001 9 8 5 11 6

2009 12 12 5 14 3

See Table 1 for habitat description

Biodivers Conserv

123

Page 7

We used Markov chain Monte Carlo (MCMC) and Gibbs sampling for fitting models and

parameter estimation (Lunn et al. 2000; Spiegelhalter et al. 2003). For each model sim-

ulation, MCMC were run for 300,000 iterations, including a 50,000 iteration burn. We

computed three independent MCMC chains using as reference different starting parameter

values for allowing convergence (Brooks and Gelman 1998). Uninformative prior distri-

butions were assigned to parameters in the model because little information is known about

how environmental variables influence bird density.

Density was normalized by log transformation. We used a corner constraint approach to

compare density between habitats, where the effect of native forest is set equal to zero in

order to have a baseline or reference category for the habitat factor (Ntzoufras 2009).

Population sizes were calculated by scaling MCMC density estimates to the entire surface

of each habitat type (Table 1; Hahn et al. 2006). We also used the Bayesian framework to

calculate a Habitat Selection Index (HSI). Basically a HSI represents the likelihood that a

particular habitat type would be used given its relative availability in the landscape (Manly

et al. 1993). Since all habitat types were sampled extensively within the island (Table 2),

we estimated HSI on the basis of the expected number of individuals of each species

occupying each habitat type. Thus, the HSI of the ith habitat (HSIi) was calculated as:

HSIi ¼ni

Pki niwi

where ni is the expected density in habitat i obtained from MCMC simulations (see below), wi

is the weight of the ith habitat in terms of their relative surface area and k represents the total

number of available habitats. Thus, if habitat i is used at random, then HSI will be 1.0 for this

habitat type whereas significant departures from 1.0 indicate that either habitat use is

selective (if HSI [ 1) or habitat is underused (if HSI \ 1). However, HSI, as estimated from

the above expression, could be biased to the extent that habitats with a very high relative area

are included (e.g. grassland habitat, which covers more than 50% of the island area). For

avoiding this problem, and considering that bird records of all species were scant in grassland

habitat, we did not include this habitat type in our HSI estimation. The exclusion of grassland

from this analysis did not affect our conclusions about the habitat selectivity pattern of birds

since we are confident that all species underused this habitat type (i.e. HSI * 0).

Density differences between habitat types and years were evaluated by comparing

MCMC coefficients and their 95% Bayesian credibility intervals. A coefficient can be

considered as ‘‘significant’’ when their 95% credibility intervals did not overlap zero (i.e.

the 2.5 and 97.5th percentile). We also used the above MCMC approach to determine if

HSI selection indices differed significantly from 1.0 and for computing population size for

each sampling year. Model fit was assessed using the proportion of explained deviance (r2)

and by testing deviance differences with null models containing only nuisance variables

(e.g. Thogmartin et al. 2004). We used R2WinBUGS (Sturtz et al. 2005) as a platform for

programming and running MCMC.

Results

Bird density

For all four species, the goodness of fit tests indicated that hierarchical models fit the data

adequately, as shown by r2 and deviance chi-square test (Table 3). Although species

Biodivers Conserv

123

Page 8

densities decreased significantly in grassland, model results showed that the density of all

bird species was affected differently by habitat perturbations (Table 4). The endemics juan

fernandez firecrown and juan fernandez tit-tyrant had lower densities in exotic scrub than

in native forest, whereas the density of tit-tyrants also was significantly lower in settlement

areas (Table 4). Conversely, the green-backed firecrown and the austral thrush had sig-

nificantly higher density in the human-disturbed exotic scrub and ridge scrub, respectively

(Table 4). In addition, the density of austral thrushes was lower in settlement areas than in

native forest (Table 4).

Habitat selection

Terrestrial bird species showed different habitat selection patterns, with endemics selecting

for native forest whereas the green-backed firecrowns and austral thrushes selecting for

perturbed habitats, or using them at random (Fig. 2). The endemic juan fernandez tit-tyrant

underused (HSI \ 1) all perturbed habitats, including exotic scrub, ridge scrub and set-

tlement areas (Fig. 2). The juan fernandez firecrown also underused exotic scrub and ridge

scrub, but the use of settlement areas was not significantly different from expectations

under random distribution (Fig. 2). The green-backed firecrown selected settlement areas,

in addition to exotic scrub, but also underused ridge scrub and used native forest at random

(Fig. 2). Austral thrushes used all habitats at random, with the exception of settlement

areas, which were underused (Fig. 2).

Temporal pattern

Terrestrial bird species experienced temporal trends in their overall population sizes over

the study time period (Fig. 3). However, only green-backed firecrowns and austral thrushes

displayed habitat-dependent changes in density over time, as shown by significant inter-

actions between year and habitat types (Table 4).

The endemic juan fernandez tit-tyrant and the green-backed firecrown show a declining

population trend (Fig. 3), but only juan fernandez tit-tyrant evidenced a significant

reduction in density between 1994 and 2009, as shown by the MCMC coefficient for the

year 2009 (Table 4). This population decline implied a reduction of 63% from its initial

population size, from 4198.0 (2.5–97.5th percentiles = 2322.9–6361.6) individuals in

1994 to 1551.8 (1005.3–1848.4) individuals in 2009. Although not significant, the

reduction in the size of the population of green-backed firecrowns was about 50%, from

5843.7 (1841.3–8377.4) individuals in 1994 to 2685.2 (1889.2–3215.9) individuals in

2009. The population size of austral thrushes also decreased 45%, from 6275.7

Table 3 Summary of goodness-of-fit statistics for hierarchical models, showing the deviance values(including SD and 95% credibility intervals), proportion of explained deviance (r2) and P-value from adeviance chi-square test

Species Deviance r2 P-value

Mean SD 2.5% 97.5%

Juan fernandez tit-tyrant 55.9 7.1 42.9 71.1 0.74 \0.001

Juan fernandez firecrown 97.3 8.1 81.4 113.7 0.45 \0.001

Green-backed firecrown 116.2 6.3 106.0 130.3 0.46 \0.001

Austral thrush 109.8 8.3 94.3 128.5 0.54 \0.001

Biodivers Conserv

123

Page 9

Ta

ble

4M

ean

s,st

andar

dd

evia

tio

ns

(SD

),an

d9

5%

cred

ible

inte

rval

(95

%C

I)v

alu

eso

fM

CM

Cco

effi

cien

tsfr

om

Bay

esia

nH

iera

rch

ical

Mo

del

ses

tim

atin

gd

ensi

tyo

ffo

ur

terr

estr

ial

bir

dsp

ecie

sin

the

fiv

eh

abit

atty

pes

pre

sen

tin

Ro

bin

son

Cru

soe

Isla

nd

du

ring

the

thre

ece

nsu

sy

ears

of

19

94

,2

00

1,

and

20

09

Coef

fici

ent

Juan

fern

andez

tit-

tyra

nt

Juan

fern

andez

fire

crow

n

Mea

nS

D2

.5%

97

.5%

Mea

nS

D2

.5%

97

.5%

Ex

oti

csc

rub

-0

.93

0.2

4-

1.3

9-

0.4

7-

0.6

40

.31

-1

.23

-0

.01

Rid

ge

scru

b-

0.4

10

.35

-1

.10

0.2

9-

0.8

40

.44

-1

.70

0.0

2

Set

tlem

ent

area

-1

.55

0.2

7-

2.0

7-

1.0

4-

0.1

00

.36

-0

.76

0.6

2

Gra

ssla

nd

-1

.72

0.2

9-

2.3

0-

1.1

6-

1.1

40

.39

-1

.87

-0

.32

Yea

r2

00

1-

0.3

00

.17

-0

.64

0.0

3-

0.2

00

.21

-0

.61

0.2

2

Yea

r2

00

9-

0.5

00

.16

-0

.82

-0

.19

-0

.05

0.2

0-

0.4

50

.35

Ex

oti

csc

rub

9Y

ear

20

01

0.1

40

.26

-0

.36

0.6

50

.08

0.3

2-

0.5

40

.70

Ex

oti

csc

rub

9Y

ear

20

09

0.1

40

.25

-0

.35

0.6

2-

0.1

60

.30

-0

.77

0.4

2

Rid

ge

scru

b9

Yea

r2

00

1-

0.5

90

.41

-1

.41

0.2

30

.20

0.5

1-

0.8

01

.20

Rid

ge

scru

b9

Yea

r2

00

9-

0.4

10

.39

-1

.17

0.3

50

.04

0.4

8-

0.9

10

.98

Set

tlem

ent

9Y

ear

20

01

0.2

10

.24

-0

.25

0.6

8-

0.2

60

.29

-0

.83

0.3

2

Set

tlem

ent

9Y

ear

20

09

0.4

00

.23

-0

.05

0.8

5-

0.4

40

.28

-0

.99

0.1

3

Gra

ssla

nd

9Y

ear

20

01

0.3

40

.29

-0

.23

0.9

10

.21

0.3

5-

0.4

90

.91

Gra

ssla

nd

9Y

ear

20

09

0.1

51

00

.18

-1

95

.60

19

6.0

00

.15

10

0.1

8-

19

5.6

01

96

.00

Alt

itu

de

-0

.001

40

.00

06

-0

.00

26

-0

.00

03

-0

.00

10

0.0

00

8-

0.0

02

50

.000

8

Coef

fici

ent

Gre

en-b

acked

fire

crow

nA

ust

ral

thru

sh

Mea

nS

D2

.5%

97

.5%

Mea

nS

D2

.5%

97

.5%

Ex

oti

csc

rub

0.9

60

.32

0.3

31

.59

0.3

00

.33

-0

.36

0.9

6

Rid

ge

scru

b0

.53

0.5

0-

0.4

51

.52

1.2

30

.47

0.3

02

.15

Set

tlem

ent

area

0.4

30

.35

0.2

61

.12

-0

.88

0.4

0-

1.6

4-

0.0

8

Gra

ssla

nd

-1

.08

0.3

9-

1.8

5-

0.3

0-

1.3

90

.41

-2

.23

-0

.58

Yea

r2

00

10

.19

0.2

4-

0.2

80

.66

0.1

60

.22

-0

.28

0.6

0

Biodivers Conserv

123

Page 10

Ta

ble

4co

nti

nu

ed

Coef

fici

ent

Gre

en-b

acked

fire

crow

nA

ust

ral

thru

sh

Mea

nS

D2

.5%

97

.5%

Mea

nS

D2

.5%

97

.5%

Yea

r2

00

90

.14

0.2

3-

0.3

10

.58

0.5

90

.21

0.1

71

.01

Ex

oti

csc

rub

9Y

ear

20

01

-0

.92

0.3

7-

1.6

2-

0.2

0-

0.0

70

.34

-0

.73

0.5

9

Ex

oti

csc

rub

9Y

ear

20

09

-0

.97

0.3

5-

1.6

6-

0.2

9-

0.9

10

.33

-1

.56

-0

.28

Rid

ge

scru

b9

Yea

r2

00

1-

1.0

10

.59

-2

.16

0.1

5-

1.3

30

.54

-2

.40

-0

.26

Rid

ge

scru

b9

Yea

r2

00

9-

1.3

90

.55

-2

.48

-0

.31

-2

.53

0.5

1-

3.5

4-

1.5

3

Set

tlem

ent

9Y

ear

20

01

-0

.04

0.3

3-

0.7

00

.62

-0

.12

0.3

1-

0.7

30

.50

Set

tlem

ent

9Y

ear

20

09

-0

.57

0.3

2-

1.2

00

.06

-0

.56

0.3

0-

1.1

50

.04

Gra

ssla

nd

9Y

ear

20

01

-0

.16

0.4

0-

0.9

50

.64

-0

.03

0.3

8-

0.7

80

.74

Gra

ssla

nd

9Y

ear

20

09

0.1

51

00

.18

-1

95

.60

19

6.0

00

.15

10

0.1

8-

19

5.6

01

96

.00

Alt

itu

de

-0

.000

70

.00

07

-0

.00

21

0.0

00

7-

0.0

02

20

.000

9-

0.0

04

1-

0.0

00

4

Bold

coef

fici

ents

are

consi

der

edas

signifi

cant

bec

ause

thei

rcr

edib

ilit

yin

terv

als

do

not

over

lap

zero

(see

text)

.C

oef

fici

ents

for

nat

ive

fore

stw

ere

fix

edto

zero

Biodivers Conserv

123

Page 11

0.0

0.5

1.0

1.5

2.0

0.0

1.0

2.0

2.5

0.0

0.5

1.0

1.5

2.0

2.5

0.0

0.5

1.0

1.5

2.0

2.5

Hab

itat s

elec

tion

Exotic scrub Nativeforest

Ridge scrub SettlementExotic scrub Nativeforest

Ridge scrub Settlement

Exotic scrub Nativeforest

Ridge scrub Settlement Exotic scrub Nativeforest

Ridge scrub Settlement

green-backed firecrown austral thrush

juan fernandez firecrownjuan fernandez tit-tyrantin

dex

Hab

itat s

elec

tion

inde

xH

abita

t sel

ectio

nin

dex

Hab

itat s

elec

tion

inde

x

Fig. 2 Box plots showing the MCMC estimates of HSI with a 95% credible interval (staples) for fourterrestrial bird species of Robinson Crusoe Island using four different habitat types. The horizontal line atscale segment 1.0 corresponds to the average abundance value across all habitat types. Thus, HSI valuesequal to 1 indicate random habitat use. For this analysis grasslands was not included because species overalloccur here only by exception (see text)

1994 2001 2009 1994 2001 2009

1994 2001 20091994 2001 2009

020

0060

00

juan fernandez tit-tyrant

Pop

ulat

ion

(n)

010

0030

0050

00

juan Fernandez firecrown

Pop

ulat

ion

(n)

040

0080

0012

000

040

0080

0012

000green-backed firecrown

Pop

ulat

ion

(n)

austral thrush

Pop

ulat

ion

(n)

Fig. 3 Box plots showing the annual MCMC estimates of population size (n = number total of individuals)with a 95% credible interval (staples) for four terrestrial bird species of Robinson Crusoe Island. Boxescorrespond to the 25 and 75% quartiles of the mean population size of each species whereas whiskers extendup to 1.5 times the inter-quartile range of the box

Biodivers Conserv

123

Page 12

(3016.0–9635.4) individuals in 1994 to 3397.1 (2554.5–4139.6) individuals in 2009

(Fig. 3; Table 4). The population size of juan fernandez firecrown tended to be relatively

stable, ranging from 1270 (820.5–3046.4) to 1980 (545.6–1979.3) individuals (Fig. 3).

Green-backed firecrown and austral thrush showed a temporal reduction in their den-

sities in perturbed scrub habitats (Table 4; Fig. 4). In the case of the green-backed fire-

crown, there was a significant negative interaction between: (1) exotic scrub and years

2001 and 2009, and (2) ridge scrub and year 2009 (Table 4). The density of austral

thrushes was negatively affected by the interaction between: (1) exotic scrub and year

2009, and (2) ridge scrub and years 2001 and 2009 (Table 4). These interactions imply that

the density of green-backed firecrowns decreased more steeply in these scrub habitats than

in native forest, where it remained stable (Fig. 4). However, the density of austral thrushes

in native forest exhibited the opposite pattern than in scrub habitats, increasing in the

former and decreasing in the latter case (Fig. 4).

Discussion

The density and habitat use pattern shown by terrestrial birds confirms our first two

hypotheses. The more even spatial distribution of the green-backed firecrown and austral

thrush among available habitats and/or their preference for degraded habitats identifies

them as habitat generalists. This ecological strategy had made them becoming similarly

successful and widespread on the mainland (Johnson 1965, 1967; Vergara and Armesto

2009). Endemic birds might increase their population or distribution after anthropogenic

habitat change (Catry et al. 2000; Trainor 2007). However, this does not appear to be the

austral thrush

0

2

4

6

8

10

1994 2001 2009

Bird

den

sity

(n/

ha) Ridge scrub

Native forest

green-backed firecrown

0

2

4

6

8

10

12

1994 2001 2009

Bird

den

sity

(n/

ha)

Exotic scrub

Native forest

Fig. 4 Significant interactionsbetween habitat types and year(MCMC estimates of mean andtheir 95% credible intervals) forthe density of two invasiveterrestrial bird species ofRobinson Crusoe Island (see text)

Biodivers Conserv

123

Page 13

case for Robinson Crusoe’s endemics because they are forest specialists, showing a clear

habitat preference for the relict forests, such as described for most paleo-tropic island birds

(Riley 2003).

Endemic species used perturbed habitats little, probably due to their evolved morpho-

logical and behavioural attributes making them specialized in using native forest recourses, as

suggested for island bird species (e.g. Blondel 2000). In addition, scrub stands probably offer

few resources for endemics, since juan fernandez firecrowns could be less efficient in

exploiting exotic flowering resources compared to the generalist green-backed firecrown.

Although not assessed in this study, perturbed habitats also may act as barriers for endemic

species, reducing the connectivity between native habitat patches. However, historic habitat

loss in Juan Fernandez has not resulted in severe isolation of forest remnants because pro-

tected native areas are located mostly in connected upland areas of the island (Fig. 1).

Green-backed firecrowns and austral thrushes used native forests at random, indicating

that natural habitats in Robinson Crusoe can be utilized by these species. Green-backed

firecrowns profit from native food resources like nectar of the Rhaphithamnus venustus and

arthropods from Myrceugenia fernandeziana; austral thrushes not only forage in native

forest areas, but also use this habitat for nesting (Hahn, unpubl.). However, the higher

densities and selectivity of green-backed firecrowns and thrushes for exotic scrub suggest

that this habitat could provide them with abundant feeding or nesting resources.

The high abundance of green-backed firecrowns and their preference for settlement

areas likely result from the high flower availability in these sites during most of the year.

Garden habitats could ensure landscape supplementation for several flower-dependent

species in oceanic islands (Waltert et al. 2004; Trainor 2007). Some flowering plants in

settlement areas have been established for conservation reasons, such as the endemic

Dendroseris litoralis, planted to provide juan fernandez firecrowns with flower nectar.

This, in turn, could explain the random use of this habitat by endemic firecrowns. In exotic

scrub, the green-backed firecrown feeds on the flowers of Rubus ulmifolius (an invasive

shrub), being much more efficient in exploiting this resource than the juan fernandez

firecrown (Hagen 2009). Similarly, in ridge and exotic scrub, the austral thrush consumes

berries of the invasive mainland shrubs Aristotelia chilensis, Ugni molinae and R. ul-mifolius, dispersing their seeds over the island. The active use of exotic plants by green-

backed firecrowns and austral thrushes in scrub habitats can be explained by the fact that

these plant–animal interactions have evolved for a long time in the mainland habitats

(Armesto et al. 1987; Rozzi et al. 1996; Amico and Aizen 2005).

The hypothesis that endemic species exhibit declining population sizes was confirmed

for the juan fernandez tit-tyrant whose population dropped dramatically during the study

period. Assuming that the current rate of decline remains constant and that the changes are

not part of oscillating fluctuations, the size of the tit-tyrant population in 20 years will be

less than 100 individuals. Although early loss of forest probably was an important factor

causing population declination, the current rate of forest loss in the island is negligible, and

therefore, we suspect other detrimental effects resulting from human perturbations may be

affecting demographic rates. It is likely that factors like predation by mammals or birds,

competition with invasive species, hunting, and climate change play an important role in

reducing nesting success and adult survival of tit-tyrants in native forest. However, the

small forest area available for the tit-tyrant can lead its small population to extinction due

to increased demographic stochasticity (Lande 1993). As a practical consequence in

conservation classification, the juan fernandez tit-tyrant should be upgraded in the IUCN

red list from the status near threatened to endangered. Although the population of juan

fernandez firecrown had not experienced a sharp decline during our study, several studies

Biodivers Conserv

123

Page 14

suggest that the juan fernandez firecrown population decline occurred mostly during the

twentieth century (Brooke 1987; Stiles 1987; Colwell 1989; Roy et al. 1999). Thus, future

conservation guidelines should modify current management activities, such as the addition

of flowering plants in settlement areas. Increased cat predation has been observed in

gardens as the flowers are placed much lower than in native forests (e.g. Hahn et al. 2010).

It is possible that these flowers attract firecrowns from native forest, and hence settlement

areas could function as trap habitats for endemic hummingbirds (Bourne et al. 1992; Hahn

and Romer 2002).

Our results were conclusive in supporting the hypothesis that green-backed firecrowns

and austral thrushes exhibit temporal changes in habitat use, which could result from changes

in population size and the spatial distribution of resources (Mobæk et al. 2009). The

reduction in the overall population size of both species over the study time period was related

to a decrease in the density of these two species in their preferred scrub habitats. Assuming

that invasive species select habitats that favour their survival and chances of reproduction, an

important question is therefore why habitat suitability did decline in scrub habitats.

Austral thrushes have suffered intensive hunting since mid 1990s to reduce the dis-

persion of invasive plants. Thus, it is possible that hunting pressure was greater in scrub

habitats where thrushes were more abundant and seed dispersal is apparently more intense.

On the contrary, density of thrushes in native forest doubled between 1994 and 2009,

probably as a behavioural response to avoid risky scrub habitats or as the result of a

temporal increase in foraging recourses in native forest. The reduction of green-backed

firecrown density in exotic scrub, however, was not accompanied by an increase in density

in native forest. Due to their feeding habits, green-backed firecrowns are less general in

their habitat requirements than thrushes, and hence firecrowns could be more vulnerable to

density dependent competition for flowering recourses. Therefore, thrushes apparently

were able to compensate the sharp decrease in quality of their preferred habitats by using

native forests, a safer habitat offering few resources. This temporal change in habitat use

could be detrimental for the conservation of the breeding populations of juan fernandez tit-

tyrant and juan fernandez firecrown in native forest, since austral thrushes are recognized

as important nest predators in island forests (Hahn, unpubl.) and mainland habitats (e.g.

Vergara 2007). Therefore, we conclude that endemic bird species behave as specialists

whereas the mainland originating green-backed firecrown and austral thrush must be

treated as invasive generalists on Robinson Crusoe Island. Thus, conservation strategies in

Robinson Crusoe should be to improve habitat conditions for endemics while applying

intensive control for the invasive bird species.

Acknowledgments Two anonymous reviewers were sincerely thanked for their enriching improvementsof contents and English writing. This study was supported by Alexander von Humboldt Foundation (AvH)and FONDECYT 11080085. We thank CONAF for allowing us to work in protected areas of the JuanFernandez Archipelago National Park: special thanks go to M. Galvez, J. Reyes, J. Mesa, G. Gonzalez, C.Diaz, I. Leiva, and the park rangers Alfonso, Bernardo, Danilo, Guillermo, Manuel, Nino, Oscar, Jorge, andRamon. We thank also the Rojas and Lopez families on Juan Fernandez for hospitality and friendship. Thescientists M. Fernandez, R. Schlatter, H. Mattes, W. Beisenherz, and especially Thomas Schmitt werevaluable discussion partners.

References

Allen D (1984) Threatened ‘protected natural areas’ of the world. Environ Conserv 12:76–77Amico GC, Aizen MA (2005) Dispersion de semillas por aves en un bosque templado de Sudamerica

austral: >quien dispersa a quien? Ecol Austral 15:89–100

Biodivers Conserv

123

Page 15

Anderson C, Rozzi R (2000) Bird assemblages in the southernmost forests of the world: methodologicalvariations for determining species composition. Anal Inst Patagonia Ser Cien Nat 28:89–100

Araya B, Millie G, Bernal M (1998) Guıa de campo de las aves de Chile. Editorial Universitaria, Santiago deChile

Armesto JJ, Rozzi R, Miranda P, Sabag C (1987) Plant/frugivore interactions in South American temperateforests. Rev Chil Hist Nat 60:321–336

Audouin MM (1830) Notice sur l`Histoire naturelle de l`ıle Juan Fernandez. Ann des Siences Nat1(1830):58

Blondel J (2000) Evolution and ecology of birds on islands: trends and prospects. Vie et Milieu 50:205–220Blondel J, Chessel D, Frochot B (1988) Bird species impoverishment, niche expansion, and density inflation

in Mediterranean island habitats. Ecology 69:1899–1917Bourne WRP, Brooke MdeL, Clark GS, Stone T (1992) Wildlife conservation problems in the Juan

Fernandez Archipelago, Chile. Oryx 26:43–51Brooke MDL (1987) The birds of the Juan Fernandez Islands, Chile. ICBP Press, CambridgeBrooks SP, Gelman A (1998) General methods for monitoring convergence of iterative simulations.

J Comput Graph Stat 7:434–455Byers JE (2002) Impact of non-indigenous species on natives enhanced by anthropogenic alteration of

selection regimes. Oikos 97:449–458Carrascal LM, Palomino D, Seoane J, Alonso CL (2008) Habitat use and population density of the houbara

bustard Chlamydotis undulata in Fuerteventura (Canary Islands). Afr J Ecol 46:291–302Castro SA, Jaksic FM (2008) How general are global trends in biotic homogenization? Floristic tracking in

Chile, South America. Glob Ecol Biogeogr 17:524–531Catry P, Mellanby R, Suleiman KA, Salim KH, Hughes M, McKean M, Anderson N, Constant G, Heany V,

Martin G, Armitage M, Wilson M (2000) Habitat selection by terrestrial birds on Pemba Island(Tanzania), with particular reference to six endemic taxa. Biol Conserv 95:259–267

Chen J, Wang XM, Zhang SY (2008) Dietary shifts in relation to fruit availability among masked palmcivets (Paguma larvata) in central China. J Mammal 89:435–447

Clark JS (2005) Why environmental scientists are becoming Bayesians. Ecol Lett 8:2–14Colautti RI, MacIsaac HJ (2004) A neutral terminology to define invasive species. Divers Distrib

10:135–141Colwell RK (1989) Hummingbirds of the Juan Fernandez Islands: natural history, evolution and population

status. Ibis 131:548–566Crawley MJ (1986) The population biology of invaders. Philos Trans R Soc Lond B 314:711–729Cuevas J, van Leersum G (2001) Project ‘‘conservation, restoration, and development of the Juan Fernandez

islands, Chile’’. Rev Chil Hist Nat 74:899–910Diamond JM (1975) Assembly of species communities. In: Cody ML, Diamond JM (eds) Ecology and

evolution of communities. Belknap Press, Cambridge, pp 342–444Duncan RP, Blackburn TM, Sol D (2003) The ecology of bird introductions. Annu Rev Ecol Evol Syst

34:71–98Estades CF (1997) Bird-habitat relationships in a vegetational gradient in the Andes of central Chile. Condor

99:719–727Fordham DA, Brook BW (2010) Why tropical island endemics are acutely susceptible to global change.

Biodivers Conserv 19:329–342Fretwell SD, Lucas HL (1970) On territorial behavior and other factors influencing habitat distribution in

birds—I. Theoretical development. Acta Biotheor 19:16–36Greimler J, Lopez P, Stuessy TF, Dirnbock T (2002) The vegetation of Robinson Crusoe Island (Isla

Masatierra), Juan Fernandez Archipelago, Chile. Pac Sci 56:263–284Grigera D, Pavic C (2007) Ensambles de aves en un sitio quemado y en un sitio no alteradoen un area

forestal del noroeste de la Patagonia Argentina. Hornero 22:29–37Hagen E (2009) Castaways on Robinson Crusoe Island: influences of introduced species on an endemic

hummingbird. Dissertation, University of WashingtonHahn I (2006) Biogeographical isolation and bioacoustics: the Juan Fernandez Firecrown Sephanoides

fernandensis (Aves: Trochilidae) (King, 1831) of Robinson Crusoe Island, Chile. Bonner Zool Beitrage55:101–103

Hahn I, Romer U (2002) Threatened avifauna of the Juan Fernandez Archipelago, Chile: the impact ofintroduced mammals and conservation priorities. Cotinga 17:66–72

Hahn I, Romer U, Schlatter R (2005) Distribution, habitat use, and abundance patterns of land bird com-munities on the Juan Fernandez Islands, Chile. Ornitol Neotrop 16:371–385

Hahn I, Romer U, Schlatter R (2006) Population numbers and status of land birds of the Juan FernandezArchipelago, Chile. Senckenb Biol 86:109–125

Biodivers Conserv

123

Page 16

Hahn I, Romer U, Vergara P, Walter H (2009) Biodiversity and biogeography of the birds of the JuanFernandez Islands, Chile. Vertebr Zool 59:103–114

Hahn I, Vergara P, Romer U (2010) Forest structures of the Juan Fernandez Islands, Chile: significance forbird community composition and conservation. Geo-Oeko 31:1–10

Hahn I, Vergara PM, Romer U (2011) Importance of environmental nest attributes in the conservation ofJuan Fernandez endemic landbirds. Bird Conserv Int 21:119–135

Holt RD (1993) Ecology at the mesoscale: the influence of regional processes on local communities. In:Ricklefs R, Schluter D (eds) Species diversity in ecological communities. University of Chicago Press,Chicago, pp 77–88

IUCN (2011) 2010.4 IUCN Red List of threatened species, Gland. http://www.iucnredlist.org. Accessed 24May 2011

Jaramillo A (2003) Birds of Chile. University of Princeton Press, PrincetonJimenez JE (2000) Effect of sample size, plot size, and counting time on estimates of avian diversity and

abundance in a Chilean rainforest. J Field Ornithol 71:66–87Johnson AW (1965) The birds of Chile and adjacent regions of Argentina, Bolivia and Peru, vol 1. Platt

Establecimientos Graficos S A, Buenos AiresJohnson AW (1967) The birds of Chile and adjacent regions of Argentina, Bolivia and Peru, vol 2. Platt

Establecimientos Graficos S A, Buenos AiresJohnson TH, Stattersfield AJ (1990) A global review of island endemic birds. Ibis 132:167–180Kuschel G (1963) Composition and relationship of the terrestrial faunas of Easter, Juan Fernandez, Des-

venturadas and Galapagos Islands. Occas Pap Calif Acad Sci 44:79–95Lande R (1993) Risks of population extinction from demographic and environmental stochasticity and

random catastrophes. Am Nat 142:911–927Latta SC, Faaborg J (2002) Demographic and population responses of Cape May warblers wintering in

multiple habitats. Ecology 83:2502–2515Link WA, Sauer JR (2002) A hierarchical analysis of population change with application to cerulean

warblers. Ecology 83:2832–2840Lunn DJ, Thomas A, Best N, Spiegelhalter D (2000) WinBUGS—a Bayesian modelling framework: con-

cepts, structure, and extensibility. Stat Comput 10:325–337Magura T, Tothmeresz B, Elek Z (2003) Diversity and composition of carabids during a forestry cycle.

Biodivers Conserv 12:73–85Manly BFJ, McDonald LL, Thomas DL (1993) Resource selection by animals: statistical design and

analysis for field studies. Chapman and Hall, LondonMarvier M, Kareiva P, Neubert MG (2004) Habitat destruction, fragmentation, and disturbance promote

invasion by habitat generalists in a multispecies metapopulation. Risk Anal 24:869–878Medel RG, Jaksic FM (1988) Ecologıa de los Canidos sudamericanos: una revision. Rev Chil Hist Nat

61:67–79Mobæk R, Mysterud A, Loe LE, Holand Ø, Austrheim G (2009) Density dependence and temporal vari-

ability in habitat selection by a large herbivore: an experimental approach. Oikos 118:209–218Ntzoufras I (2009) Bayesian modeling using WinBUGS. Wiley, HobokenOwens IPF, Bennett PM (2000) Ecological basis of extinction risk in birds: habitat loss versus human

persecution and introduction predators. Proc Nat Acad Sci USA 97:12144–12148Palomino D, Carrascal LM (2005) Birds on novel island environments. A case study with the urban avifauna

of Tenerife (Canary Islands). Ecol Res 20:611–617Ricci M (2006) Conservation status and ex situ cultivation efforts of endemic flora of the Juan Fernandez

Archipelago. Biodivers Conserv 15:3111–3130Riley J (2003) Population sizes and the conservation status of endemic and restricted-range bird species on

Karakelang, Talaud Islands, Indonesia. Bird Conserv Int 13:59–74Rosenzweig ML (1985) Some theoretical aspects of habitat selection. In: Cody M (ed) Habitat selection in

birds. Academic Press, New York, pp 517–540Roy MS, Torres-Mura JC, Hertel F (1998) Evolution and history of hummingbirds (Aves: Trochilidae) from

the Juan Fernandez Islands, Chile. Ibis 140:265–273Roy MS, Torres-Mura JC, Hertel F, Lemus M, Sponer R (1999) Conservation of the Juan Fernandez

firecrown and its island habitat. Oryx 33:223–232Rozzi R, Martinez D, Willson MF, Sabag C (1996) Avifauna de los bosques templados de Sudamerica. In:

Armesto JJ, Villagran C, Arroyo MTK (eds) Ecologıa de los bosques nativos de Chile. EditorialUniversitaria, Santiago de Chile, pp 135–152

Sakai AK, Weller SG, Allendorf FW, Holt JS, Lodge DM, Molofsky J, With KA, Baughman S, Cabin RJ,Cohen JE, Ellstrand NC, McCauley DE, O’Neil P, Parker IM, Thompson JN (2001) The populationbiology of invasive species. Annu Rev Ecol Syst 32:305–320

Biodivers Conserv

123

Page 17

Skalski JR, Ryding KE, Millspaugh JJ (2005) Wildlife demography: analysis of sex, age, and count data.Elsevier Academic Press, Boston

Skottsberg C (1953) The vegetation of the Juan Fernandez Islands. In: Skottsberg C (ed) The Natural Historyof Juan Fernandez and Easter Islands, vol 2. Almquist Wiksells Boktryckeri, Uppsala, pp 793–960

Soh MCK, Sodhi NS, Lim SLH (2006) High sensitivity of montane bird communities to habitat disturbancein Peninsular Malaysia. Biol Conserv 129:149–166

Sol D, Santos DM, Feria E, Clavell J (1997) Habitat selection by the Monk Parakeet during colonization of anew area in Spain. Condor 99:39–46

Spiegelhalter DJ, Thomas A, Best NG, Lunn D (2003) WinBUGS, version 1.4. Imperial College and MRCBiostatistics Unit, London

Stattersfield AJ, Crosby MJ, Long AJ, Wege DC (1998) Endemic bird areas of the world: priorities forbiodiversity conservation. Burlington Press, Cambridge

Stiles FG (1987) Observaciones sobre la situacion actual de Picaflor Rojo de Juan Fernandez (Sephanoidesfernandensis), con recomendaciones para un estudio integral de su ecologia y biologia poblacional.FAO/PNUMA, Santiago de Chile

Stuessy TF, Ono M (1998) Evolution and speciation of islands plants. Cambridge University Press,Cambridge

Sturtz S, Ligges U, Gelman A (2005) R2WinBUGS: a package for running WinBUGS from R. J Stat Softw12:1–16

Thogmartin WE, Sauer JR, Knutson MG (2004) A hierarchical spatial model of avian abundance withapplication to cerulean warblers. Ecol Appl 14:1766–1779

Trainor CR (2007) Changes in bird species composition on a remote and well-forested Wallacean Island.Biol Conserv 140:373–385

Vergara PM (2007) Effects of nest box size on the nesting and renesting pattern of Aphrastura spinicaudaand Troglodytes aedon. Ecol Austral 17:133–141

Vergara PM, Armesto JJ (2009) Responses of Chilean forest birds to anthropogenic habitat fragmentationacross spatial scales. Landsc Ecol 24:25–38

Vergara PM, Simonetti JA (2004) Avian responses to fragmentation of the Maulino Forest in central Chile.Oryx 38:383–388

Vergara PM, Smith C, Delpiano CA, Orellana I, Gho D, Vazquez I (2010) Frugivory on Persea lingue intemperate Chilean forests: interactions between fruit availability and habitat fragmentation acrossmultiple spatial scales. Oecologia 164:981–991

Vergara PM, Pizarro J, Castro SA (2011) An island biogeography approach for understanding changes incompositional similarity. Ecol Model 222:1964–1971

Waltert MA, Mardiastuti A, Muhlenberg M (2004) Effects of land use on bird species richness in Sulawesi,Indonesia. Conserv Biol 18:1339–1346

Willson MF, de Santo TL, Sabag C, Armesto JJ (1994) Avian communities of fragmented south-temperaterainforests in Chile. Conserv Biol 8:508–520

Biodivers Conserv

123