Until lately, it has commonly been assumed that the last major reorganization of the North Atlantic oceaneatmosphere system, the Younger Dryas climatic reversal, spread synchronously on continental tohemispheric scales. This assumption arose because reliable chronologies, which would allow capturingthe complexity surrounding local responses to abrupt climate change, were lacking. To better understandthe temporal structure at the inception of the Younger Dryas across the North Atlantic, we revised,updated and compared the chronological framework of four Northern European sediment sequences(Lake Kråkenes, Lake Madtj€arn, Lake Gammelmose, Sluggan Bog) by applying classical Bayesianmodelling. We found distinct and spatially consistent age differences between the inferred ages of theAllerød interstadial e Younger Dryas stadial pollen zone boundaries among the four sites. Our resultssuggest an earlier vegetation response at sites along latitude 56e54�N as compared to sites located at 60e58�N. We explain this time lag by a gradual regional cooling that started as early as c. 12,900e13,100 cal. BP. This phenomenon was probably linked to cooling around the Nordic Seas as a result ofenhanced iceberg calving from the Fennoscandian Ice Sheet during the final stage of the Allerød inter-stadial. By contrast, vegetation shifts at sites located further north occurred significantly later and inconcert with the establishment of full stadial climate conditions (c. 12,600e12,750 cal. BP). Our studyemphasizes the need to develop solid regional 14C chronologies and to employ the same age modellingapproach to determine the temporal and spatial response to a climatic shift.
The Younger Dryas stadial (YD) is the most recent and wide-spread abrupt climate oscillation, which occurred towards the endof the last deglaciation. It is recorded in numerous paleoclimatearchives around the North Atlantic region as a c. 1000-year longcold interval (Lowe et al., 2008). The onset of the YD is oftenattributed to a reduction in the North Atlantic meridional over-turning circulation (Broecker,1998;McManus et al., 2004) resultingfrom a major meltwater pulse into the North Atlantic (Duplessyet al., 1992; Bard et al., 2000; Bradley and England, 2008).
The pollen-stratigraphic transition associated with the onset ofthe YD has long been used as a common and synchronous strati-graphic boundary in the region (Mangerud et al., 1974; Bj€orck et al.,1996, 1998a,b; Wohlfarth, 1996; Lowe et al., 2008). Related shiftsobserved in diverse proxy records in locations far from the North
F. Muschitiello).
Atlantic region are often assumed to reflect the same climatic event(e.g. Cheng et al., 2009).
In the Greenland NGRIP ice core, the onset of the regionalcounterpart of the YD is defined by a marked increase in deuteriumexcess (Steffensen et al., 2008), a proxy for far-field changes in theprecipitation source region. This distinct shift occurs over 1e3 yearsand marks the onset of a cold phase that is referred to as GreenlandStadial 1 (GS-1) (Steffensen et al., 2008).
Various problems commonly arise when attempting to correlatethe start of the YD/GS-1 between terrestrial records and ice cores(Lohne et al., 2013). Since the majority of terrestrial sedimentaryarchives do not provide the time resolution that would allow for aprecise correlation to annually resolved ice cores. Moreover, thevarious proxies that are used to infer past climatic shifts havedifferent environmental sensitivities, and these do not necessarilychange in phase across major climate transitions. Due to theselimitations, it has been commonly assumed that abrupt climatechanges occurred more or less synchronously on continental tohemispheric scales, and for simplification proxy records are oftentuned to template sequences such as the Greenland ice core records(e.g. Schwander et al., 2000; Bakke et al., 2009).
F. Muschitiello, B. Wohlfarth / Quaternary Science Reviews 109 (2015) 49e5650
However, a recent study has convincingly shown that althoughclimate transitions may appear locally abrupt, this feature is not aprecondition for assuming synchronicity on regional or largerscales (Lane et al., 2013). Rather, it seems that rapid local shifts canbe part of a time-transgressive propagation of a large-scale atmo-spheric phenomenon. These findings have encouraging prospectsand stress the importance of establishing a robust regional networkof densely dated, chronologically reliable, and temporally wellresolved proxy-based climate reconstructions. These qualificationsare critical not only for capturing the complexity that surroundslocal-to-regional responses to abrupt climate change, but are alsoimportant for understanding themechanisms driving the inceptionand propagation of abrupt climate shifts.
The latest advancement of the IntCal radiocarbon calibrationcurve (IntCal13) (Reimer et al., 2013) offers the opportunity tobetter constrain the age of the transition from the warm Allerød(AL) interstadial to the cold YD stadial. By using new tree-ring 14Crecords, the gap in the European tree-ring radiocarbon chronolo-gies prior to and during the early YD, has been filled, replacing theocean-based 14C data from the Cariaco Basin (Hughen et al., 2000).The new IntCal13 chronology, which is based on tree-ring data,constitutes a more reliable representation of atmospheric radio-carbon variations and has minimised previous uncertainties asso-ciated with the age reservoir fluctuations in the marine Cariacorecord (Hua et al., 2009).
In order to better constrain the timing of the terrestrial envi-ronmental responses to the onset of the YD, and to examine theuncertainties associated with radiocarbon dating, we revised,updated and compared the chronological framework of four non-varved Northern European sediment sequences from Lake Krå-kenes (Birks et al., 2000), Lake Madtj€arn (Bj€orck et al., 1996), LakeGammelmose (Andresen et al., 2000), and Sluggan Bog (Lowe et al.,2004), the chronologies of which are underpinned by a largenumber of AMS 14C dates. By employing classical Bayesian agemodelling with two independent Bayesian softwares, we hereprovide new calibrated ages for the ALeYD pollen-stratigraphicboundary defined in each of the four sequences. The results arediscussed in the light of potential mechanisms that may havecaused a difference in timing among ecosystem shifts in Northern
Fig. 1. Location of the sites for which new age-depth models were constructed (red dots). (Fothe web version of this article.)
Europe. We also broach some considerations that must be takeninto account when assessing the true duration and timing ofclimate events from terrestrial sedimentary data sets.
2. Methods
2.1. Selection criteria
The four non-varved Northern European Lateglacial sedimen-tary sequences (Fig. 1; Table 1) were selected on the basis of i) thenumber of published 14C measurements on terrestrial plant mac-rofossils. We preferred records characterised by densely spaced,continuous and evenly distributed sequences of radiocarbon dates,with the ability to provide a highly resolved age model. Moreimportantly, in order to restrict the uncertainty on the determi-nation of the ALeYD pollen-zone transition, we established that ii)the pollen-stratigraphic boundary should be closely constrained bymeans of one radiocarbon measurement on either side of thetransition (Table 1), and that iii) the resolution of the pollen samplemust not exceed a one-cm interval in close proximity to the tran-sition between the pollen zones.
2.2. Site settings and definition of the ALeYD transition
Lake Kråkenes is located on thewest coast of Norway (Fig.1) andcontains a continuous sedimentary sequence that records theenvironmental history of the catchment since the early AL. Birkset al. (2000) and Birks and Ammann (2000) suggested that Late-glacial biotic and abiotic changes within the small catchment wereprimarily affected by changes in temperature, which in turn wereinfluenced by the formation and melting of the cirque glacierlocated above the lake. However, recently Lohne et al. (2013) couldshow that the glacier formed 20e40 years after the AL/YD transi-tion, and that it did not influence the environmental changesrecorded in the sediments until later. The ALeYD pollen-zoneboundary was defined using a rate-of-change analysis approach(Birks et al., 2000), which identifies statistically significant rates ofchange (lithostratigraphy, changes in aquatic and terrestrial biota)based on Monte Carlo permutation tests.
r interpretation of the references to colour in this figure legend, the reader is referred to
Table 1Geographical, chronological, and pollen sampling information for the sites presented in this study.
Site Lat Lon Alt(m a.s.l.)
Numberof 14C dates(plant macro)
Distance (cm) betweenconsecutive 14C measurementsacross the ALeYD pollen boundary
Pollen samplingerror (cm) at theALeYD pollenboundary
Reference
Kråkenes 62� N 5� E 38 52 1 ±0.5 Birks et al. (2000); Lohne et al. (2014, 2013)Madtj€arn 58� N 12� E 135 28 2 ±0.5 Bj€orck and M€oller (1987); Bj€orck et al. (1996)Lake
Gammelmose56� N 9� E 34 16 3 ±0.25 Andresen et al. (2000)
Sluggan Bog 54� N 6� W 91 29 2 ±0.5 Lowe et al. (2004); Walker et al. (2012)
F. Muschitiello, B. Wohlfarth / Quaternary Science Reviews 109 (2015) 49e56 51
Lake Madtj€arn is located in southwestern Sweden (Fig. 1) andbecame isolated from the sea during the last 100e200 years of theAL pollen zone. At that time, the lake was located on an island,surrounded by a cold ocean, and in close proximity to the southernmargin of the Fennoscandian Ice Sheet (FIS) (Bj€orck et al., 1998a,b).The lake sediments contain a pollen-stratigraphic record of theregional vegetation development spanning the Late AL to EarlyHolocene (Bj€orck et al., 1998a,b). The ALeYD pollen-zone transitionis placed at the mid-point of the first increase in Artemisia pollenpercentages and decrease in pollen concentrations (Bj€orck et al.,1996).
Lake Gammelmose is a small lake situated in a lowland area ineastern Denmark (Fig. 1) and likely formed as a kettle hole whenstagnant icemelted (Andresen et al., 2000). The pollen-stratigraphyfor this site covers the interval between the Late AL and the earlyYD (Andresen et al., 2000). The ALeYD pollen-zone boundary isdefined by the first decrease in Betula pollen percentage and pollenconcentration values (Andresen et al., 2000).
Sluggan Bog is part of a raised bog complex located in thelowlands of Northern Ireland (Fig. 1). The basin is a regional refer-ence site that contains a full Lateglacial and Holocene pollen-stratigraphic sequence (Walker et al., 2012). The ALeYD pollenzone transition is characterised by the first marked increase inArtemisia and Caryophyllaceae pollen values, and by a sharp lith-ological boundary (Walker et al., 2012).
2.3. Age-depth modelling
The published radiocarbon data sets for the four sites, basedexclusively on terrestrial plant macro remains, were modelled us-ing OxCal4.2 (Bronk Ramsey, 2010) and Bacon2.2 (Blaauw andChristen, 2011) after calibration with the IntCal13 calibrationcurve (Reimer et al., 2013). The comparison of the age output fromtwo of the most widely used age-modelling routines allows testingthe validity of the age derivation process. It also provides a morerobust control on the age derivation that highlights the estimationperformance of the two Bayesian age-modelling methods.
Age-modelling with OxCal was performed using the P_Sequencedeposition procedure (Bronk Ramsey, 2008). The P_Sequencefunction leans on the correct choice of k, a parameter that definesthe stiffness of the model upon the dating sequences whenreproducing the sedimentation process (Bronk Ramsey, 2008).Once an appropriate k parameter has been chosen, the OxCal pro-gram performs millions of Markov Chain Monte Carlo (MCMC)sampling steps that calculate the highest probability density rangefor each depositional event, which results in the generation of anage-depth model.
For the sequences of Kråkenes and Sluggan Bog, the k factorshad already been defined with values of k ¼ 0.4 cm�1 (Lohne et al.,2013) and k ¼ 0.5 cm�1 (Walker et al., 2012), respectively. The kparameter for Kråkenes was estimated by calculating the variabilityin distance between the sedimentary units from multiple cores asdescribed in Bronk Ramsey (2008). For the remaining sequences
(Madtj€arn and Lake Gammelmose), we opted for the default valueof k ¼ 1.0 cm�1 as a reasonable compromise that would constrainthe radiocarbon sequences with a sizable degree of flexibility of themodel upon the data set (Bronk Ramsey, pers. comm.). The keylithostratigraphic units were included in the sequences to definethe step changes in depositional rates. Sequences that containedtephra horizons, which have also been identified in Greenland icecores (i.e. Vedde Ash, Saksunarvatn Ash), were included as ice corecalendar years in the models (after converting the b2k age into BP)(Abbott and Davies, 2012). Clear outliers among the radiocarbondates were detected and removed by applying the Outlier Analysisuntil robust and coherent agemodels were generated as defined byhigh agreement indices with values higher than a threshold of 60%(Bronk Ramsey, 2009). The Outlier_Model analysis was performedwith the General setting and the prior probability fixed to 0.05,which weights down the radiocarbon measurements that have astatistical probability of more than 5% of being outliers (BronkRamsey, 2009).
In a similar way to OxCal, Bacon splits the sequence into severalsections, and in line with the prescribed prior information, thesoftware calculates the accumulation rates for each of these sec-tions by means of millions of MCMC iterations (Blaauw andChristen, 2013).
The thickness of the sections is set by the thick parameter, uponwhich the flexibility of the age-depth model depends, with a largenumber of thin sections resulting in a smoother model. Within therange of acceptable values suggested by the software, we adoptedthick values of 2 cm for the sequences longer than 1m (i.e. Kråkenesand Madtj€arn) and values of 1 cm for the sequences shorter than1 m (i.e. Lake Gammelmose and Sluggan Bog).
For each sequence, we prescribed specific prior information foraccumulation rates (acc.mean) suggested by Bacon, and defaultparameters of memory prior distribution (mem.strength ¼ 4 andmem.mean ¼ 0.7), which according to Blaauw and Christen (2013)allow for a broad range of depositional conditions.
Unlike in OxCal, lithological changes cannot be prescribed inBacon and the information regarding the accumulation history of asequence is generally approached by means of an adequate choiceof prior information. However, tephra horizonswere included as icecore calendar years in the same way as for OxCal.
The error treatment in Bacon is substantially similar to that fromthe General Outlier Analysis approach in OxCal. Priors for outliers inboth routines draw from a long-tailed Student's t-distribution.However, the t-modelling error analysis in Bacon differs from thatin OxCal in terms of the number of parameters employed in theprocess (Christen and P�erez, 2009). This makes Bacon generallymore conciliatory towards potential outlying dates than OxCal andalso enables the model to account for possible unknown orunderestimated errors associated with the 14C determinations(Christen and P�erez, 2009). A detailed description of the t-model-ling error analysis and a technical comparison between OxCal andBacon is given in Christen and P�erez (2009) and Blaauw andChristen (2011). The size of the error distribution in Bacon is
F. Muschitiello, B. Wohlfarth / Quaternary Science Reviews 109 (2015) 49e5652
dictated by two parameters, t.a and t.b, which are by default set at 3and 4 (Christen and P�erez, 2009). However, for the sequences witha high number of accurate dates (i.e. Kråkenes, Madtj€arn, and LakeGammelmose), we decided to set t.a and t.b at relatively highervalues (7 and 8), thus employing slightly less spread out t distri-butions (Blaauw and Christen, 2013).
A full list of the radiocarbon dates for each site used here and thelist of dates used in the final age-depth models are available assupplementary material (Tables S1eS4).
2.4. Bayesian analysis
To compare the age results of the ALeYD transition in a statis-tical fashion and to integrate the age uncertainties associated withthe pollen sampling resolution in our analysis, we performed aBayesian estimation of the age-model output. The Bayesian esti-mation is based on aMonte Carlo method (Kruschke, 2013), which -assuming a Gaussian behaviour e provides a probability distribu-tion over the mean age of the transition from the parameter valuesof the two consecutive sampling levels that straddle the pollen-stratigraphic boundary. Specifically, we randomly sampled 1000age values at each of the two adjacent pollen-sampling levels from
Fig. 2. Age-depth model output for the four radiocarbon-dated sequences using OxCal (Bronthe IntCal13 radiocarbon calibration curve (Reimer et al., 2013). Agreement indices associatepollen-stratigraphic units and grey circles indicate the position of possible outliers in the seRed stars show the position of the local AllerødeYounger Dryas pollen-stratigraphic bounmodels. A full list of the dated events included in the age models is presented as supplementalegend, the reader is referred to the web version of this article.)
the related pool of MCMC iterations derived from the age-modelling process. Therefore, the 1000 þ 1000 random valueswere used as starting parameter values for the Bayesian estimation.Finally, using 10,000 Monte Carlo sampling steps, the numericalmethod was applied on the starting parameter values to produce aposterior predictive distribution of the mean age between the twopollen-sampling levels. This approach yields the most crediblemean age and the related standard deviation over the depth un-certainty interval that includes the ALeYD transition.
3. Results and discussion
3.1. Age-depth models and outlier analysis
After removal of the outlying dates, the OxCal-derived age-depth models produce agreement indices higher than 70e80%(Fig. 2). To further test the goodness of fit of the individual dates, weperformed additional runs for each model prescribing additionalminor lithological changes. However, this approach did notimprove the models, and no changes in the number of outliers or inthe age estimates for the ALeYD boundaries were observed. Wethus adopt the most conservative approach by keeping the number
k Ramsey, 2009) and Bacon (Blaauw and Christen, 2011). 14C dates were calibrated usingd with output from OxCal are shown. The colours indicate the prescribed major localquences as detected with the Outlier Analysis procedure in OxCal (see text for details).dary. The uncertainty envelopes represent the 95.4% confidence intervals of the agery material in Tables S1eS4. (For interpretation of the references to colour in this figure
Fig. 3. Results of the new AllerødeYounger Dryas age estimates. Ages are plottedagainst the latitude of the corresponding site (K ¼ Kråkenes; M ¼ Madtj€arn; LG ¼ LakeGammelmose; SB ¼ Sluggan Bog). Median age of the corresponding pollen-stratigraphic transition as modelled with OxCal and Bacon, respectively, is alsoshown. For reference, the results are plotted together with the AllerødeYounger Dryaspollen boundary and the onset of cooling (Rach et al., 2014) as inferred from theMeerfelder Maar (MFM) varve chronology (Brauer et al., 1999; Brauer et al., 2000). Barsrepresent age estimates of the pollen-stratigraphic boundaries accounting for addi-tional pollen sampling errors with their median age as modelled with OxCal (red) andBacon (yellow). Dating uncertainties due to sampling resolution are here simplisticallypresented using the uppermost 2s age error associated with the sampling level abovethe ALeYD transition and the lowermost 2s age error associated with the samplinglevel below the ALeYD transition. (For interpretation of the references to colour in thisfigure legend, the reader is referred to the web version of this article.)
F. Muschitiello, B. Wohlfarth / Quaternary Science Reviews 109 (2015) 49e56 53
of lithological changes in themodels to a minimum and prescribingonly those boundaries that are associated with the major localpollen-stratigraphic units (e.g. Older Dryas, Early AL, Late AL, YD,Early Holocene). The new age-depth models are in good agreementwith previous studies, which used OxCal to construct age-depthcurves (Kråkenes: Lohne et al., 2013, 2014; Sluggan Bog: Walkeret al., 2012).
The Bacon-derived age-depth models generally show muchnarrower uncertainty envelopes for each sequence than thosemodelled with OxCal (Fig. 2). All age models produced with Baconstem from stable runs characterised by a solid structure of thedistribution among the MCMC iterations. However, the sequencefrom Madtj€arn failed to produce reliable runs when default mem-ory parameters were prescribed. This is potentially due to a markedchange in sedimentation rate, which demands different parametersof prior memory distribution (Blaauw and Christen, 2013). We thusdetermined an optimal mem.mean value of 0.4 and a mem.strengthvalue of 3 for this sequence. This is justified by an improvement ofthe goodness of the runs after multiple tests.
Bacon is also superior in constraining the confidence intervalsfor sections dated at lower resolution (e.g. Sluggan Bog) (Fig. 2).OxCal, on the other hand, seems to generate output that better fitsthe sequence of 14C dates (e.g. Kråkenes and Lake Gammelmose)owing to the prescription of lithostratigraphic information (Fig. 2).These minor differences in the fitting performance do not bias theage derivation process around the analysed level. The age estimatesfor the ALeYD transition, as calculated by the two routines sepa-rately, compare well to each other in terms of median values(Table 2), which indicates that the results are precise.
Furthermore, the age-depth models generated with both soft-wares show an overall solid structure of the dated sequences and,more importantly, are robust across the ALeYD boundary. In otherwords, none of the 14C measurements that constrain the ALeYDboundaries in the different sequences is recognized as a possibleoutlier (Fig. 2; Table S1eS4), giving no indication for the presence ofreworked material or age reversals. This gives us confidence thatthe models are accurate at the analysed stratigraphic boundaries.
3.2. New age estimates for the AllerødeYounger Dryas pollen-stratigraphic boundaries
The age estimates for the ALeYD pollen-zone transition inKråkenes and Madtj€arn are the least uncertain of the data set andare in good agreement with each other when model errors aretaken into account: OxCal-derived median ages are 12,722 and12,685 cal. BP, respectively, and Bacon-derived median ages are12,735 and 12,677 cal. BP, respectively (Fig. 3, Table 2). Although theestimated ages for Lake Gammelmose and Sluggan Bog showrelatively higher standard deviations for the ALeYD pollen-zoneboundary, modelled ages are very close at the 95.4% confidenceinterval: OxCal-derived median ages are 12,919 and 13,021 cal. BP,respectively, and Bacon-derived median ages are 12,957 and13,017 cal. BP, respectively (Fig. 3, Table 2). Larger age uncertaintiesfor sequences where the ALeYD transition falls on the radiocarbon
Table 2New ages of the AllerødeYounger Dryas pollenstratigraphic boundary as modelled with
plateau are not surprising since the age interval in the IntCal13chronology that is part of the 14C plateau is relatively hard totranslate efficiently into calendar years, while ages that occur alongthe sloping section of the calibration curve are reasonably wellconstrained.
The modelled ages for the ALeYD pollen-zone transition - asdefined by the respective authors for each site - thus suggest a timedifference of a few hundred years between Kråkenes/Madtj€arn onone side and Gammelmose/Sluggan Bog on the other side (Fig. 3).This in turn would mean that the vegetation around Lake Gam-melmose and Sluggan Bog crossed critical environmental thresh-olds associated with the YD reversal earlier than at Kråkenes andMadtj€arn.
3.3. Bayesian age estimation and comparison
To obtain a statistical estimate of the ALeYD ages that in-corporates both age-modelling and sampling uncertainties, weused the Bayesian approach described in Section 2.4. This Bayesiananalysis provides i) uncertainties on the age derivations, and ii)estimates of the probability distribution over the difference inmean ages between Kråkenes and Madtj€arn, and between LakeGammelmose and Sluggan Bog. We performed the Bayesian anal-ysis on the most robust age derivations obtained with Bacon asoutlined in Section 2.4. (Fig. 4, Table 3).
Table 3Ages of the AllerødeYounger Dryas pollenstratigraphic boundary and related 1standard deviation accounting for both age-modelling and sampling resolution.
F. Muschitiello, B. Wohlfarth / Quaternary Science Reviews 109 (2015) 49e5654
The models generate a mean ALeYD age of 12,736 ± 32 cal. BPfor Kråkenes, of 12,671 ± 78 cal. BP for Madtj€arn, of12,957 ± 62 cal. BP for Lake Gammelmose, and of 13,016 ± 98 cal. BPfor Sluggan Bog. The mean ALeYD age is significantly differentbetween Kråkenes and Madtj€arn, and between Lake Gammelmoseand Sluggan Bog (Fig. 4). Specifically, Kråkenes and Sluggan Boglead Madtj€arn and Lake Gammelmose by about 60 years, respec-tively. It could be argued that the statistical significance of the agedifference stems from the arbitrarily high accuracy of the method,which approximates the posterior distribution by assembling alarge MCMC sample (Kruschke, 2013). However, significant differ-ences are observed even when using a less sophisticated c2 test ofindependence on a restricted number of parameters to testwhether the distribution of the samples stem from the same mean.Therefore, from a statistical point of view, we can conclude that theALeYD transition, as recorded in the respective pollen stratigra-phies, occurred at different times at different locations. We here-after refer to the estimates obtained from the Bayesian estimationas the most credible ages of the ALeYD transition, envisaging bothage-modelling and pollen sampling errors.
3.4. Duration of the vegetation shifts
Differences in the duration of vegetation shifts in response torapid climate changes are an expression of diverse local climatic
Fig. 4. a) Predictive posterior distributions of the mean age differences betweenKråkenes and Madtj€arn (left), and between Sluggan Bog and Lake Gammelmose (right)derived using the Bayesian estimation described in Section 2.4. Both distributions ofthe mean age differences have a high probability of being greater than 0. b) Predictiveposterior distributions of the mean ALeYD age for each site (K ¼ Kråkenes;M ¼ Madtj€arn; LG ¼ Lake Gammelmose; SB ¼ Sluggan Bog) obtained from theBayesian estimation, with black bars indicating the 95% highest density interval. Pre-dictive posterior distributions of the mean ALeYD ages are plotted together withGreenland climate events (GS-1 ¼ Greenland Stadial 1; GI-1a ¼ Greenland Interstadial1a; GI-1b ¼ Greenland Interstadial 1b). c) Dust abundance and d18O proxy records fromGreenland ice core NGRIP (Rasmussen et al., 2006; Ruth et al., 2007). The dashedstraight line shows the rapid increase in dust concentrations in the atmosphere at c.12,700 cal. BP. The shaded area shows the 0.5 maximum counting error (MCE) from theGreenland Ice Core Chronology 2005 (GICC05) (Rasmussen et al., 2006) equivalent to1s uncertainty (68% confidence interval). Ice core records and events are presented onthe GICC05 age scale after converting the b2k age to years BP.
and environmental conditions. The relationship between climatechange and environmental shifts, likewise the concept of criticalbiotic thresholds, is however complex and problematic to disen-tangle. Apart from major climate parameters (e.g. temperature andprecipitation), other factors such as soil characteristics, plant di-versity, and plant types define the sensitivity and the stability of theinteraction between vegetation and climate. Conceptual simula-tions have for example demonstrated that within an ecosystem thedistribution of plant species sensitive to changes in precipitationversus more resilient plant types can attenuate or enhance thestability of the vegetation with respect to climate change, driving,to a large extent, the evolution of a terrestrial ecosystem (e.g.Claussen et al., 2013). Additionally, the trade off between biologicalproductivity and population immigration is another importantfactor that determines the time response of an ecosystem to rapidclimate shifts (Ammann et al., 2000).
Even though the diverse aspects regulating the stability ofterrestrial ecosystems are important pacing factors across majorclimate boundaries, differences in the duration of the vegetationshifts associated with these factors are probably relatively smalland blended within the resolution of the records used in this study,and accordingly within the uncertainties that accompany our age-modelling approach. In fact, the observed duration of the transitionin the pollen stratigraphies seems to be rather a function of thepollen sampling resolution, which depends upon the specificsedimentation rates, with lower accumulation resulting in a longerapparent duration of the transition (Fig. 5). By comparing theduration of the transition at each site with the relative
Fig. 5. Maximum duration of the ALeYD pollen-zone transition for each site plottedagainst mean sedimentation rate (dots), showing the effects of sedimentary processeson the observed duration of the transitions. Mean sedimentation rates were derivedfrom Bacon-based output information and were calculated for each site between thetwo adjacent pollen sampling levels that straddle the ALeYD boundary. Maximumdurations refer to the error margins accounting for both age-modelling and pollensampling uncertainties. Bars show the range of estimated accumulation rate betweenthe two sampling levels crossing the transition.
F. Muschitiello, B. Wohlfarth / Quaternary Science Reviews 109 (2015) 49e56 55
accumulation rate, we observe that a hypothetical annuallyresolved record would probably yield a duration of the vegetationtransition on the order of several decades (Fig. 5). This would stillmake it difficult to estimate the true temporal dependency of theecosystem shift to secondary catchment-contingent factors. Weconclude that depositional factors, in concert with the specific ageuncertainty associated with each position of the 14C dates along thecalibration curve (e.g. larger uncertainties on the radiocarbonplateau), disguise the exact duration of environmental transitionsin the sedimentary record and therefore result in a relatively largeapparent duration of climate events.
3.5. Phasing of the vegetation shifts
The new age determinations for the ALeYD pollen-zoneboundary indicate that environmental shifts in Northern Europeoccurred in the following chronological and spatial order: at13,016 ± 98 cal BP in Northern Ireland (56�N); at 12,957 ± 62 cal. BPin eastern Denmark (54�N); at 12,736 ± 32 cal. BP in westernNorway (60�N); and at 12,671 ± 78 cal. BP in southern Sweden(58�N). Thus, the results point at a mean temporal lag of c. 290years between sites located at 56e54�N and sites located at60e58�N.
Such a time discrepancy in vegetation response could beexplained in terms of the resilience of the vegetation to changes inseasonality. Ecosystems that were located farther north were underthe influence of relatively colder climatic background conditionsand experienced a shorter growing season during the AL intersta-dial as compared to ecosystems located farther south. For instance,the catchment at Kråkenes and Lake Madtj€arn was influenced bycoastal sea ice and cold katabatic winds coming off the ice sheetduring most of the year. The FIS probably acted as a climatic buffer,subduing the seasonal cycle and the amplitude of temperaturefluctuations over the year. By contrast, Lake Gammelmose andSluggan Bog, which are located farther to the south, were notinfluenced by a nearby ice sheet, but were rather exposed to theinfluence of seasonal shifts in storm track trajectories over theNorth Atlantic Ocean. We thus hypothesize that a shortening of thewarm season that accompanied the transition from AL into YD hada stronger impact upon the vegetation at 56e54�N than at60e58�N. Local biotic thresholds would therefore have beencrossed earlier at Lake Gammelmose and Sluggan Bog than atKråkenes and Madtj€arn.
3.6. Mechanisms for the initiation of the Younger Dryas in NorthernEurope
The relatively early vegetation response in Northern Ireland at13,016 ± 98 cal. BP and in eastern Denmark at 12,957 ± 62 cal. BP isintriguing, but seems to compare in time to the late AL temperaturedecrease observed in other terrestrial sites in the British Isles(Mayle et al., 1999; Elias and Matthews, 2013). Paleoceanographicreconstructions from the North Sea (Koç Karpuz and Jansen, 1992;Klitgaard-Kristensen et al., 2001) and the Skagerrak-Kattegat (Jiangand Nordberg, 1996; Bod�en et al., 1997; Jiang et al., 1998) show awidespread and strong increase in ice rafting and incursion ofrelatively fresher water around 13,000 cal. BP. This has beenattributed to enhanced melting of the southern margin of the FIS(Jiang and Nordberg, 1996; Bod�en et al., 1997; Jiang et al., 1998), andalso to the drainage of the Baltic Ice Lake, which took place a fewhundred years before the end of the AL (e.g. Bj€orck, 1979; Bj€orck,1981, 1995).
We suggest that the early change in vegetation in NorthernIreland and eastern Denmark reflects an “early” cooling signinduced by the input of freshwater, which was triggered by ice
sheet instability at the end of the relatively warm interstadialcomplex, i.e. Late AL pollen zone/Greenland Interstadial 1a (GI-1a)(Figs. 3 and 4).
At sites located farther north critical summer thresholds werecrossed later, at c. 12,600e12750 cal. BP, when the ice marginexpanded and a large-scale atmospheric reorganization tookplace (Rach et al., 2014). This second phase of environmentalshifts occurred concomitantly with an increase in long-rangemineral dust concentrations in Greenland ice cores (Figs. 3 and4), which suggests the establishment of full stadial climateconditions and a stadial atmospheric circulation regime (Ruthet al., 2007).
A scenario characterised by an early stage of cooling that headstowards a climate reversal is in line with proxy evidence fromwestern Norway and southern Sweden, indicating a gradual, butsteady trend towards lower temperatures during the Late AL pollenzone, followed by a shift in vegetation several decades later. This isevident at Kråkenes, where early signs of summer cooling havebeen observed about 100 years earlier than the start of the local YD(Birks et al., 2000). Similarly, palaeobotanical records from south-ern Sweden indicate that climate cooling started already during thelate AL pollen zone (Bj€orck and M€oller, 1987). This gives support tothe idea that the cooling that begun at c. 12,900e13100 cal. BP wasnot sufficiently pronounced north of 56e54�N to allow criticalthresholds of summer temperatures to be crossed, at least until c.12,600e12750 cal. BP.
4. Conclusions
Assuming that the definitions of the local/regional pollen zoneboundaries at each site are reliable, the present analysis indicatesthat the timing of local vegetation responses at the ALeYD transi-tion across Northern Europe was asynchronous. We identified atime-transgressive and geographically consistent signal of changesthat we ascribe to distinct and separate physical events that tookplace across Northern Europe and potentially across parts of theNorth Atlantic. We suggest that the early phase of vegetationchanges recorded at 56e54�N (c. 12,900e13100 cal. BP) was causedby a gradual regional cooling that occurred in response to enhancedfreshwater outflow into the Nordic Seas. This freshwater forcingwas probably associated with calving at the southern margin of theFIS during the final warm stage of the AL interstadial (GI-1a inGreenland). In contrast, the vegetation shifts recorded at 60e58�Noccurred significantly later (c. 12,600e12750 cal. BP) and wererelated to further cooling and ice sheet expansion associated with alarge-scale climate reorganization. We expect that future well-dated proxy reconstructions from intermediate sites between c.60�N and 55�N will corroborate our findings and will help to betterunderstand the chronological sequence of events that took place inthis region. Our results show the importance of establishing robustterrestrial radiocarbon chronologies and coherent age models, tobetter understand spatial and temporal differences at the inceptionof a cold climatic regime.
Acknowledgements
The authors gratefully acknowledge the helpful comments of JanMangerud and an anonymous reviewer.
Appendix A. Supplementary data
Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.quascirev.2014.11.015.
F. Muschitiello, B. Wohlfarth / Quaternary Science Reviews 109 (2015) 49e5656
References
Abbott, P.M., Davies, S.M., 2012. Volcanism and the Greenland ice-cores: the tephrarecord. Earth-Sci. Rev. 115, 173e191.
Ammann, B., Birks, H., Brooks, S.J., Eicher, U., von Grafenstein, U., Hofmann, W.,Lemdahl, G., Schwander, J., Tobolski, K., Wick, L., 2000. Quantification of bioticresponses to rapid climatic changes around the Younger Dryas d a synthesis.Palaeogeogr. Palaeoclimatol. Palaeoecol. 159, 313e347.
Andresen, C., Bj€orck, S., Bennike, O., Heinemeier, J., Kromer, B., 2000. What do D14Cchanges across the Gerzensee oscillation/Gl-1b event imply for deglacial os-cillations? J. Quat. Sci. 15, 203e214.
Bakke, J., Lie, Ø., Heegaard, E., Dokken, T., Haug, G.H., Birks, H.H., Dulski, P., Nilsen, T.,2009. Rapid oceanic and atmospheric changes during the Younger Dryas coldperiod. Nat. Geosci. 2, 202e205.
Bard, E., Rostek, F., Turon, J.-L., Gendreau, S., 2000. Hydrological impact of Heinrichevents in the subtropical northeast Atlantic. Science 289, 1321e1324.
Birks, H.H., Ammann, B., 2000. Two terrestrial records of rapid climatic changeduring the glacialeHolocene transition (14,000e9,000 calendar years BP) fromEurope. Proc. Natl. Acad. Sci. 97, 1390e1394.
Birks, H.H., Battarbee, R., Birks, H., 2000. The development of the aquatic ecosystemat Kråkenes Lake, western Norway, during the late glacial and early Holocene-asynthesis. J. Paleolimnol. 23, 91e114.
Bj€orck, S., 1979. Late Weichselian Stratigraphy of Blekinge, SE Sweden, and WaterLevel Changes in the Baltic Ice Lake (thesis 7). University of Lund, Departmentof Quaternary Geology, Lund, p. 248.
Bj€orck, S., 1981. A stratigraphic study of Late Weichselian deglaciation, shoredisplacement, and vegetation history in south-eastern Sweden. Foss. Strat. 14,1e93.
Bj€orck, S., 1995. A review of the history of the Baltic Sea, 13.0-8.0 ka BP. Quat. Int. 27,19e40.
Bj€orck, S., M€oller, P., 1987. Late Weichselian environmental history in southeasternSweden during the deglaciation of the Scandinavian ice sheet. Quat. Res. 28,1e37.
Bj€orck, S., Kromer, B., Johnsen, S., Bennike, O., Hammarlund, D., Lemdahl, G.,Possnert, G., Rasmussen, T., Wohlfarth, B., Hammer, C., 1996. Synchronizedterrestrial-atmospheric deglacial records around the North Atlantic. Science274, 1155e1160.
Bj€orck, S., Bennike, O., Possnert, G., Wohlfarth, B., Digerfeldt, G., 1998a. A high-resolution 14C dated sediment sequence from southwest Sweden: age com-parisons between different components of the sediment. J. Quat. Sci. 13, 85e89.
Bj€orck, S., Walker, M.J., Cwynar, L.C., Johnsen, S., Knudsen, K.L., Lowe, J.J.,Wohlfarth, B., 1998b. An event stratigraphy for the Last Termination in theNorth Atlantic region based on the Greenland ice-core record: a proposal by theINTIMATE group. J. Quat. Sci. 13, 283e292.
Blaauw, M., Christen, J.A., 2011. Flexible paleoclimate age-depth models using anautoregressive gamma process. Bayesian Anal. 6, 457e474.
Blaauw, M., Christen, J.A., 2013. Bacon Manualev2. 2. http://chrono.qub.ac.uk/blaauw/.
Bod�en, P., Fairbanks, R.G., Wright, J.D., Burckle, L.H., 1997. High-resolution stableisotope records from southwest Sweden: the drainage of the Baltic Ice Lake andYounger Dryas ice margin oscillations. Paleoceanography 12, 39e49.
Bradley, R.S., England, J.H., 2008. The Younger Dryas and the sea of ancient ice. Quat.Res. 70, 1e10.
Brauer, A., Endres, C., Günter, C., Litt, T., Stebich, M., Negendank, J.F., 1999. Highresolution sediment and vegetation responses to Younger Dryas climate changein varved lake sediments from Meerfelder Maar, Germany. Quat. Sci. Rev. 18,321e329.
Brauer, A., Endres, C., Zolitschka, B., Negendank, J.F.W., 2000. AMS radiocarbon andvarve chronology from the annually laminated sediment record of Lake Meer-felder Maar, Germany. Radiocarbon 42, 355e368.
Broecker, W.S., 1998. Paleocean circulation during the last deglaciation: a bipolarseesaw? Paleoceanography 13, 119e121.
Bronk Ramsey, C., 2008. Deposition models for chronological records. Quat. Sci. Rev.27, 42e60.
Christen, J.A., P�erez, S., 2009. A new robust statistical model for radiocarbon data.Radiocarbon 51, 1047e1059.
Claussen, M., Bathiany, S., Brovkin, V., Kleinen, T., 2013. Simulated climate-vegetation interaction in semi-arid regions affected by plant diversity. Nat.Geosci. 6, 954e958.
Duplessy, J.-C., Labeyrie, L., Arnold, M., Paterne, M., Duprat, J., van Weering, T.C.,1992. Changes in surface salinity of the North Atlantic Ocean during the lastdeglaciation. Nature 358, 485e488.
Elias, S.A., Matthews, I.P., 2013. A comparison of reconstructions based on aquaticand terrestrial beetle assemblages: Late glacialeEarly Holocene temperaturereconstructions for the British Isles. Quat. Int. 341, 69e79.
Hua, Q., Barbetti, M., Fink, D., Kaiser, K.F., Friedrich, M., Kromer, B., Levchenko, V.A.,Zoppi, U., Smith, A.M., Bertuch, F., 2009. Atmospheric 14C variations derivedfrom tree rings during the early Younger Dryas. Quat. Sci. Rev. 28, 2982e2990.
Hughen, K.A., Southon, J.R., Lehman, S.J., Overpeck, J.T., 2000. Synchronous radio-carbon and climate shifts during the last deglaciation. Science 290, 1951e1954.
Jiang, H., Nordberg, K., 1996. Late Weichselian environmental changes of thesouthern Kattegat, Scandinavia, inferred from diatom records. In: GeologicalSociety, London, Special Publications 111, pp. 245e260.
Jiang, Hui, Nils-Olof, Svensson, Svante, Bj€orck, 1998. Meltwater discharge to theSkagerrakeKattegat from the Baltic Ice Lake during the Younger Dryas interval.Quat. Res. 49, 264e270.
Karpuz, Nalan Koç, Jansen, Eystein, 1992. A high-resolution diatom record of the lastdeglaciation from the SE Norwegian Sea: documentation of rapid climaticchanges. Paleoceanography 7, 499e520.
Klitgaard-Kristensen, D., Sejrup, H., Haflidason, H., 2001. The last 18 kyr fluctuationsin Norwegian Sea surface conditions and implications for the magnitude ofclimatic change: evidence from the North Sea. Paleoceanography 16, 455e467.
Kruschke, J.K., 2013. Bayesian estimation supersedes the t test. J. Exp. Psychol.General 142, 573.
Lane, C.S., Brauer, A., Blockley, S.P., Dulski, P., 2013. Volcanic ash reveals time-transgressive abrupt climate change during the Younger Dryas. Geology.G34867. 34861.
Lohne, Ø.S., Mangerud, J., Birks, H.H., 2013. Precise 14C ages of the Vedde andSaksunarvatn ashes and the Younger Dryas boundaries from western Norwayand their comparison with the Greenland Ice Core (GICC05) chronology. J. Quat.Sci. 28, 490e500.
Lohne, Ø.S., Mangerud, J.A.N., Birks, H.H., 2014. IntCal13 calibrated ages of theVedde and Saksunarvatn ashes and the Younger Dryas boundaries from Krå-kenes, western Norway. J. Quat. Sci. 29, 506e507.
Lowe, J., Walker, M., Scott, E., Harkness, D., Bryant, C., Davies, S., 2004. A coherenthigh-precision radiocarbon chronology for the Late-glacial sequence at SlugganBog, Co. Antrim, Northern Ireland. J. Quat. Sci. 19, 147e158.
Lowe, J.J., Rasmussen, S.O., Bj€orck, S., Hoek, W.Z., Steffensen, J.P., Walker, M.J.,Yu, Z.C., 2008. Synchronisation of palaeoenvironmental events in the NorthAtlantic region during the Last Termination: a revised protocol recommendedby the INTIMATE group. Quat. Sci. Rev. 27, 6e17.
Mayle, F., Bell, M., Birks, H.H., Brooks, S., Coope, G., Lowe, J., Sheldrick, C., Shijie, L.,Turney, C., Walker, M., 1999. Climate variations in Britain during the Last Gla-cialeHolocene transition (15.0e11.5 cal ka BP): comparison with the GRIP ice-core record. J. Geol. Soc. 156, 411e423.
Mangerud, J., Andersen, S.T., Berglund, B.E., Donner, J.J., 1974. Quaternary stratig-raphy of Norden, a proposal for terminology and classification. Boreas 3,109e126.
McManus, J., Francois, R., Gherardi, J.-M., Keigwin, L., Brown-Leger, S., 2004.Collapse and rapid resumption of Atlantic meridional circulation linked todeglacial climate changes. Nature 428, 834e837.
Rach, O., Brauer, A., Wilkes, H., Sachse, D., 2014. Delayed hydrological response toGreenland cooling at the onset of the Younger Dryas in western Europe. Nat.Geosci 7, 109e112.
Rasmussen, S.O., Andersen, K.K., Svensson, A., Steffensen, J.P., Vinther, B.M.,Clausen, H.B., Siggaard-Andersen, M.L., Johnsen, S.J., Larsen, L.B., Dahl-Jensen, D., 2006. A new Greenland ice core chronology for the last glacialtermination. J. Geophys. Res. Atmos. (1984e2012) 111.
Reimer, P.J., Bard, E., Bayliss, A., Beck, J.W., Blackwell, P.G., Ramsey, C.B.,Grootes, P.M., Guilderson, T.P., Haflidason, H., Hajdas, I., 2013. IntCal13 andMarine13 radiocarbon age calibration curves 0e50,000 years cal BP. Radio-carbon 55, 1869e1887.
Ruth, U., Bigler, M., R€othlisberger, R., Siggaard-Andersen, M.-L., Kipfstuhl, S., Goto-Azuma, K., Hansson, M.E., Johnsen, S.J., Lu, H., Steffensen, J.P., 2007. Ice coreevidence for a very tight link between North Atlantic and east Asian glacialclimate. Geophys. Res. Lett. 34, L03706.
Schwander, J., Eicher, U., Ammann, B., 2000. Oxygen isotopes of lake marl at Ger-zensee and Leysin (Switzerland), covering the Younger Dryas and two minoroscillations, and their correlation to the GRIP ice core. Palaeogeogr. Palae-oclimatol. Palaeoecol. 159, 203e214.
Steffensen, J.P., Andersen, K.K., Bigler, M., Clausen, H.B., Dahl-Jensen, D., Fischer, H.,Goto-Azuma, K., Hansson, M., Johnsen, S.J., Jouzel, J., 2008. High-resolutionGreenland ice core data show abrupt climate change happens in few years.Science 321, 680e684.
Wohlfarth, B., 1996. The chronology of the Last Termination: a review ofradiocarbon-dated, high-resolution terrestrial stratigraphies. Quat. Sci. Rev. 15,267e284.
