Non-linear regime shifts in Holocene Asian monsoon variability

33 Pages·2015·8.13 MB·English

Clim.Past,11,709–741,2015
www.clim-past.net/11/709/2015/
doi:10.5194/cp-11-709-2015
©Author(s)2015.CCAttribution3.0License.
Non-linear regime shifts in Holocene Asian monsoon variability:
potential impacts on cultural change and migratory patterns
J.F.Donges1,2,R.V.Donner1,3,N.Marwan1,S.F.M.Breitenbach4,*,K.Rehfeld1,5,andJ.Kurths1,6,7
1PotsdamInstituteforClimateImpactResearch,TelegrafenbergA31,14473Potsdam,Germany
2StockholmResilienceCentre,StockholmUniversity,Kräftriket2B,11419Stockholm,Sweden
3DepartmentofBiogeochemicalIntegration,MaxPlanckInstituteforBiogeochemistry,Hans-Knöll-Str.10,
07745Jena,Germany
4InstituteforGeology,MineralogyandGeophysics,RuhrUniversityBochum,Universitätsstr.150,44801Bochum,Germany
5AlfredWegenerInstituteforPolarandMarineResearch,TelegrafenbergA43,14473Potsdam,Germany
6DepartmentofPhysics,HumboldtUniversity,Newtonstr.15,12489Berlin,Germany
7InstituteforComplexSystemsandMathematicalBiology,UniversityofAberdeen,Aberdeen,AB243FX,UK
*formerlyat:GeologicalInstitute,DepartmentofEarthSciences,ETHZurich,8092Zurich,Switzerland
Correspondenceto:J.F.Donges([email protected])
Received:10February2014–PublishedinClim.PastDiscuss.:6March2014
Revised:1March2015–Accepted:2March2015–Published:7May2015
Abstract. The Asian monsoon system is an important tip- changes in mean monsoon intensity and other climatic pa-
ping element in Earth’s climate with a large impact on hu- rameters, regime shifts in monsoon complexity might have
man societies in the past and present. In light of the poten- playedanimportantroleasdriversofmigration,pronounced
tiallysevereimpactsofpresentandfutureanthropogeniccli- culturalchanges,andthecollapseofancienthumansocieties.
mate change on Asian hydrology, it is vital to understand
theforcingmechanismsofpastclimaticregimeshiftsinthe
Asian monsoon domain. Here we use novel recurrence net-
work analysis techniques for detecting episodes with pro- 1 Introduction
nounced non-linear changes in Holocene Asian monsoon
dynamics recorded in speleothems from caves distributed Relationships between past climate change and societal re-
throughoutthemajorbranchesoftheAsianmonsoonsystem. sponsesinthehistoricalandarchaeologicalrecordhavefre-
Anewlydevelopedmulti-proxymethodologyexplicitlycon- quently been reported, e.g. increased frequencies of war
siders dating uncertainties with the COPRA (COnstructing (Zhang et al., 2007), societal conflicts and crises (Hsiang
Proxy Records from Age models) approach and allows for etal.,2011,2013;Zhangetal.,2011),migrations(Büntgen
detection of continental-scale regime shifts in the complex- et al., 2011), and collapse of complex societies such as the
ity of monsoon dynamics. Several epochs are characterised Akkadian empire (Gibbons, 1993; Cullen et al., 2000), the
bynon-linearregimeshiftsinAsianmonsoonvariability,in- EgyptianOldKingdom(Stanleyetal.,2003),Mayanurban
cluding the periods around 8.5–7.9, 5.7–5.0, 4.1–3.7, and centres(Haugetal.,2003;Kennettetal.,2012),andChinese
3.0–2.4kaBP. The timing of these regime shifts is consis- dynasties (Yancheva et al., 2007). Those societal responses
tentwithknownepisodesofHolocenerapidclimatechange are generally acknowledged to be driven by multiple fac-
(RCC) and high-latitude Bond events. Additionally, we ob- tors and, additionally, societies differ in their vulnerability
serve a previously rarely reported non-linear regime shift to changing environmental conditions (Tainter, 1990). Nev-
around7.3kaBP,atimingthatmatchesthetypical1.0–1.5ky ertheless,investigatingclimateasonepossiblekeydriveris
return intervals of Bond events. A detailed review of previ- of great interest in the face of recent anthropogenic climate
ouslysuggestedlinksbetweenHoloceneclimaticchangesin change(Stockeretal.,2014).Deeperinsightsinthisfieldare
theAsianmonsoondomainandthearchaeologicalrecordin- urgently needed to assess theadaptive capacity and dynam-
dicatesthat,inadditiontopreviouslyconsideredlonger-term icsofcurrentsocieties(Widloketal.,2012)underglobalen-
PublishedbyCopernicusPublicationsonbehalfoftheEuropeanGeosciencesUnion.
710 J.F.Dongesetal.:Non-linearregimeshiftsinHoloceneAsianmonsoonvariability
vironmental change within the co-evolving planetary socio-
environmentalsystem(Schellnhuber,1998,1999).
In our contribution here, we focus on regime shifts in 30°N Jiuxian
Tianmen Heshang
Asian summer monsoon dynamics during the last 10ky Lianhua
and discuss their potential societal impacts such as cultural Hoti Mawmluh Dongge
change or migratory patterns. Investigating the Asian mon- 15°N Qunf EASM
soondomainisrelevantforthreereasons:(i)theAsianmon- Dimarshim BB
S
soon is a highly dynamic, vulnerable, and multistable sys- A
theams b(eZeinckifdeeldntiefitedal.a,s2a00p5o;teLnteivaelrcmliamnnatiecttiaplp.,in2g00e9le)mthenatt 0 45°E 60°E 75°E 90°E 105°E 12A0°IES M 135°E
(Lentonetal.,2008),(ii)ca.60%oftheworld’spopulation Liang Luar
are directly affected by the Asian monsoon, the failures of
Figure1.MapofsouthernAsiashowingthemainflowdirections
which have large potential consequences for food supply in
of moist air masses associated with different monsoon branches:
these regions (Wu et al., 2012), and (iii) there are multiple
Arabian Sea (AS) and Bay of Bengal (BB) branches of the In-
knownexamplesforthecollapseofearlycomplexsocieties
diansummermonsoon,EastAsiansummermonsoon(EASM),and
intheAsianmonsoonrealm,includingtheHarappanculture
Australian–Indonesiansummermonsoon(AISM).Furthermore,the
in the Indus Valley (Staubwasser and Weiss, 2006), and ex- locations of the caves where the speleothem records used in this
amples of the impact of climate change on socio-political workhavebeenobtainedfromaredisplayed(seeTable1).
developments, e.g. war frequencies or dynastic changes in
China (Zhang et al., 2007; Yancheva et al., 2007). Thus,
a deeper understanding of past changes in Asian monsoon presentstudyisarobustchronologyoftheincludedarchives,
dynamicsandtheirimpactonsocietieswillcontributetoim- a requirement that is met by many speleothem records. Fo-
provedcapacitiesforanticipatingpotentialconsequencesof cussingononetypeofarchiveandproxy,thisworkextends
futureclimatechangeintheregion. upon and complements several related re-assessment stud-
TheAsiansummermonsoonsystemisaseasonallyrecur- ies (Morrill et al., 2003; Hu et al., 2008; Maher, 2008; Re-
ring wind pattern related to the migration of the Intertrop- hfeld et al., 2013). In contrast to earlier work, we focus not
ical Convergence Zone and is active from June to October. only on the intensity of monsoon rainfall per se, but aim to
It is nominally separated into the Indian summer monsoon identifychangesinthecomplexityofmonsoonvariationsas
(ISM) and the East Asian summer monsoon (EASM). The important higher-order information contained in the avail-
ISMisdividedintotheArabianSea(AS)branchandtheBay ablerecords.Therationalebehindthisapproachisthatreg-
of Bengal (BB) branch, which transport moisture from the ular and, thus, predictable monsoon variations are crucial
IndianOceantowardstheArabianPeninsulaandtheIndian for sustained socio-economic development, while irregular
subcontinentduringthesummerwetseason(Fig.1).TheAS variationsofseasonalrainfallandclimaticinstabilitieshave
branchreachesNEAfricaandtheArabianPeninsulabefore been shown to have acted as triggers for social unrest and
turningeasttowardsthewestcoastofIndia.TheBBbranch as drivers of societal changes (Hsiang et al., 2013). There-
of the ISM receives much of its moisture from the Arabian fore,identifyingepochsofregimeshiftsinthecomplexityof
Sea,crossessouthernIndia,andreloadsovertheBayofBen- palaeoclimaticvariabilityisofgreatinterestforinvestigating
galbeforemovingnorthwarduntilitreachestheHimalayan the role of Asian summer monsoon dynamics as a potential
mountainrange.Unabletocrossthisbarrier,itsplitsintotwo driverofculturalchangeormigratorypatternsinthehuman
branches,onemovingnorth-westwardalongtheHimalayas, realm.
theotherextendingnorth-eastwardintoTibetandtherestof Fromamethodologicalpointofview,thisworkintroduces
China, where it contributes greatly to summer rainfall. The several new aspects to the study of palaeoclimate variabil-
EASMtransportsmoisturefromtheadjacentseasintoChina ity: (i) we focus on non-linear aspects of monsoon dynam-
andalsoontotheTibetanPlateau.Complexandtime-varying ics such as regime shifts in the regularity of monsoon vari-
interdependencies have been demonstrated to exist between ations, extending upon previous work on climatic regime
thedifferentbranchesoftheAsiansummermonsoonsystem shifts in linear time series properties (Mudelsee, 2000; Ro-
duringthelateHolocenebasedonpalaeoclimatedata(Feld- dionov,2004)suchasmeanmonsoonintensity.Themethod
hoff et al., 2012; Rehfeld et al., 2013) as well as during the of choice, recurrence network (RN) analysis of time series,
periodofinstrumentalobservations(Bakeretal.,2015). isparticularlyusefulfordetectingqualitativechangesinthe
Our approach is to integrate information on decadal- to dynamics of complex systems (Marwan et al., 2009; Don-
centennial-scaleAsianpalaeomonsoonvariabilityduringthe neretal.,2010b)andhasbeensuccessfullyappliedinfields
Holocenefromhigh-resolutionoxygenisotoperecordsfrom rangingfromfluiddynamicstoelectrochemistrytophysiol-
multiplecaves,asspeleothemsarerecognisedashigh-quality ogy(Donneretal.,2014).RNanalysisisspecificallysuitable
palaeoclimatearchivesfortheconsideredtimescalesandge- for studying palaeoclimate records – unlike other methods
ographical region (Table 1). Of special importance for the thereareonlyimpliciteffectsofnon-uniformsamplinginthe
Clim.Past,11,709–741,2015 www.clim-past.net/11/709/2015/
J.F.Dongesetal.:Non-linearregimeshiftsinHoloceneAsianmonsoonvariability 711
Table1.Listofspeleothemδ18Ooxygenisotoperecordsusedinthisstudyincludingcavename,speleothemID,cavelocation,numberof
datapointsN,averagesamplingtime(cid:104)(cid:49)T(cid:105),andclimatologicalinterpretation(IOM,IndianOceanmonsoon;ISM,Indiansummermonsoon;
EASM, East Asian summer monsoon; AISM, Australian–Indonesian summer monsoon. From the Mawmluh record, all data points prior
to 9kaBP were discarded, since the older part of the record was recently found to be insufficiently dated (M. Berkelhammer, personal
communication,2013).
Cavename Speleothem Latitude Longitude N (cid:104)(cid:49)T(cid:105) Interpretation Reference
(years)
1 Dimarshim,Yemen D1 12◦33(cid:48)N 53◦41(cid:48)E 530 8.3 IOMstrength Fleitmannetal.(2007)
2 Qunf,Oman Q5 17◦10(cid:48)N 54◦18(cid:48)E 1412 7.3 IOMstrength Fleitmannetal.(2003,2007)
3 Hoti,Oman H5 23◦05(cid:48)N 57◦21(cid:48)E 832 4.3 IOMstrength Neffetal.(2001)
4 Mawmluh,India KM-A 25◦16(cid:48)N 91◦53(cid:48)E 889 6.0 ISMstrength Berkelhammeretal.(2012)
5 Tianmen,Tibet,China TM-18 30◦55(cid:48)N 90◦40(cid:48)E 1005 4.9 ISMstrength Caietal.(2012)
6 Dongge,China DA 25◦17(cid:48)N 108◦5(cid:48)E 2124 4.2 EASMstrength Wangetal.(2005)
7 Lianhua,China A1 29◦29(cid:48)N 109◦32(cid:48)E 819 8.1 EASMstrength Cosfordetal.(2008)
8 Heshang,China HS4 30◦27(cid:48)N 110◦25(cid:48)E 1223 7.8 EASMstrength Huetal.(2008)
9 Jiuxian,China C996-1 33◦34(cid:48)N 109◦6(cid:48)E 828 10.5 EASMstrength Caietal.(2010)
10 LiangLuar,Indonesia LR06-B1 8◦32(cid:48)S 120◦26(cid:48)E 1289 9.7 AISMstrength Griffithsetal.(2009)
time domain and minor dating uncertainties (Donges et al., Our results on qualitative changes in monsoon dynamics as
2011a, b). (ii) We study continental-scale climatic changes recorded in speleothem records are reported in Sect. 4 and
through the integration of information from proxy records arediscussedinthecontextofpotentialimpactsoncultural
frommultiple(Table1)sitesdistributedovertheAsianmon- changeandmigratorypatternsinSect.5.Finally,conclusions
soon domain (Rehfeld et al., 2013). (iii) We explicitly con- aredrawninSect.6.
sider and propagate dating uncertainties (Mudelsee et al.,
2012; Goswami et al., 2014) in the available cave records
2 Dataandmethods
usingtheCOPRAframework(COnstructingProxyRecords
fromAgemodels,Breitenbachetal.,2012b),and(iv)weem- In this section, we explain the selection of palaeoclimate
ploystatisticaltestsfortheco-occurenceofthesignaturesof records that are suitable for studying the stability of Asian
monsoonalregimeshiftsatdifferentsites. monsoon dynamics during the Holocene within the frame-
Applying this methodology, we find that previously re- workofthepresentedanalyticalstrategy.Specifically,weare
ported high-latitude Bond events (Bond et al., 1997, 2001) interested in shifts in the dynamic regime beyond changes
andrapidclimatechange(RCC)episodes(Mayewskietal., inmeanintensityduringthelastabout10ka,whichcallfor
2004; Fleitmann et al., 2008; Wanner et al., 2008, 2011) high-resolution(i.e.sub-decadal)recordsspanningasmuch
wereoftenaccompaniedbyepochsoflarge-scalenon-linear oftheHoloceneaspossible.Subsequently,themethodology
regimeshiftsinmonsoondynamics,e.g.pronouncedchanges employed for quantitatively evaluating the effects of irreg-
in the regularity of interannual to inter-decadal monsoon ular sampling and dating uncertainties, both common prob-
variations. Furthermore, we are able to robustly identify at lems in time series analysis of palaeoclimate records, is in-
least one previously rarely reported regime shift in Asian troduced.Recurrenceanalysisisthenpresentedasourstatis-
summer monsoon dynamics during the Holocene (Wang ticaltechniqueofchoicefordetectingepochsofregular(sta-
et al., 2005), which manifests itself as an epoch of signif- bleorperiodic)andmore-erratic(andhencelesspredictable)
icantly increased regularity of monsoon variations around dynamics as well as transitions between such episodes in
7.3kaBP when considering dating uncertainties with the palaeomonsoon variability. See Fig. 2 for a summary of the
COPRA approach. Examining our results in context with a methodologicalworkflowemployedinthisstudy.
thorough review of the previously reported archaeological
record,wefindthatsomeofthedetectedepochsofnon-linear
2.1 SpeleothemrecordsoftheAsianpalaeomonsoon
regimeshiftsinAsiansummermonsoondynamicsmayhave
contributed to triggering major periods of migration, pro- Temporallywell-resolvedandpreciselyandaccuratelydated
nouncedculturalchanges,andthecollapseofancienthuman proxy records of palaeoclimate variability are indispens-
societiesinAsia. able for the study of spatially and temporally disperse
This article is structured as follows: after introducing decadal- to centennial-scale climatic episodes. Speleothems
the data and methods used (Sect. 2), an overview is pro- (secondarycavedepositssuchasstalagmites)constituteter-
vided on possible imprints of Holocene RCC episodes and restrial archives potentially covering several hundred thou-
Bond events within the Asian monsoon system (Sect. 3). sand years of environmental variability. Current U–Th dat-
ing techniques allow the establishing of robust age models
www.clim-past.net/11/709/2015/ Clim.Past,11,709–741,2015
712 J.F.Dongesetal.:Non-linearregimeshiftsinHoloceneAsianmonsoonvariability
backto600–800kaBP(Chengetal.,2013).Stalagmitesfre- The record database includes speleothem data from the
quently allow for high-resolution sampling at sub-decadal Arabian Peninsula (influenced by the AS), India, and Tibet
andevensub-annualscales(Johnsonetal.,2006).Stalagmite (mainly under the influence of the BB), and eastern China
records offer a multitude of geochemical and petrographic (with governing EASM) (Fig. 1). One additional record
proxies(e.g.oxygenandcarbonstableisotoperatios,major (Liang Luar cave) from Indonesia has also been included,
andminorelementratiosandconcentrations,andfluidinclu- as this record indicates that the climatic episodes and tran-
sionwaterisotopes;FairchildandBaker,2012),thoughoften sitionsdiscussedbelowarefoundalsointheAISMdomain.
onlystableisotoperatiosareusedtoinferchangesinrainfall TheselectedproxyrecordsarelistedinTable1togetherwith
amount,source,orintensity. information on sampling location, the number of samples,
Availablelacustrine(e.g.Pontonetal.,2012)andmarine temporalresolution,andtheirinterpretationandcorrespond-
(vonRadetal.,1999;Kudrassetal.,2001;Staubwasserand ingreferences.
Weiss,2006)sedimentrecordsfromthesouthernAsiando- Many speleothem records, especially those from China,
main – although valuable for the study of long-term trends havebeenreportedoverthelastdecadegivingdeepinsights
and millennial- to centennial-scale climate episodes – often intothehistoryoftheEASM(Wangetal.,2005,2008;Dong
lackbothsufficientlyhighsamplingratesandchronological etal.,2010)foruptoapproximatelythepast380ky(Cheng
controltoallowstatisticallysignificantcomparisonofmulti- et al., 2009). Unfortunately, only a few terrestrial records
ple reconstructions at decadal timescale. Available ice core are available so far from the Indian subcontinent that meet
records from the Himalaya and Tibet regions (Thompson our criteria (Sinha et al., 2005, 2007, 2011a, b; Berkelham-
et al., 1997, 2000, 2003) are unfortunately either too short meretal.,2010,2012).Theavailableproxyreconstructions
(coveringonlythepastfewthousandyears)orlackalsotem- fromIndianstalagmitesreflectchangesinthestrengthofthe
poralresolution.Treeringscangenerallyaddvaluableinfor- ISM,buttheyalsoshowregionalvariabilityinISMdynam-
mation for reconstructing moisture and/or temperature dy- ics.SuchdifferencesreflectthecomplexityoftheISMover
namics(Cooketal.,2010;Borgaonkaretal.,2010;Treydte India,ratherthancontradictingthegeneralunderstandingof
etal.,2006,2009).However,veryfewrecordsfromAsiaex- ISMdynamics(Breitenbachetal.,2012a).Theinfluenceof
tend beyond the last 1000 years and, thus, these cannot be theISMondifferentregionsinIndiahasbeennotedbymete-
usedtostudyqualitativechangesofmonsoondynamicsover orologicalaswellaspalaeoclimaticstudies(HoyosandWeb-
thecourseoftheHolocene,whicharethefocusofthisstudy. ster, 2007; Sinha et al., 2011b; Rehfeld et al., 2013). Moni-
Given the considerable number and reasonable spatial toringofrainwaterinNEIndiarevealsthatitsisotopiccom-
distribution of available high-resolution speleothem records position(δ18O )dependsnotsimplyontheamountofrain-
rw
from the Asian monsoon domain, and for permitting bet- fall, but also on moisture source changes such as increased
ter comparability of the results to be obtained, in this study melt water flux, the pathway length that an air mass moves
we restrict our attention to such cave archives. Specifi- from the source to the sampling site and related Rayleigh
cally,weselectpublishedspeleothemoxygenisotope(δ18O) fractionation,andchangesoftheisotopiccompositionofthe
proxy records from the Asian monsoon domain that fulfil source (Breitenbach et al., 2010). Still, a clear ISM signal
the following criteria: (i) coverage of a significant part of is detected, whichallows us to use oxygenisotope ratios as
the Holocene (at least several thousand years), (ii) at least tracersforISMintensity.
decadal resolution (i.e. a resolution of 10 years would re- SeveralrecordsareavailablefromtheArabianSearealm
sultin100datapointspermillennium–areasonablenumber (Qunf, Hoti, and Dimarshim caves, Table 1, Fig. 1). These
forobtainingreliablestatistics–whichisaboutthecoarsest areinterpretedasreconstructionsoftheArabianSeabranch
scaleonwhichweexpectthedynamicalregimeshiftsstudied oftheISM(Neffetal.,2001;Fleitmannetal.,2003,2007).
inthisworktobedetectable),and(iii)withageuncertainties However, these records might not be representative for the
nogreaterthanafewcenturies(RehfeldandKurths,2014). dynamicsoftheISMovertheIndiansubcontinent,andaddi-
Theserequirementsarenecessarytoreliablydetectshiftsin tionalrecordsmustberecoveredforbetterspatialcoveragein
thedynamicalregimes(i.e.thenon-linearvariability)ofthe theheartoftheISMdomain.ThestalagmitefromIndonesia
Asiansummermonsoonbeyondsimplechangesinamplitude (Griffithsetal.,2009)representsthetropicalAsianmonsoon
orvariance.Inthecaseofmultiplerecordsfromthesamesite domainandisusedforcomparisontoexamineourresultsin
(i.e. multiple speleothems from the same cave), we choose thebroaderclimatologicalcontext.
thedatasetwiththehighesttemporalresolutionandlongest
time interval covered. Furthermore, we discard other types
2.2 Treatmentofuncertaindepth–agemodels
ofproxies,suchasδ13C,greyscalevalues,etc.availablefor
somerecords,toobtainaconsistentdatabaseincludingonly Chronologiesforthepalaeoclimateproxyrecordsusedinthis
one variable (which may however be interpreted differently study have been established by means of U series dating.
in different regions, see below). Other types of records are The associated depth–age models are usually obtained by
considered for comparison in the discussion of our results interpolationbetweenthedatingpoints.However,radiomet-
whereverappropriate. ricdatingscomewithuncertainties,suggestingthatdifferent
Clim.Past,11,709–741,2015 www.clim-past.net/11/709/2015/
J.F.Dongesetal.:Non-linearregimeshiftsinHoloceneAsianmonsoonvariability 713
Earth system Speleothem
A
g
e

1
7
Graph theoretical analysis
2
Proxy record
B)
6 D
P
V
Recurrence network ‰
O (
18δ
Age (ka BP)
3
5
4.0
3.8
P)
a B 3.6
k
ge ( 3.4
A Z
3.2 4
3.0
X
3.0 3.2 3.4 3.6 3.8 4.0
Age (ka BP) Y
Recurrence plot Phase space trajectory
Figure 2.Workflowofrecurrencenetworkanalysisofpalaeoclimaterecords(herefromspeleothems).Step1indicatesthedepositionof
chemical or physical information on past climate fluctuations in growing speleothems. In step 2, this information is extracted from the
speleothem in the form of a proxy record (here, the Dongge DA δ18O record; Wang et al., 2005, is used as an example). Subsequently,
instep3,asectionoftheobtainedproxyrecord(selectedtimeintervalindicatedbygreybar)isembeddedinphasespacetounravelthe
fluctuationsinducedbyvariationsofthemultiplicityofdifferentrelevantclimaticparameters.Insteps4and5,thestructureofrecurring
palaeoclimatestatesintheproxyrecordisrepresentedasarecurrenceplotandrecurrencenetwork(nodecolourindicatesageofpalaeoclimate
statesincreasingfromred(younger)toblue(older)),respectively.Instep6,thestructureoftherecurrencenetworkcorrespondingtoacertain
epochisquantifiedbygraph-theoreticalmeasuressuchastransitivityT oraveragepathlengthL.Finally,thisanalysisprovidesinsightsinto
non-linearpalaeoclimatevariabilitythatcanbeinterpretedtakingtheunderlyingEarthsystemdynamicsintoaccountinstep7.Theoriginal
contributionofthisstudyliesinsteps3–7.
www.clim-past.net/11/709/2015/ Clim.Past,11,709–741,2015
714 J.F.Dongesetal.:Non-linearregimeshiftsinHoloceneAsianmonsoonvariability
chronologiesmightbevalidfortheproxyrecordinquestion
0
(Telfordetal.,2004;BuckandMillard,2004;Blaauw,2010). 40
Moreover,intercomparisonbetweenproxyrecordswithdif- 100
ferentdatingstrategies(and,hence,timescaleuncertainties) 60
200
anddifferentnon-uniformsamplingsdemonstratesthatcon-
sidering a single depth–age model limits the reliability of p) 300 80
o
all results. Therefore, we use the recently introduced CO- m t
PRA framework for the calculation of ensembles of con- o 400 100
r
sistent chronologies (depth–age models) for the used proxy m f 500 0.5 1
records (Breitenbach et al., 2012b). This strategy allows er- m
rorpropagationthroughsubsequentstatisticaltreatmentsand h ( 600
comparisonsofmultiplerecords. pt
Within COPRA, dating uncertainties are considered by De 700
a Monte Carlo (MC) simulation. To obtain an ensemble of
800
agemodels,firstarandomnumberdrawnfromanormaldis-
tributionofthestandarddeviationasgivenbythe1σ errorof 900
thedatingisaddedateachindividualdatingpoint.Apiece-
wise interpolation is then applied on these modified dating 1000
0 2 4 6 8 10
pointstoobtainagesforallproxydatapoints.Thisprocedure Age (ka BP)
isrepeated100times(MCsimulation)producinganensem-
ble of 100 possible age models (Fig. 3), where inconsistent Figure3.Exemplarydepth–agemodelfortheDonggeDArecord
(Wang et al., 2005) represented as an ensemble of 1000 different
realisationsviolatingthestratigraphicconstraintarerejected
chronologies.Theinsetshowsanenlargedviewofaspecifictime
beforehand. In the next step, these age models are used to
interval illustrating the spread of depth–age models due to dating
interpolatethemeasuredproxyvaluestoanequidistanttime
uncertainties. Discrete dating points are indicated by black dots,
axis(regularsampling),resultinginadistributionofpossible
whiletheassociated2σ datinguncertaintiesaredisplayedbyerror
proxyvaluesateachgiven(orrequired)timepoint.Nowthe bars.
time axis is a truly absolute and comparable reference sys-
temforalldifferentproxyrecordsbecausetheuncertainties
withinthetimedomainaretransferredtouncertaintiesinthe
proxydomain(Breitenbachetal.,2012b). ofinformationonthedynamicalsystemunderstudyandcan
Notethatinterpolatingtheproxysignaltoanabsolutetime bemathematicallyrepresentedandquantifiedbyrecurrence
axis is a post-processing step particular to this study that is plots(Marwanetal.,2007)orrecurrencenetworks(Marwan
not included in the COPRA framework by default. It is in- etal.,2009;Donneretal.,2010b,2011;Dongesetal.,2012)
troduced here to additionally test the robustness of the re- (seeFig.2).
sults of recurrence analysis obtained below with respect to Recurrenceanalysishasbeensuccessfullyappliedtoanal-
the effects of irregular sampling displayed by the available yse climatological and palaeoclimatological data, e.g. for
proxyrecords.Hence,inSect.4wewillcomparetheresults aligning the timescales of rock magnetic data from sedi-
obtained from the original irregularly sampled records with mentcores(Marwanetal.,2002)andsearchingforrelation-
thosecomputedfromtheCOPRAensemblesignalsthathave shipsbetweentheElNiño–SouthernOscillation(ENSO)and
beeninterpolatedtoaregularlysampledreferenceframe. South American palaeoprecipitation (Marwan et al., 2003;
Trauth et al., 2003). Recently and relevant to this study, it
has been shown that recurrence network analysis (Marwan
2.3 Recurrenceanalysis
etal.,2009;Dongesetal.,2011a,b;Marwanetal.,2013)and
Recurrence analysis comprises a class of non-linear meth- relatedtechniques(Maliketal.,2012)areparticularlywell-
ods of time series analysis that are sensitive to dynamical suitedfordetectingsubtlequalitativechangesinthedynam-
features beyond what can be captured by commonly used ics recorded by palaeoclimate proxy series with relatively
linear statistics, such as power spectra and auto- or cross- fewdatapoints(comparedtotypicalexperimentalandmod-
correlationfunctions(Marwanetal.,2007).Itisbasedonthe ernobservationaltimeseries).Inthisstudy,weareinterested
fundamental observation that dynamical systems in nature intransitionsornon-linearregimeshiftsinclimatevariability
tend to recur close to their previously assumed states after that are characterised by pronounced changes in dynamical
afinitetime(Poincaré,1890).Forexample,theweatherob- complexity, i.e. changes between regular and rather erratic
servations(temperature,precipitation,pressure,etc.)madeat climatevariations(Dongesetal.,2011b).Wearguethatthese
somemeteorologicalstationonagivendaymaybeverysim- transitionsaresubtleinthesensethattheycannotalwaysbe
ilar, but not the same, as those recorded a few years earlier. easilyandunambiguouslyidentifiedbyvisualinspection,de-
Thetemporalstructureoftheserecurrencescontainsawealth spitecontradictingclaims(Wunsch,2007).
Clim.Past,11,709–741,2015 www.clim-past.net/11/709/2015/
J.F.Dongesetal.:Non-linearregimeshiftsinHoloceneAsianmonsoonvariability 715
Sinceweareinterestedinthecomplexityofmonsoonfluc-
tuationsduringtheHoloceneondecadaltocentennialscales,
ratherthanmillennialandlonger-termtrends,wedetrendthe
1.0
proxy records using a 1000-year running window (Donges 0
-1.0
et al., 2011a) and study the residual oxygen isotope sig-
1.0
nals (cid:49)δ18O as a time series {x }N (Fig. 4). Below, we de-
i i=1 0
scribetheanalyticaltechniquesusedfordetectingandquan-
-1.0
tifyingqualitativechangesinHolocenemonsoondynamics, 1.0
0
namelytime-delayembeddingandrecurrencenetworkanal-
-1.0
ysis.Generally,weapplythesameanalysisstepstoboththe
1.0
original irregularly sampled proxy records and the interpo- 0
lated COPRA ensemble signals. Minor methodological ad- -1.0
2.0
justmentsarenotedbelowwhereverappropriate.Theanalyt-
0
icalworkflowemployedinthisstudyisvisualisedinFig.2. -2.0
1.0
2.3.1 Time-delayembedding 0
2.0
The palaeoclimate records that we use in this study are
0
univariate (cid:49)δ18O time series containing information on
-1.0
the impacts of different relevant variables from the higher- 1.0
0
dimensional climate system in the recording archive. For
-1.0
example, an oxygen isotope record may be dominated by
2.0
a precipitation amount signal, but precipitation co-evolves 0
-2.0
and is coupled with many other variables such as tempera-
ture, wind strength, and moisture source (Lachniet, 2009). 0.5
0
For performing a meaningful recurrence analysis, the time -0.5
evolutionofthehigher-dimensionalunderlyingclimatesys-
temneedstobereconstructedfromtheavailabledata(Mar-
wan et al., 2007). Under quite general assumptions (Tak- Figure4.Residualoxygenisotoperecords(cid:49)δ18Oanalysedinthis
ens,1981;KantzandSchreiber,2004),thiscanbeachieved study (all measured in units of ‰ VPDB). The original records
through time-delay embedding of a given detrended record listed in Table 1 have been detrended using a 1000-year moving
{x }N (seestep3inFig.2).Forobtainingthereconstructed window(asinDongesetal.,2011a).Bondevents(violetlines,Bond
i i=1
trajectory etal.,1997)andRCCepisodes(greybars,Mayewskietal.,2004)
aredisplayedforreference.
yi =(xi,xi+τ,...,xi+(m−1)τ), (1)
theparametersembeddingdelayτ andembeddingdimension
mhavetobechosenasfreeparameters. Table 2. Decorrelation time τd of speleothem oxygen isotope
recordsdeterminedasthefirstzero-crossingoftheauto-correlation
Foreachrecord,wesettheembeddingdimensiontom=3
function(ACF).TheACFwascomputedfromthedetrendedtime
asatrade-offbetweenresultsofthefalsenearestneighbours
seriesusingaGaussiankernelestimatorappliedtotheoriginalir-
criterionindicatingtheminimumnumberofdynamicallyrel- regularly sampled records (kernel bandwidth h=(cid:104)(cid:49)T(cid:105)/4 as sug-
evantsystemvariables(Kenneletal.,1992)andtherelatively
gestedbyRehfeldetal.,2011).Asliding-windowdetrendingwas
small number N of available data points (Table 1). Moder- appliedwitha1000-yearbandwidthforallrecords.
atelyincreasingmtypicallyleadstoqualitativelysimilar,yet
statisticallylessreliableresults(Dongesetal.,2011a). Cavename Decorrelationtimeτd(years)
Theembeddingdelayischosentoequalthedecorrelation
1 Dimarshim,Yemen 100
timescale τ (Table 2), hereafter defined as the first zero-
d 2 Qunf,Oman 216
crossingoftheauto-correlationfunction,inunitsoftheaver-
3 Hoti,Oman 57
agesamplingtime(cid:104)(cid:49)T(cid:105),i.e.τ =(cid:98)τ /(cid:104)(cid:49)T(cid:105)(cid:99)(Dongesetal.,
d 4 Mawmluh,India 146
2011a),where(cid:98)x(cid:99)denotesthelargestintegernotlargerthan
5 Tianmen,Tibet,China 90
x.Onetechnicalprobleminproperlydeterminingthisvalue 6 Dongge,China 185
is that proxy records, like the original speleothem records 7 Lianhua,China 193
discussedhere,aregenerallycharacterisedbyirregularsam- 8 Heshang,China 60
pling. This either requires an interpolation of the data (pos- 9 Jiuxian,China 73
sibly leading to biased results, cf. Rehfeld et al., 2011) or 10 LiangLuar,Indonesia 135
specificmethodswhichareabletocopewithsuchirregular
www.clim-past.net/11/709/2015/ Clim.Past,11,709–741,2015
716 J.F.Dongesetal.:Non-linearregimeshiftsinHoloceneAsianmonsoonvariability
bustwithrespecttovariationsinbothwindowandstepsize
(Dongesetal.,2011a).
Decorrelation time For each window, we build a recurrence network (RN)
from the corresponding set of state vectors (Marwan et al.,
2009; Donner et al., 2010b; Donges et al., 2012) (steps 4
and 5 in Fig. 2). Nodes of this network (Boccaletti et al.,
2006; Newman, 2010; Cohen and Havlin, 2010) represent
state vectors y from the reconstructed climate trajectory,
i
whilelinksareestablishedbetweenstatevectorsthatarere-
current, i.e. closer to each other in the phase space of the
reconstructed variables than a prescribed threshold distance
ε. Formally, the RN is represented by its adjacency matrix
(which can be alternatively visualised as a recurrence plot,
cf.Fig.2):
(years) Aij =(cid:50)(ε−(cid:107)yi−yj(cid:107))−δij. (2)
Figure 5. Auto-correlation function (ACF) for the Dongge resid- Here,(cid:50)(·)istheHeavisidefunction,(cid:107)·(cid:107)denotesthesupre-
ual oxygen isotope record (black solid line) computed using the mum norm, and δij is Kronecker’s delta ensuring that state
GaussiankernelestimatorbyRehfeldetal.(2011)andRehfeldand vectors are not connected to themselves forming self-loops
Kurths(2014)aftersubtractingmillennial-scaletrendsbymeansof in the RN (Donner et al., 2010b). Following generally sug-
asliding-windowdetrendingwithabandwidthof1000years.The gested good practice for recurrence network analysis (Don-
effectofdatinguncertaintiesisillustratedbythefullspreadofauto- neretal.,2010a;Marwan,2011;Dongesetal.,2012;Eroglu
correlationateachtimedelayobtainedfromtheregularlysampled
etal.,2014),thethresholddistanceεisadaptivelychosenfor
Dongge COPRA ensemble (Sect. 2.2) using a standard ACF esti-
eachwindowtoensurethatapproximately5%ofalltheoret-
mator (grey shading, Brockwell and Davis, 2006). The decorrela-
icallypossiblelinksarepresentinthenetwork.
tiontimeinferredusingtheGaussiankernelestimatorisindicated
The RN’s structure contains information on climate dy-
byaverticalline.
namicsduringthetimeintervalcoveredbythecorresponding
window.Weusethestatisticalnetworkquantifierstransitiv-
ityT andaveragepathlengthLtocapturethisinformation,
samplingforestimatingcorrelationfunctions(Scargle,1989;
which can be interpreted as measures of climate regularity
BabuandStoica,2009).Inthisstudy,weuseaGaussianker-
and abrupt dynamical change, respectively (Donges et al.,
nel based estimator for the auto-correlation functions (Re-
2011a, b). High T values indicate epochs with regularly
hfeldetal.,2011;RehfeldandKurths,2014)todeterminethe
varyingclimateondecadalandcentennialscales,e.g.adom-
typical values for τ . Although these values can vary when
d
inatingperiodiccomponentintheproxysignalortimeinter-
chronological uncertainties are taken into account using the
valswithstationaryorslowlychangingclimate,whilelowT
COPRAframework(Fig.5,cf.Sect.2.2),weusethedecor-
valuesimplyepochswithmore-erratic(i.e.lesspredictable)
relationtimeobtainedfromtheoriginalrecordforreference.
climatefluctuations.Incontrast,extremeLvalueshighlight
timeintervalsincludingpronouncedshiftsbetweendifferent
climaticregimes,whileintermediateLvalues,asdefinedby
2.3.2 Recurrencenetworkanalysis
thestatisticaltestdescribedbelow,pointtoamorestationary
Our aim is to detect qualitative changes in palaeomonsoon climateduringthecorrespondingepoch.
dynamicsreflectedintemporalvariationsofcomplexitymea- Torobustlydetectepisodesofstability,aswellasqualita-
sures computed from the proxy records. For this purpose, tivechangesinmonsoondynamicsthatareunlikelytoarise
weslideawindowoverthereconstructedclimatetrajectory from statistical fluctuations alone, we apply the stationarity
containingW statevectorsy ,withastepsizeof(cid:49)W state test proposed in Donges et al. (2011a, b) for each proxy
i
vectors.Forrenderingtheresultsoftheanalysiscomparable record. From this test we obtain 90% confidence bounds
between records with differing average sampling intervals for both quantifiers T and L indicating which range of val-
(cid:104)(cid:49)T(cid:105)(thisisonlyanissuefortheoriginalrecords,itdoesnot ues are typical when information from the whole record is
affectCOPRAensembles),bothparametersarechosensuch takenintoaccount.Thus,fortheoriginalirregularlysampled
thatthecorrespondingtimescalesW∗ and(cid:49)W∗ areapprox- recordsweareabletoidentifystatisticallysignificantepochs
imately the same for all records, i.e. W =(cid:98)W∗/(cid:104)(cid:49)T(cid:105)(cid:99) and ofnon-stationarityinclimatedynamicsasthosewhereT or
(cid:49)W =(cid:98)(cid:49)W∗/(cid:104)(cid:49)T(cid:105)(cid:99).WeselectthewindowsizeW∗=750 L lie outside the estimated confidence bounds. For evaluat-
years for all records. The step size is chosen as (cid:49)W∗=50 ing the effects of dating uncertainties for a certain record,
years for all records. It has been shown in earlier work that 90% confidence bounds for both network measures are es-
the results of recurrence network analysis (RNA) are ro- timated based on combining information from all members
Clim.Past,11,709–741,2015 www.clim-past.net/11/709/2015/
J.F.Dongesetal.:Non-linearregimeshiftsinHoloceneAsianmonsoonvariability 717
ofthecorrespondingCOPRAensemble.Inthenextstep,we changearoundthe8.2kaeventareweakerintheAISMdo-
identifysignificantepochsofunusualmonsoonvariabilityas main(Partinetal.,2007;Griffithsetal.,2009).
thosewhereatleast5%oftheCOPRAensemblesignalsdis- WhilethenotionofBondeventsrefersprimarilytoNorth
playT orLvalueslyingoutsidethoseconfidencebounds. Atlanticclimatevariability,Mayewskietal.(2004)compiled
information on global Holocene climate changes and iden-
tified several periods of significant large-scale fluctuations
3 HoloceneclimatevariabilityintheAsianmonsoon based on ca. 50 globally distributed climate records from
domain:currentstateofknowledge different kinds of archives (glacier fluctuations, ice cores,
marinesediments),forwhichtheycoinedthetermrapidcli-
For a long time, the Holocene has been thought of as a rel- matechange(RCC)episodes.Thesechangesappearedglob-
atively stable climatic period, since extremely strong tem- allyinacoherentwayandweresufficientlyabrupttoaffect
perature fluctuations comparable to those occurring during earlyhumansocieties(deMenocal,2001;HaberleandDavid,
glacialperiodshavebeenabsent.Overapproximatelythelast 2004).Infact,severaloftheidentifiedRCCepisodescanbe
twodecades,thispicturehas,however,distinctivelychanged attributed to the timing of major disruptions of civilisation
due to the identification of large-scale regional and even (Cullen et al., 2000; Drysdale et al., 2006; Fleitmann et al.,
global climate episodes, which have interrupted the gener- 2008;Baldinietal.,2002;Berkelhammeretal.,2012).
ally reduced climate variability and have lasted for several The six major Holocene RCC episodes identified by
decadestocenturies(Wanneretal.,2008).Theclimaticper- Mayewski et al. (2004) are listed in Table 4. Notably, these
turbationsfoundduringtheHolocenearepotentiallyrelevant episodespartlycoincidewiththeBondeventsB0–B6.
factorsforthedevelopmentofcomplexhumansocietiesand The earliest Holocene RCC episode (RCC5) has been
theirbehaviour(deMenocal,2001). termed “Glacial Aftermath” by Mayewski et al. (2004) and
Investigating marine records from the North Atlantic has been originally attributed to a time interval between
Ocean, Bond et al. (1997, 2001) were the first to provide about 9.0 and 8.0kaBP. Notably, this episode includes the
evidence of a persistent cycle of ice-rafted debris in sedi- 8.2ka event (Bond event B5), which was probably initiated
mentary sequences. Specifically, Bond et al. (1997) identi- by a large meltwater pulse reducing the Atlantic thermoha-
fiedmorethaneightoftheseBondevents(cf.Table3),with line circulation (Broecker et al., 2010), as well as several
events B1–B8 originally dated to ca. 1.4, 2.8, 4.2, 5.9, 8.1, earliermeltwaterpulses,oneofwhichpossiblytriggeredan-
9.4, 10.3, and 11.1kaBP. Some but not all of these events otherwidespreadclimateanomalyatabout9.2kaBP(Fleit-
sharply coincide with periods of marked high-latitude cool- mannetal.,2008)closetoBondeventB6.Inbothcases,the
ing and/or low-latitude aridification and possibly associated corresponding large-scale temperature drop in mid- to high
cultural changes and large-scale migration patterns (Gupta northernlatitudeshasbeenaccompaniedbyasignificantdry-
et al., 2003; Wang et al., 2005; Parker et al., 2006) due to ing in the northern tropics reflected in several records from
distinctlocalresponsessuchasdroughts,increasingstormi- the Asian monsoon domain (Mayewski et al., 2004; Fleit-
ness,orseasonalitychanges. mannetal.,2008).
Mostprominently,the8.2kaeventiswidelyrecognisedas The following four RCC episodes (RCC4–RCC1) varied
themostpronouncedlarge-scaleNorthernHemispherecool- in strength and geographical extension, but share the same
ing episode during the mid-Holocene (Alley et al., 1997). common pattern of high-latitude cooling and low-latitude
Regional climate changes associated with this episode in- aridity. The most extensive episode lasted from about 6.0
cludeastrengthenedatmosphericcirculationovertheNorth to 5.0kaBP (RCC4) and featured North Atlantic ice rafting
Atlantic and Siberia, resulting in more frequent winter out- (5.9ka or B4 event), alpine glacier advances, strengthened
breaks of polar air masses (Mayewski et al., 2004). In the westerlies, and pronounced aridity in the Arabia Peninsula
lowernorthernlatitudes,thereispalaeoclimaticevidenceof (Parker et al., 2006). In the low latitudes, it coincided with
widespread aridity, for example, in terms of an intermittent thebeginningdeclineoftheAfricanHumidPeriod(deMeno-
interruptionoftheAfricanHumidPeriod(deMenocaletal., caletal.,2000;Francusetal.,2013),butpossiblydisplayed
2000), dramatically weakened summer monsoons over the sustained moist conditions in north-western India and Pak-
ArabianSeaandtropicalAfrica(Fleitmannetal.,2003;Al- istan (Enzel et al., 1999). The 4.2–3.8kaBP period (RCC3,
leyandÁgústsdóttir,2005;MorrillandJacobsen,2005),and falling together with the 4.2ka or B3 event) is commonly
persistent drought in Pakistan (Mayewski et al., 2004). In weakerandlesswell-expressedinitsglobal-scalecharacter-
turn, precipitation in the Middle East increased during the istics and displayed weaker winds over the North Atlantic
8.2ka event (Bar-Matthews et al., 2000; Arz et al., 2003), and Siberia and generally dryer conditions in the low lati-
which indicates a possible southward displacement and in- tudes(fortheISMseeBerkelhammeretal.,2012).Between
tensification of westerlies, associated with changes of the 3.5and2.5kaBP,anotherRCCepisode(RCC2)withNorth
NorthAtlanticOscillation.Moreover,regardingtheregional Atlanticicerafting(BondeventB2)andstrengthenedwest-
focusofthisstudy,wenotethatincomparisonwiththeISM erlies is found, whereas the signature of the 1.2–1.0kaBP
and EASM domains, available indications for rapid climate episode(RCC1)againwidelyresemblesthatofRCC3.
www.clim-past.net/11/709/2015/ Clim.Past,11,709–741,2015
718 J.F.Dongesetal.:Non-linearregimeshiftsinHoloceneAsianmonsoonvariability
Table3.ListofBondeventsandpotentiallyrelatedculturalchanges(selection).
No. Timing Notesandrelatedevents
(kaBP)
B0 ≈0.5 LittleIceAge(Esperetal.,2003,2007)
B1 ≈1.4 MigrationPeriod(Büntgenetal.,2011;Esperetal.,2003,2007)
B2 ≈2.8 InitiationofIronAgeColdEpoch(vanGeeletal.,1996;Swindlesetal.,2007;PlunkettandSwindles,2008),
earlythirdmillenniumBPdroughtintheeasternMediterranean(Weiss,1982;Kaniewskietal.,2008,2010),
possiblytriggeringthecollapseofLateBronzeAgecultures
B3 ≈4.2 4.2kaevent,collapseoftheAkkadianEmpire
(Gibbons,1993;Weissetal.,1993;Cullenetal.,2000;Stanleyetal.,2003;Drysdaleetal.,2006),
endofEgyptianOldKingdom
B4 ≈5.9 5.9kaevent(Parkeretal.,2006)
B5 ≈8.1 8.2kaevent(Alleyetal.,1997;Baldinietal.,2002)
B6 ≈9.4 ErdaleneventofglacieractivityinNorway(Dahletal.,2002;Fleitmannetal.,2008),
coldeventinChina(Zhouetal.,2007;Houetal.,2012)
B7 ≈10.3 (Houetal.,2012;Amigoetal.,2013)
B8 ≈11.1 TransitionfromtheYoungerDryastotheboreal(Mayewskietal.,2004)
Table 4. List of Holocene rapid climate change (RCC) episodes plexityandpossibleregionalvarietyofsuchchanges.While
(Mayewskietal.,2004)includingpossibleforcingmechanismsand recent studies have almost exclusively focussed on changes
temporallyassociatedBondevents(Table3). in the amplitudes of palaeoclimate proxies, this work ex-
amines more subtle changes in the non-linear dynamics of
No. Timing Possible Associated palaeoclimatevariabilityandsearchesforconsistentpatterns
(kaBP) mechanisms Bond relatedtothisspecificaspect.
events
Notably,theHoloceneRCCepisodes(althoughrelatively
RCC0 0.6–0.15 Reducedsolarforcing, B0 small in magnitude, Mayewski et al., 2004) coincide with
volcanism periods of dramatic changes in some ecosystems and hu-
RCC1 1.2–1.0 Reducedsolarforcing B1 mancivilisations.Forexample,theshort-livedRCC1episode
RCC2 3.5–2.5 Reducedsolarforcing B2 appears synchronously with the collapse of the Maya civil-
RCC3 4.2–3.8 Reducedsolarforcing B3 isation, which has possibly been substantially aggravated
RCC4 6.0–5.0 Reducedsolarforcing B4
by multiple prolonged droughts (Haug et al., 2003; Kennett
RCC5 9.0–8.0 Orbitalchanges, B5,B6
etal.,2012).TheRCC3episodesawthedemiseofsomeof
meltwaterpulse,
theworld’sfirsthighlydevelopedcomplexsocietiessuchas
volcanism
theAkkadianEmpire,Egypt’s Old Kingdom,andtheIndus
Valley (Harappan) civilisation (Gibbons, 1993; Weiss et al.,
1993;Cullenetal.,2000;Stanleyetal.,2003;Drysdaleetal.,
2006). The time period around the 8.2ka event was charac-
WhiletheexistenceofepisodesofHolocenerapidclimate
terisedbyanabruptabandonmentofflourishingsettlements
changes is meanwhile widely accepted, there is an ongoing
in Anatolia and a subsequent spread of early farmers into
debate about the timing and spatial coherency of these pat-
southernEurope,whichwaspossiblytriggeredbyincreased
terns.Forexample,Wanneretal.(2008)usedadifferentset
aridity in the Middle East (Weninger et al., 2006). We will
of palaeoclimate time series and “did not find any time pe-
furtherelaborateonsomeimportantexamplesandtheirsig-
riodforwhicharapidordramaticclimatictransitionappears
natures within the Asian monsoon system in the following
even in a majority of the time series” with the exception of
sections.
twolarge-scaleshiftsatabout5.2kaBPandbetween3.1and
2.5kaBP, which partly coincide with the RCC4 and RCC2
episodes reported by Mayewski et al. (2004). Another re- 4 Results
centstudycametoasimilarresultreportingthatthespatio-
temporalpatternsoftemperatureandhumidity/precipitation In this section, we report our results of RN analysis for the
exhibitedverystrongvariabilityduringsomeofthemostpro- chosen set of Asian speleothem records, which reveal sev-
nounced Holocene cold episodes (Wanner et al., 2011). In eral epochs of significant non-linear climatic change during
our opinion, these differences among recent studies do not the Holocene. First, the results obtained by using original
necessarilycontradicttheexistenceofconsistentlarge-scale age models are evaluated for their robustness with respect
responsesoftheclimatesystem,butratherhighlightthecom- to dating uncertainties. For this purpose, the well-studied
Clim.Past,11,709–741,2015 www.clim-past.net/11/709/2015/