Climate impacts on Mediterranean blue tit survival: an investigation across seasons and spatial...

15
Climate impacts on Mediterranean blue tit survival: an investigation across seasons and spatial scales VLADIMIR GROSBOIS *w , PIERRE-YVES HENRY * 1 , JACQUES BLONDEL *, PHILIPPE PERRET * , JEAN-DOMINIQUE LEBRETON *, DONALD W. THOMAS *z and MARCEL M. LAMBRECHTS * *Centre d’Ecologie Fonctionnelle et Evolutive, CNRS, 1919 Route de Mende, 34293, Montpellier cedex 5, France, wDepartment of Biology, Campus Drie Eiken, University of Antwerp, B-2610 Antwerp, Belgium, zGroupe de Recherche en Ecologie, Nutrition et Energe ´tique, Centre de Recherche en Biologie Forestie `re, Universite ´ de Sherbrooke, Sherbrooke, Que ´bec JIK 2R1, Canada Abstract In some hole nesting passerine species, long-term monitoring data are available for several geographically independent populations. Climate forcing can then be documen- ted and predictions made on the scale of distribution ranges. Several demographic studies of Paridae report dramatic impacts of wintertime climatic factors. However, these studies were undertaken in populations located in the northern parts of the species’ ranges. Studies on the survival of Paridae in their southern ranges are necessary in order to assess potential latitudinal variation in climate forcing on survival. Based on monitoring of individual adult blue tits (Parus caeruleus), the effects of climatic factors on annual survival were assessed in three distinct Mediterranean populations. In these regions, climatic conditions in early summer might be expected to have a strong impact because they can be extremely hot and dry and because at this time of year Paridae are subjected to intrinsic constraints that stem from energetically costly postbreeding moult, recovery from reproductive costs, and from population densities inflated by the new cohort of fledglings. The impact of climatic conditions in early summer was, thus, addressed in addition to that prevailing in winter. In order to consider a large number of local climatic variables while limiting statistical power loss, integrative indices of local climate were built using multivariate techniques. In addition, the NAO and three large-scale factors that are closely linked with atmospheric and oceanic circulation in the intertropical zone were considered as potentially influential factors in winter and early summer. Relationships between blue tit survival and indices of local temperature and precipitation in winter and in early summer were detected. Adult survival also correlated with a large-scale tropical index in early summer: rainfall in the Sahel. This is one of the first quantitative indications that fluctuations in summer climatic conditions explain a significant part of the temporal variation in adult survival in unconnected populations of a sedentary European vertebrate. Furthermore, the results support the hypothesis that summertime local climates in Western Europe are closely linked with atmospheric and oceanic circulation in the intertropical zone. Keywords: Mediterranean basin, Parus caeruleus, Sahel rainfall, summer climate, survival Received 3 February 2006; revised version received 17 July 2006 and accepted 1 August 2006 Introduction Climate change over the past century has had important ecological consequences (Parmesan & Yohe, 2003), but predictions concerning the impact of future climate change on biodiversity remain subject to large uncer- tainties (Thomas et al., 2004). A necessary, although not sufficient, step for investigating ecological conse- Correspondence: Vladimir Grosbois, Centre d’Ecologie Fonctionnelle et Evolutive, CNRS, 1919 Route de Mende, 34293, Montpellier Cedex 5, France, e-mail: [email protected] 1 Present address: UMR 5173, Departement d’Ecologie et Gestion de la Biodiversite ´, Muse ´um National d’Histoire Naturelle, 55 rue Buffon, 75005 Paris, France. Global Change Biology (2006) 12, 2235–2249, doi: 10.1111/j.1365-2486.2006.01286.x r 2006 The Authors Journal compilation r 2006 Blackwell Publishing Ltd 2235

Transcript of Climate impacts on Mediterranean blue tit survival: an investigation across seasons and spatial...

Climate impacts on Mediterranean blue tit survival:an investigation across seasons and spatial scales

V L A D I M I R G R O S B O I S *w , P I E R R E - Y V E S H E N R Y *1 , J A C Q U E S B L O N D E L *,

P H I L I P P E P E R R E T *, J E A N - D O M I N I Q U E L E B R E T O N *, D O N A L D W. T H O M A S *zand M A R C E L M . L A M B R E C H T S *

*Centre d’Ecologie Fonctionnelle et Evolutive, CNRS, 1919 Route de Mende, 34293, Montpellier cedex 5, France,

wDepartment of Biology, Campus Drie Eiken, University of Antwerp, B-2610 Antwerp, Belgium, zGroupe de Recherche en Ecologie,

Nutrition et Energetique, Centre de Recherche en Biologie Forestiere, Universite de Sherbrooke, Sherbrooke, Quebec JIK 2R1, Canada

Abstract

In some hole nesting passerine species, long-term monitoring data are available for

several geographically independent populations. Climate forcing can then be documen-

ted and predictions made on the scale of distribution ranges. Several demographic

studies of Paridae report dramatic impacts of wintertime climatic factors. However, these

studies were undertaken in populations located in the northern parts of the species’

ranges. Studies on the survival of Paridae in their southern ranges are necessary in order

to assess potential latitudinal variation in climate forcing on survival. Based on

monitoring of individual adult blue tits (Parus caeruleus), the effects of climatic factors

on annual survival were assessed in three distinct Mediterranean populations. In these

regions, climatic conditions in early summer might be expected to have a strong impact

because they can be extremely hot and dry and because at this time of year Paridae are

subjected to intrinsic constraints that stem from energetically costly postbreeding moult,

recovery from reproductive costs, and from population densities inflated by the new

cohort of fledglings. The impact of climatic conditions in early summer was, thus,

addressed in addition to that prevailing in winter. In order to consider a large number

of local climatic variables while limiting statistical power loss, integrative indices of

local climate were built using multivariate techniques. In addition, the NAO and three

large-scale factors that are closely linked with atmospheric and oceanic circulation in the

intertropical zone were considered as potentially influential factors in winter and early

summer. Relationships between blue tit survival and indices of local temperature and

precipitation in winter and in early summer were detected. Adult survival also correlated

with a large-scale tropical index in early summer: rainfall in the Sahel. This is one of the

first quantitative indications that fluctuations in summer climatic conditions explain a

significant part of the temporal variation in adult survival in unconnected populations of

a sedentary European vertebrate. Furthermore, the results support the hypothesis that

summertime local climates in Western Europe are closely linked with atmospheric and

oceanic circulation in the intertropical zone.

Keywords: Mediterranean basin, Parus caeruleus, Sahel rainfall, summer climate, survival

Received 3 February 2006; revised version received 17 July 2006 and accepted 1 August 2006

Introduction

Climate change over the past century has had important

ecological consequences (Parmesan & Yohe, 2003), but

predictions concerning the impact of future climate

change on biodiversity remain subject to large uncer-

tainties (Thomas et al., 2004). A necessary, although not

sufficient, step for investigating ecological conse-

Correspondence: Vladimir Grosbois, Centre d’Ecologie

Fonctionnelle et Evolutive, CNRS, 1919 Route de Mende, 34293,

Montpellier Cedex 5, France,

e-mail: [email protected]

1Present address: UMR 5173, Departement d’Ecologie et Gestion

de la Biodiversite, Museum National d’Histoire Naturelle, 55 rue

Buffon, 75005 Paris, France.

Global Change Biology (2006) 12, 2235–2249, doi: 10.1111/j.1365-2486.2006.01286.x

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd 2235

quences of climate change is to study the impact of past

climatic variation from direct or indirect population

dynamics indices. For marine ecosystems, such indices

are sometimes available for extremely large time scales,

including millenia (Roy et al., 1996; Graham et al., 2003)

and centuries (Montevecchi & Myers, 1997; Finney et al.,

2002). To our knowledge, such lengthy indirect indices

are not available for land birds. However, direct mon-

itoring data collected over several decades have re-

vealed climatic influences on the demography and

dynamics of land bird populations (e.g. Clobert et al.,

1988; Peach et al., 1994; Perdeck et al., 2000).

Species for which such data are available for several

geographically independent populations are of particu-

lar interest because climate forcing can be documented

and predictions made on the scale of distribution ranges

(e.g. Forchhammer et al., 1998; Parmesan & Yohe, 2003).

Several hole-nesting passerine birds such as tits and

flycatchers have been subjected to long-term monitor-

ing programmes in several populations scattered over

their distribution ranges, providing the opportunity to

investigate climate impacts at large geographic scales.

This approach has revealed, for example, latitudinal

variation in the influence of winter and spring climatic

conditions on the phenology of reproduction (Visser

et al., 2002; Sanz et al., 2003; Both et al., 2004) and on

population growth rates (Sæther et al., 2003) of great tits

Parus major, blue tits Parus caeruleus and pied flycatchers

Ficedula hypoleuca.

Several studies have reported on the influence of

climatic conditions on adult survival in populations of

hole-nesting Paridae in the northern part of the species’

ranges (great tit; Clobert et al., 1988; Perdeck et al., 2000,

willow tit, P. montanus; Lahti et al., 1998, tufted titmouse,

P. bicolor Doherty & Grubb, 2002). In all these studies,

correlations between adult survival and winter and/or

early spring climatic conditions have been detected.

Furthermore, the authors of some of these studies

(Nilsson, 1987; Clobert et al., 1988; Lahti et al., 1998)

addressed influences of climatic conditions at other

times of year but failed to detect any. It could, conse-

quently, be concluded that survival of resident Paridae

is particularly sensitive to the combination of low

temperatures and food shortage in winter (Newton,

1998). However, to our knowledge, the relationships

between climate and adult survival of these Paridae

have never been addressed in populations located in

the southern part of their ranges. In these regions,

climatic conditions in summer might be expected to

have a strong impact because they can be extremely hot

and dry (Blondel & Aronson, 1999a) and because at this

time of year Paridae are subjected to important post-

breeding intrinsic constraints (Sanz, 1999; Dhondt,

2001). Studies integrating the impacts of winter and

summer climatic factors on the survival of Paridae in

their southern ranges are necessary in order to assess

potential latitudinal variation in climate forcing on

survival.

Investigating the impact of local climate on demogra-

phy is complicated by the fact that local climate is

comprised of many components. Furthermore, the me-

chanisms through which climate influences the survival

of hole-nesting passerines are still poorly known. This

precludes the definition a priori of a restricted set of

potentially influential climate variables. This situation

has sometimes led to the selection of a large set of

climatic covariates (Newton et al., 1992; Peach et al.,

1994; Franklin et al., 2000). However, without correction

for multiple tests, such a strategy risks detecting spur-

ious relationships (Garcia, 2004). Furthermore, different

components of local climate such as rainfall and tem-

peratures are not independent (Graham, 2003). One

way to overcome these issues is to build indices that

combine several correlated variables (Moss et al., 2001).

Alternatively, one can use large-scale integrative cli-

matic patterns, such as the North Atlantic oscillation

(NAO) or El Nino southern oscillation, as proxies for

local climate (Stenseth et al., 2003). Many studies have

used only these large scale climate proxies in studies of

demographic variation (Sanz, 1999, 2002; Sillet et al.,

2000; Sæther et al., 2003), reviewed in Stenseth et al.

(2002), a shortcoming of which is that the link between

the surrogate variable and the local climate factors that

actually influence vital rates can be weak. Ideally,

one should consider simultaneously the effects of

both large-scale and local integrative climate factors

(Almaraz & Amat, 2004) because these two types oper-

ate at different levels of the chain of causation (Stenseth

et al., 2003).

Here, the influence of climate on adult survival of

blue tits near the southern limit of the species range, in

the Mediterranean basin, is addressed. As monitoring

data were available in three distinct populations, it was

possible to investigate whether these populations were

similarly influenced by the same climatic variables. This

comparison is all the more interesting because despite

their close proximity, these populations differ greatly in

terms of demography and habitat (Tremblay et al.,

2003). Furthermore, two of these populations are

located on the island of Corsica where they are specia-

lized to their local habitat (Blondel et al., 1999b; Thomas

et al., 2001). The influence of winter and summer climate

variables were examined. Synthetic indices of local

climate derived using multivariate statistics (Draper &

Smith, 1981; Graham, 2003), and large-scale climate

factors were considered. The NAO is thought to control

local winter climates in Europe (Hurrell et al., 2001).

Winter NAO was, thus, selected as a factor potentially

2236 V. G R O S B O I S et al.

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2235–2249

influencing adult survival. The large-scale climatic in-

fluencing on European summers are still poorly docu-

mented, but recent work implicates tropical oceanic and

atmospheric circulation phenomena (Hurrell et al., 2002;

Cassou et al., 2004). These tropical influences include

rainfall in the Sahel (Folland et al., 1986; Raicich et al.,

2003), India monsoon (Raicich et al., 2003), and the

southern oscillation index (Allan et al., 1996; Holmgren

et al., 2001). Thus, the impact of these large-scale tropi-

cal climate factors was also addressed in this attempt to

explain adult survival variation.

Methods

Capture-mark-recapture data

Adult blue tits were captured and ringed annually at

nest boxes during the nestling feeding period (March–

June). Three sedentary populations were studied in

the Mediterranean region, two from Corsica (Pirio

and Muro) and one from mainland southern France

(La Rouviere). The populations of Muro and La Rouviere

occupied broad-leaved deciduous oak (Quercus humilis)

habitat while the population of Pirio was settled in

evergreen oak (Q. ilex) habitat. Because stone walls

offered many alternative breeding sites at Muro, breed-

ing pairs’ density there was at least twofold higher than

at the two other locations. Individual monitoring data

were available over distinct but overlapping time per-

iods for the three populations: 1985–2000 (n 5 692 in-

dividual capture histories) at Pirio, 1993–2000 (n 5 253)

at Muro, and 1991–2000 (n 5 501) at La Rouviere. Sex,

status (resident vs. immigrant) and age (1-year old

vs.41-year old) of ringed individuals were determined

using standard criteria, well established since 1976

(Blondel et al., 1993, 1999b).

Capture histories were analysed using specific proce-

dures designed to provide robust estimates of survival

rates (hereafter referred to as F), while accounting for

potential biases due to variation in recapture probabil-

ities (hereafter referred to as p; Lebreton et al., 1992). A

logit link function was used in order to constrain the

estimates of F and p between 0 and 1 (Lebreton et al.,

1992). Goodness of fit tests were conducted using

U-CARE (Choquet et al., 2001). A preliminary analysis

(not described here) was conducted to address the effects

of the factors Sex, Status, Age, Population (hereafter

referred to as Pop) and Year on survival and recapture

probabilities. This preliminary analysis allowed defin-

ing a departure model that captured the main factors of

variation. The departure model included for both sur-

vival and recapture probabilities the effects of Pop, Year,

and the interaction between these two factors; and, for

recapture probability an additional factor Sex. It fitted

the data adequately (omnibus test: P 5 0.94). In this

departure model [denoted F(Pop 1 Year 1 Pop �Year);

p(Sex 1 Pop 1 Year 1 Pop �Year)] recapture probability

was higher for females than for males. All subsequent

models considered were nested in this starting model

and differed from it by the survival terms only. All

capture-mark-recapture analyses were performed with

the statistical package MARK 3.0 (White & Burnham,

1999).

Before testing the covariates that might underlie

temporal variation in adult survival, the pattern of

temporal and spatial variation in survival was explored

using the subset of models nested in the starting model,

and using Akaike’s Information Criterion corrected for

data sparseness (hereafter AICc) for model selection

(Burnham & Anderson, 1998). Models with low AICc

were considered as achieving a good compromise be-

tween parsimony (number of parameters in the model)

and fit to the data (deviance). A difference of two AICc

points was considered to be significant. When the

difference was less than two AICc points, the model

with the lowest number of parameters was preferred

(Lebreton et al., 1992). After identifying the spatio-

temporal pattern of variation in survival, we assessed

which climatic covariates could explain it.

Local climate indices

Monthly records of rainfall in millimetres (RF), number

of days with more than 0.1 mm rainfall (DR), number of

days with more than 1 mm rainfall (DHR), averages of

the daily minimum (MinT), maximum (MaxT), and

mean temperature (AvT) in 1C, and daily records of

the speed (S in ms�1) and direction (y1; 01: north, 901:

east, 1801: south, 2701: west) of the strongest wind gust,

were provided by Meteo France for the meteorological

stations of Calvi in Corsica, and Montpellier in main-

land southern France over the period 1985–2000. Wind

variables were used to build for each of the two me-

teorological stations four daily wind strength measures

(one per cardinal point, denoted NW, SW, EW, WW).

NW ¼S cosðyÞ if cosðyÞ � 0;

0 if cosðyÞ < 0:

�; SW ¼

�S cosðyÞ if cosðyÞ � 0;

0 if cosðyÞ > 0:

EW ¼S sinðyÞ if sinðyÞ � 0;

0 if sinðyÞ < 0:

�; WW ¼

�S sinðyÞ if sinðyÞ � 0;

0 if sinðyÞ > 0:

For each year from 1985 to 2000 and each meteorologi-

cal station, the monthly precipitation and temperature

variables and the daily wind variables were averaged

over a summertime and a wintertime period that were

considered as potentially critical. These two periods

were defined according to prior climatological and

biological knowledge. In the mediterranean regions

where the study populations were located, the warmest

M E D I T E R R A N E A N B L U E T I T S U R V I VA L 2237

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2235–2249

and driest period of year extends from June to August

(see http://www.worldclimate.com/for a description of

climate in Montpellier and Ajaccio, Corsica). In addi-

tion, the June–July period immediatly follows the chick

rearing period in the study populations (Thomas et al.,

2001). During this post breeding period blue tits

are subjected to intrinsic constraints that stem from

population densities inflated by the new cohort

of fledglings, energetically costly postbreeding moult

(Siikamaki et al., 1994; Sanz, 1999), and recovery from

reproductive costs (Nur, 1984; Nilsson & Svensson,

1996; Dhondt, 2001). In the light of these climatological

and biological knowledge, the June–July period was

considered as potentially constraining for the survival

of blue tits in the study populations. The potentially

critical wintertime period chosen was longer and was

defined in the light of climatological information only

because, to our knowledge, there is no relatively short

period in winter when particular intrinsic constraints

exert on blue tits. The four coldest months of year

(December through March; see http://www.worldcli-

mate.com/) were, thus, selected. Principal components

analysis (PCA) was used to characterize local climatic

conditions in winter and early summer in the form of

integrative indices of minimal dimensionality (Draper

& Smith, 1981; Graham, 2003). Four PCA were under-

taken (early summer and winter for each of Montpellier

and Calvi) using procedure PRINCOMP of SAS 8.01

(SAS Institute Inc., Cary NC, USA).

Large scale climate indices

All links to large-scale climatic variables were found on

the Climate Diagnostics Centre web-site (http://

www.cdc.noaa.gov/ClimateIndices/). The selected vari-

ables were December through March and monthly NAO,

(http://www.cgd.ucar.edu/� jhurrell/nao.html#seasonal;

referred to as NAO), monthly Indian Monsoon (http://

www.cdc.noaa.gov/Correlation/indiamon.data; referred

to as INDM), monthly standardized Sahel rainfall

(http://jisao.washington.edu/data/sahel/; referred to as

SRF), and monthly Multivariate ENSO Index (http://

www.cpc.noaa.gov/data/indexs/soi; referred to as

MEI). Early summer indices for these variables were

computed as the sum of the June and July indices.

Correlation between blue tit survival and climate indices

The relationships between adult survival rates and the

10 candidate climate indices (Table 1) were explored in

Table 1 Climate Indices

Name Label Definition

Winter (December–March); large scale

North Atlantic oscillation W-NAO �: low atmospheric pressure differential between the high-pressure

centre near the Azores and the low-pressure centre near Iceland

1: high pressure differential

Winter (December–March); local scale

Precipitation and temperature

maxima

W-P-Tmax �: wet; low daily temperature maxima

1: dry with high daily temperature maxima

Wind and temperature

minima

W-W-Tmin �: northern (northeastern in Calvi) wind; low daily temperature

minima

1: southern wind; high daily temperature minima

Early summer (June-July); large scale

Standardized Sahel rainfall SRF �: low rainfall in the Sahel region

1: high rainfall in the Sahel region

Indian Monsoon INDM �: low rainfall in the core Indian Monsoon region

1: high rainfall in the core Indian Monsoon region

Multivariate El Nino

southern oscillation

MEI �: cold ENSO phase (La Nina)

1: warm ENSO phase (El Nino)

North Atlantic oscillation S-NAO �: low atmospheric pressure differential between the high-pressure

centre near the Azores and the low-pressure centre near Iceland

1: high pressure differential

Early summer (June–July); local scale

Precipitation and temperature S-P-T �: wet and cold

1: dry and warm

Wind S-W �: northern (northeastern for Calvi) wind

1 : southern wind

2238 V. G R O S B O I S et al.

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2235–2249

three steps: (1) reducing the number of candidate cli-

mate indices, (2) building and comparing models that

account simultaneously for the effects of several climate

indices and for different survival responses to these

indices among populations, and (3) producing survival

rate estimates based on the best supported models. In

the first step of the analysis, only models including the

effect of one candidate climatic index at a time and in

which this effect was similar across populations were

considered. Two types of relationship were examined

for each climatic index, linear and quadratic on the logit

scale, and significance levels for these correlations were

assessed using ANODEV (Skalski, 1996). ANODEV tests rely

on three models. In the general model, temporal varia-

tion is accounted for by a Year factor where one para-

meter is used for each year of the study. The reduced

model is built under the hypothesis of no temporal

variation. In the covariate model, temporal variation is

accounted for by the relationship with the focal climate

index. ANODEV determines whether the focal climate

index accounts for a significant fraction of the difference

in deviance between the reduced and the general

model. In the covariate models containing effects of

large-scale climate indices, the pattern of interannual

variation was constrained to parallelism on a logit scale

among populations. The general model used for

ANODEV was thus the model F (Pop 1 Year) that gener-

ates similar constraint. In the covariate models contain-

ing effects of the local climate indices the pattern of

interannual variation in the continental population dif-

fered from that in the the two Corsican populations.

This dissimilarity was accounted for in the general

model used for ANODEV by adding to this general model

an interaction between the factor Year and a two-level

factor distinguishing the continental population from

the two Corsican populations. Given that the effect

of 10 climate indices was tested, the threshold P-value

should have been set at 0.05/10 5 0.005 (Bonferroni

correction for multiple tests; Sokal & Rohlf, 1995).

Application of such a correction would guarantee that

the probability of detecting spurious climate indices

effects remains low (i.e. low type-I error). However,

given the short time series available for survival rates

(at best 15 years), it would also result in high probability

of not detecting genuine effects of climate indices (i.e.

high type-II error). In order to evaluate the likelihood of

not detecting genuine climate effects using a threshold

P-value of 5%, the minimum value of a deviance

equivalent of the coefficient of determination (R2) for

the effect of a climate index to be considered as sig-

nificant at the 5% level was computed. First the value of

the deviance of a covariate model leading to a value of

P-ANODEV equal to 5% was determined. The corre-

sponding R2 was then computed as the ratio of the

difference in deviance between the reduced and the

general model to the difference in deviance between

the reduced and the covariate model (Skalski, 1996).

This minimum R2 varied between 35% and 62%

depending on the scale at which the climate index

operates (because, as mentioned above, the total

amount of deviance that had to be considered for the

computation of R2 differs between local and large scale

climate indices) and on the shape of the relationship

with survival (because the number of parameters re-

quired to model the effect of the climate index differs

between linear and quadratic relationships). It seems

reasonable to hypothesize that variation in survival

probability in natural populations of blue tits arise from

the influence of multiple endogenous and exogenous

factors and that a single climate index is unlikely to

account for more than 35% of this variation. We, thus,

decided to rely on the usual 5% threshold P-value. In

the second step, models including more than one cli-

matic index and interactions between climate indices

and the factor Pop were considered. Only the climate

indices selected at the first step were considered in this

second step. For each of the models considered at this

step, the deviance equivalent of the coefficient of

determination (R2) was computed. Finally, estimates

from the lowest AICc model and from the two

models differing from it by less than two AICc points

were combined, using model averaging (Burnham &

Anderson, 1998), to obtain survival rate estimates

and associated unconditional standard errors that

accounted for model selection uncertainty.

Results

Local climate indices

The PCA analysis produced almost identical eigenvec-

tors for Montpellier and Calvi, implying that similar

types of weather were prevailing in these two locations.

The correlation circles (Fig. 1) indicate for each season

and each location the correlations between the original

local climate variables and the two first PCA compo-

nents. For winter, the PCA first components for both

Calvi and Montpellier distinguish years that are wet

with low daily temperature maxima vs. dry with high

daily temperature maxima; hence, our reference to it as

the winter precipitation and temperature maxima index

(W-P-Tmax; Table 1). The second component provided a

northern (northeastern for Calvi) wind and low daily

temperature minima vs. southern wind and high daily

temperature minima index, subsequently referred to as

the winter wind and temperature minima index (W-W-

Tmin; Table 1). For summer, the PCA first component

provided a wet and cold vs. dry and warm index,

M E D I T E R R A N E A N B L U E T I T S U R V I VA L 2239

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2235–2249

subsequently referred to as the summer precipitation

and temperature index (S-P-T; Table 1). The second

component provided a northern (northeastern for

Calvi) wind vs. southern wind index, referred to

subsequently as the summer wind index (S-W; Table 1).

Correlation among climate indices

Correlations among climate indices for early summer

and winter are listed in Table 2. In early summer, the

S-P-T index for Montpellier was positively correlated to

that for Calvi (Table 2, Fig. 2). Hot and dry or cold and

wet June–July conditions thus tended to occur in the

same years on mainland and in Corsica. The correlation

between Montpellier and Calvi for the S-W index was

lower, indicating different wind regimes on mainland

and Corsica (Table 2, Fig. 2). Sahel rainfall correlated

positively with the S-W index for Calvi but not for

Montpellier (Table 2, Fig. 2). Finally a positive correla-

tion was detected between Indian Monsoon and the

NAO (Table 2, Fig. 2). In winter, the W-P-Tmax index for

Montpellier was positively correlated with that for

Calvi (Table 2, Fig 3). Wet or dry winter conditions,

thus, tended to occur in the same years on mainland

and in Corsica. The correlation between Montpellier

and Calvi for the W-W-Tmin index was low, indicating

WW

WW

WW

WW

DHR

DHR

DHR

DHR

RF

RF

RF

RF

MaxT

MaxT

MaxT

MaxT

NW

NWNW

NW

DR

DR

DR

DR

EW

EW

EW

EW

SW

SW SW

SW

AvT

AvT

AvT

AvT

MinT

MinT

MinT

MinT

DryHigh temperature maxima

WetLow temperature maxima

Low

tem

pera

ture

min

ima

Hig

h te

mpe

ratu

re m

inim

a

DryWarm

WetCold

Stro

ng N

orth

erlie

s

Stro

ng N

orth

erlie

s

Stro

ng S

outh

erlie

s

Stro

ng S

outh

erlie

s

Dec

embe

r–M

arch

W-W

-Tm

inJu

ne–

July

S-W

Montpellier (Mainland) Calvi (Corsica)

S-P-T

W-P-Tmax

(a)

(b)

Fig. 1 Correlation circles derived from principal components analysis (PCA) analysis of local climatic variables in (a) winter and

(b) early summer for mainland and Corsica. The x- and y-axes are the two first PCA components ranging from�1 to 1 1. See the text for

the definition of the local weather variables represented on the correlation circles.

2240 V. G R O S B O I S et al.

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2235–2249

different wind regimes on mainland and Corsica (Table

2, Fig. 3). As expected, the NAO correlated positively

and strongly with the W-P-Tmax indices for Montpel-

lier and Calvi (Table 2, Fig. 3). A weaker negative

correlation was detected between NAO and W-W-Tmin

index for Montpellier only (Table 2, Fig. 3).

General description of the spatial and temporal variationin blue tit adult survival

The model including additive Year and Pop effects

explained much better the spatio-temporal variations

in adult survival than models including either only a

Pop or only a Year effect (Table 3). Therefore, adult

survival differed considerably both among years and

among populations (Fig. 4). However, the model in-

cluding just the additive effects of Year and Pop

achieved a much better compromise between parsi-

mony and fit to the data than did the model that

included in addition the interaction between Year and

Pop (Table 3). This indicated that the pattern of inter-

annual variation in adult survival was similar among

populations (Fig. 4). These results indicate that adult

survival in these blue tit populations was influenced by

environmental factors, such as climate, that operate at a

relatively large spatial scale.

Effects of climate indices on blue tit adult survival

Based on the statistical significance of the global effects

on blue tit survival in the three study populations of

each candidate climate index (i.e. the statistical signifi-

cance of the climate index in models where it and the

factor Pop had additive effects), the set of climate effects

was narrowed to the logit-linear effects of the W-P-

Tmax and Sahel rainfall indices (P-ANODEV 5 0.005 and

0.015, respectively; Table 4) and the logit-quadratic

effect of the S-P-T index (P-ANODEV 5 0.025; Table 4).

In the second step of the model selection procedure, all

possible models including one, two, or all of these

climate indices’ effects, as well as all possible two-way

interactions between these and the factor Pop were built

(Table 5). The three models with the lowest AICc,

among the 26 possible models, described the data

equally well (i.e. their AICc differed by less than two

points). Two of these models included the three climate

indices’ effects. One of these two models included in

addition the interaction between Pop and W-P-Tmax

and the other the interaction between Pop and Sahel

rainfall (Table 5). The third model included the effects of

S-P-T, that of Sahel rainfall, and the interaction between

Pop and Sahel rainfall (Table 5). The fraction of the

interannual variation in blue tit adult survival in the

three study populations accounted for by each of these

three models ranged from 66% to 73% (Table 5). As

considered globally, the slope estimates obtained from

the three selected models indicated that blue tits in the

northwestern part of the Mediterranean basin survived

poorly in years characterized by extreme local climatic

conditions in early summer (either cold and wet or hot

and dry; Table 6, Fig. 5c) and/or by a deficit of rainfall

in the Sahel in early summer (Table 6, Fig. 5b) and/or by

Table 2 Pearson correlation coefficients among climate indices

Corsica S-W Mainland S-P-T Mainland S-W SRF MEI INDM S-NAO

(a) Early summer (June–July) climate indicesCorsica S-P-T 0 0.51** �0.05 0.26 �0.35 0.14 �0.11

Corsica S-W �0.31 0.41 0.49* �0.02 �0.02 0.01

Mainland S-P-T 0 0.04 �0.40 0.39 0.14

Mainland S-W 0.22 �0.07 0.06 �0.07

SRF 0.09 0.24 �0.28

MEI �0.16 �0.31

INDM 0.53**

Corsica W-W-Tmin Mainland W-P-Tmax Mainland W-W-Tmin W-NAO

(b) Winter (December–March) climate indicesCorsica W-P-Tmax 0 0.68*** �0.04 0.65***

Corsica -W-W-Tmin 0.54** 0.33 0.19

Mainland W-P-Tmax 0 0.72***

Mainland W-W-Tmin �0.51**

*Po0.10, **Po0.05, ***Po0.01.

M E D I T E R R A N E A N B L U E T I T S U R V I VA L 2241

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2235–2249

high precipitations and low daily temperature maxima

in winter (Table 6, Fig. 5a). However, the slope estimates

obtained from the two selected models including inter-

actions between climatic indices and Pop reveal notice-

able variation among populations in the impact of

climatic conditions. The effect of winter climatic condi-

tions was strong in the Corsican population of Muro,

moderate in the Corsican population of Pirio and absent

in the mainland population of La Rouviere (Table 6, Fig.

5a). Moreover, the effect of Sahel rainfall was more

pronounced in the Corsican population of Muro than

in the two other populations (Table 6, Fig. 5b). The

descriptions provided by the three selected models of

the relationships between the variation in climatic fac-

tors and variation in survival rates in the three study

populations generated (using model averaging) survi-

val rate estimates that matched closely those obtained

from a general model where one parameter was used

per year and per population (Fig. 6).

−2.5−2

−1

00.51

1.5

0.5

1

1.5

22.53

−3

−2

−1

0

1

2

3

−3

−2

−1

0

1

2

3

−3−4−5

−2−101234

0

0.5

1

1.5

2

1986

1988

1991

P+

P−

S-W

S-P

-TNW

SW

Sah

el r

ain

fall

ME

I

−1.5

−0.5

−2

−1

0

−1.5

−0.5

−1.5

−0.5

−1

1985

1987

1989

1990

1992

1993

1994

1995

1996

1997

1998

1999

1986

1988

1991

1985

1987

1989

1990

1992

1993

1994

1995

1996

1997

1998

1999

1986

1988

1991

1985

1987

1989

1990

1992

1993

1994

1995

1996

1997

1998

1999

Ind

ian

mo

nso

on

S-N

AO

Year (June Y through July Y)

Fig. 2 Time series of local and large-scale climate indices for

early summer. Dashed lines represent mainland, and solid lines

Corsican indices.

NW

SW

NA

O

Year(December Y-1 through March Y)1986

1988

1991

1987

1989

1990

1992

1993

1994

1995

1996

1997

1998

1999

2000

−3−4−5

−2−10123

−3−4−5

−2−10123

45

−3−4−5

−2−1012345

6

WW

-Tm

inW

P-T

max

Tmax

+Tmax

−Tmin

−Tmin+

P+

P−

Fig. 3 Time series of the local and large-scale climate indices for

winter. Dashed lines represent local mainland, and solid lines

Corsican indices.

Table 3 General models for describing temporal and spatial

variation in blue tit adult survival

Model

Number of

parameters Deviance AICc

Pop 1 Year 48 1305.4 4217.5

Pop 35 1343.5 4228.7

Pop 1 Year 1 Pop �Year 60 1295.6 4232.7

Year 46 1331.6 4239.6

Constant 33 1374.9 4256.0

2242 V. G R O S B O I S et al.

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2235–2249

Discussion

This study explored the relationship of various climate

indices to regional blue tit adult survival through the

examination of correlations between 28 adult survival

estimates from three distinct populations and 10 climate

indices considered as potentially influential. The ratio of

number of statistical units per relationship examined

was, thus, low. However, in the absence of realistic

assumptions on the mechanisms through which climate

influences the survival of adult blue tits, it was not

possible to narrow the set of candidate climate indices.

We established that in this context, the application of

proper corrections for multiple tests to threshold P-

values would have resulted in low probability of detect-

ing genuine effects of climate indices. As advised by

Roback & Askins (2005) for studies relying on a limited

amount of data, we consequently chose to retain the

effects of climate indices showing significant correla-

tions with blue tit adult survival at the traditional 5%

threshold P-value. Admittedly, this strategy exposed us

to a risk of retaining spurious relationships (Garcia,

2004). Using the Binomial distribution (Garcia, 2004),

we quantified this risk by computing the probability of

detecting at least one, at least two, and at least three

significant relationships arising by chance only when

the effects of 10 covariates are tested at the 5% level.

This risk took values of 40%, 9%, and 1% respectively. It

is, thus, likely that one of the three relationships de-

tected here arose by chance only. However, more than

one of or all of these three relationships arising by

chance only were unlikely events. We, thus, believe that

the results obtained here essentially reflect genuine

climate effects and that a highly plausible picture

emerges from the modelling work we undertook: early

summer local climate and large-scale climate correlated

with survival in three blue tit populations from the

northwestern part of the Mediterranean basin, while

winter local climate did so only in the two Corsican

populations. In addition, survival in the Corsican po-

pulation of Muro, which was characterized by the

highest densities, was more sensitive to climatic fluc-

tuations than in the two other populations. We believe

that this preliminary conclusion allows generating ori-

ginal and interesting hypotheses on the geographic

variation of the impact of climate on the demography

of passerine birds that we develop.

Critical periods of year

Evidence for the impact of climate on the mortality of

resident adult passerine birds is to date mainly limited to

the influence of winter and early spring climatic condi-

tions (but see Robinson et al., 2004; see Newton, 1998 for

a review). Correlations between adult survival and win-

ter climatic conditions have been detected in several bird

species, (e.g. great tit in the UK; Clobert et al., 1988 and in

the Netherlands; Perdeck et al., 2000, nuthatch, Sitta

europaea in Sweden; Nilsson, 1987, tufted titmouse in

Ohio, USA; Doherty & Grubb, 2002, and willow tit in

Finland; Lahti et al., 1998). Furthermore, in three of the

above-mentioned studies either several recapture ses-

Year

Blu

e ti

t ad

ult

su

rviv

al

00.10.20.30.4

0.50.60.70.8

0.9

1998–1999

1997–1998

1996–1997

1995–1996

1994–1995

1993–1994

1992–1993

1991–1992

1990–1991

1989–1990

1988–1989

1987–1988

1986–1987

1985–1986

Fig. 4 Time series of blue tit adult survival estimates obtained

from model F(Pop�Year) where one parameter was used per

year and population }, Muro, Corsica; � , Pirio, Corsica; &, La

Rouviere, mainland southern France.

Table 4 Regression models for assessing the effect of climatic

indices on blue tit adult survival

Model

Number of

parameters Deviance

P-

ANODEV

Large-scale climate indices

Pop 1 SRF_L 36 1328.2 0.015

Pop 1 SRF_Q 37 1326.29 0.04

Pop 1 W-NAO_L 36 1339.3 0.24

Pop 1 -S-NAO_L 36 1340.85 0.36

Pop 1 W-NAO_Q 37 1339.17 0.51

Pop 1 S-NAO_Q 37 1340.65 0.65

Pop 1 Indian Monsoon_Q 37 1340.88 0.67

Pop 1 MEI_L 36 1343.02 0.70

Pop 1 Indian Monsoon_L 36 1343.2 0.74

Pop 1 MEI_Q 37 1342.5 0.86

Local weather indices

Pop 1 W-P-Tmax_L 36 1328.9 0.005

Pop 1 W-P-Tmax_Q 37 1327.85 0.01

Pop 1 S-P-T_Q 37 1329.47 0.025

Pop 1 S-W_Q 37 1336.82 0.22

Pop 1 S-P-T_L 36 1341.83 0.39

Pop 1 W-W-Tmin_L 36 1342.2 0.45

Pop 1 W-W-Tmin_Q 37 1340.61 0.53

Pop 1 S-W_L 36 1343.49 0.90

P-values are given for the effects of climate indices and were

derived using ANODEV (see text for details).

Linear and quadratic relationships are indicated using _L and

_Q, respectively.

M E D I T E R R A N E A N B L U E T I T S U R V I VA L 2243

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2235–2249

sions per year allowed defining directly the period of the

year when most adult mortality occurred (Nilsson, 1987;

Lahti et al., 1998), or the impact of climatic conditions at

different times of year was addressed (Clobert et al.,

1988). As expected, adult mortality peaked during winter

and was chiefly influenced by winter climatic conditions

in these populations located at relatively high latitudes.

It could consequently be concluded that an increase

in winter temperatures as a result of global warming

(Luterbacher et al., 2004) could result in reduced blue tit

mortality in populations located at relatively high lati-

tudes. By contrast, in two studies of passerine bird

demography in low-latitude populations where climatic

conditions were relatively mild during winter: serins,

Serinus serinus in northeastern Spain (Conroy et al., 2002),

long-tailed wagtail, Motacilla clara in South Africa (Piper,

2002); the impact of wintertime environmental condition

was not stronger than that of summertime environmen-

tal conditions. Our results corroborate the view that

winter is not the only constraining time of year for the

survival of blue tits in the northwestern Mediterranean.

Indeed, in addition to the detrimental effects of abundant

precipitation and low temperatures in winter we de-

tected influences of precipitation and temperature in

early summer.

Biogeographic and evolutionary implications

Blue tits in our study populations incurred higher mor-

tality when conditions in early summer were either

Table 5 Final selection among models combining selected climate indices

Model

Number of

parameters Deviance AICc R2

1 covariate/no interaction with Pop

Pop 1 SRF 36 1328.2 4215.4 0.32

Pop 1 W-P-Tmax_L 36 1328.9 4216.2 0.304

Pop 1 S-P-T_Q 37 1329.5 4218.8 0.293

1 covariate/1 interaction with Pop

Pop�W-P-Tmax_L 38 1319.6 4210.9 0.5

Pop� SRF_L 38 1322.3 4213.6 0.44

Pop� S-P-T_Q 41 1322.3 4219.8 0.44

2 covariates/no interaction with Pop

Pop 1 S-P-T_Q 1 SRF_L 38 1318.4 4209.8 0.52

Pop 1 W-P-Tmax_L 1 SRF_L 37 1322.1 4211.4 0.45

Pop 1 W-P-Tmax_L 1 S-P-T_Q 38 1323 4214.3 0.43

2 covariates/1 interaction with Pop

S-P-T_Q 1 Pop� SRF_L 40 1311.9 4207.4 0.66

Pop�W-P-Tmax_L 1 S-P-T_Q 40 1313.4 4208.9 0.63

Pop� S-P-T_Q 1 SRF_L 42 1311 4210.6 0.68

Pop�W-P-Tmax_L 1 SRF_L 39 1317.5 4210.9 0.54

W-P-Tmax_L 1 Pop� SRF_L 39 1319.5 4212.9 0.50

W-P-Tmax_L 1 Pop� S-P-T_Q 42 1318.5 4217.8 0.52

2 covariates/2 interactions with Pop

Pop� S-P-T_Q 1 Pop� SRF_L 44 1307.4 4211.2 0.75

Pop� W-P-Tmax_L 1 Pop� SRF_L 41 1314.9 4212.5 0.60

Pop� W-P-Tmax_L 1 Pop� S-P-T_Q 44 1310.3 4214.1 0.69

3 covariates/no interaction with Pop

W-P-Tmax_L 1 SRF_L 1 S-P-T_Q 39 1316.1 4209.5 0.57

3 covariates,/1 interaction with Pop

Pop�W-P-Tmax_L 1 S-P-T_Q 1 SRF_L 41 1308.7 4206.2 0.73

W-P-Tmax_L 1 Pop� S-P-T_Q 1 SRF_L 43 1310.3 4212 0.69

W-P-Tmax_L 1 S-P-T_Q 1 Pop� SRF_L 41 1310.6 4208.2 0.69

3 covariates/2 interactions with Pop

Pop�W-P-Tmax_L 1 S-P-T_Q 1 Pop� SRF_L 43 1307 4208.7 0.76

Pop�W-P-Tmax_L 1 Pop� S-P-T_Q 1 SRF_L 45 1304.8 4210.7 0.81

W-P-Tmax_L 1 Pop� S-P-T_Q 1 Pop� SRF_L 45 1307.1 4212.9 0.76

3 covariates/3 interactions with Pop

Pop�W-P-Tmax_L 1 Pop� S-P-T_Q 1 Pop� SRF_L 47 1304.4 4214.4 0.82

Lowest AICc models highlighted in bold.

2244 V. G R O S B O I S et al.

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2235–2249

extremely wet and cold or extremely dry and hot. In the

southern part of their range, blue tits could, thus, suffer

measurably from the predicted increased frequency of

summer extreme climatic conditions, such as heat waves

and droughts (Easterling et al., 2000) or heavy rainfall

(Christensen & Christensen, 2002). Along with other

studies that addressed the impact of climatic conditions

on tits’ reproductive parameters, such as laying dates, at

large geographical scales (Visser et al., 2002), our results

suggest that global warming could differently affect the

demography of tits in various parts of their range which

could ultimately lead to range shifts.

Given that detrimental effects of hot and dry summer

climatic conditions on survival have been detected here

and under the reasonable assumption that summer

climate in Mediterranean regions has started to and

will continue to evolve towards hotter and drier condi-

tions in the course of 21st century (IPCC, 2001), selective

pressures for adaptations to increased temperatures

and dryness in summer should be expected. However,

as long as the mechanisms underlying the impact of

summer climatic conditions on blue tit survival remain

unknown, the phenological, behavioural, morphologi-

cal, or physiological traits upon which these selective

pressures are likely to exert an influence cannot be

determined and it is impossible to evaluate whether

these unknown traits are likely to evolve under the

influence of selective pressures induced by the summer

warming trend.

Mechanisms through which climate variation results invariation in survival rates

Climate variation can directly induce physiological

and/or behavioural modifications that may have im-

mediate effects on demographic traits such as adult

survival. Aridity in early summer could induce a phy-

siological stress resulting in enhanced mortality. In fact,

Sanz et al. (2003) showed that under hot conditions in

May, adult-pied flycatchers reduce their daily energy

expenditure resulting in poor fledging success. One

hypothesis to explain these results is that adults show

a plastic response to hot climatic conditions that con-

sists in reducing the allocation of resources directed

towards reproduction. In the light of these results it

seems reasonable to hypothesize that hot and dry cli-

matic conditions in early summer induce a physiologi-

cal stress for adult blue tits. However, because too few

studies have so far addressed the influence of summer

climate variation on the physiology of passerine birds,

the support for this hypothesis can presently not be

considered as solid. We believe that longitudinal mon-

itoring of physiological indicators such as body mass,

field metabolic, and water influx rates at the within and

between summers scales, should be undertaken in

order to understand the impact of extreme summer

climatic conditions on important physiological pro-

cesses such as the demand to delivery oxygen, energy

and water balances (Speakman, 1997; Portner, 2002).

Local climatic conditions could also have an indirect

impact on blue tit survival through their influence on

food availability. It would be extremely difficult to

demonstrate the existence of such indirect impacts on

blue tits. Indeed, blue tits have a diversified diet (Betts,

1955) and it seems almost impossible to define one or

few key resources critical to adult survival. However,

0.8

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

−3 −2 −1 0 1 2 3

0.1

0.2

0.3

0.4

0.5

0.6

0.7

−2.5 −2 −1.5 −1 −0.5 0 0.5 1 1.5 2

−2.5 −2 −1.5 −1 −0.5 0 0.5 1 1.5 2

Blu

e ti

t ad

ult

surv

ival

WetLow temperature maxima

DryHigh temperature maxima

WetCold

DryWarm

W-P-Tmax index

Sahel rainfall index

S-P-T index

(a)

(b)

(c)

Fig. 5 Relationships between blue tit adult survival estimates

obtained from a model F(Pop�Year) where one parameter was

used per year and population (}, Muro, Corsica; � , Pirio,

Corsica; &, La Rouviere, mainland southern France) and three

standardized climate indices. (a) local winter precipitation and

maximum temperature index, (b) June–July Sahel rainfall index

(c) local June–July precipitation and temperature index. For the

sake of readability the confidence intervals of survival estimates

are not shown.

M E D I T E R R A N E A N B L U E T I T S U R V I VA L 2245

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2235–2249

we argue speculatively that early summer climatic

conditions could reflect an interaction between climate

and food availability, as follows. Blue tits in Mediterra-

nean habitats are known to consume ripe fruits almost

throughout the year (Herrera, 1984). In these habitats,

the abundance of ripe fruits is at its lowest in spring and

early summer (Herrera, 1984). Ripe fruit shortage is

probably not limiting in spring because caterpillars

provide an abundant source of food. However, this

resource could be limiting in early summer when

caterpillars are no longer available and a new cohort

of fledglings increase local population densities consid-

erably. Under such a nutritional stress, the direct (phy-

siological or behavioural) impact of extreme climatic

conditions could be enhanced. Supplemental feeding

experiments have been undertaken in tit populations

from Northern Europe in order to assess the importance

of winter food supply for adult survival (Lahti et al.,

1998; Perdeck et al., 2000). We believe that it would be

valuable to undertake similar experiments for assessing

the importance of summer food supply on the adult

survival of Mediterranean blue tits.

It has been found that climate change can affect

reproductive parameters of passerines birds by indu-

cing mismatching between periods of peak resource

demand and availability (Sanz et al., 2003). The detri-

mental effect of extreme summer climatic conditions on

survival detected here could result from low resource

availability during an energetically demanding phase of

blue tit annual cycle: the postbreeding moult. A possi-

ble way of addressing this hypothesis would be to test

whether laying date, a routinely recorded phenological

trait correlated with the timing of postbreeding moult

(Sanz, 1999), modulates at the individual scale the effect

of extreme summer climatic condition on survival.

Variation of climate influences among populations

Our results indicate that geographically close popula-

tions can respond differently to climate fluctuations.

The response of adult survival to fluctuations of cli-

matic conditions was steeper in the population where

density was the highest. This pattern could reflect

synergistic detrimental effects of population

density and harsh climatic conditions (e.g. Barbraud &

Weimerskirch, 2003). Furthermore, winter climatic

conditions seemed to have a stronger influence on adult

survival in the two insular populations than in the

mainland population. It has been hypothesized that in

the mainland population, but not in the island popula-

tions, adaptation to Mediterranean environmental con-

ditions is hampered by a net gene flow originating from

populations located at more northern latitudes (Blondel

et al., 1999b; Thomas et al., 2001). The absence of impact

of winter climatic conditions on survival in the main-

land population would make sense under this hypoth-

Table 6 Estimates of the parameters of the relationships between the logit of blue tit adult survival in the three study populations

and standardized climate indices obtained from the three lowest AICc models

Muro Pirio Rouviere

Estimate 95% CI Estimate 95% CI Estimate 95% CI

Model: Pop�W-P-Tmax_L 1 S-P-T_Q 1 SRF_L; AICc 5 4206.2; R2 5 0.73

Intercept �0.41 �0.68; �0.16 0.42 0.28; 0.56 0.16 �0.05; 0.37

W-P-Tmax 0.56 0.14; 0.98 0.13 �0.01; 0.27 �0.08 �0.28; 0.12

S-P-T �0.14 �0.27; �0.02

S-P-T2 �0.11 �0.20; �0.02

SRF 0.14 �0.01; 0.26

Model: S-P-T_Q 1 Pop� SRF_L; AICc 5 4207.4; R2 5 0.66

Intercept �0.52 �0.80; �0.25 0.46 0.32; 0.60 0.17 �0.04; 0.38

S-P-T �0.18 �0.31; �0.05

S-P-T2 �0.14 �0.23; �0.05

SRF 1.19 0.42; 1.95 0.17 0.04; 0.31 0.17 �0.08; 0.42

Model: W-P-Tmax_L 1 S-P-T_Q 1 Pop� SRF_L; AICc 5 4208.2; R2 5 0.69

Intercept �0.53 �0.80; �0.25 0.43 0.28; 0.58 0.12 �0.11; 0.35

W-P-Tmax 0.07 �0.05; 0.20

S-P-T �0.16 �0.29; �0.02

S-P-T2 �0.12 �0.21; �0.02

SRF 1.06 0.28; 1.84 0.14 �0.01; 0.29 0.18 �0.07; 0.43

When only one figure is displayed in a row, the same estimate applies to the three populations.

CI, confidence interval.

2246 V. G R O S B O I S et al.

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2235–2249

esis since winter in the Mediterranean is much milder

than in more temperate environments with which these

mainland populations are supposed to be connected

through gene flow.

Large-scale climatic factors driving local Europeanclimates in summer

In addition to the effects of local climate indices in

winter and early summer, we detected a relationship

between adult survival and a large scale climate index:

Sahel rainfall, that is known to reflect oceanic and

atmospheric circulation phenomena in the inter tropical

zone (Hurrell et al., 2002; Raicich et al., 2003; Cassou

et al., 2004). The impact of Sahel rainfall on environ-

mental conditions in the staging and wintering grounds

of sub-Saharan migrant bird species and, as a conse-

quence, on the survival of these migrant bird species is

well established (Peach et al., 1991; Barbraud et al.,

1999). Our study suggests that this tropical climatic

factor also correlates with demographic parameters in

nonmigratory European bird populations. Furthermore,

the Sahel rainfall index and the early summer local

climate index (integrating temperature and precipita-

tion) seemed to capture distinct pieces of information

on the climatic conditions influencing blue tit adult

survival. Firstly, these two climate indices were not

correlated. Secondly, the comparison of the statistical

relevance of alternative models for adult survival re-

vealed that a model including both the effect of June–

July Sahel rainfall and the June–July local climate index

explained more of the variation in adult survival in the

three study populations than models including only

one of these effects. The Sahel rainfall index, thus,

probably reflected local climatic factors influencing blue

tit survival that were not reflected by our local summer

climate index. Future collaborations with climatologists

could hopefully help us better understand the mechan-

isms underlying the remote influence of Sahel rainfall

on the demography of Mediterranean birds. Although

these mechanisms remain unknown, our results suggest

that large-scale climatic factors reflecting oceanic and

atmospheric circulation in the intertropical zone might

constitute useful explorative proxies in the studies of

the impact of summer climate on European ecosystems,

especially in the Mediterranean basin.

Acknowledgements

We are grateful to Paula C. Dias, Marie Maistre and manystudents for their assistance in the field; Marie-Jose Galan,Mireille Son for helping in data base management. Many thanksto Tom Sherry, Andre Dhondt, Jim Hurrell, Erik Matthysen,Marcel M. Visser, Nils Christian Stenseth, and to the membersof the ECOCLIM project for their helpful comments. We are alsograteful to two anonymous referees for their thoughtful com-ments and criticisms on the manuscript. The study was fundedby grants from the European Commission (project METABIRD),from the GICC/IFB (Climate changes impact project) and fromthe CNRS (ACI CLIM-POP). The CRBPO provided ringingmaterial.

References

Allan RJ, Lindesay J, Parker D (1996) El Nino Southern Oscillation

and Climatic Variability. Csiro, Australia.

Almaraz P, Amat JA (2004) Complex structural effects of two

hemispheric climatic oscillatiors on the regional spatio-tem-

poral expansion of a threatened bird. Ecology Letters, 7,

547–556.

Year

Blu

e ti

t ad

ult

surv

ival

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1985

–1986

1986

–198

7

1987

–198

8

1988

–198

9

1989

–199

0

1990

–199

1

1991

–199

2

1992

–199

3

1993

–199

4

1994

–199

5

1995

–199

6

1996

–199

7

1997

–199

8

1998

–199

9

(a)

(b)

(c)

Fig. 6 Blue tit adult survival as estimated from model

F(Pop�Year) where one parameter was used per year and

population [(a) &, La Rouviere, mainland southern France; (b)

}, Muro, Corsica; (c) � , Pirio, Corsica with 95% confidence

interval (CI)] and from model F(Pop�W-P-Tmax_L 1 S-P-

T_Q 1 Pop� SRF_L) where parameters were determined

through relationships with three climate indices: the local winter

precipitation and maximum temperature index, the local June–

July precipitation and temperature index, and the June–July

Sahel rainfall index: (estimates: solid lines; 95% CI: dashed lines).

M E D I T E R R A N E A N B L U E T I T S U R V I VA L 2247

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2235–2249

Barbraud C, Barbraud JC, Barbraud M (1999) Population dy-

namics of the white Stork Ciconia Ciconia in western France.

Ibis, 141, 469–479.

Barbraud C, Weimerskirch H (2003) Climate and density shape

population dynamics of a marine top predator. Proceedings

of the Royal Society of London B, 270, 2111–2116.

Betts MM (1955) The food of titmice in oak woodland. Journal of

Animal Ecology, 24, 282–323.

Blondel J, Aronson J (1999a) Biology and Wildlife of the Mediterra-

nean Region. Oxford University Press, Oxford.

Blondel J, Dias PC, Maistre M et al. (1993) Habitat heterogeneity

and life history variation of Mediterranean tits. The Auk, 110,

511–520.

Blondel J, Dias PC, Perret P et al. (1999b) Selection-based biodi-

versity at a small spatial scale in a low-dispersing insular bird.

Science, 285, 1399–1402.

Both C, Artemyev AV, Blaauw B et al. (2004) Large-scale geo-

graphical variation confirms that climate change causes birds

to lay earlier. Proceedings of the Royal Society of London Series

B-Biological Sciences, 271, 1657–1662.

Burnham KP, Anderson DR (1998) Model Selection and Inference,

a Practical Information-Theoretic Approach. Springer, New York.

Cassou CCD, Terray L, Hurrell JW et al. (2004) Summer sea

surface temperature conditions in the North Atlantic and their

impact upon the atmospheric circulation in early winter.

Journal of Climate, 17, 3349–3363.

Choquet R, Reboulet A-M, Pradel R et al. (2001) U-care (Utilities –

CApture-REcapture) user’s guide. CEFE, CNRS, Montpellier,

France.

Christensen JH, Christensen OB (2002) Severe summertime

flooding in Europe. Nature, 421, 805–806.

Clobert J, Perrins CM, Mc Cleery RH et al. (1988) Survival rate in

the great tit Parus major in relation to sex, age and immigration

status. Journal of Animal Ecology, 57, 287–306.

Conroy MJ, Senar JC, Domenech J (2002) Analysis of individual-

and time-specific covariate effects on survival of Serinus

serinus in north-eastern Spain. Journal of Applied Statistics, 29,

125–142.

Dhondt AA (2001) Trade-offs between reproduction and survival

in tits. Ardea, 89, 155–166.

Doherty PF, Grubb TC (2002) Survivorship of permanent-resi-

dent birds in a fragmented forested landscape. Ecology, 83,

844–857.

Draper NR, Smith H (1981) Principal component regression.

Applied Regression Analysis, 2nd edn. John Wiley & Sons Inc.,

New York.

Easterling DR, Meehl GA, Parmesan C et al. (2000) Climate

extremes: observations, modeling, and impacts. Science, 289,

2068–2074.

Finney BP, Gregory-Eaves I, Douglas MSV et al. (2002) Fisheries

productivity in the northeastern Pacific Ocean over the past

2200 years. Nature, 416, 729–733.

Folland CK, Palmer TN, Parker DE (1986) Sahel rainfall and

worldwide sea temperatures. Nature, 320, 602–607.

Forchhammer MC, Post E, Stenseth NC (1998) Breeding phenol-

ogy and climate. Nature, 391, 29–30.

Franklin AB, Anderson DR, Gutierrez RJ et al. (2000) Climate,

habitat quality, and fitness in northern spotted owl popula-

tions in northwestern California. Ecological Monographs, 70,

539–590.

Garcia LV (2004) Escaping the Bonferroni iron claw in ecological

studies. Oikos, 105, 657–663.

Graham MH (2003) Confronting multicollinearity in ecological

multiple regression. Ecology, 84, 2809–2815.

Graham MH, Dayton PK, Erlandson JM (2003) Ice ages and

ecological transitions on temperate coasts. Trends in Ecology

and Evolution, 18, 33–40.

Herrera CM (1984) A study of avian frugivores, bird-dispersed

plants, and their interaction in Mediterranean scrublands.

Ecological Monographs, 54, 1–23.

Holmgren M, Scheffer M, Ezcurra E et al. (2001) El Nino effects

on the dynamics of terrestrial ecosystems. Trends in Ecology &

Evolution, 16, 89–94.

Hurrell JW, Deser C, Folland CK et al. (2002) The relationship

between tropical Atlantic rainfall and the summer circulation

over the North Atlantic. In: U.S. CLIVAR Atlantic meeting

(ed.Legler D) pp. 111–114. Boulder, CO.

Hurrell JW, Kushnir Y, Visbeck M (2001) The North Atlantic

Oscillation. Science, 291, 603–605.

IPCC. (2001) Climate Change 2001 Third Assessment Report. Cam-

bridge University Press, Cambridge, UK.

Lahti K, Orell M, Rytkonen S et al. (1998) Time and food

dependence in Willow Tit winter survival. Ecology, 79,

2904–2916.

Lebreton J-D, Burnham KP, Clobert J et al. (1992) Modeling

survival and testing biological hypothesis using marked ani-

mals: a unified approach with case studies. Ecological Mono-

graphs, 62, 67–118.

Luterbacher J, Dietrich D, Xoplaki E et al. (2004) European

seasonal and annual temperature variability, trends, and ex-

tremes since 1500. Science, 303, 1499–1503.

Montevecchi WA, Myers RA (1997) Centurial and decadal ocea-

nographic influences on changes in northern gannet popula-

tions and diets in the north-west Atlantic: implications for

climate change. ICES Journal of Marine Science, 54, 608–614.

Moss R, Oswald J, Baines D (2001) Climate change and breeding

success: decline of the capercaillie in Scotland. Journal of

Animal Ecology, 70, 47–61.

Newton I (1998) Chapter 11: weather. Population Limitation In

Birds. Academic Press, London.

Newton I, Wyllie I, Rothery P (1992) Annual survival of Sparro-

whawks Accipiter nisus breeding in three areas of Britain. Ibis,

135, 49–60.

Nilsson J-A, Svensson E (1996) The cost of reproduction: a new

link between current reproductive effort and future reproduc-

tive success. Proceedings of the Royal Society of London B, 263,

711–714.

Nilsson SG (1987) Limitation and regulation of population

density in the nuthatch Sitta Europea (Aves) breeding in

natural cavities. Journal of Animal Ecology, 56, 921–937.

Nur N (1984) The consequences of brood size for breeding blue

Tits.1. Adult survival, weight change and the cost of reproduc-

tion. Journal of Animal Ecology, 53, 479–496.

Parmesan C, Yohe G (2003) A globally coherent fingerprint of

climate change impacts across natural systems. Nature, 421,

37–42.

2248 V. G R O S B O I S et al.

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2235–2249

Peach W, Baillie S, Underhill L (1991) Survival of British sedge

warblers Acrocephalus schoenobaenus in relation to west African

rainfall. Ibis, 133, 300–305.

Peach WJ, Thompson PS, Coulson JC (1994) Annual and long-

term variation in the survival rates of British lapwings Vanellus

vanellus. Journal of Animal Ecology, 63, 60–70.

Perdeck AC, Visser ME, Van Balen JH (2000) Great tit Parus major

survival and the beech-crop index. Ardea, 99, 99–108.

Piper SE (2002) Survival of adult, territorial longtailed wagtails

Moticilla clara: the effects of environmental factors and indivi-

dual covariates. Journal of Applied Statistics, 29, 107–124.

Portner HO (2002) Climate variations and the physiological basis

of temperature dependent biogeography: systemic to molecu-

lar hierarchy of thermal tolerance in animals. Comparative

Biochemistry and Physiology Part A, 132, 739–761.

Raicich F, Pinardi N, Navarra A (2003) Teleconnections between

Indian monsoon and Sahel rainfall and the Mediterranean.

International Journal of Climatology, 23, 173–186.

Roback PJ, Askins RA (2005) Judicious use of multiple hypoth-

esis tests. Conservation biology, 19, 261–267.

Robinson RA, Green RE, Baillie SR et al. (2004) Demographic

mechanisms of the population decline of the song thrush

Turdus philomelos in Britain. Journal of Animal Ecology, 73,

670–682.

Roy K, Valentine JW, Jablonski D et al. (1996) Scales of climatic

variability and time averaging in Pleistocene biotas: implica-

tions for ecology and evolution. Trends in Ecology and Evolution,

11, 458–463.

Sæther B-E, Engen S, M�ller AP et al. (2003) Climate variation

and regional gradients in population dynamics of two hole-

nesting passerines. Proceedings of the Royal Society of London B,

270, 2397–2404.

Sanz JJ (1999) Seasonal variation in reproductive success and

post-nuptial moult of blue tits in southern Europe: an experi-

mental study. Oecologia, 121, 377–382.

Sanz JJ (2002) Climate change and breeding parameters of great

and blue tits throughout the western Palearctic. Global Change

Biology, 8, 409–422.

Sanz JJ, Potti J, Moreno J et al. (2003) Climate change and fitness

components of a migratory bird breeding in the Mediterra-

nean region. Global Change Biology, 9, 461–472.

Siikamaki P, Hovi M, Ratti O (1994) A trade-off between current

reproduction and moult in the Pied Flycatcher – an experi-

ment. Functional Ecology, 8, 587–593.

Sillet TS, Holmes RT, Sherry TW (2000) Impacts of global climate

cycle on population dynamics of a migratory songbird. Science,

288, 2040–2042.

Skalski JR (1996) Regression of abundance estimates from mark-

recapture surveys against environmental covariates. Canadian

Journal of Fisheries and Aquatic Sciences, 53, 196–204.

Sokal RR, Rohlf FJ (1995) Biometry, 3rd edn. W. H. Freeman and

Company, New York.

Speakman JR (1997) Doubly-labelled water. Chapman & Hall,

London.

Stenseth NC, Mysterud A, Ottersen G et al. (2002) Ecological

effects of climate fluctuations. Science, 297, 1292–1296.

Stenseth NC, Ottersen G, Hurrel JW et al. (2003) Studying climate

effects on ecology through the use of climate indices: the North

Atlantic Oscillation, El Nino Southern Oscillation and beyond.

Proceedings of the Royal Society of London B, 270, 2087–2096.

Thomas CD, Cameron A, Green REG et al. (2004) Extinction risk

from climate change. Nature, 427, 145–148.

Thomas DW, Blondel J, Perret P et al. (2001) Energetic and fitness

costs of mismatching resource supply and demand in season-

ally breeding birds. Science, 291, 2598–2600.

Tremblay I, Thomas DW, Lambrechts MM et al. (2003) Variation

in Blue Tit breeding performance across gradients in habitat

richness. Ecology, 84, 3033–3043.

Visser ME, Adriaensen F, Van Balen JH et al. (2002) Variable

responses to large-scale climate change in European parus

populations. Proceedings of the Royal Society of London B, 270,

367–372.

White GC, Burnham KP (1999) Program MARK: survival esti-

mation from populations of marked animals. Bird Study, 46

(Suppl.), 120–129. http://www.cnr.colostate.edu/gwhite/

software.html.

M E D I T E R R A N E A N B L U E T I T S U R V I VA L 2249

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2235–2249