Multiproxy comparison of oceanographic temperature during Heinrich Events in the eastern subtropical Atlantic

Earth and Planetary Science Letters 310 (2011) 45–58

Contents lists available at ScienceDirect

Earth and Planetary Science Letters

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

Multiproxy comparison of oceanographic temperature during Heinrich Events in theeastern subtropical Atlantic

Genna M. Patton a,⁎, Pamela A. Martin a, Antje Voelker b, Emilia Salgueiro b

a University of Chicago, Department of the Geophysical Sciences, 5734 S. Ellis Ave, 60637 Chicago, USAb Unidade de Geologia Marinha, Laboratorio Nacional de Energia e Geologia, Estrada da Portela, Zambujal, 2610-143 Amadora, Portugal

⁎ Corresponding author.E-mail address: [email protected] (G.M. Patton

0012-821X/$ – see front matter © 2011 Elsevier B.V. Adoi:10.1016/j.epsl.2011.07.028

a b s t r a c t

Article history:Received 9 February 2011Received in revised form 21 July 2011Accepted 28 July 2011Available online 30 August 2011

Editor: P. DeMenocal

Keywords:AtlanticforaminiferaMg/Ca-paleothermometryHeinrich Events

Abrupt climate change episodes associated with catastrophic calvings of the Laurentide and European IceSheets during the last glacial, Heinrich Events (HE), had far-reaching effects in the Atlantic. We havemeasured minor and trace element concentrations in planktonic foraminiferal calcite in a core from theIberian Margin (MD95-2040) and a core from the Gulf of Cadiz (MD99-2339) during HE 1, 4 and 5 to explorethe use of the Mg–temperature proxy for constraining regional temperature changes during these uniqueperiods of ice-rafting. High sedimentation rates throughout the length of both cores allow a detailedreconstruction of temperature during abrupt events. Mg/Ca-derived temperatures reveal ~5–13 °C amplitudeof cooling during HE1, the largest and most well defined event in this region. The northern site most likelyexperienced the higher end of this amplitude range. Cooling implied by the Mg/Ca data is within the range ofassemblage-derived sea surface temperature (SST) changes from the same cores showing ~10 °Cperturbations for each HE but the Mg/Ca data imply smaller changes during HE4 and HE5. Cooling estimatedfrom unsaturated alkenone ratios (~3 °C) represent the smallest changes. The prominence of HE1 in thisregion is consistent with the changes over thewider North Atlantic that show colder temperatures during HE1vs. HE4. We use Cd/Ca to reconstruct nutrient concentrations and find very low values (0.02–0.04 μmol/mol)at both locations during HE. These low values are consistent with core top planktonic Cd/Ca from each regionsuggesting that the cooling during HE was unlikely to be caused by more intense upwelling than in themodern. Temperature comparisons of other proxies, alkenone (U37

K′) and assemblage data, from the same coresuggest that Mg/Ca paleothermometry is sensitive to the relative size of each HE. This sensitivity is capable ofhighlighting other variables that influence the impact of Heinrich Events on SST, including the backgroundclimate state.

).

ll rights reserved.

© 2011 Elsevier B.V. All rights reserved.

1. Introduction

During the last glacial and deglacial periods, Greenland experi-enced rapid changes in surface air temperature (Dansgaard–Oeschgeror “D–O” Events), characterized by an abrupt, decadal scale warmingfollowed by a gradual, millennial scale cooling (Dansgaard et al., 1993;Grootes and Stuiver, 1997). The coldest stadials of these events werepunctuated by catastrophic ice sheet calvings, especially of theLaurentide Ice Sheet, approximately every 7.2 kyr (Sarnthein et al.,2000). These massive calving episodes, Heinrich Events (HE), areidentified by layers of ice rafted debris as far south as the northernsubtropics and are associated with strong sea surface temperaturefluctuations in the North Atlantic (Cortijo et al., 1997; Van Kreveldet al., 2000). The climatic significance of HE is also evidenced by far-reaching geographic locations that show correlative events along the

coasts of Brazil and Florida and in the Mediterranean Sea (Arz et al.,1998; Broecker, 2006; Cacho et al., 1999; Grimm et al., 2006).

SST for HE have been derived mostly from faunal assemblage datawith some additional data from alkenone U37

K′ and Mg/Ca. SST estimatesderived from planktonic foraminiferal assemblage data suggest thatduring HE, summer SSTs dropped to 3–6 °C in the subarctic NorthAtlantic (Chapman andMaslin, 1999; Cortijo et al., 1997) and to 5–10 °Calong the IberianMargin in the subtropicalNorthAtlantic (deAbreu et al.,2003; McManus et al., 2004; Salgueiro et al., 2010; Schoenfeld and Zahn,2000). A curious feature of the subtropical SSTs implied by the faunalassemblage data along the Iberian Margin is that the data imply nearlythe same large cooling during each HE, despite disparate thicknessesof ice-rafted debris (IRD) layers in the North Atlantic for different HE.In contrast, the dramatic shifts in subtropical SSTs of as much as10 °C, specifically during HE1 (~17–15 ka), HE4 (~40–38 ka) andHE5 (~47–45 ka) (de Abreu et al., 2003; Voelker et al., 2006), are notconsistent with SSTs derived from alkenone U37

K′ which imply littlecooling during the glacial HE in this region (Bard et al., 2000; Paillerand Bard, 2002).

46 G.M. Patton et al. / Earth and Planetary Science Letters 310 (2011) 45–58

Lower SSTs throughout the North Atlantic during HE are consistent,however, with modeling experiments in which freshwater released intoareasof deepwater formation in theNorthAtlantic results in aweakening,or complete shutdown of North Atlantic DeepWater (NADW) circulation(Ganopolski and Rahmstorf, 2001). Model runs that explore a range offreshwater forcing scenarios predict lower SSTs throughout the NorthAtlantic that vary with both the amplitude and timing of the freshwaterinput. The degree of North Atlantic cooling during these events and theeffects on theAtlanticMeridionalOverturningCirculation (AMOC) systemare likely also to depend on the background climate state (Prange et al.,2004). Proxy data can help in evaluating the environmental changesassociated with HE during both the glacial and deglacial states and mayhelp to constrain sensitivity of the AMOC to the background climate.

Here, we use Mg/Ca data from planktonic foraminifera to furtherexplore SST changes during HE. Mg/Ca has been widely applied inpaleotemperature reconstructions, particularly over the last glacialcycle, but have been only looked at during HE in a limited number ofstudies (Peck et al., 2008; Skinner and Elderfield, 2005). Complicationsin applying Mg-paleothermometry, in general, are the potentialinfluence of dissolution and salinity on shell Mg/Ca as well as thenecessity for species-specific Mg–temperature equations. Dissolution isunlikely to be amajor influence in the cores in this study given that bothcores are located above the regional lysocline and contain well-preserved foraminifer throughout. In these two cores, shell weightsper individual are high (10–14 μg), indicating well preserved speci-mens, however there is variation in normalized shell weight duringHE1in core MD95-2040 (see Section 4.1.1). Any potential salinity effect,however, could have a larger impact in temperature reconstructionsgiven thatHE could be accompanied by significant local salinity changesdue to the melting of ice; yet, the magnitude of the salinity effect onMg/Ca is notwell defined. Herewedealwith uncertainties in theMg–temperature relationship, including salinity, by using multiple calibra-tion equations to constrain a range of temperatures and compare therange to other temperature proxies. We derive one of the calibrationequations from published core top data for the North Atlantic.

Comparing new foraminiferal minor and trace element data to thesuite of existing paleoceanographic proxy data including foraminiferalδ18O, assemblage data and alkenone U37

K′ during HE1, HE4 and HE5 inthis region helps to constrain the climate fluctuations during HE. Weuse a published ~15 ka record of Mg/Ca variability from the sameregion to put the temperature changes during HE into the context ofclimate variability during the last glacial and deglacial (Skinner andElderfield, 2005). We also compile SST proxy data for HE1 and HE4 todevelop a better picture of the impact of these two events in thebroader North Atlantic. Finally, we present the first measurements ofplanktonic Cd/Ca from HE. We investigate whether planktonic Cd/Ca,a proxy for nutrient concentration, shows evidence for changes inupwelling, which is one mechanism by which SSTs could changeamong the coastal records as well as in comparison to changes in theNorth Atlantic, more generally, during HE.

We show that, in this region, each temperature proxy impliesdifferent amplitudes of change during Heinrich Events. Despite thedifferences among the proxies, the dominant trend among the proxydata suggests that the climate changes in the North Atlantic were largerduringHE1,which occurs over the transition from glacial to interglacial,than during HE4 and HE5, which are solidly in the glacial. Temperaturechanges in the coastal region we examine here appear to have beenlarger than changes in the open ocean during bothHE1 andHE4, but areunlikely to be attributed to more intense coastal upwelling.

2. Methods

2.1. Core locations, age models and sampling

We obtained samples from two cores, MD95-2040 (40.58°N,9.86°W; 2465 m) along the western Iberian Margin (the northwest

margin) and MD99-2339 (35.88°N, 7.53°W; 1170 m) along thesouthern Iberian Margin, in the Gulf of Cadiz (Fig. 3). Analysis of theseIMAGES project sediment cores reveals IRD during each of the fivemostrecent HE, indicating the presence of icebergs in this region (de Abreuet al., 2003; Voelker et al., 2006). The presence of IRD allows tightcorrelation of regional oceanographic changeswith theHeinrich Events.

Sample depths from cores MD95-2040 and MD99-2339 wereconverted to age using published age models (Salgueiro et al., 2010;Voelker et al., 2006). In both cores, Heinrich Events were sampled toyield a resolution of ~250 yr, which was approximately every 9 cm incore MD95-2040 and every 6 cm in core MD99-2339. Averagesedimentation rates during each HE at site MD95-2040 were calculatedto be 30–40 cm/kyr, while between HE sedimentation rates wereapproximately 20 cm. Sedimentation rate at site MD99-2339 during HEranges from30 to45 cm/kyr andoutsideofHE ranges anywhere from20to 70 cm/kyr.ModernMg/Ca-based SSTs are reconstructed from surfacesamples near each of the core sites (Salguiero et al., in prep).

To provide additional context for the Mg/Ca variations we reportfor HE in this region, we plot our Mg/Ca data from HE1 from MD95-2040 and MD99-2339 with a published high resolution Globigerinabulloides record from core MD99-2334 (37°47′N, 10°10′W, 3146 m;Skinner and Elderfield, 2005) for 10 ka to 22 ka. Mg/Ca for G. bulloidesdata are presented graphically versus depth in Skinner and Elderfield(2005). We plot the data using a published age model for core MD99-2334 based on radiocarbon dates to convert depths to ages (Skinnerand Shackleton, 2004).

2.2. Samples and sample preparation

Our reconstructions in both cores are based on the planktonicspecies, G. bulloides. Analysis of surface sediment data suggest thatG. bulloides are related to upwelling and thus most abundant duringsummers in this region (Salgueiro et al., 2008). G. bulloides are presentthroughout the length of core MD95-2040 (de Abreu et al., 2003),however, small available sample sizes prevented replicate analyses.The site of MD99-2339 recorded a more diverse planktonic forami-nifera fauna than core MD95-2040 and abundance of specific speciesvaried significantly between HE (Voelker, unpubl. data). Althoughscarce in most of the core, G. bulloides, makes up approximately 50% ofthe planktonic assemblage during HE (Voelker, unpubl. data). In coreMD99-2339, we combined intervals prior to and after HE (i.e.intervals with low abundance of G. bulloides) with one or twoadjacent depths to yield samples large enough to clean and analyze.Samples from combined intervals before and after HE have inconsis-tent resolution and span, on average, 10 cm (200–500 yr).

Approximately 30 individual G. bulloideswere picked from the 255–350 μm size fraction of disaggregated sediment yielding sample sizesbetween 0.3 and 0.4 mg. Samples from both cores were crushed beforecleaning for trace element analysis such that the inner chamber wallswere exposed. Any obvious mineral contaminants were removed.Cleaning followed the protocol of Boyle (1981) with modifications asoutlined in Lea andMartin (1996). Themethod involves 5 steps: n-pureand methanol rinses, reduction to remove authigenic contaminants,oxidation to remove organics, heating and final acid leach. The cleaningprocedure was modified for samples ~0.1 mg. Samples of this size areone third of theoptimalweight forwhich the standard cleaningprotocolis designed. A significant amount of sample is lost during the reductionstep; therefore, for small samples, we reduced the amount of bufferedhydrazine in the reduction step to 33 μl instead of the typical 100 μl(Bian and Martin, 2010).

After cleaning, samples were dissolved in a spiked acid solution andanalyzed by single collector magnetic sector-ICPMS to determine tracemetal to Ca ratios at the analytical lab of the Marine Science Institute attheUniversity of California, Santa Barbara, following amethod similar tothat outlined in Lea and Martin (1996) for Quadrapole ICP-MS butmodified for the magnetic sector instrument. In addition to Mg/Ca,

47G.M. Patton et al. / Earth and Planetary Science Letters 310 (2011) 45–58

Sr/Ca, and Cd/Ca, we measured Al/Ca, Mn/Ca and Fe/Ca as potentialindicators of authigenic contaminants. Only Mg/Ca and Cd/Ca arepresented in detail here. Precision of measurements estimated from thestandard deviation of replicate analyses of a consistency standard is 0.7%for Mg/Ca and 7% for Cd/Ca (one outlying Cd/Ca value was excluded).Precision of theMn/Ca, Fe/Ca, and Al/Ca is better than 5%. Mn/Ca valueswere all lower than 55 μmol/mol. Fe/Ca and Al/Ca were generallylower than 100 μmol/mol, although about 15% were higher than100 μmol/mol. Samples with high Fe/Ca and Al/Ca were well correlatedbut there was no obvious correlation between samples with high Fe/Ca(and high Al/Ca) and Mg/Ca or Cd/Ca values; thus, we did not excludethose points from the dataset.

3. Results

3.1. Mg/Ca

Mg/Ca of G. bulloides during all three HE are lower than Holocenevalues at the two sites investigated. The lowest values are found at thenorthernmost site (MD95-2040) during HE1 (Fig. 1). HE1 shows themost well-defined and structured cooling of HE1, HE4, and HE5 in thiscore. Near the site of MD95-2040, Mg/Ca of surface G. bulloides samplesare 2.6 mmol/mol; just prior to HE1 (~17.6 ka) Mg/Ca values are

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

Mg/

Ca

mm

ol/m

ol

0

0.02

0.04

0.06

0.08

0.10

Cd/

Ca

umol

/mol

0

10

20

30

40

IRD

(#

> 2

50um

/g)

10 12 14 16 18 20 220

20

40

60

80

100

% p

achy

(s)

A)

B)

C)

D)

Holocene value

HE1

MD95-2040 (40.58

+/− 2σ

+/− 2σ

Fig. 1. Trace metal data for MD95-2040 plotted versus age. (A) Mg/Ca, (B) Cd/Ca, (C) IRD and(A) and (B) represent ±2 standard deviations of consistency standard replicates. Core topMD95-2040 denoted by a horizontal dashed line provide a Holocene baseline for each trace e(2010) age model for MD95-2040 is based on a combination of δ18O stratigraphy, 14C datinget al. (2003). Salgueiro et al. (2010) adjusted HE2 ages and at the base of Greenland Inters

2.0 mmol/mol (Fig. 1A). There is a distinctdrop in theMg/Ca at the onsetof HE1 that reaches a minimum value of 0.9 mmol/mol. Following HE1,the Mg/Ca gradually increases until it reaches a value of 1.7 mmol/mol(~12.7 ka). The dip in Mg/Ca associated with HE1 is well correlatedwith the intervals of elevated IRD. HE4 and HE5 in MD95-2040 areless well-defined than HE1. Prior to HE4, there is a brief minimum inMg/Ca approximately equivalent to the lowest measured value(1.3 mmol/mol) during HE4. Values leading into HE4 are comparableto the lowest values duringHE4, but theMg/Ca values showa steady riseto a value of 1.9 mmol/mol correlated with the end of HE4. The ampli-tude of Mg/Ca change is much smaller than for HE1. Similarly, HE5 atthis location shows only slightly lower values compared with theintervals prior to, and after the event.

Mg/Ca values are generally higher at the site of MD99-2339 andshow smaller changes during the HE relative to MD95-2040 (Fig. 2A).Similar to the data fromMD95-2040, lowMg/Ca values aremost clearlydefined during HE1. Surface G. bulloides samples near this site are2.8 mmol/mol. Prior to HE1, Mg/Ca are ~3.4 mmol/mol, higher than theHolocene surface samples from nearby the core site. A one-point spikeimmediately before the onset of HE1may be an outliner, although thereis no analytical or statistical reason to eliminate the data point in thissmall data set. During HE1, theMg/Ca values are significantly lower andextremely constant with an average of 1.8 mmol/mol (Fig. 2A). Values

36 38 40 42

Age (ka)43 45 47 49

HE4 HE5

°N, 9.86°W, 2465m)

(D)% abundance of N. pachyderma sinistral (de Abreu et al., 2003). Error bars in panelsvalues from MD04-2814 (40.06°N, 9.86°W, 2449) (courtesy of Emilia Salguiero) nearlement. The colored vertical bars represent Heinrich Event intervals. The Salgueiro et al.and synchronization of the assemblage SST record and GISP2 δ18O data, after de Abreu

tadial 11.

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

Mg/

Ca

mm

ol/m

ol

0

0.02

0.04

0.06

0.08

0.10

Cd/

Ca

umol

/mol

0

2

4

6

8

10

IRD

(#

> 2

50um

/g)

10 12 14 16 18 20 220

2

4

6

8

10

% p

achy

(s)

38 39 40Age (ka)

43 45 47 49

A)

B)

C)

D)

HE1 HE4 HE5

Holocene

value

MD99-2339 (35.88°N, 7.53°W, 1170m)

+/− 2σ

+/− 2σ

Fig. 2. Tracemetal data for MD99-2339 plotted versus age. (A)Mg/Ca, (B) Cd/Ca, (C) IRD and (D)% abundance ofN. pachyderma sinistral (C and D fromVoelker et al., 2006). Error barsin panels (A) and (B) represent ±2 standard deviations of consistency standard replicates. Core top values fromM39029-6 (36.05°N, 8.23°W, 1918 m) (courtesy of Emilia Salguiero)near MD99-2339 denoted by a horizontal dashed line provide a Holocene baseline for each trace element. The age model of core MD99-2339 is based on calibrated 14C ages and twotuning points to the GISP 2 chronology in the vicinity of HE5 (Voelker et al., 2006). Note: in the middle vertical set of panels (HE4) the length of the x-axis is not proportional to time.N. pachyderma sinistral maximum abundance reaches 15.6% during HE4.

48 G.M. Patton et al. / Earth and Planetary Science Letters 310 (2011) 45–58

return to pre-HE1 values by at least 13.7 ka, the first data point after theHE. Due to difficulties of low abundances, additional depths youngerthan13.7 ka (more recent)were not robust enough to performanalyses.DuringHE4 there is a higher degree of variability,withminimumMg/Cavalues averaging 1.7 mmol/mol. Pre-HE5Mg/Ca is 2.4 mmol/mol. Thereis a highdegree of variabilityduring this period and asa result there is noclearly defined event associated with HE5 evidenced by the Mg/Ca.

We place the variations into a better temporal and regional frame-work by comparison of our data for HE1 to the published G. bulloidesMg/Ca data of Skinner and Elderfield (2005) for the last 10 to 22 ka(Fig. 3). Each core location shows an abrupt decrease during HE1consistent with lower temperature changes. The latitudinal gradientcreated by these three locations fits precisely into the context ofcolder (lower Mg/Ca) higher latitudes and progressively warmer(higher Mg/Ca) lower latitudes.

3.2. Cd/Ca

We were able to generate a small amount of Cd/Ca data from thesmall samples available for this study. Cd/Ca values for HE1, 4, and 5are all relatively low in both cores (Figs. 1B and 2B), consistent withlow Cd/Ca values measured in G. bulloides from surface samples nearthe core site (i.e. Holocene values of 0.02–0.04 μmol/mol). Prior to and

during HE1, values vary between 0.02 and 0.03 μmol/mol at siteMD95-2040. After HE1 (15 ka), values increase to slightly less than0.04 μmol/mol. HE4 at MD95-2040 has consistently low values thatoscillate between 0.01 and 0.03 μmol/mol prior to, during and afterthe event. Cd/Ca values are similarly low during HE5 but with twoanomalously high values at the end of the event.

The Cd/Ca values in core MD99-2339 are slightly lower than coreMD95-2040. Cd/Ca shows the most variability during HE1 with valuesranging between 0.01 and 0.04 μmol/mol (Fig. 2B). Values are slightlylower in HE4 than HE5 relative to HE1. The primary patterns thatemerge from both sets of data are: (1) small amplitude, but consistentlyhigher values surrounding HE1 (though not significantly higher) and(2) low values during all three HE, comparable to Holocene values. Thevariability is larger than the analytical reproducibility/precision.

4. Discussion

4.1. Proxy temperature calibration equations

4.1.1. Calculating a range of temperature estimates from Mg/CaWe begin our discussion with an examination of the set of Mg-

temperature equations that we use to constrain temperatures duringHE. Several Mg/Ca–temperature equations have been developed for

10 12 14 16 18 20 220.5

1.0

1.5

2.0

2.5

3.0

3.5

Age (ka)

Skinner (reductively adjusted)MD95-2040 (Most Northerly)MD99-2339 (Most Southerly)

usted)erly)

herly)

Mg/

Ca

(mm

ol/m

ol)

-12 -10 -8 -6 -4 -2 034

36

38

40

42

44

Longitude (°W)

Latit

ude

(°N

)

36°N38°N40°N

MD95-2040

MD99-2339A) B)

MD95-2040

MD99-2339

Fig. 3. Comparison of three Mg/Ca datasets during HE1. (A) Locations of MD99-2339 and MD95-2040 (this study) and MD99-2334 (Skinner and Elderfield, 2005). (B) Mg/Ca dataduring HE1 for each location in panel A. Core top values of Mg/Ca for each location are represented by a small dash on the y-axis. Skinner and Elderfield (2005) Mg/Ca data wasadjusted to account for different cleaning methods (Bian and Martin, 2010).

49G.M. Patton et al. / Earth and Planetary Science Letters 310 (2011) 45–58

G. bulloides using core top, sediment trap and culturing data (Clérouxet al., 2008; Elderfield and Ganssen, 2000; Lea et al., 1999; McConnelland Thunell, 2005). Calibrations for G. bulloides yield a Mg–T responseof ~6–10% increase per °C, withmost of the equations defining aMg–Tresponse of ~8–10% increase per °C (Table 1). As one estimate oftemperature we apply the Lea et al. (1999) calibration equation,which is based on the culturing work of Mashiotta et al. (1999):

Mg=Ca = 0:53 exp 0:10⁎ SSTAnnð Þ ð1Þ

where SSTAnn is themean annual sea surface temperature. Applicationof this equation to surface sediments from nearby our core locationsyields Holocene valuesmost consistentwith spring SSTs in this region,~15.9 °C vs. 15.4 °C for core MD95-2040 and ~16.6 °C vs. 17.9 °C forcore MD99-2339 (Salguiero, in prep). The Lea et al. (1999) equationhas a nearly identical Mg-T sensitivity (0.1% increase per °C) as themulti-species equation determined by Elderfield and Ganssen (2000),with the primary difference in the pre-exponential constant.Differences in the pre-exponential constant primarily reflect whetherSST or the calcification temperature is used in the regression. Thechoice between these two equations does not significantly change theamplitude of our estimated temperature oscillations (relative to coretop values in the region), only the absolute value of the temperatures.

For comparison with the published calibration equations, weexplored the regional Mg/Ca–temperature relationship for G. bulloidesusing published core top data we assembled from the North Atlantic(Cléroux et al., 2008; Yu et al., 2008). While application of a global

Table 1Coefficients for the calibration equations and core top temperatures.

Mg/Ca=A exp (B∗SST) Aa Bb MD95-2040c MD99-2339d

Cléroux et al. (2008) 0.78 0.082 14.6 15.6Lea et al. (1999) 0.53 0.1 15.9 16.6Mashiotta et al. (1999) 0.474 0.107 15.9 16.6McConnell and Thunell (2005) 1.2 0.057 13.5 14.9Elderfield and Ganssen (2000) 0.56 0.1 15.3 16.1Elderfield and Ganssen (2000)e 0.81 0.081 14.4 15.3Regional North Atlantic data(this study)

0.97 0.066 14.9 16.1

a Pre-exponential constant.b Exponential constant.c Core top temperatures determined from each proxy for MD95-2040.d Core top temperatures determined from each proxy for MD99-2339.e Modified by Anand et al. (2003).

calibration ought to be, in general, the most robust approach toquantifying past temperatures, the regional temperature responsemight vary from the global response due to parameters that have notyet been identified as an influence on Mg/Ca for this species in thelimited existing calibration data (for example, different regionalseasonality of G. bulloides or a salinity effect). In addition, the existingpublished calibration equations do not include the range of temper-atures found in the North Atlantic. Thus, it is useful to examineregional core top data. We examined isotopic temperatures (Tiso),seasonal SSTs, SSTAnn and mean annual salinities (SSSAnn). SSTs andSSSs were extracted from the 2005 World Ocean Atlas (WOA05)gridded dataset and are highly correlated (e.g. r2=0.81 for the annualdata) (Antonov et al., 2006; Locarnini et al., 2006). We performed asuite of linear multiple regressions using a combination of SSTs,calcification temperature derived from published shell δ18O (Tiso),bottom water calcite saturation state (ΔCO3

=calcite), SSSs and ln

(Mg/Ca). The data necessary for the calculation of ΔCO3=

calcite wereextracted from the GLODAP bottle dataset (Key et al., 2004) usingOcean Data View (Schlitzer, 2002) from surface water and core toplocations. We performed a suite of linear regressions and multiplelinear regressions using combinations of seasonal SST, Tiso, SSS andΔCO3

=calcite (Table 2) and assessed the significance of exponential

constants (t-value and t-critical value).First, we explored regressions of regional North Atlantic Mg/Ca

onto seasonal SST, annual SST and T. The linear regression of ln(Mg/Ca) onto Tiso yields the best correlation of the three data sets:

Mg=Ca = 0:97 exp 0:066⁎Tisoð Þ ð2Þ

with n=52 and r2 of 0.82. Warm season SSTs yield an r2 of 0.68 whileannual SSTs yield an r2 of 0.67. The regional calibration implies achange in Mg/Ca per °C significantly lower than the Lea et al. (1999)equation, but similar to that of the McConnell and Thunell (2005)calibration that was based on G. bulloides from sediment traps in theGulf of California (Fig. 4A). Applied to a given change of Mg/Ca, thesmaller change in Mg/Ca per degree relative to Eq. (1) above (~6.6%per °C vs. 10% per °C) implies a significantly larger temperaturechange. Using this equation yields a higher bound for estimating theamplitude of HE temperature changes (i.e. yields lower temperaturesand a larger amplitude relative to Holocene values calculated from thesame equation). Using this regionally derived calibration equationcaptures the variation in the Mg/Ca–temperature relationship amongthe published calibration equations (Fig. 4).

Table 2Table of regression coefficients and t-values.

Of the form: Mg/Ca=Aexp(B∗T+C∗SSS+D∗ΔCO3=)

Variable(s) regressed ontoln(Mg/Ca)

A Ba

All significantCb

All not significantDb

All not significantr2

Tiso 0.97 0.066 (15.28) 0.82SST warm season 0.96 0.050 (10.24) 0.68SST an avg 0.99 0.058 (10.07) 0.67Tiso, SSS 0.2 0.061 (7.86) 0.046 (0.66) 0.82SST warm season, SSS 32.56 0.060 (3.57) −0.104 (−0.63) 0.68SST ann avg, SSS 1208.33 0.082 (3.50) −0.209 (−1.07) 0.68Tiso, SSS, ΔCO3

= 0.31 0.063 (8.00) 0.031 (0.50) 0.001 (0.95) 0.83SST warm season, SSS, ΔCO3

= 1171.45 0.073 (3.76) −0.213 (−1.16) 0.003 (1.29) 0.69SST ann avg, SSS, ΔCO3

= 211081.59 0.10 (3.74) −0.36 (−1.63) 0.0029 (1.40) 0.69

a Regression coefficients of temperature. Values in parentheses are the t-values for each constant. Critical t-value from approximately 2.009 to 2.01, depending on the number ofvariables in the regression. All values in this column are above the critical value, and therefore significant.

b All values in this column are below the critical value, and therefore not significant.

50 G.M. Patton et al. / Earth and Planetary Science Letters 310 (2011) 45–58

Next, we explored whether there is a relationship between Mg/Caand dissolution in G. bulloides. Equations have been developed forsome species that take into account a post-depositional decrease ofMg/Ca due to dissolution by adding a second term to the equation,either depth or calcite saturation state. The amplitude of thecorrection factor varies significantly among species but has not beenquantified for G. bulloides. Therefore, we explore the dissolution effectusing the same published G. bulloides data from the North Atlantic asabove. We performed a multiple linear regression of the ln(Mg/Ca),Tiso, seasonal SSTs, SSTAnn, and bottom water ΔCO3

=calcite. The North

Atlantic region generally shows high values of ΔCO3= and is relatively

shallow. The compiled core top data have an associated range inΔCO3

= from 2.7 to 55.8 μmol/kg, with values generally above20 μmol/kg. In the context of a regional calibration, we feel thesevalues are suitable for a regression exercise and would expect thatif there were a strong dissolution effect on the Mg/Ca content ofG. bulloides, it would emerge from the regression. We did not find astatistically significant dissolution effect on the Mg/Ca. However, wealso recognize that a much larger range would be ideal for bettercharacterization of the dissolution effect. Examination of G. bulloidesdata from a depth transect along the Ontong Java Plateau that spanslower saturation state values, ΔCO3

=−3 to−22 μmol/kg, also showsno obvious decrease in Mg/Ca with decreasing saturation (Mekik etal., 2007). This lack of a large dissolution effect on G. bulloidesMg/Cais consistent with the relatively small correction factors of spinosesurface dwelling species Globigerinoides ruber and Globigerinoidessacculifer (0.5 and 0.4% decrease per μmol/kg decrease of carbonateion, respectively) (Dekens et al., 2002). Given that the cores used inour reconstructions are also well above the regional lysocline and themultiple linear regression does not yield a statistically significantrelationship with ΔCO3

=, we do not consider a dissolution effect in ourbest-fit North Atlantic calibration equation. During HE1 in core MD95-2040, we do see decreased shell weights. But, we contend that this isadditional support for dramatic and substantial cooling during thisevent at the location of this core, and is most likely not a signal ofdissolution. Decreased shell weights of G. bulloides have been shown tobe associated with decreased temperature during calcification(Marr et al., 2011; Mekik et al., 2007).

Finally, we explored the potential salinity effect in G. bulloides inthe North Atlantic core top data. In contrast to dissolution, a salinityeffect could potentially have a non-negligible impact on ourtemperature reconstructions, particularly if the ice rafting eventswere accompanied by surface melting of ice. Culturing and core topcalibration studies have reported a range of potential salinity effectson planktonic foraminiferal Mg/Ca, with the magnitude of the effectvarying by species as well as range of environmental conditions(Dueñas-Bohórquez et al., 2009; Lea et al., 1999; Mathien-Blard andBassinot, 2009). The largest effect has been reported for the

planktonic foraminiferal species, G. ruber. While culturing studiesdemonstrate a salinity effect of ~4% per salinity unit for G. ruber,recently published core top studies imply a 15 to nearly 30% change inMg/Ca per salinity unit (Arbuszewski et al., 2010; Mathien-Blard andBassinot, 2009). At the higher end of these estimates, the salinityeffect can have a significant influence on the calculation oftemperatures from Mg/Ca.

We explored regressions of the G. bulloides Mg/Ca data againsttemperature (summer SSTs, mean annual SSTs and and Tiso) andsalinity (summer SSS and mean annual SSS), and also exploredmultiple regressions using ΔCO3

=calcite. In addition, we investigated

Mg/Ca corrected for a presumed ΔCO3=

calcite effect ranging from anexponential constant of ~0.001 (the best fit, though statistically notsignificant) to 0.005 per μmol/kg (the coefficient from the G. rubercalibration). The exponential constant on temperature remained~0.050 and 0.066. Regressions of ln(Mg/Ca), SST, SSS and ΔCO3

=calcite

did not yield a significant pre-exponential constant for ΔCO3=calcite,

nor did any scenarios yield a statistically significant relationship withsalinity. While we conclude from our analysis that salinity does notappear to be a significant factor in Mg/Ca variations of G. bulloides inthe existing North Atlantic G. bulloides dataset, we also recognizeisolating a salinity effect in the core top data can be difficult when SSTand SSS are highly correlated. Arbuszewski et al. (2010) was able toavoid this problem in their investigation of G. ruber by targeting aregion that spanned a wide range of temperatures (16.7 °C to 27.6 °C)and salinities (35.5–37.3) that were not highly correlated. In theNorth Atlantic dataset we compiled from published data, thecorrelation is 0.82 for the annual data and even higher for theseasonal SST and SSS. A main goal of the calibration aspect of thisstudy was to examine a regional relationship of salinity, temperature,ΔCO3

= and Mg/Ca. The fact that salinity and temperature are well-correlated is an artifact of the North Atlantic region.We agree that is itmore rigorous to have a global calibration, but felt that it is alsovaluable to examine the regional data, particularly in light of theArbuszewski et al. (2010) results that highlight a salinity effect.

Given the potential complication of isolating a salinity effect withthe existing North Atlantic G. bulloides data, we use the equation fromArbuszewski et al. (2010) for G. ruber in which there is a strongdependence on salinity (~27%/salinity unit) as a third equation toestimate temperature changes during HE relative to the Holocene.While it is less than ideal to apply a calibration equation from adifferent species, previous studies have suggested that the exponen-tial constants associated with the temperature response of G. ruberand G. bulloides are typically similar (ranging between 0.08 and 0.1)(Dekens et al., 2002). The usefulness in applying the Arbuszewski etal. (2010) equation is that the dataset covers themost extensive rangeof temperatures and salinities and applying the calibration equationmay be useful for exploring how a large salinity effect might impact

51G.M. Patton et al. / Earth and Planetary Science Letters 310 (2011) 45–58

the temperature estimates during HE. The Arbuszewski et al. (2010)equation can be written in the same form as Eqs. (1) and (2):

Mg=Ca= exp −9:7+0:05⁎ SSTAnn+0:27⁎ SSSAnn+0:005⁎ΔCO=3 calciteÞ

�

ð3Þ

SST (°C)

0 5 10 15 20 25 300

1

2

3

4

5

6

7

8

Mg/

Ca

mm

ol/m

ol

0 5 10 15 20 25 300

2

4

6

8

10

Mg/

Ca

mm

ol/m

ol

SST (°C)

Lea (1999)Arbuszewski (2010) [G. ruber]McConnell and Thunell (2005)NA calibration

0 5 10 15 20 25 300

5

10

15

20

25

30

SST (°C)

SS

T-d

eriv

ed (

°C)

NA calibration

A)

B)

C)

where SST is the mean annual SST, SSS is the mean annual sea surfacesalinity, and ΔCO3

2− is the bottom water ΔCO3=calcite. However, to

calculate temperatures based on the relationship in Eq. (3), we use theequation given by Arbuszewski et al. (2010) that relies on arelationship between modern δ18O and salinity and correct the δ18Ofor changes in ice volume in our records. We apply the equation to themeasured Mg/Ca and δ18O values from G. bulloides measured in bothcores as a third estimate of temperature; in most cases, the equationyields estimates slightly larger than those of Eqs. (1) and (2), whichdo not include a salinity effect but have different exponential factorsassociated with temperature.

4.1.2. Calculating temperature from downcore δ18OAn estimate of temperature changes can also be derived from the

δ18Oshell record (Bemis et al., 1998). While the δ18Oshell is dependenton temperature, it is also dependent on the isotopic composition ofthe seawater such that, even on the short timescales of HeinrichEvents, may be influenced by meltwater as well as any climaticallyinduced salinity changes. Because ameltwater influencewould lead toisotopically lighter values whereas cooling would lead to isotopicallyheavier values, simultaneous decrease of salinity from a meltwaterinfluence and a decrease of temperature could reduce the ampli-tude of oxygen isotope excursions during Heinrich Events. Under suchconditions, temperature estimates from δ18O likely represent mini-mum amplitude of applied temperatures by this proxy. In contrast,given the freshwater effects on Mg/Ca, this estimate of the amplitudeof temperature change should be a maximum. Therefore, the estima-tion of temperature changes using δ18Oshell is an important conceptualaspect of these events. Interpretation of temperature during HE1,which occurs over the deglaciation, is further complicated by meanocean changes in δ18O associatedwith themelting of ice sheets. HE1 isduring the deglaciation, with δ18O of the seawater (δ18Osw) becomingmore isotopically light during the deglaciation. We limit ourdiscussion of δ18O derived temperature changes to HE4 and HE5.

4.2. Comparison of temperature estimates along the Iberian Margin

The Iberian Margin has been extensively studied due to the highsedimentation rates associated with this seasonal upwelling region(Salgueiro et al., 2008), enabling us to compare multiple temperatureproxies with our Mg-derived temperatures and trace elementanalyses as well as the planktonic δ18O from the same cores. Wecompare the Mg/Ca based temperatures, Mg/Ca and δ18O basedtemperatures, SSTs estimated from the planktonic foraminifera faunalanalysis and alkenone U37

K′ (Pailler and Bard, 2002). Estimates basedon the modern analog technique for core MD99-2339 are fromVoelker et al. (2009) for HE1 and Voelker et al. (2006) for HE4 and 5.For core MD95-2040 we use the Salgueiro et al. (2010) data.

One complication in comparing SSTs from multiple proxies, par-ticularly in the subtropics, is seasonality. Assemblage-derived SSTs havebeen calibrated for the four different seasons but in general onlysummer and winter temperatures are discussed (Pflaumann et al.,2003). Estimates are calibrated to reflect temperatures at 10 m water

Fig. 4. (A) Calibration equations used in this study represented as functions of temperature.Solid lines denote the range of temperatures used in the calibration equation. The verticalgray bar represents the range of temperatures over the regional North Atlantic Calibration.In order to plot on the same axes, the pre-exponential constant of Arbuszewski et al. (2010)(calibration equation forG. ruber)was removed. (B) TheNorth Atlantic calibration equationrepresented by the solid line, surrounded by curves of ± the standard error of the pre-exponential constant and exponential constant. Scatter points are Tiso andmeasuredMg/Ca(mmol/mol) frompublished studies used in the regression. Error bars reflect± the standarderror of the pre-exponential constant and exponential constant. (C) The North Atlanticcalibration applied to the dataset against Tiso. Error bars of temperature are included inaddition to deviation from the 1:1 line (dashed).

52 G.M. Patton et al. / Earth and Planetary Science Letters 310 (2011) 45–58

depth. In addition, assemblage-derived SSTs are a statistical consider-ation of deep and shallow dwellers and therefore assumptions ofhydrologic structure are part of these statistical correlations. Alkenonesas well can be seasonally influenced if the majority of productivityoccurs in one season, resulting in biased temperatures indicative ofspring or summer (Bard, 2001) or winter as it is the case at present onparts of thePortuguesemargin (Abrantes et al., 2009).Mg/Ca can alsobeseasonally biased ifwe are analyzing a species associatedwithparticularseasonal hydrologic conditions. G. bulloides, for example is typically anupwelling indicator (Salgueiro et al., 2008) in the Iberian Margin in themodern, having a preference for high nutrient waters which areassociated with spring and summer temperatures. Comparison of coretop data in this region to modern sea surface temperatures supportsinterpretation of theproxydata as representative of summer conditions.While we plot reconstructed temperatures, we discuss temperatures interms of the amplitude of the relative to reconstructed values for theHolocene (from surface samples) to minimize artifacts of multi-proxyreconstruction of the actual temperature (ΔT summarized in Table 3).

The temperature estimates based solely on Mg/Ca data showpronounced and abrupt cooling during HE1 and little change intemperature during HE4 and HE5 at both core locations relative to theHolocene, regardless of the calibration equation applied to thereconstruction. Here the Holocene value is determined from a nearbycore top using the relevant calibration (reconstructed SSTs in Table 1).At the more northerly site (MD95-2040), the estimate of theamplitude cooling during HE1 (i.e. ΔT) using the Mg/Ca data rangesfrom ~8 °C (estimated from the Lea et al., 1999) to ~13 °C (estimatedfrom the regional calibration). The estimate assuming temperatureand salinity sensitivities comparable to G. ruber yields the warmestimplied SSTs but a ΔT between these estimates, ~10 °C. Further south(at the site of MD99-2339), the Mg/Ca values during HE1 implyslightly warmer temperatures, although after correcting for a slightlywarmer Holocene temperature in this region, there is a smallerestimated range of temperature changes (~4 °C to 8 °C). For HE4 andHE5, the range of cooling implied by the Mg/Ca is much less at bothsites (from little discernable cooling during HE5 at the southern site to6–10 °C, during HE4 at the northern site). Taking theMg/Ca variabilityin core MD99-2339 at face value, it suggests that HE4 and HE5 weremore hydrologically (and climatically) variable events than HE1.

The four SST proxies (δ18Oshell, Mg/Ca, assemblages and alke-nones) imply different patterns of relative cooling intensity duringeach of the three Heinrich Events (HE1, HE4 and HE5) at our twolocations (also in Figs. 5 and 6). Similar to the Mg/Ca data, thealkenone data for site MD95-2040 imply different amplitudes oftemperature change for the three Heinrich Events but with warmingduring HE1. The higher temperature for HE1 is surprising given thepronounced coolings from the other proxies as well as additionalalkenone data from the region showing 2–8 °C decrease in SSTdependent on location and calibration equation (Bard et al., 2000).There are several hypotheses for high latitude anomalous tempera-tures of U37

K′ (Rosell-Mele and Comes, 1999). Low temperature, lowsalinity environments are not well described by U37

K′ in the Nordic Seas.

Table 3Table of HE comparisons.

Lea et al. (1999) ΔT,(°C)

Regional ΔT,(°C)

MD95-2040 HE1a–Holocene 8.2 12.5HE4–Holocene 5.6 8.5HE5–Holocene 5.1 7.8

MD99-2339 HE1b–Holocene 4.5 6.8HE4–Holocene 1.5 2.3HE5–Holocene 1.3 1.9

a Core top minus average temperature surrounding pachy (s) abundance peak during HE1value is 18.6 °C and the alkenone derived Holocene value is 18.3 °C.).

b Core top minus average temperature surrounding pachy (s) abundance peak during HE

Additionally, IRD may provide a source of allochtonous alkenones,which can influence the derived temperature signal. The IRD inputinto the Iberian Margin during HE is small relative to the rest of theNorth Atlantic, however, a combination of these effects may result inthe increased temperatures from U37

K′ during HE1. It is difficult toimagine a hydrologic scenario in which decreased temperaturesduring HE1 were completely obliterated by horizontal transport.Based on the discrepancy of the alkenone proxy compared with thefaunal and Mg/Ca estimates, this multiproxy comparison agrees withthe Bard (2001) assessment stating that low alkenone SSTs are moreuncertain. The alkenone-derived SSTs (Pailler and Bard, 2002) showchanges of 3–4 °C in HE4 and HE5, in line with the Mg/Ca changes atthe location of core MD95-2040.

There is considerable uncertainty in calculating the amplitude ofthe temperature change for HE1 from δ18Oshell given the sea levelchange over this interval is related to the deglaciation; however, sealevel influences were likely smaller over HE4 and HE5 (~20 m)(Siddall et al., 2008). The δ18O shifts over HE4 and HE5 were ~1.5‰ atthe site of MD95-2040. Attributing this entire change to temperaturewould yield 6–7.5 °C change depending on the fractionation factor asa function of temperature (Rohling and Cooke, 1999). Allowing anadditional 0.2‰ as a result of a meltwater influence during HE5,would increase the estimate of this temperature change to 6.8–8.5 °C.The δ18Oshell is of slightly smaller amplitude at the southern site withapproximately 1‰ increase over the course of each event. Attributingthis change entirely to temperature yields 4–5 °C change as aminimum estimate. Allowing for 0.2‰ meltwater influence on theδ18Oshel, another estimate of this temperature change ranges from 4.8to 6 °C. At both sites, the three proxies imply much smaller tem-perature changes for HE4 and HE5 than the temperatures implied bythe assemblage data.

The assemblage data imply large temperature swings of 10–12 °Cfor all three of the events indicated (de Abreu et al., 2003). Thesimilarity in the reconstructed temperatures could be the result of thehigh Neogloboquadrina pachyderma (sinistral) abundance having aparticularly significant impact on the assemblage-derived SST giventhat it is the dominant species in modern polar assemblages. In allthree events, N. pachyderma (s) abundances exceed 68%. The nearlyidentical amplitude of temperature responsewith each event suggeststhat the dominance ofN. pachyderma (s) in the fauna is driving the SSTestimation. Mg/Ca and alkenone estimates of temperature are moresensitive to (summer) temperature oscillations during HE in com-parison to assemblage-derived temperature, which could be “bottom-ing out” in the high% N. pachyderma (s) range.

There are complex differences between proxies as a result of seasonalbiases, water column stratification and other hydrographic traits that caninfluence the resulting SSTs. Despite that HE are “no-analog” situations,the differences between these proxies are as informative of the past astheir quantitative similarities. In core MD95-2040, the comparison ofalkenone, faunal and Mg/Ca temperature estimates is particularlyinformative. The alkenone temperature reconstruction is very similar tothe reconstruction using Mg/Ca and δ18Oshell applied in the Arbuszewski

Arbuszewski et al. (2010) ΔT,(°C)

Assemblage ΔT,(°C)

Alkenone ΔT,(°C)

10.8 12.3 5.99.2 11.2 8.28.1 9.9 6.76.2 7.14.7 5.35.3 4.9

for MD95-2040. (Mg-derived Holocene values in Table 1. Assemblage derived Holocene

1 for MD99-2339. (Assemblage derived Holocene value is 20.3 °C.).

0.5

1.0

1.5

2.0

2.5

3.0

3.5

0

5

10

15

20

SS

T (

°C)

0

5

10

15

20

SS

T (

°C)

10 12 14 16 18 20 220

5

10

15

20

SS

T (

°C)

36 38 40 42

Age (ka)43 45 47 49

A)

B)

C)SST=f(Assemblage)

D)SST=f(Alkenone)

Age (ka)

HE1 HE4 HE5

3.5

3.0

2.5

2.0

1.5

1.0

0.5

10 15 20 25 30 35 40 45 50

HE1 HE4 HE5

G. b

ullo

ides

δ18

O (

‰)

δ18O

δ O18

0

-3

-6

-9

ΔT

Mg/CaMg/Ca and

MD95-2040 (40.58°N, 9.86°W, 2465m)

δ O18

Fig. 5. (A) The MD95-2040 oxygen isotopic composition from G. bulloides (note secondary y-axis represents 1 °C cooling per 0.23‰) is compared with SST reconstructions derivedfrom (B)Mg/Ca (black lines, using Lea et al. (1999) represented by the thinner black line and the North Atlantic calibration represented by the thicker black line) andMg/Ca and δ18O(gray line, Arbuszewski et al. 2010). These lines denote a range of temperatures implied by the suite of equations. Horizontal dash on the y-axis represents the Holocene SST derivedby each equation. (C) Foraminiferal assemblages (summer; Salgueiro et al., 2010) and (D) alkenones-U37

K′ (Pailler and Bard 2002). HE intervals denoted by the vertical colored bars.

53G.M. Patton et al. / Earth and Planetary Science Letters 310 (2011) 45–58

et al. (2010) equation. HE4 and HE5 are nearly identical in temperatureminima and amplitude of temperature change (Fig. 5). Increasedstratification of the water column during HE1 (Voelker et al., 2009),coupled with slow down of overturning circulation seen in 231Pa/230Thmeasured in the IberianMargin (Gherardi et al., 2005), is anotherpossibleexplanation of the anomalous alkenone temperatures during this event.Meltwater eventsmay suppress surface exchangewith colder, subsurfacewaters. Increased exposure of surfacewaters due to stratification and the

slowdown of overturning circulation may result in the higher temper-atures associated with alkenones in core MD95-2040 (Peck et al., 2008).The behavior of the proxies relative to each other during HE1 highlightsthe uniqueness of this deglacial event. The assemblage data informs usqualitatively of abrupt changes to colder temperatures, evident in thelarge swings in temperature seen during HE. This signal, however, ispotentially overly influenced by the dominance of polar species and thusthe reduction in diversity. Overall, assemblage data for the broader North

0

5

10

15

20

SS

T (

°C)

10 12 14 16 18 20 220

5

10

15

20

SS

T (

°C)

38 39 40

Age (ka)43 45 47 49

A)

B)

C)SST=f(Assemblage)

3.5

3.0

2.5

2.0

1.5

1.0

0.5

10 15 20 25 30 35 40 45 50Age (ka)

HE1 HE4 HE5

HE1 HE4 HE5

δ O18

MD99-2339 (35.88°N, 7.53°W, 1170m)

δ O18Mg/CaMg/Ca and

0.5

1.0

1.5

2.0

2.5

3.0

3.5

δ18O

0

-3

-6

-9

ΔT

G. b

ullo

ides

δ18

O (

‰)

Fig. 6. (A) The MD99-2339 oxygen isotopic composition from G. bulloides (note secondary y-axis represents 1 °C cooling per 0.23‰) is compared with SST reconstructions derived from(B)Mg/Ca (black lines, using Lea et al. (1999) represented by the thinner black line and the North Atlantic calibration represented by the thicker black line) andMg/Ca and δ18O (gray line,Arbuszewski et al. 2010). These lines denote a range of possible temperatures. Horizontal dash on the y-axis represents the Holocene SST derived by each equation (C) Foraminiferalassemblages (summer). HE intervals denoted by the vertical colored bars. Note: in the middle vertical set of panels the length of the x-axis is not proportional to time.

54 G.M. Patton et al. / Earth and Planetary Science Letters 310 (2011) 45–58

Atlantic revealmore variation in reconstructed SSTs (Fig. 7) (Cortijo et al.,1997; Maslin et al., 1995; Sarnthein et al., 2000; Voelker et al., 2006).

A compilation of the SSTs reconstructed for the broader NorthAtlantic for HE helps to put the temperature changes we reconstructfor this marginal region into a broader perspective (Fig. 7) (Chapmanand Shackleton, 1998; Martrat et al., 2007; Sánchez-Goñi et al., 2008;Shackleton, 2000; Sierro et al., 2005; Skinner and Elderfield, 2007;Vautravers and Shackleton, 2006; Voelker and de Abreu, in press).Two notable features emerge from comparison of the reconstructedSSTs for HE1 and HE4 relative to modern SSTs (left panels), which arehighlighted by shifting temperatures to match the average of thebroader North Atlantic proxy data. First, the temperatures during HEalong the Iberian Margin are generally colder than reconstructedchanges over the wider North Atlantic. Second, in the context of

examining HE relative to the background climate, HE1 occurs duringthe deglaciation while HE4 and HE5 occur solidly in the glacial.Sufficient data exist to compare HE1 and HE4. In both the broaderNorth Atlantic and the Mg/Ca in the Iberian margin, HE1 stands out asa colder event than HE4. We will first discuss the colder temperaturesalong the Iberian Margin, then return to the larger pattern of a colderHE1 in paragraph 4.5.

4.3. Constraining the influence of upwelling using Cd/Ca

Assuming the Mg/Ca changes during HE are driven by decreases intemperature, potential causes for dramatic shifts in temperature inthis marginal setting include increased upwelling, local meltwaterinfluence, and changes in surface hydrography. The contrast between

60° W 30° W 0°

30° N

45° N

60° N

60° W 30° W 0°

30° N

45° N

60° N

60° W 30° W 0°

30° N

45° N

60° N

60° W 30° W 0°

30° N

45° N

60° N

60° W 30° W 0°

30° N

45° N

60° N

60° W 30° W 0°

30° N

45° N

60° N

HE1 contour cartoon

HE4 on Modern Summer

HE1 on Modern Summer

A)

C)

E)

B)

Tem

p (°C)

25

20

15

10

5

0Coretops on Modern Winter

HE4 contour cartoon

Tem

p (°C)

25

20

15

10

5

0

D)

25

20

15

10

5

0

Tem

p (°C)

F)

Coretops on Modern Summer

Fig. 7. (A) Map of the Northern Atlantic core top SSTs plotted on the modern contoured summer SST. Colors of each data point and the SST contours share the same color bar. Circlesrepresent assemblage-derived SSTs; triangles, alkenones; and, squares, Mg/Ca. Assemblage-derived SSTs denote summer temperatures. (B) Map of the Northern Atlantic core topSSTs plotted on the modern contoured winter SST. (C) HE1 SST data plotted on the modern contoured summer SST. (D) Cartoon of contoured HE1 summer SSTs based on data pointsfrom Panel C. (E) HE4 SST data plotted on the modern contoured summer SST (F) Cartoon of contoured HE4 summer SSTs based on data points from Panel E.

55G.M. Patton et al. / Earth and Planetary Science Letters 310 (2011) 45–58

marginal and open ocean temperatures could be consistent withmoreintense coastal upwelling. The Cd/Ca of the G. bulloides from thesecores can address this hypothesis.

Cd can be used an indicator of upwelled nutrients, given its depthprofile and behavior as a micronutrient (Boyle, 1988). Previous workin upwelling regions shows increased dissolved Cd during upwellingevents and consequently increased Cd/Ca in the foraminiferal lattice(van Geen et al., 1992). In addition, culturing studies have shown thatthe shell chemistry of G. bulloides records variations in the Cd contentof the waters in which they calcified (Mashiotta et al., 1997). Thus, theshell Cd/Ca of G. bulloides should provide estimates of variations in

nutrient concentration and can be used to test whether lowermarginal temperatures during HE could be linked to upwelling.These datawill be useful for placing future Cd/Cameasurements into alarger framework of a proxy for past nutrient availability.

The G. bulloides Cd/Ca data in our cores on the Iberian Margin andin the Gulf of Cadiz imply low nutrient values during HE. The constantlow Cd/Ca values in our cores (~0.02 μmol/mol during all three HE inMD95-2040 from the SW Iberian Margin and slightly lower in HE4and HE5 in the Gulf of Cadiz) suggest that upwelling is at least nomore intense, and perhaps lower, during HE than during the Holocenein this region. A detailed comparison of Cd/Ca values during HE and in

56 G.M. Patton et al. / Earth and Planetary Science Letters 310 (2011) 45–58

intervals between HE reveals that values are not significantlydifferent. A core top calibration set (Salgueiro et al., in prep) ofCd/Ca for G. bulloides ranges from 0.02 to 0.04 μmol/mol along thesouthwest Iberian Margin and a slightly lower range of values 0.02to 0.03 μmol/mol in the Gulf of Cadiz region, where there is lessupwelling. At the termination of HE5, there are two data points ofincreased value (Fig. 1). These could be the result of increasedupwelling associated with circulation changes at the end of HE5;however, few conclusions can be drawn from two data points.Quantitative analysis of the Cd/Ca values in this region is necessary toestablish estimation of nutrient availability and potential upwellingchanges. Assemblage changes examined for productivity changes(Salgueiro et al., 2010) suggest decreased export production duringthe same periods in the Iberian Margin (MD95-2040), also implyingthat upwelling was no more intense during the HE. The low HE Cd/Cavalues are similar to low glacial Cd/Ca values from the Subantarctic(Mashiotta, 1998). An alternative explanation of the low Cd/Ca duringHE is that the low Cd/Ca values result from a temperature effect on theincorporation of Cd/Ca as proposed by Rickaby and Elderfield (1999);however, that the variations in our Cd/Ca do not directly track Mg/Cachanges, particularly just prior to and during HE1, suggests thattemperature is not the primary control on the Cd/Ca (Figs. 1 and 2).

4.4. Mechanisms of hydrologic change during Heinrich Events

The broad pattern of colder HE1 relative to HE4 could reflect thesensitivity of the North Atlantic to HE forcing related to backgroundstate. The trigger for the calving of icebergs leading to HE is stillunknown, but independent of that trigger are the hydrologic changesthat occur as a result. There have been a number of modeling studiesto investigate the effects of freshwater dumping in areas of NorthAtlantic Deep Water formation (Ganopolski and Rahmstorf, 2001;Seidov andMaslin, 1999; Stouffer et al., 2006). These models predict ashutdown of the thermohaline circulation given a sufficient freshwa-ter input in critical areas of deep water formation (Seidov and Maslin,1999). In addition, sedimentary evidence in the form of 231Pa/230Thalso indicates slow down, or complete shut down of the oceanicconveyor during HE1 (Gherardi et al., 2005; McManus et al., 2004).Gherardi et al. (2005) directly demonstrates that the cores in theIberian Margin are subject to the circulation changes resulting fromperturbations of the AMOC coherent with a southward shift of thePolar Front (Eynaud et al., 2009).

Intermediate complexity models reveal that a conveyor-beltshutdown generates the strongest temperature effects in the sub-tropical Atlantic andMediterranean (Ganopolski and Rahmstorf, 2001);this is consistent with the large temperature fluctuations we see in thisstudy during HE1 (for a compilation of SST during HE1 in the Atlantic,see Fig. 7C). Stouffer et al. (2006) assembled model data resulting fromhosing experiments as part of the Climate Model IntercomparisonProject (CMIP). The ensemblemean suggests SST decreases just south ofGreenland by3 °C (Stouffer et al., 2006). This perturbation is in the samedirection as that seen by assemblage data and Mg/Ca in the Iberianmargin, but with smaller amplitude in the north. The mechanism ofNorth Atlantic cooling explored in the model is predominantly one ofchanges in the advection of heat from lower latitudes. Comparison ofdata fromour studywithpublished estimates forHE1 andHE4, place theSST calculated from this study into a larger climate context (Fig. 7)(Maslin et al., 1995; Sarnthein and Altenbach, 1995).

An aspect of the hosing experiments that has not been fullyexplored with models is the sensitivity of NADW shutdown de-pendent upon the initial climate state, glacial or deglacial. HE5 andHE4 occur solidly in the glacial period, while HE1 occurs during thedeglaciation. It is well known that HE1 and HE4 are considered to bethe largest ice-rafting events due to the highest IRD counts in theNorth Atlantic. In our trace element data for HE1, HE4 and HE5,however, HE1 is the most clearly delineated in temperature and

duration while HE4 only shows slight minima. It is possible that thesensitivity of the climate to HE1 is determined by the deglacial stateand its potentially unstable mode of ocean circulation, relative to theglacial and more stable mode. A study of the climate response to the8.2 ka event proposed an initial intermediate climate state due to abaseline input of freshwater from the LIS (Bauer et al., 2004). Theyfind that the stability of the intermediate state is weaker than that ofthe “turned on” deep water formation state and “turned off” deepwater formation state.

We propose a similar climate state during the deglaciation with abaseline melt from the LIS that makes this climate state more sensitiveto hosing events, such asHE1.HE4does not result in the sameamplitudeof climate perturbation. The size and shape of the pulses are not knownand theAMOC is a highly non-linear system.Many variables could affectits response to a freshwater input. For example, in intermediatecomplexity models, the shape of the pulse, as well as the ratio of ice towater in the pulse, can have a large impact on the SST of the NorthAtlantic (Friedrich et al. in prep). Intermediate complexity models thatsimulate HE with different freshwater-to-ice ratios find differentamplitudes of SST change for different HE realizations/simulations(personal communication T. Friedrich). This is consistent with ourMg/Ca data and the reconstructions of SSTs for the broader NorthAtlantic however not consistent with temperatures derived from theother two proxies. On one hand, foraminiferal assemblages imply lowtemperatures that are nearly identical during HE1, HE4 and HE5,particularly in the more northerly core. On the other hand, somealkenone reconstructions imply a smallwarmingduringHE1 (Naughtonet al., 2009; Pailler and Bard, 2002; Penaud et al., 2010). Comparativestudies of the intensity of deepwater formation using 231Pa/230ThduringHE1 and HE4 would help to determine the relative sensitivity of thesedifferent climate states to freshwater hosing events, as investigated inintermediate complexity models.

5. Conclusion

A comparison of temperature proxies outside of HE4 andHE5 showsthat Mg/Ca-derived temperatures and alkenone (U37

K′)-derived SSTsagree very well. Both of these reconstructed temperatures are lowerthan assemblage-derived SSTs and the amplitudes of HE are larger thanthe difference of the proxy comparison. During HE, alkenone andMg/Caproxies reveal similar amplitudes of change (relative to core topreconstructions) while assemblage data shows very large temperaturefluctuations. HE1 is the most dramatic and clearly delineated HE inMg/Ca in the Iberian Margin, a pattern that also emerges fromcomparison of reconstructed SSTs for HE1 and HE4 from the broaderNorth Atlantic. The pattern of cooling during HE is evident regardless ofthe calibration equation applied to the reconstruction. The largertemperature change during HE1 may arise due to the climate state atthe time of the ice-rafting event (Bauer et al., 2004).

Cd/Ca values are very lowduringHEs, varying around 0.02 μmol/molat both locations, slightly lower than core top values in this region(Salgueiro et al. in prep). This implies that at both locations, nutrientavailability does not vary significantly over the timescale of HE and thatthe cooler temperatures during HE at these coastal sites are unlikely toresult from more intense upwelling.

Modeling studies of freshwater input (hosing experiments)predict decreases of ~3 °C in North Atlantic temperatures due to lessadvection of heat from lower latitudes (Stouffer et al., 2006). Mg/Caand alkenone SSTs are consistent with this result during HE4 and HE5.HE1 shows a much larger SST decrease, suggesting a potentialsensitivity of the AMOC during this deglacial ice-rafting event. Giventhat each proxy shows a different amplitude response to HE, inves-tigating the behavior of each proxy will better constrain these climateevents and may speak to the sensitivity of overturning circulation tofreshening events.

57G.M. Patton et al. / Earth and Planetary Science Letters 310 (2011) 45–58

Acknowledgments

G.P. wrote the manuscript with substantial input from P.M. andA.V. A.V. also provided samples. E.S. provided regional surfacesediment data and help with speciation as well as participating indiscussion. We would like to thank Georges Paradis and Kate Stegerfor analysis of our cleaned samples by magnetic sector ICP-MS.Samples were analyzed at the University of California, Santa Barbara;we thank D. Lea for making the analytical facility available. Discussionwith Tobias Friedrich was very useful. Nanxi Bian, Jennifer Fehrenbacher,Christina Belanger and Sophie McCoy provided helpful comments onthe manuscript. Jennifer Arbuzsewski provided essential instructionfor the extraction of GLODAP and WOA 05 data. We would also like tothank reviewers and the editor who greatly improved the quality of thismanuscript.

Appendix A. Supplementary data

Supplementary data to this article can be found online at doi:10.1016/j.epsl.2011.07.028.

References

Abrantes, F., Lopes, C., Rodrigues, T., Gil, I., Witt, L., 2009. Proxy calibration toinstrumental data set: implications for paleoceanographic reconstructions. Geo-chem. Geophys. Geosyst. 10 (9), Q09U07. doi:10.1029/2009GC002604.

Anand, P., Elderfield, H., Conte, M., 2003. Calibration of Mg/Ca thermometry inplanktonic foraminifera from a sediment trap time series. Paleoceanography 18 (2),1050. doi:10.1029/2002PA000846.

Antonov, J.I., Locarnini, R.A., Boyer, T.P., Mishonov, A.V., Garcia, H.E., 2006. World OceanAtlas 2005, 62 ed. Salinity, NOAA Atlas NESDIS, Volume 2. U.S. Government PrintingOffice, Washington, D. C.

Arbuszewski, J., deMenocal, P., Kaplan, A., Farmer, E., 2010. On the fidelity of shell-derived δ18Oseawater estimates. Earth Planet. Sci. Lett. 185–196. doi:10.1016/j.epsl.2010.10.035.

Arz, H., Pätzold, J., Wefer, G., 1998. Correlated millennial-scale changes in surfacehydrography and terrigenous sediment yield inferred from last-glacial marinedeposits off Northeastern Brazil. Quat. Res. 50, 157–166 QR981992.

Bard, E., 2001. Comparison of alkenone estimates with other paleotemperature proxies.Geochem. Geophys. Geosyst. 2 (1) 2000GC00050.

Bard, E., Rostek, F., Turon, J., Gendreau, S., 2000. Hydrological impact of Heinrich Eventsin the subtropical Northeast Atlantic. Science 289 (5483), 1321–1324. doi:10.1126/science.289.5483.1321.

Bauer, E., Ganopolski, A., Montoya, M., 2004. Simulation of the cold climate event8200 years ago by meltwater outburst from Lake Agassiz. Paleoceanography 19(PA3014). doi:10.1029/2004PA001030.

Bemis, B., Spero, H., Bijma, J., Lea, D., 1998. Reevaluation of the oxygen isotopiccomposition of planktonic foraminifera: experimental results and revised paleo-temperature equations. Paleoceanography 13 (2), 150–160.

Bian, N., Martin, P., 2010. Investigating the fidelity of Mg/Ca and other elemental datafrom reductively cleaned planktonic foraminifera. Paleoceanography 25 (PA2215).doi:10.1029/2009PA001796.

Boyle, E., 1981. Cadmium, zinc, copper, and barium in foraminifera tests. Earth Planet.Sci. Lett. 53, 11–35. doi:10.1016/0012-821X(81)90022-4.

Boyle, E., 1988. Cadmium: chemical tracer of deepwater paleoceanography. Paleocea-nography 3 (4), 471–489.

Broecker, W., 2006. Abrupt climate change revisited. Glob. Planet. Chang. 54, 211–215.doi:10.1016/j.gloplacha.2006.06.019.

Cacho, I., Grimalt, Pelejero, 1999. Dansgaard–Oeschger and Heinrich event imprints inAlboran Sea paleotemperatures. Paleoceanography 14 (6), 698–705. doi:10.1029/1999PA900044.

Chapman, M., Maslin, M., 1999. Low-latitude forcing of meridional temperature andsalinity gradients in the subpolar North Atlantic and the growth of glacial icesheets. Geology 27 (10). doi:10.1130/0091-7613(1999)027.

Chapman, M., Shackleton, N., 1998. Millennial-scale fluctuations in North Atlantic heatflux during the last 150,000 years. Earth Planet. Sci. Lett. 57–70.

Cléroux, C., Cortijo, E., Anand, P., Labeyrie, L., 2008. Mg/Ca and Sr/Ca ratios in planktonicforaminifera: proxies for upper water column temperature reconstruction.Paleoceanography 23 (PA3214). doi:10.1029/2007PA001505.

Cortijo, E., Labeyrie, L., Vidal, L., Vautravers, M., Chapman, M., Duplessy, J., Elliot, M.,Arnold, M., Turon, J., Auffret, G., 1997. Changes in sea surface hydrology associatedwith Heinrich event 4 in the North Atlantic Ocean between 40 and 60 N. EarthPlanet. Sci. Lett. 146 (1–2), 29–45. doi:10.1016/S0012-821X(96)00217-8.

Dansgaard, W., Johnsen, S., Clausen, H., Dahl-Jensen, D., 1993. Evidence for generalinstability of past climate from a 250-kyr ice-core record. Nature 364.

de Abreu, L., Shackleton, N., Schönfeld, J., Hall, M., Chapman, M., 2003. Millennial-scaleoceanic climate variability off the Western Iberian margin during the last twoglacial periods. Mar. Geol. 196 (1–2), 1–20. doi:10.1016/S0025-3227(03)00046-X.

Dekens, P., Lea, D., Pak, D., 2002. Core top calibration of Mg/Ca in tropical foraminifera:refining paleotemperature estimation. Geochem. Geophys. Geosyst. 3 (4).doi:10.1029/2001GC000200.

Dueñas-Bohórquez, A., da Rocha, R., Kuroyanagi, A., Bijma, J., Reichart, G., 2009. Effect ofsalinity and seawater calcite saturation state on Mg and Sr incorporation incultured planktonic foraminifera. Mar. Micropaleontol. 73 (3–4), 178–189.doi:10.1016/j.marmicro.2009.09.002.

Elderfield, H., Ganssen, G., 2000. Past temperature and δ18O of surface ocean watersinferred from foraminiferal Mg/Ca ratios. Nature 405 (6785).

Eynaud, F., de Abreu, L., Voelker, A., Schoenfeld, J., Salgueiro, E., Turon, J.-L., Penaud, A.,Toucanne, S., Naughton, F., Goni, M.F.S., Malaize, B., Cacho, I., 2009. Position of thePolar Front along the western Iberian margin during key cold episodes of the last45 ka. Geochem. Geophys. Geosyst. 10 (Q07U05). doi:10.1029/2009GC002398.

Ganopolski, A., Rahmstorf, S., 2001. Rapid changes of glacial climate simulated in acoupled climate model. Nature 409, 153–158.

Gherardi, J.M., Labeyrie, L., McManus, J., Francois, R., Skinner, L., Cortijo, E., 2005.Evidence from the Northeastern Atlantic basin for variability in the rate of themeridional overturning circulation through the last deglaciation. Earth Planet. Sci.Lett. 240 (3–4), 710–723. doi:10.1016/j.epsl.2005.09.061.

Grimm, E., Watts, W., Jacobson Jr., G., Hansen, B., Almquist, H., Dieffenbacher-Krall, A.,2006. Evidence for warmwet Heinrich events in Florida. Quat. Sci. Rev. 25 (17–18),2197–2211. doi:10.1016/j.quascirev.2006.04.008.

Grootes, P., Stuiver, M., 1997. Oxygen 18/16 variability in Greenland snow and ice with103- to 105-year time resolution. J. Geophys. Res. 102, 26455–26470.

Key, R.M., Kozyr, A., Sabine, C.L., Lee, K., Wanninkhof, R., Bullister, J.L., Feely, R.A.,Millero, F.J., Mordy, C., Peng, T.-H., 2004. A global ocean carbon climatology: resultsfrom Global Data Analysis Project (GLODAP). Glob. Biogeochem. Cycles 18 (4),GB4031. doi:10.1029/2004GB002247.

Lea, D., Martin, P., 1996. A rapid mass spectrometric method for the simultaneousanalysis of barium, cadmium, and strontium in foraminifera shells. Geochim.Cosmochim. Acta 60 (16), 3143–3149. doi:10.1016/0016-7037 (96)00184-6.

Lea, D., Mashiotta, T., Spero, H., 1999. Controls on magnesium and strontium uptake inplanktonic foraminifera determined by live culturing. Geochim. Cosmochim. Acta63 (16), 2369–2379. doi:10.1016/S0016-7037(99)00197-0.

LeGrande, A., Schmidt, G., 2006. Global gridded data set of the oxygen isotopiccomposition in seawater. Geophys. Res. Lett. 33 (12), L12604. doi:10.1029/2006GL026011.

Locarnini, R.A., Mishonov, A.V., Antonov, J.I., Boyer, T.P., Garcia, H.E., 2006. World OceanAtlas 2005, 61 ed. Temperature, NOAA Atlas NESDIS, Volume 1. U. S. GovernmentPrinting Office, Washington, D.C.

Marr, J., Baker, J., Carter, L., Allan, A., 2011. Ecological and temperature controls on Mg/Caratios ofGlobigerinabulloides fromthe southwest PacificOcean. Paleoceanography26(PA2209). doi:10.1029/2010PA002059.

Martrat, B., Grimalt, J.O., Lopez-Martinez, C., Cacho, I., Sierro, F.J., Flores, J.A., Zahn, R.,Canals, M., Curtis, J.H., Hodell, D.A., 2007. Four climate cycles of recurring deep andsurface water destabilizations on the Iberian Margin. Science 317, 502–507.doi:10.1126/science.1139994.

Mashiotta, T., 1998. The Trace Element Geochemistry of Planktonic Foraminifera. Ph.D.Thesis, University of California, Santa Barbara.

Mashiotta, T., Lea, D., Spero, H., 1997. Experimental determination of cadmium uptakein shells of the planktonic foraminifera Orbulina universa and Globigerina bulloides:implications for surface water paleoreconstructions. Geochim. Cosmochim. Acta 61(19), 4053–4065.

Mashiotta, T., Lea, D., Spero, H., 1999. Glacial-interglacial changes in Subantarctic seasurface temperature and δ18O-water using foraminiferal Mg. Earth and PlanetaryScience Letters 170 (4), 417–432. doi:10.1016/S0012-821X(99)00116-8.

Maslin, M., Shackleton, N., Pflaumann, U., 1995. Surface water temperature, salinity, anddensity changes in the northeast Atlantic during the last 45,000 years: Heinrich events,deep water formation, and climatic rebounds. Paleoceanography 10 (3), 527–544.

Mathien-Blard, E., Bassinot, F., 2009. Salinity bias on the foraminiferaMg/Ca thermometry:correction procedure and implications for past ocean hydrographic reconstructions.Geochem. Geophys. Geosyst. 10 (12), Q12011. doi:10.1029/2008GC002353.

McConnell, M., Thunell, R., 2005. Calibration of the planktonic foraminiferal Mg/Capaleothermometer: sediment trap results from the Guaymas Basin, Gulf ofCalifornia. Paleoceanography 20 (PA2016). doi:10.1029/2004PA001077.

McManus, J., Francois, R., Gherardi, J., Keigwin, L., Brown-Leger, S., 2004. Collapse andrapid resumption of Atlantic meridional circulation linked to deglacial climatechanges. Nature 428 (6985), 834–837.

Mekik, F., François, R., Soon, M., 2007. A novel approach to dissolution correction of Mg/Ca-based paleothermometry in the tropical Pacific. Paleoceanography 22 (PA3217).doi:10.1029/2007PA001504.

Naughton, F., Goñi, M.F.S., Kageyama, M., Bard, E., Duprat, J., Cortijo, E., Desprat, S.,Malaizé, B., Joly, C., Rostek, F., Turon, J.-L., 2009. Wet to dry climatic trend in north-western Iberia within Heinrich events. Earth Planet. Sci. Lett. 284 (3–4), 329–342.doi:10.1016/j.epsl.2009.05.001.

Pailler, D., Bard, E., 2002. High frequency palaeoceanographic changes during the past140000 yr recorded by the organic matter in sediments of the Iberian Margin.Palaeogeography 181, 431–452. doi:10.1016/S0031-0182(01)00444-8.

Peck, V., Hall, I., Zahn, R., Elderfield, H., 2008. Millennial-scale surface and subsurfacepaleothermometry from the northeast Atlantic, 55–8 ka BP. Paleoceanography 23(PA3221). doi:10.1029/2008PA001631.

Penaud, A., Eynaud, F., Turon, J., Blamart, D., 2010. Contrasting paleoceanographicconditions off Morocco during Heinrich events (1 and 2) and the Last GlacialMaximum. Quat. Sci. Rev. 29, 1923–1939. doi:10.1016/j.quascirev.2010.04.011.

Pflaumann, U., et al., 2003. Glacial North Atlantic: sea-surface conditions reconstructedby GLAMAP 2000. Paleoceanography 18 (3), 1065. doi:10.1029/2002PA000774.

58 G.M. Patton et al. / Earth and Planetary Science Letters 310 (2011) 45–58

Prange, M., Lohmann, G., Romanova, V., Butzin, M., 2004. Modelling tempo-spatialsignatures of Heinrich Events: influence of the climatic background state. Quat. Sci.Rev. 23 (5–6), 521–527. doi:10.1016/j.quascirev.2003.11.004.

Rickaby, R., Elderfield, H., 1999. Planktonic foraminiferal Cd/Ca: paleonutrients orpaleotemperature? Paleoceanography 14 (3), 293–303. doi:10.1029/1999PA900007.

Rohling, E., Cooke, S., 1999. Stable Oxygen and Carbon Isotopes in ForaminiferalCarbonate Shells. Kluwer Academic Publishers, Dordrecht/Boston/London.

Rosell-Mele, A., Comes, P., 1999. Evidence for a warm Last Glacial Maximum in theNordic seas or an example of shortcomings in UK

37′ and UK37′ to estimate low sea

surface temperature? Paleoceanography 14 (6), 770–776.Salgueiro, E., Voelker, A., Abrantes, F., Meggers, H., 2008. Planktonic foraminifera from

modern sediments reflectupwellingpatternsoff Iberia: insights fromaregional transferfunction. Mar. Micropaleontol. 66, 135–164. doi:10.1016/j.marmicro.2007.09.003.

Salgueiro, E., Voelker, A., de Abreu, L., 2010. Temperature and productivity changes offthe western Iberian margin during the last 150 ky. Quat. Sci. Rev. 29 (5–6),680–695. doi:10.1016/j.quascirev.20 09.11.013.

Sánchez-Goñi, M.F., Landais, A., Fletcher, W.J., Naughton, F., Desprat, S., Duprat, J.,2008. Contrasting impacts of Dansgaard–Oeschger events over a westernEuropean latitudinal transect modulated by orbital parameters. Quat. Sci. Rev.27, 1136–1151.

Sarnthein, M., Altenbach, A., 1995. Late quaternary changes in surface water and deepwater masses of the Nordic Seas and north-eastern North Atlantic: a review. Geol.Rundsch. 84 (1), 89–107.

Sarnthein, M., Stattegger, K., Dreger, D., Erlenkeuser, H., Grootes, P., Haupt, B., Jung, S.,Kiefer, T., Kuhnt, W., Pflaumann, U., Schäfer-Neth, C., Schulz, M., Seidov, D.,Simstich, J., van Kreveld-Alfane, S., Vogelsang, E., Voelker, A., Weinelt, M., 2000.Fundamental modes and abrupt changes in North Atlantic circulation and climateover the last 60 ky-numerical modelling and reconstruction. In: Schäfer, P., Ritzrau,W., Schlüter, M., Thiede, J. (Eds.), The Northern North Atlantic: A ChangingEnvironment. Springer Verlag, Heidelberg, Berlin, pp. 365–410.

Schlitzer, R., 2002. Interactive analysis and visualization of geoscience data with OceanData View. Comput. Geosci.-Uk 28 (10), 1211–1218.

Schoenfeld, J., Zahn, R., 2000. Late glacial to holocene history of the Mediterraneanoutflow. Evidence from benthic foraminiferal assemblages and stable isotopes atthe Portuguese margin. Palaeogeogr. Palaeoclimatol. Palaeoecol. 159 (1–2),85–111. doi:10.1016/S0921-8181(02)00197-2.

Seidov, D., Maslin, M., 1999. North Atlantic deep water circulation collapse duringHeinrich events. Geology 27 (1), 23–26.

Shackleton, N., 2000. The 100,000-year ice-age cycle identified and found to lagtemperature, carbon dioxide, and orbital eccentricity. Science 289 (5486), 1897.doi:10.1126/science.289.5486.1897.

Siddall, M., Rohling, E.J., Thompson, W.G., Waelbroeck, C., 2008. Marine isotope stage 3sea level fluctuations: data synthesis and new outlook. Rev. Geophys. 46 (4)2007RG000226.

Sierro, F., Hodell, D., Curtis, J., Flores, J., Reguera, I., 2005. Impact of iceberg melting onMediterranean thermohaline circulation during Heinrich events. Paleoceanogra-phy 20 (PA2019). doi:10.1029/2004PA001051.

Skinner, L., Elderfield, H., 2005. Constraining ecological and biological bias in planktonicforaminiferal Mg/Ca and δ18Occ: a multispecies approach to proxy calibrationtesting. Paleoceanography 20 (PA1015). doi:10.1029/2004PA001058.

Skinner, L., Shackleton, N., 2004. Rapid transient changes in northeast Atlantic deepwater ventilation age across Termination I. Paleoceanography 19 (PA2005).doi:10.1029/2003PA000983.

Skinner, L.C., Elderfield, H., 2007. Rapid fluctuations in the deep North Atlantic heatbudget during the last glacial period. Paleoceanography 22 (PA1205). doi:10.1029/2006PA001338.

Stouffer, R., Yin, J., Gregory, J., Dixon, K., Spelman,M., Hurlin,W.,Weaver, A., Eby, M., Flato,G., Hasumi, H., 2006. Investigating the causes of the response of the thermohalinecirculation to past and future climate changes. J. Clim. 19 (8), 1365–1387.

van Geen, A., Luoma, S., Fuller, C., Anima, R., Clifton, H., Trumbore, S., 1992. Evidencefrom Cd/Ca ratios in foraminifera for greater upwelling off California 4000 yearsago. Nature 358 (6381), 54–56.

Van Kreveld, S., Sarnthein, M., Erlenkeuser, H., 2000. Potential links between surging icesheets, circulation changes, and the Dansgaard–Oeschger cycles in the IrmingerSea, 60–18 kyr. Paleoceanography 15 (4), 425–442. doi:10.1029/1999PA000464.

Vautravers, M., Shackleton, N.J., 2006. Centennial-scale surface hydrology off Portugalduring marine isotope stage 3: insights from planktonic foraminiferal faunavariability. Paleoceanography 21 (PA3004). doi:10.1029/2005PA001144.

Voelker, A.H.L., de Abreu, L., in press. A review of abrupt climate change events in theNortheastern Atlantic Ocean (Iberian margin): latitudinal, longitudinal and verticalgradients. In: Rashid, H., Polyak, L., Mosley-Thompson, E. (Eds.), Understanding theCauses, mechanisms and extent of the Abrupt Climate Change. GeophysicalMonograph. AGU, Washington D.C. doi:10.1029/2010GM001021.

Voelker, A., de Abreu, L., Schönfeld, J., Erlenkeuser, H., Abrantes, F., 2009. Hydrographicconditions along the western Iberian margin during marine isotope stage 2.Geochem. Geophys. Geosyst. 10 (12), Q12U08. doi:10.1029/2009GC002605.

Voelker, A., Lebreiro, S., Schönfeld, J., Cacho, I., 2006. Mediterranean outflowstrengthening during northern hemisphere coolings: a salt source for the glacialAtlantic? Earth Planet. Sci. Lett. 245, 39–55. doi:10.1016/j.epsl.2006.03.014.

Yu, J., Elderfield, H., Jin, Z., Booth, L., 2008. A strong temperature effect on U/Ca inplanktonic foraminiferal carbonates. Geochim. Cosmochim. Acta 72 (20),4988–5000. doi:10.1016/j.gca.2008.07.011.