A map of autumn precipitation for the third millennium BP in the Eastern Iberian Peninsula from charcoal carbon isotopes

This article appeared in a journal published by Elsevier. The attachedcopy is furnished to the author for internal non-commercial researchand education use, including for instruction at the authors institution

and sharing with colleagues.

Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third party

websites are prohibited.

In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further information

regarding Elsevier’s archiving and manuscript policies areencouraged to visit:

http://www.elsevier.com/copyright

http://www.elsevier.com/copyright

Author's personal copy

A map of autumn precipitation for the third millennium BP in the Eastern IberianPeninsula from charcoal carbon isotopes

M. Aguilera a, C. Espinar b, J.P. Ferrio b,1, G. Pérez c, J. Voltas b,⁎a Unitat de Fisiologia Vegetal, Facultat de Biologia, Universitat de Barcelona, Avgda. Diagonal 645, E-08028 Barcelona, Spainb Dep. of Crop and Forest Sciences, University of Lleida, Rovira Roure 191, E-25198 Lleida, Spainc Dep. Prehistòria i Arqueologia, Universitat de València, Avgda. Blasco Ibáñez 28, E-46010 Valencia, Spain

a b s t r a c ta r t i c l e i n f o

Article history:Received 13 June 2008Accepted 17 November 2008Available online 27 November 2008

Keywords:Quercus ilexDroughtMediterranean climateLate HoloceneIsoscapeGISIron Age Cold Epoch

Carbon isotope composition (δ13C) in tree-rings has become routinely used in palaeoclimatic research for theassessment of changes in plant water availability in seasonally dry climates. However, the distribution of longtree-ring records around the world is very limited. Alternatively, the original climate signal of wood δ13C iswell preserved in fossil charcoal and, accordingly, charcoal δ13C can be used to quantify past changes in wateravailability (e.g. precipitation). We report a case study on spatial palaeoclimate reconstruction which aims tocharacterize the transition between Bronze and Iron Ages, the so-called Iron Age Cold Epoch (ca. 900–300 BCE), using charcoals of Quercus ilex/coccifera from a set of 11 contemporary archaeological sites ofeastern Spain. Climatic inferences were obtained after calibrating a linear model predicting seasonalprecipitation from δ13C of Q. ilex wood samples obtained across a rainfall gradient. The best regression modelcorresponded to September–December (autumn) precipitation (Paut), in agreement with the fact that Q. ilexis able to exploit previous-year water reserves thanks to very effective water uptake. Subsequently, weestimated Paut from the δ13C of fossil charcoal to infer spatial patterns in water availability. Overall, estimatedpast Paut was about 19% higher (296 mm) than present-time values averaged across archaeological sites(249 mm). However, a clear geographic pattern of differences in precipitation could be observed in which theinner continental regions of eastern Spain were characterized by more humid conditions in the past, whereasthe coastal strip of the Mediterranean Sea barely differed in past and present Paut values. The quite uniformdistribution of archaeological sites over eastern Spain allowed development of contour maps of absolute andrelative (to present) past Paut using gridded interpolation methods implemented in a GIS, highlighting thepotential of this approach for reconstructing high-resolution spatial patterns of past climate.

© 2008 Elsevier B.V. All rights reserved.

1. Introduction

In drought-prone areas, precipitation is the most significantclimate element impacting ecosystems. Unlike temperature, precipi-tation regimes show relatively poor spatial correlations and canexhibit contrasting associations with global climate trends in differentareas (Rodó et al., 1997). Thus, evidences of precipitation changes overthe recent past (i.e. the Holocene) are still scarce and mainly of aqualitative nature (IPCC, 2001) because the signal-to-noise ratio islower than for glacial climate variability (EuroCLIMATE, 2007). Thisholds true for vast areas such as the Iberian Peninsula (NWMediterranean Basin), where the climate is defined by complicatedinteractions between Atlantic, continental, Mediterranean, and sub-tropical influences. Reliable precipitation proxies for this region

should allow for high temporal and spatial resolutions. Indeed, abroad spatial heterogeneity in response to global climate trends in thewestern Mediterranean is supported by present climatic data (Rodóet al., 1997) and by the palaeoenvironmental record (Magny andRichard, 1992; Davis, 1994; Riera, 1994; Jalut et al., 2000; Magny et al.,2003).

Measurements of carbon isotope composition (δ13C) in tree-ringshave become routinely used in palaeoclimatic research over the lastdecades (McCarroll and Loader, 2004). Indeed, δ13C has been relatedto various environmental variables such as light intensity, tempera-ture or air pollution (Elsik et al., 1993; Berry et al., 1997; Andersonet al., 1998), and to several indicators of water availability in regionswhere evaporation exceeds precipitation (Leavitt, 1993; Warren et al.,2001; Ferrio and Voltas, 2005). However, although some exceptionaltree-ring chronologies may cover almost the whole Holocene, thedistribution of such long tree-ring records around the world is verylimited. In this context, the analysis of δ13C in wood charcoal, a plantremain routinely recovered in the course of archaeological excava-tions, has been proposed as proxy for in situ palaeoenvironmentalreconstructions (February and Van der Merwe, 1992; Vernet et al.,

Journal of Geochemical Exploration 102 (2009) 157–165

⁎ Corresponding author. Tel.: +34 973 702855; fax: +34 973 232864.E-mail address: [email protected] (J. Voltas).

1 Present address: Chair of Tree Physiology, University of Freiburg, Georges-KöhlerAllee 53/54, 79085 Freiburg im Breisgau, Germany.

0375-6742/$ – see front matter © 2008 Elsevier B.V. All rights reserved.doi:10.1016/j.gexplo.2008.11.019

Contents lists available at ScienceDirect

Journal of Geochemical Exploration

j ourna l homepage: www.e lsev ie r.com/ locate / jgeoexp


1996). Nevertheless, the potential effect of charring on δ13C has to beconsidered, as it seems species-specific. Experimental carbonizationsperformed in conifer wood fragments (Jones and Chaloner 1991;Czimczik et al., 2002; Ferrio et al., 2006) have shown that, in a rangefrom ca. 300 °C to 600 °C, charcoal δ13C tends to decrease (from 1‰ to2‰) as temperature increases. A weaker decrease (up to 0.6‰) hasalso been reported for one hardwood species (birch) (Czimczik et al.,2002). Despite such changes, Ferrio et al. (2006) found that the δ13C ofAleppo pine charred at 300 °C, 400 °C and 500 °C was still stronglycorrelated with the original δ13C of wood. These authors proposed tocorrect for the effect of charring on δ13C using charcoal carbon

concentration (%C) (which ranged from ca. 50% to 90%) as an indicatorof the degree of carbonization, which resulted in comparablerelationships with climate variables as those found for intact wood.In another study performed on four different Mediterranean oaks(both evergreen and deciduous) and four species from the genusPistacia, Ferrio et al. (2007) found no significant δ13C changes aftercharring (300–500 °C). In this case, the association between controland charred material followed a 1:1 linear relationship, indicatingthat, at least for these species, δ13C in charcoal might be directlycomparable with the values of intact wood. In summary, although theeffect of carbonization on δ13C varies with the species and charring

Fig. 1. Geographic distribution of sampling sites for reference wood cores of Quercus ilex, along with archaeological sites where fossil charcoal was collected. For reference, a contourmap of present-time autumn precipitation (September–December period), adapted from the Digital Climatic Atlas of the Iberian Peninsula (Ninyerola et al., 2005), is presented.Detailed descriptions of sampling and archaeological sites are provided in Table 1 and in Ferrio et al. (2003), and in Table 2, respectively.

158 M. Aguilera et al. / Journal of Geochemical Exploration 102 (2009) 157–165


temperature, the original environmental signal seems to be preservedin charcoal and can be enhanced, if necessary, after correcting for %C.

Thewidespread presence of charcoal remains in the archaeologicalrecord of the Mediterranean Basin offers the possibility to studychanges in the spatial distribution of precipitation for particular timeperiods. A significant Holocene climate event is the Iron Age ColdEpoch or Iron Age Neoglaciation (ca. 900–300 BCE), which ischaracterized by an abruptly cooling climate in northern Europe(Van Geel et al., 1996). For the Iberian Peninsula, however, climateevidence is relatively scarce and contradictory, supporting either acold but dry period (Magny and Richard, 1992; Davis, 1994; Gutiérrez-Elorza and Peña-Monné, 1998) or a colder/wetter climate (Davis,1994; Gutiérrez-Elorza and Peña-Monné,1998).We hypothesized thatsuch apparent inconsistency would be partly due to the particulargeographic position of the Iberian Peninsula in the interphase ofEuropean and African climatic influences, as reported for other climateevents (e.g. Rodó et al., 1997; Jalut et al., 2000; Magny et al., 2003). Inthis regard, a network of palaeoenvironmental records in this areawould assist in locating this interphase. To shed light on this issue, wepresent here the first attempt to spatially map past precipitation usingcharcoal δ13C data from Quercus ilex/coccifera, a species complexmade up of sclerophyllous Mediterranean oaks sensitive to changes inwater availability, particularly during autumn–winter (Ferrio et al.,2003). The main objective of this work was therefore to characterizethe spatial distribution of precipitation in the eastern IberianPeninsula during the Iron Age Cold Epoch, as compared withpresent-day conditions. For this purpose, we specifically aimed to (i)obtain an empirical model relating changes in seasonal (presumably,autumn–winter) precipitation to δ13C records of extant Q. ilex andapply this model to fossil charcoal of Q. ilex/coccifera from a networkof archaeological sites; (ii) create a map of precipitation distributionin the region during the third millennium BP based on charcoal δ13Cdata; and (iii) assess whether the inferred patterns of precipitationdiffered across eastern Iberian Peninsula.

2. Materials and methods

2.1. Reference material: tree-ring dating and sample preparation

We sampled wood cores of Holm oak (Q. ilex L.) from 16 locationsin northeast Spain representing most of the range of variation inthermal and precipitation regimes along which this species is found inSpain (Jiménez et al.,1996) (Fig.1; Table 1). The sampling strategywasintended to (i) provide referencematerial growing as close as possibleto a subset (7) of the archaeological sites used for palaeoclimatereconstruction and (ii) complement the field survey reported in Ferrioet al. (2003) and based on the current division in five provenanceregions for northeast Spain, which defines adaptive units according tothe identification of homogeneous ecogeographical zones (Martínet al., 1998). In winter of 2007, we selected eight mature and healthy,dominant trees per location far from sources of water other thanprecipitation such as springs, gullies, streams, etc., and a single core

was taken from the south side of each. The cores were obtained with aTrephor tool (Rossi et al., 2006). They were 2 mm in diameter andabout 15 mm in length, and contained the most recently formed treerings. Visual tree-ring dating was performed with a binocularmicroscope, with the number of tree rings included in the segmentsaveraging 27±8 (standard deviation, SD). Fragments were oven-dried at 60 °C for 48 h and milled separately to a fine powder using aball mill (Retsch MM301, Haan, Germany) for carbon isotope analysis.

2.2. Archaeobotanical material: dating and sample preparation

One hundred forty charcoal remains of Q. ilex/coccifera wererecovered from 11 contemporary archaeological sites located in theeastern Iberian Peninsula (western Mediterranean basin) (Fig. 1).Q. ilex (Holm oak) and Q. coccifera (Kermes oak) are sclerophyllousevergreen Mediterranean oaks, but they are indistinguishable basedon the anatomical features of charcoal. This may introduce uncertain-ties related to the sampling of reference material, but Q. ilex, amedium-sized tree, is nowadays more abundant in eastern Spain thanthe shrubbyKermes oak, having a potentiallywider variety of end-usesfor ancient communities. The archaeological sites were selected basedon charcoal availability for this species complex through the periodunder study (the transition between Bronze and Iron Ages or ‘Iron AgeCold Epoch’: ca. 900–300 BCE). Currently, the sites are characterized bydifferent Mediterranean bioclimates (Table 2). The samples werecollected from various archaeological contexts, such as cooking ovens,domestic fires, storage jars, the floors of dwellings, and levels of rubblefrom housing structures and pits. Presumably, ancient communitiesgathered Q. ilex/coccifera close to their settlements according to thedominance of this species complex in arboreal pollen records (Alonso,1999; Carrión and Van Geel, 1999; Fernández et al., 2007).

For the sites from the Late Bronze and First Iron periods, radiocarbonages were determined at Beta Analytical Inc. (Miami, FL, USA) and at theUniversitat de Barcelona (Barcelona, Spain), and dates were calibratedwith thesoftwareCALIBREV4.3 (Stuiveret al.,1998). Radiocarbondatingof individual charcoal fragments was not feasible, so we used averagedates for each stratigraphic unit. For the remaining sites, a combinationof stratigraphic and archaeological datingwas used (Table 2). To recoverplant remains, soil samples were treated using a standard flotation tankin the field with 5, 2 and 0.5 mm sieves. The average size of charcoalfragments used for isotope analysiswas 13mm×10mm×7mmand theestimated average number of tree rings per fragment was 15±8 (SD).Each charcoal fragment was soaked separately with HCl 6 M for 24 h atroom temperature to remove carbonate crusts (DeNiro and Hastorf,1985), which is a crucial step to avoid shifts in δ13C, and then rinsedrepeatedly with distilled water. Finally, charcoal fragments were ovendried and milled separately for carbon isotope analysis.

2.3. Meteorological data

Current meteorological data were obtained from the DigitalClimatic Atlas of the Iberian Peninsula (http://opengis.uab.es/wms/

Table 1Main geographical and environmental variables of sampling sites for reference wood cores.

Sampling site (province) Lat. Long. Alt. (m) Pan (mm) Tan (°C) Köppen classificationa

La Fatarella (Tarragona) 41°09′N 0°28′E 487 503 14.3 Csa, Pure MediterraneanBot (Tarragona) 41°00′N 0°20′W 511 493 14.0 Csa, Pure MediterraneanHorta de San Joan (Tarragona) 40°57′N 0°20′E 546 666 14.1 Csa, Pure MediterraneanEl Figaró (Barcelona) 41°43′N 2°15′E 575 917 12.5 Cfb, MarineCaldes de Montbui (Barcelona) 41°38′N 2°09′W 284 735 14.0 Cfb, MarineViladesens (Girona) 42°06′N 2°54′E 136 766 14.4 Cfb, MarineSant Martí Vell (Girona) 41°58′N 2°53′E 392 740 13.0 Cfb, MarineCiutadilla (Lleida) 41°34′N 1°09′W 477 509 13.6 BSk, Cool semiaridSant Mateu de Bages (Barcelona) 41°47′N 1°37′E 658 600 12.2 Cfa, Humid subtropical

Lat., latitude; Long., longitude; Alt., altitude a.s.l.; Pan, annual precipitation; Tan, mean annual temperature.a Köppen and Geiger (1936).

159M. Aguilera et al. / Journal of Geochemical Exploration 102 (2009) 157–165


iberia/index.htm) implemented inMiraMon-GIS (spatial resolution of200 m) with available monthly maps of air mean temperature(minimum, mean and maximum), precipitation and solar radiation.These maps combine climate estimates obtained from regressingmeteorological data of the Meteorological National Institute network(1951–1999 period) on geographic factors affecting climate (latitude,longitude, continentality, among others) (i.e. predicted climaticsurface) with the spatial interpolation of the regression residuals ateach meteorological station (i.e. residual surface) (Ninyerola et al.,2005). We took monthly values of precipitation from 35 gridreferences to characterize present climate at the archaeological sites(11) and at the sampling sites for reference wood cores (31 sites, butseven located close to particular archaeological sites).

2.4. Carbon isotope discrimination

The 13C/12C ratios were determined by mass spectrometry at Iso-Analytical (Sandbach, Cheshire, UK) and referred to the PeeDeeBelemnite (PDB) standard as carbon isotope composition (δ13C, ‰):

δ13C xð Þ = Rsample

Rstandard

� �− 1

� �× 1000 ð1Þ

The accuracy of analyses (standard deviation of working stan-dards) was 0.05‰ (for reference material) and 0.06‰ (for fossilcharcoal). To account for changes in δ13C of atmospheric CO2 (δ13Cair)during the Holocene, we calculated carbon isotope discrimination inwood (Δ13Cw) from δ13Cair and wood δ13C (δ13Cw), as described byFarquhar et al. (1982):

Δ13Cw =δ13Cair − δ13Cw

1 + δ13Cwð2Þ

δ13Cair was inferred by interpolating a range of data from Antarcticice-core records, together with modern data from two Antarcticstations (Halley Bay and Palmer Station) of the CU-INSTAAR/NOAA-CMDL network for atmospheric CO2 measurements, as described inFerrio et al. (2005). The resulting smoothed δ13C curve of atmosphericCO2 from 16100 BCE to present is available on the internet (http://web.udl.es/usuaris/x3845331/AIRCO2_LOESS.xls). According tothese data, the δ13Cair value applied to extant wood fragments was−8.0‰, varying between −6.5‰ and −6.4‰ for fossil charcoal.Previous results on the effect of experimental carbonization in woodfragments of Mediterranean oaks sampled across rainfall gradientsindicated no changes in δ13C values as compared with referencematerial (Ferrio et al., 2007). Therefore, any correction for the effect ofcarbonization on δ13C was deemed unnecessary.

2.5. Statistical analysis

The relative strength of the between-site common variance signalfor Δ13C data (either current or charcoal) was evaluated using theconcept of signal-to-noise ratio (SNR) as defined for dendroclimatol-ogy (e.g. Wigley et al., 1984):

SNR =N − 1ð Þ × SSBS − SSE

SSEð3Þ

where SSBS and SSE represent the sum of squares between sites andthe error (or intra-site) sum of squares, respectively, that appear in aone-way analysis of variance, and N is the average number of intra-site samples. The ratio [SNR/(1+SNR)], or ‘expressed populationsignal’ statistic (EPS; Wigley et al., 1984), was also calculated as SNRbehaves in a markedly non-linear fashion and the latter is easier tointerpret.

Relationships between Δ13Cw (as independent variable) andannual and monthly precipitation for extant wood of Q. ilex wereTa

ble2

Mainge

ograph

ical

anden

vironm

entalv

ariables,d

ating,

cultural

period

andresu

ltsof

thean

alysis

ofδ1

3Cch

arco

alforea

charch

aeolog

ical

site.

Site

(provinc

e)Lat.

Long

.Cu

rren

tclim

ateco

nditions

Archa

eologicalsamples

Alt.

(m)

T an

(°C)

P an

(mm)

P aut

(mm)

P aut(±

SD)

(199

5-05

)(m

m)

Köp

pen

classification

aN

Calend

arye

ar(B

CE)

Mea

nInterval

Cultural

period

δ13C a

ir(‰

)δ1

3C(‰

)Δ13C(‰

)Estimated

P aut

(±SD

)(m

m)

Cabe

zode

laCruz

(Zarag

oza)

41°28′N

0°54

′W60

014

386

138

129±

52BS

k,Co

olsemiarid

986

990

7–83

1cal.

Late

Bron

zeAge

−6.53

−24

.92±

0.99

18.95±

1.04

287±

68Ca

nRo

quetaII(B

arcelona

)b

41°32′N

2°8′E

173

14.9

655

272

245±

141

Cfb,

Marine

1286

295

0−77

5cal.

Late

Bron

zeAge

−6.53

−26

.05±

0.70

20.13±

0.72

353±

40Sa

ntJaum

e(Tarrago

na)

40°33′N

0°29

′W16

516

.259

829

822

2±

75Cs

a,Pu

remed

it.

1061

565

0−57

5Ea

rlyIron

Age

−6.50

−26

.55±

0.23

20.65±

0.24

383±

13Olèrdola(B

arcelona

)b

41°18′N

1°42

′E28

914

.566

028

823

8±

134

Cfb,

Marine

660

062

5–57

5Ea

rlyIron

Age

−6.50

−24

.96±

0.39

19.00±

0.42

290±

24Molíd

'Esp

ígol

(Lleida)

b41

°42′N

1°4′E

305

13.4

406

158

129±

56BS

k,Co

olsemiarid

1060

062

5–57

5Ea

rlyIron

Age

−6.50

−24

.14±

0.34

18.12±

0.36

241±

20Tu

róde

laFo

ntde

laCa

nya

(Barcelona

)b

41°23′N

1°46

′E21

415

.363

329

622

5±

127

Cfb,

Marine

2060

062

5–57

5Ea

rlyIron

Age

−6.50

−24

.81±

0.97

18.90±

0.99

285±

55

Altde

Benimàq

uia(A

lican

te)

38°49′N

0°4′E

225

17.3

727

376

407±

157

Csa,

Pure

med

it.

1460

062

5–57

5Ea

rlyIron

Age

−6.50

−25

.78±

0.44

19.84±

0.45

337±

25Lo

sVillares

(Valen

cia)

39°33′N

1°17

′W80

013

.346

118

818

0±

55BS

k,Co

olsemiarid

1257

560

0–55

0Ea

rlyIron

Age

−6.49

−23

.97±

1.02

17.95±

1.07

232±

60Mas

Castellar(G

iron

a)b

41°12′N

2°54

′E117

14.4

769

266

349±

125

Cfb,

Marine

1450

060

0–40

0Late

Iron

Age

−6.47

−25

.27±

0.82

19.32±

0.86

308±

48Pu

igde

Sant

And

reu(G

iron

a)b

42°00′N

3°5′E

2215

.469

430

636

2±

153

Cfb,

Marine

1448

552

5–45

0Late

Iron

Age

−6.47

−25

.25±

0.85

19.28±

0.88

306±

50ElsVila

rs(Lleida)

b41

°34′N

0°57

′E30

214

.239

815

915

6±

64BS

k,Co

olsemiarid

1948

558

0–39

0Late

Iron

Age

−6.42

−24

.60±

0.59

18.62±

0.62

269±

35

Lat.,

latitude

;Lon

g.,lon

gitude

;Alt.,altitude

a.s.l.;

T an,m

eanan

nual

tempe

rature;P

an,P

aut,an

nual

andSe

ptem

ber–Decem

berprecipitation(D

igital

Clim

atic

Atlas);

P aut(199

5–20

05),Se

ptem

ber–Decem

berprecipitation(closest

meteo

rological

station)

;Ca

lend

arye

ar,m

eanva

lues

andda

ting

interval

(2σfor14Ccalib

rated–cal.BC

E–da

tes;

estimated

rang

efortherest

ofsites)

forea

chof

thesets

ofsamples

stud

ied;

BCE,

before

curren

tera;

N,n

umbe

rof

samples;δ1

3C a

ir,δ

13Cof

atmosph

eric

CO2;δ1

3C,

δ13Cva

lues

ofwoo

dch

arco

al;Δ13C,

carbon

isotop

ediscriminationof

woo

dch

arco

al;Estimated

P aut,estimates

ofpa

stSe

ptem

ber–Decem

berprecipitationcalculated

from

charco

alΔ13C.

aKöp

penan

dGeige

r(193

6).

bSa

mplingsitesforreferenc

ewoo

dco

res.



first explored by fitting simple linear regression models. First-orderpolynomial exponentials were also fitted to the data to account for apossible trend towards Δ13Cw saturation as precipitation increased(Warren et al., 2001; Ferrio et al., 2003). Based on this information, thefamily of models (either linear or exponential) that overall bestpredicted monthly and annual precipitation from Δ13Cw was identi-fied and the best monthly precipitation model was retained accordingto coefficient of determination (R2) values. This model served asstarting point for a two-step process to identify a particularcombination of consecutive months for which realistic inferences ofprecipitation could be drawn from Δ13Cw data. Firstly, differentcombinations of consecutive months were tested starting with atwo-month combination that included the best monthly precipitationmodel, following by a three-month combination that included thebest two-month precipitation model based also on R2 values, and soforth. The procedure stopped when the combination of consecutivemonths included all months of the year. Secondly, a leave-one-outcross-validation process was applied to each of the previouslydetected month combinations using as validation data only thosesampling sites (7) located close to archaeological sites. In this case, avalidation point was dropped from the data and the remaining pointswere used to fit themodel, whichwas then used to predict the value atthe dropped point and, consequently, the cross-validated residualsand root mean square error (RMSE). The cross-validated RMSEprovided a measure of the predictive capability of each model but,in order to compare RMSEs having different order of magnitude, theRMSEs were normalized to the range of the validation dataset. Thebest combination of consecutive months was chosen as the one thatsimultaneously displayed a good model fitting (high R2) and a lownormalized RMSE (relative to the rest of models). This model wassubsequently applied to Δ13Cw values of archaeological charcoal inorder to infer spatial patterns in past precipitation.

2.6. Development of precipitation contour maps

A contour map of past precipitation was generated for theSeptember–December period, which was the most informativemonth combination according to the model fitted for currentΔ13Cw data (Section 2.5). The map stemmed from 11 precipitationestimates corresponding to the complete set of archaeological sites.Firstly, gridded interpolations (spatial resolution of 200 m) weregenerated using the GIS program MiraMon (Pons, 2004) implemen-ted with a spline methodology (parameters: tension=400,smooth=0) that did not take into account geographic factorsother than the location of sites, as described in Ninyerola et al.(2007). Secondly, as altitude is strongly correlated with precipitationin the modern climate, we applied a particular correction for thisgeographic element consisting of an 8% increment per 100 mincrease (Gandullo, 1994), assuming that the present-time relation-ship applies to past climate. This was done wherever the altitude ofeach gridded point (obtained by the interpolation of contour linesdigitized from 1:200,000 topographic maps; Pons, 2004) differedfrom a map of interpolated altitudes based on the set of archae-ological sites used as reference points. For purposes of smoothcontouring, the final precipitation map was created with a spatialresolution of 6 km. To obtain precipitation estimates at thisresolution, a precipitation range (made at intervals of 25 mm) wasassigned to each smaller (200×200 m) grid cell, and the mode ofprecipitation ranges was obtained for the set of smaller grid cellsincluded in each 6×6 km grid. In order to delimit the area for whichthe precipitation inferences were likely valid, we used a radius ofhalf the maximum distance between two contiguous archaeologicalsites (98.5 km) as geographic threshold for precipitation estimation.It should be bear in mind that the map displayed precipitation valuesbetween 175 and 400 mm, but also higher than 400 mm. The lattervalues should therefore be interpreted cautiously as they were

interpolated beyond the range of autumn precipitation values foundin the field survey of Q. ilex (Section 2.1).

A map of current autumn precipitation (September–Decemberperiod) was obtained from the Digital Climatic Atlas of the IberianPeninsula (Section 2.3), which was smooth-contoured to 6×6 kmgrids as well. We also generated a map of differences between pastand present conditions, expressed as percentages to current pre-cipitation, to compare precipitation trends. Those regions where pastprecipitation estimates fell outside the range 175–400 mm weremasked in the map. In addition, an alternative map of differencesbetween estimated (i.e. based on Δ13Cw from reference samples) andrecorded current precipitation was created to assess the predictiveability of the palaeoenvironmental reconstruction over wider spatialareas. For this purpose, the samemethodology described above for thedevelopment of the past precipitation map was undertaken.

3. Results and discussion

3.1. Δ13C signal strength in extant wood and fossil charcoal

SNR values forΔ13C of extant wood and fossil charcoal were 5.1 and10.5, respectively, while their corresponding EPS values were 0.84 and0.91. These results indicate a strong coherence of site-to-site variationsofΔ13C records across the region covered by the samples. In particular,signal strength was two-fold higher for the case of fossil charcoal assuggested by SNR, which agrees with the effect of the number ofextant (N=5) and fossil samples (average N=11.7) on Eq. (3).Remarkably, the EPS statistic for extant and fossil wood took similarvalues to those recently reported for δ13C chronologies (e.g. 0.80–0.90,Gagen et al., 2004; 0.83, Tardif et al., 2008; 0.90, Kirdyanov et al.,2008), which suggests that temporal and spatial variations in externalfactors (e.g. climate forcing) may have a similar impact on carbonisotope ratios of tree-rings. In particular, the extent by whichbetween-site differences in precipitation variability are responsiblefor the magnitude of the signal strength recorded in extant samples isassessed in the next section.

3.2. Δ13C response to annual, monthly and seasonal precipitation in Q. ilex

Δ13Cw varied among modern sites from 16.8‰ to 19.9‰, with amean value of 18.1‰, being significantly related to annual precipita-tion using either a linear or first-order exponential model (Fig. 2).Earlier studies have shown that Δ13C is more sensitive to wateravailability under water-limited conditions, becoming progressivelyless sensitive to changes in precipitation when water is available inexcess, i.e. when precipitation exceeds evaporation (Korol et al., 1999;Warren et al., 2001). Under such conditions, other environmentalfactors (e.g. temperature or solar radiation) emerge as main sources of

Fig. 2. Coefficients of determination (R2) for the relationships between monthly orannual precipitation and Δ13C of extant Quercus ilex wood fitted using either linear orfirst-order polynomial exponential models. The dotted line indicates the thresholdvalue for significant regressions.



variation for Δ13C (Warren et al., 2001). In our case, however, thebetter fit of both annual and monthly linear models as compared withfirst-order exponentials (Fig. 2) indicates that the sensitivity ofΔ13C toprecipitation changes was constant over the precipitation range and,for example, alternative variables such as mean temperature were notsignificantly related to Δ13Cw changes (results not shown).

Most regression models relating monthly precipitation to Δ13Cwwere also significant, in particular for the August–April period and,especially, for October andMarch (Fig. 2). Becausemost of thewesternMediterranean is characterised by the existence of two annual peaksof precipitation (autumn and late winter–spring) (Le Houerou, 2004),this finding suggests that Q. ilex, a deep-rooted tree, relies mostly ongroundwater to thrive during drought episodes. Consequently, Δ13Cwof Q. ilex is mostly sensitive to wet season precipitation (i.e. whenwater table is loaded).

The best monthly linear model (October, R2=0.65) (Fig. 2) wasused as starting point for testing different combinations of con-secutive months in the prediction of seasonal precipitation fromΔ13Cw. Two local maxima were detected in the course of changes inthe coefficient of determination (R2) as extra months were addedprogressively to the process. These corresponded to the September–December (R2=0.57) and August–March (R2=0.55) periods, respec-tively (Fig. 3a). In turn, the leave-one-out cross-validation procedureidentified two local minima (August–December and August–Marchperiods) for the normalized RMSE value of each month combination(Fig. 3b). Based on both parameters (R2 and normalized RMSE), theSeptember–December combination (autumn precipitation or Paut)was chosen as best seasonal precipitationmodel (i.e. with presumablybest predictive ability over the region surveyed forΔ13Cw) (Fig. 4). Thenon-normalized RMSE for this model, useful as indicator of theaccuracy of individual predictions from Δ13Cw, was 26.5 mm.

3.3. Estimation of past autumn precipitation and its variability fromcharcoal Δ13C of Q. ilex/coccifera

The model from Fig. 4 was subsequently applied to Δ13Cw ofarchaeological charcoal in order to infer past spatial patterns in wateravailability. Estimated autumn precipitation (Paut) ranged from 123 to432 mm, with an average value (weighted by charcoal number perarchaeological site) of 296 mm (Table 2). This amount was about 19%higher than present-time values averaged across archaeological sites(249 mm, Digital Climatic Atlas of the Iberian Peninsula).

For the subset of sites with available reference material of Q. ilex,the estimated average Paut in the past (293 mm) was also higher thanthe average Paut value obtained either from present-day Δ13Cw

(251 mm) or from current meteorological records (249 mm, DigitalClimatic Atlas of the Iberian Peninsula). In an earlier study aimed at

reconstructing past changes in annual precipitation in the Mid-Ebrobasin (NE Spain) (Ferrio et al., 2006), the period 900–300 BCEappeared as an important arid phase, based on Δ13C from Aleppo pine(Pinus halepensis Mill) charcoal. In this case, the estimated averageprecipitation was about 24% higher than present-time records, anamount that matches well the current 19% increase in Paut obtainedfor a larger area of eastern Spain embracing the Mid-Ebro depression.Current results add to a growing body of literature based on eitherΔ13C evidence (Ferrio et al., 2006; Ferrio et al., 2007; Voltas et al.,2008) or alternative palaeoclimatic proxies (Riera, 1994; Creus et al.,1996; Barriendos and Martín-Vide, 1998) indicating that presentconditions in this region and, in general, in inner Iberian Peninsula(e.g. eastern and South-eastern Spain) (Puigdefábregas, 1992; Arauset al., 1997; Ferrio et al., 2005; Aguilera et al., 2008) are the result of arecent shift towards a harsher climate (i.e. about two centuries ago),most likely enhanced by anthropogenic disturbances.

A rough approximation of temporal variability in past Paut for theIron Age Cold Epoch is provided by the standard deviation (SD) valuesof site estimates (Table 2). Because each individual charcoal integratesabout 15 (±8) tree rings, SD can be regarded as an assessment of Pautvariation for random 15-year sequences within the dating interval foreach site (approximately 100 years on average; Table 2). Thecorresponding intra-site SD values were quite uniform and averaged39.9 mm. They can be evaluated against present-day inter-annual Pautvariability obtained from meteorological records (1995–05 period)kept by stations close to each archaeological site (average SD=103.5;

Fig. 3. a) Coefficients of determination of the linear regression between accumulated precipitation for different combinations of consecutive months and Δ13C of extant Quercus ilexwood. Vertical arrows depict local maxima of the spline function defined by R2 values. b) Normalized root mean square errors (RMSE) of a leave-one-out cross-validation procedureusing as validation data a subset of sampling sites (7) located close to the archaeological sites. The procedure tested linear regression models between accumulated precipitation fordifferent combinations of consecutive months and Δ13C of Quercus ilex wood. Encircled points depict local minima of the spline function defined by RMSE values.

Fig. 4. Relationship between Δ13C of extant Quercus ilexwood and autumn precipitation(September–December period).



Table 2). Indeed, most values of present inter-annual variability werehigher than past estimates, although this does not imply that pastvariability in Paut was reduced proportionally to the decrease in SDmean values, because different time-scales are compared. However, itstresses that estimates of past Paut (based on different charcoalfragments) show a low intra-site variability over the studied region,which further supports the consistency of charcoal δ13C as palaeoen-vironmental proxy.

Although water availability is by far the most relevant elementdetermining δ13C in Q. ilex (Ferrio et al., 2003), some potential sourcesof error in estimating past precipitation from δ13C should be borne inmind. For example, cooler temperatures during the Iron Age ColdEpoch might have influenced the growing cycle, thus reducing theeffective water uptake in the past, regardless of the precipitationregime. Also, site parameters such as aspect, soil type or stand history,among others, may add additional noise to the climatic signal presentin δ13C, which can be either enhanced or decreased. In this study,however, these effects (which are extensively discussed in Ferrio et al.(2006)) were likely small as inferences for each archaeological sitewere based on a large number of charcoal fragments that may havebuffered the impact of site factors on δ13C.

3.4. Reliability of precipitation maps based on isotope networks

The ability to reconstruct spatial distribution of precipitation fromΔ13Cw values of a limited number of sites was tested by comparing amap based on meteorological data (i.e. with the high spatialresolution provided by meteorological networks) with a map ofinferred precipitation created by applying the model from Fig. 4 toextant wood samples (Fig. 5). Estimation errors (predicted minusactual values, expressed in percentage) remained below ±30% overmost of the territory, but exceeded ±50% and even ±70% for certainareas. The biggest disagreements were found in areas either isolatedfrom collection sites by important geographic barriers or subjected toother climatic influences. The first was the case of the coast line inCatalonia (NW Spain), isolated from collection sites by a mountainrange that traps a sizeable amount of precipitation, thus causing anunderestimation of coastal precipitation when predicted from innersites. In addition, the strong Atlantic influence of the mountain rangesnear the Cantabric coast (upper left corner, Fig. 5) cannot beadequately predicted from our collection sites because Q. ilex isreplaced by deciduous oaks in these areas. Overall, however, Δ13Cw-based estimates of precipitation still retained an acceptable predictivepower when interpolated over wider areas, although such powerstrongly decreased in areas with geo-climatic features distinct fromthose included in the network of sites. In this regard, the area for

effective spatial interpolations was defined based solely on horizontaldistances, but more complex algorithms, e.g. considering elevationbarriers, could delimit more precise geographic limits of inference forprecipitation. This would be useful to define potential “spatial gaps” inthe network of archaeological sites.

3.5. Spatial patterns of autumn precipitation during the Iron Age ColdEpoch

Estimates of past Paut ranged from 232 to 383 mm across sitemeans (Table 2). In comparison, the range of current Paut (138 to376 mm) was considerably wider for the same set of sites, althoughthe highest value in this range was similar to that observed for pastPaut. A simple correlation, performed between past and current Pautvalues to assess (roughly) whether inter-site variation followed asimilar geographic pattern in past and present times, was positive andsignificant (r=+0.663, P=0.019). Altogether, these results suggestthat the driest (and also coldest) areas at present (i.e. sites with coolsemiarid climate, Table 2) were characterized by more humidconditions during the Iron Age Cold Epoch, whereas current wetterareas (i.e. sites with marine and pure Mediterranean climates) barelydiffered in Paut as compared with third millennium BP conditions.

The contour precipitation maps helped to reveal past geographicpatterns of rainfall distribution and to compare past and presentclimate conditions more clearly. Overall the map of past precipitation(Fig. 6a) suggests a quite different situation to that of present-times(Fig. 1) in which past Paut tended to be considerably higher in theinner areas of eastern Spain (up to 50–100% greater) and similar alongthe coastal strip of theMediterranean Sea. Such a pattern can be betterobserved in a map of differences between past and present conditions(Fig. 6b). The differences seem particularly striking in the mid-Ebroriver basin (which runs in northwest–southeast direction south of thecentral Pyrenees), a region of great thermal contrasts and semiaridclimate.

A significant exception to the aforementioned geographic patternsin past Paut refers to the coastal area of the gulf of Valencia (easterncoast of Spain). In this area, autumn precipitation amounts during theIron Age Cold Epoch were similar or even drier (up to ca. 50% lowerPaut, see Fig. 6b) than today. This can be attributed to a differentialresponse of southern and northern Spain to that cooling event, asthese two areas often respond to different climate forces (see e.g. Rodóet al., 1997). Similarly, Magny et al. (2003) reported a contrastinghydrological response to the 8.2 ka cooling event, with northern Spainresembling central Europe (cool and wet) and southern Spaindisplaying a cool but dry period. Remarkably, the interphase betweenthe two climatic scenarios derived from our study coincides with the

Fig. 5. Contour map of differences (%) between estimated (based on Δ13Cw) and instrumental (adapted from the Digital Climatic Atlas of the Iberian Peninsula; Ninyerola et al., 2005)current autumn precipitation (September–December period). The points depict sampling sites of the field survey of Quercus ilex carried out in NE Spain.



southern limit proposed by Magny et al. (2003) for the mid-europeanwetter zone (around 38–40° N) during Holocene cooling phases.Although the temporal range covered by the sites (ca. 500 years) isrelatively wide, the observed North–South differences do not seem tobe an artifact due to different dating of the sites. Indeed, suchdifferences are still evident when comparing the six almostsynchronous sites covering the central period between ca. 650 and550 BCE (Table 2): whereas the two southern sites (b40°N) showedvalues of Paut similar to present (only 6% higher on average), the fournorthern sites (N40°N) showed a mean Paut value 20% higher thanpresent. The presence of an arid phase in southern Spain and northAfrica during this period has been confirmed by a variety of proxies(e.g. Jalut et al., 2000; Carrión, 2002), being proposed as the startingpoint for the onset of current Mediterranean climate in this area(Terral and Mengual, 1999). In this regard, a more intensive sampling,particularly in archaeological sites from southern Spain, would help tocorroborate these findings and to definemore precisely the position ofthe North–South interface. In addition, the simultaneous analysis ofother plant species more sensitive to summer precipitation wouldallow reconstruction of a complete portrait of total precipitation andits distribution (Ferrio et al., 2007). Fortunately, the relativeabundance of charcoal remains in the archaeological record for theBronze and Iron Age periods in eastern Spain, along with the risinginterest in the recovery of plant fossil remains of the Mediterraneanregion (Helbaek and Schultze, 1981), makes this task relativelystraightforward.

4. Conclusions

The present study indicates that δ13C of Q. ilex wood isparticularly sensitive to autumn–winter rainfall, providing the bestlinear fit for accumulated precipitation from September to December.Also, the results attained so far point to a differential climaticresponse to the Iron Age Cold Epoch in eastern Spain followingsouth–north and west–east precipitation gradients, with cool,semiarid areas at present characterized by more humid past

conditions, and current wetter areas barely differing in autumnprecipitation as compared with third millennium BP conditions. Webelieve that the development of δ13C-based spatially continuousmaps of water-related ecophysiological or environmental variables,which are relatively scarce compared to other isoscape applications(e.g. Cook et al., 1999; Leavitt et al., 2007), opens new prospects inpalaeoenvironmental reconstruction studies through the use ofarchaeobotanical remains.

Acknowledgements

This study was partially supported by the project DGI CGL2005-08175-C02-02/BOS. We thank N. Alonso, R. Buxó, Y. Carrión, E. Grau,D. López and J. Picazo for generously providing charcoal samples andR. Piqué for performing taxonomical determinations of fossil woodremains. M. Aguilera has a PhD fellowship from the Spanish Ministryof Science and Innovation and J.P. Ferrio has been granted by the post-doctoral fellowships Beatriu de Pinós (Generalitat de Catalunya) andMarie Curie Intra-european Fellowship (6th Framework Programm,EU). We also thank the comments from two anonymous referees,which helped us to improve significantly a former version of themanuscript.

References

Aguilera, M., Araus, J.L., Voltas, J., Rodríguez-Ariza, M.O., Molina, F., Rovira, N., Buxó, R.,Ferrio, J.P., 2008. Stable carbon and nitrogen isotopes and quality traits of fossilcereal grains provide clues on sustainability at the beginnings of Mediterraneanagriculture. Rapid Communications in Mass Spectrometry 22 (11), 1653–1663.

Alonso, N., 1999. From seed to flour: protohistoric agricultural processes in westernCatalonia. UMR 154-CNRS, Lattes (in Catalan).

Anderson, W.T., Bernasconi, S.M., McKenzie, J.A., Saurer, M., 1998. Oxygen and carbonisotopic record of climatic variability in tree ring cellulose (Picea abies): anexample from central Switzerland (1913–1995). Journal of Geophysical Research103, 31625–31636.

Araus, J.L., Buxó, R., Febrero, A., Camalich, M.D., Martín, D., Molina, F., Rodríguez-Ariza,M.O., Voltas, J., 1997. Identification of ancient irrigation practices based on thecarbon isotope discrimination of plant seeds: a case study from the south-eastIberian Peninsula. Journal of Archaeological Science 24 (8), 729–740.

Fig. 6. a) Contour map of estimated past autumn precipitation (September–December period), as inferred from Δ13C values of charcoal remains recovered from archaeological sites.b) Contour map of differences between past and present autumn precipitation (September–December period) expressed as relative changes (%) respect present climate data. Whiteareas correspond to differences (%) obtained from out-of-range past precipitation estimates (i.e. higher than 400mm). In both maps, the geographic limits for suitable inferences areindicated.



Barriendos, M., Martín-Vide, J., 1998. Secular climatic oscillations as indicated bycatastrophic floods in the Spanish Mediterranean coastal area (14th–19thcenturies). Climatic Change 38 (4), 473–491.

Berry, S.C., Varney, G.T., Flanagan, L.B., 1997. Leaf δ13C in Pinus resinosa trees andunderstory plants: variation associated with light and CO2 gradients. Oecologia 109(4), 499–506.

Carrión, J.S., 2002. Patterns and processes of Late Quaternary environmental change in amontane region of southwestern Europe. Quaternary Science Reviews 21 (18–19),2047–2066.

Carrión, J.S., Van Geel, B., 1999. Fine-resolution upper Weichselian and Holocenepalynological record fromNavarrés (Valencia, Spain) and a discussion about factorsof Mediterranean forest succession. Review of Palaeobotany and Palynology 106,209–236.

Cook, E.R., Meko, D.M., Stahle, D.W., Cleaveland, M.K., 1999. Drought reconstructions forthe continental United States. Journal of Climate 12 (4), 1145–1162.

Creus, J., Fernández-Cancio, A., Manrique-Menéndez, E., 1996. Evolución de latemperatura y precipitación anuales desde el año 1400 en el sector central de laDepresión del Ebro. Lucas Mallada 8, 9–27.

Czimczik, C.I., Preston, C.M., Schmidt, M.W.I., Werner, R.A., Schulze, E.D., 2002. Effects ofcharring on mass, organic carbon, and stable carbon isotope composition of wood.Organic Geochemistry 33, 1207–1223.

Davis, B., 1994. Palaeolimnology and Holocene environmental change from endorreiclakes in the Ebro Basin, N.E. Spain. PhD thesis, University of Newcastle upon Tyne.

DeNiro, M.J., Hastorf, C.A., 1985. Alteration of 15N/14N and 13C/12C ratios of plant matterduring the initial stages of diagenesis: studies utilizing archaeological specimensfrom Peru. Geochimica et Cosmochimica Acta 49 (1), 97–115.

Elsik, C.G., Flagler, R.B., Boutton, T.W., Ehleringer, J.R., Hall, A.E., Farquhar, G.D., 1993.Carbon isotope composition and gas exchange of loblolly and shortleaf pine asaffected by ozone and water stress. In: Ehleringer, J.R., Hall, A.E., Farquhar, G.D.(Eds.), Stable Isotopes and Plant Carbon–Water Relations. Academic Press, London.

EuroCLIMATE Programme, 2007. European Science Foundation. Available on-line(http://www.esf.org/activities/eurocores/programmes/euroclimate.html).

Farquhar, G.D., O'leary, M.H., Berry, J.A., 1982. On the relationship between carbonisotope discrimination and the inter-cellular carbon-dioxide concentration inleaves. Australian Journal of Plant Physiology 9 (2), 121–137.

February, E., Van der Merwe, N.J., 1992. Stable carbon isotope ratios of wood charcoalduring the past 4000 years— anthropogenic and climatic influences. South AfricanJournal of Science 88 (5), 291–292.

Fernández, S., Fuentes, N., Carrión, J.S., González-Sampériz, P., Montoya, E., Gil, G., Vega-Toscano, G., Riquelme, J.A., 2007. The Holocene and Upper Pleistocene pollensequence of Carihuela Cave, southern Spain. Geobios 40, 75–90.

Ferrio, J.P., Voltas, J., 2005. Carbon and oxygen isotope ratios in wood constituents ofPinus halepensis as indicators of precipitation, temperature and vapour pressuredeficit. Tellus Series B-Chemical and Physical Meteorology 57 (2), 164–173.

Ferrio, J.P., Florit, A., Vega, A., Serrano, L., Voltas, J., 2003.Δ13C and tree-ring width reflectdifferent drought responses in Quercus ilex and Pinus halepensis. Oecologia 137 (4),512–518.

Ferrio, J.P., Araus, J.L., Buxó, R., Voltas, J., Bort, J., 2005. Water management practices andclimate in ancient agriculture: inferences from the stable isotope composition ofarchaeobotanical remains. Vegetation History and Archaeobotany 14 (4), 510–517.

Ferrio, J.P., Alonso, N., Lopez, J.B., Araus, J.L., Voltas, J., 2006. Carbon isotope compositionof fossil charcoal reveals aridity changes in the NW Mediterranean Basin. GlobalChange Biology 12 (7), 1253–1266.

Ferrio, J.P., Voltas, J., Alonso, N., Araus, J.L., 2007. Reconstruction of climate and cropconditions in the past based on the carbon isotope signature of archaeobotanicalremains. In: Dawson, T.D., Siegwolf, R. (Eds.), Stable Isotopes as Indicators ofEcological Change. Academic Press, London.

Gandullo, J.M., 1994. Climatology and soil science. Fundación Conde del Valle de Salazar.E.T.S.I. Montes, Madrid (in Spanish).

Gagen, M., McCarroll, D., Edouard, J.L., 2004. Latewood width, maximum density, andstable carbon isotope ratios of pine as climate indicators in a dry subalpineenvironment, in the French Alps. Arctic, Antarctic, and Alpine Research 36 (2),166–171.

Gutiérrez-Elorza, M., Peña-Monné, J.L., 1998. Geomorphology and late Holoceneclimatic change in northeastern Spain. Geomorphology 23 (2–4), 205–217.

Helbaek, H., Schultze, M.J., 1981. The scientific work of Hans Helbaek. Kulturpflanze 29.IPCC, 2001. Climate Change 2001: the Scientific Basis. Cambridge University Press,

Cambridge.Jalut, G., Amat, A.E., Bonnet, L., Gauquelin, T., Fontugne, M., 2000. Holocene climatic

changes in theWesternMediterranean, from south-east France to south-east Spain.Palaeogeography Palaeoclimatology Palaeoecology 160 (3–4), 255–290.

Jiménez, P., Díaz-Fernández, P., Iglesias, S., De Tuero, M., Gil, L., 1996. Provenance regionsof Quercus ilex L. in Spain. ICONA, Madrid. (in Spanish).

Jones, T.P., Chaloner, W.G., 1991. Fossil charcoal, its recognition and palaeoatmosphericsignificance. Palaeogeography, Palaeoclimatology, Palaeoecology 97, 39–50.

Kirdyanov, A.V., Treydte, K.S., Nikolaev, A., Helle, G., Schleser, G.H., 2008. Climate signalsin tree-ring width, density and δ13C from larches in Eastern Siberia (Russia).Chemical Geology 252, 31–41.

Korol, R.L., Kirschbaum, M.U.F., Farquhar, G.D., Jeffreys, M., 1999. Effects of water statusand soil fertility on the C-isotope signature in Pinus radiata. Tree Physiology 19 (9),551–562.

Köppen, W., Geiger, R., 1936. Handbook of climatology. Gebrüder Bornträger, Berlin. (inGerman).

Le Houerou, H.N., 2004. An agro-bioclimatic classification of arid and semiarid lands inthe isoclimatic Mediterranean zones. Arid Land Research and Management 18 (4),301–346.

Leavitt, S.W., 1993. Seasonal C-13/C-12 changes in tree rings — species and sitecoherence, and a possible drought influence. Canadian Journal of Forest Research 23(2), 210–218.

Leavitt, S.W., Chase, T.N., Rajagopalan, B., Lee, E., Lawrence, P.J., Woodhouse, C.A., 2007.Southwestern U.S. drought maps from pinyon tree-ring carbon isotopes. Eos 88 (4),39–40.

Magny, M., Richard, H., 1992. Essai de synthèse vers une courbe de l'évolution du climatentre 500 BC et 500 AD. Les nouvelles de l'archéologie 50, 58–60.

Magny, M., Bégeot, C., Guiot, J., Peyron, O., 2003. Contrasting patterns of hydrologicalchanges in Europe in response to Holocene climate cooling phases. QuaternaryScience Reviews 22, 1589–1596.

Martín, S., Díaz-Fernández, P., DeMiguel, J., 1998. Provenance regions of Spanish foresttree species. O.A. Parques Nacionales, Madrid. (in Spanish).

McCarroll, D., Loader, N.J., 2004. Stable isotopes in tree rings. Quaternary ScienceReviews 23 (7–8), 771–801.

Ninyerola, M., Pons, X., Roure, J.M., 2005. Digital climatic atlas of the Iberian Peninsula.Methodology and applications in bioclimatology and geobotanics. UniversidadAutónoma de Barcelona, Barcelona. (in Spanish).

Ninyerola, M., Pons, X., Roure, J.M., 2007. Monthly precipitation mapping of the IberianPeninsula using spatial interpolation tools implemented in a Geographic Informa-tion System. Theoretical and Applied Climatology 89 (3–4), 195–209.

Pons, X., 2004. MiraMon. Geographic Information System and Remote Sensing software.Centre de Recerca Ecològica i Aplicacions Forestals, Barcelona.

Puigdefábregas, J., 1992. Mitos y perspectivas sobre la desertificación. Erosión ydesertificación en España. Ecosistemas 3, 18–22.

Riera, S., 1994. Holocene landscape evolution in Barcelona from pollen data. PhD. thesis,University of Barcelona (in Spanish).

Rodó, X., Baert, E., Comin, F.A., 1997. Variations in seasonal rainfall in southern Europeduring the present century: relationships with the North Atlantic Oscillation andthe El Niño Southern Oscillation. Climate Dynamics 13 (4), 275–284.

Rossi, S., Anfodillo, T., Menardi, R., 2006. Trephor: a new tool for sampling microcoresfrom tree stems. Iawa Journal 27 (1), 89–97.

Stuiver, M., Reimer, P.J., Bard, E., Beck, J.W., Burr, G.S., Hughen, K.A., Kromer, B.,McCormac, G., Van der Plicht, J., Spurk, M., 1998. INTCAL98 radiocarbon agecalibration, 24,000-0 cal BP. Radiocarbon 40 (3), 1041–1083.

Tardif, J.C., Conciatori, F., Leavitt, S.W., 2008. Tree rings, δ13C and climate in Picea glaucagrowing near Churchill, subarctic Manitoba, Canada. Chemical Geology 252,88–101.

Terral, J.F., Mengual, X., 1999. Reconstruction of Holocene climate in southern Franceand eastern Spain using quantitative anatomy of olive wood and archaeologicalcharcoal. Palaeogeography Palaeoclimatology Palaeoecology 153 (1–4), 71–92.

Van Geel, B., Buurman, H., Waterbolk, H.T., 1996. Archaeological and palaeoecologicalindications of an abrupt climate change in The Netherlands, and evidence forclimatological teleconnections around 2650 BP. Journal of Quaternary Science 11(6), 451–460.

Vernet, J.L., Pachiaudi, C., Bazile, F., Durand, A., Fabre, L., Heinz, C., Solari, M.E., Thiebault,S., 1996. Le δ13C de charbons de bois préhistoriques et historiques méditerranéens,de 35000 BP a l'àctuel. Premiers resultats. Comptes Rendus de l'Academie desSciences, série IIa, 323 (4), 319–324.

Voltas, L., Ferrio, J.P., Alonso, N., Araus, J.L., 2008. Stable carbon isotopes inarchaeobotanical remains and palaeoclimate. Contributions to Science 4 (1), 21–31.

Warren, C.R., McGrath, J.F., Adams, M.A., 2001. Water availability and carbon isotopediscrimination in conifers. Oecologia 127 (4), 476–486.

Wigley, T.M.L., Briffa, K.R., Jones, P.D., 1984. On the average value of correlated timeseries, with applications in dendroclimatology and hydrometeorology. Journal ofClimate and Applied Meteorology 23, 201–213.


A map of autumn precipitation for the third millennium BP in the Eastern Iberian Peninsula from charcoal carbon isotopes

Documents