The last termination in the central South Atlantic

Quaternary Science Reviews 123 (2015) 193e214

Contents lists available at ScienceDirect

Quaternary Science Reviews journal homepage: www.elsevier.com/locate/quascirev

The last termination in the central South Atlantic Karl Ljung a, Sofia Holmgren a, Malin Kylander b, Jesper Sjolte a, Nathalie Van der Putten a, € rck a, * Masa Kageyama c, Charles T. Porter d, 1, Svante Bjo a

Department of Geology, Quaternary Sciences, Lund University, Lund Sweden Department of Geological Sciences and the Bolin Centre for Climate Research, Stockholm University, Stockholm, Sweden c Laboratoire des Sciences du Climat et de l'Environnement, 91191 Gif-sur-Yvette Cedex, France d Patagonian Research Foundation, Puerto Williams, Chile b

a r t i c l e i n f o

a b s t r a c t

Article history: Received 9 April 2015 Received in revised form 30 June 2015 Accepted 1 July 2015 Available online 14 July 2015

Lake sediments and peat deposits from two basins on Nightingale Island (37
S), in the Tristan da Cunha archipelago, South Atlantic, have been analyzed. The studies were focused on the time period 16.2 e10.0 cal ka BP, determined by 36 14C dates from the two sequences. A wide variety of proxies were used, including pollen and diatom analyzes, biogenic silica content, C and N analyzes, stable isotopes (13C and 15 N), elemental concentrations and magnetic susceptibility measurements, to detect environmental changes that can be related to shifts of the circulation belts of the Southern Ocean. In addition, climate model simulations were carried out. We find that the sediments are underlain by a >2 cal ka BP long hiatus, possibly representing a dried-out lake bed. The climate simulations corroborate that the area might have been exposed to arid conditions as a consequence of the Heinrich 1 event in the north and a southward displacement of the ITCZ. The development on the island after 16.2 cal ka BP is determined by the position of the Subtropical Front (STF) and the Southern Hemisphere Westerlies (SHW). The period 16.2e14.75 cal ka BP was characterized by varying influence from SHW and with STF situated south of Tristan da Cunha, ending with a humidity peak and cooler conditions. The stable conditions 14.7 e14.1 cal ka BP with cool and fairly arid conditions imply that STF and SHW were both north of the islands during the first part of the Antarctic Cold Reversal. The most unstable period, 14.1e12.7 cal ka BP, indicates incessant latitudinal shifts of the zonal circulation, perhaps related to climate variability in the Northern Hemisphere and bipolar seesaw mechanisms as the strength of the Atlantic Meridional Overturning Circulation (AMOC) varied. At 12.7 cal ka BP the Holocene warming began with a gradually drier and warmer climate as a result of a dampened AMOC during the Younger Dryas cooling in the north with ITCZ, STF and SHW being displaced southwards. Peak warming seems to have occurred in the earliest part of the Holocene, but this period was also characterized by humidity shifts, possibly an effect of retraction and expansion phases of SHW during AMOC variations in the north. © 2015 Elsevier Ltd. All rights reserved.

Keywords: Multiproxy study Model simulation Last termination South Atlantic Tristan da Cunha Southern hemisphere zonal circulation Subtropical front Bipolar see-saw climate pattern

1. Introduction and background The vast oceanic areas of the Southern Hemisphere (SH), and especially the Southern Ocean (SO) surrounding Antarctica (Fig. 1A), are in contrast to the Northern Hemisphere (NH), where continents dominate, especially in the Atlantic Ocean. As a result, Quaternary terrestrial records, which have the potential to reconstruct atmospheric conditions through high-resolution dated

* Corresponding author. € rck). E-mail address: [email protected] (S. Bjo 1 Deceased. http://dx.doi.org/10.1016/j.quascirev.2015.07.003 0277-3791/© 2015 Elsevier Ltd. All rights reserved.

records, are lacking in the SH. An exception to this are island records from the vast SH oceanic areas. Continuous island archives are ideally made up by lake sediments or peat deposits, which require basin structures. The challenge outside recently glaciated areas is, however, to find such structures that have not already been filled in by pre-Quaternary deposits. Oceanic islands are often part of young, and therefore often still active, volcanic systems. In such settings, caldera/crater lakes are common landscape features and often constitute the only basin type with continuous deposition. In areas with (still) active volcanism the volcanic activity may, however, be an obstacle for coring due to the occurrence of young widespread impenetrable tephra layers covering whole landscapes (cf. € rck et al., 2006). It is therefore essential to find basins whose Bjo

194

K. Ljung et al. / Quaternary Science Reviews 123 (2015) 193e214

A

60°

40°

20°

0°

20°

Stoltenhoff Island

W12°28’39’’

20°

Tristan da Cunha

Middle Island

l front ropica Sub t

40°

Polar fro nt

Antarctic divergence

60°

W12°27’

B

Tristan da Cunha

Nightingale Island

C Younger pyroclastic sequence Trachyte

Inaccessible Island

S37°15’

Nightingale Island

Huts

S37°25’19’’

4

th

2nd 1st

3rd

Older pyroclastic sequence Paths

10 km

1000 m

Fig. 1. A) Map of the Southern Ocean and surrounding continents with its main zonal circulation systems, the position of the Tristan da Cunha island group, B) its individual islands and, C) Nightingale Island with its four overgrown “ponds”. The geology is adapted from Baker et al. (1964). Note that the westerlies are situated between the Polar Front and the Subtropical Front.

surroundings have not experienced young volcanic eruptions and ash fall-out, but at the same time are not too old to have been infilled with much older deposits. For example, on the Azores only three out of 15 cored sites were not affected by young volcanic €rck et al., 2006), and still the oldest record was only c. activity (Bjo 6 cal ka BP old. In the South Atlantic, islands are rare north of 50
S. In addition, north of 30
S the sub-tropical high-pressure cell creates very dry and warm conditions, which mean that, e.g. the volcanic island of Ascension (8
S) has a too small P/E ratio to maintain wetland/lake habitats, and St. Helena (16
S) contains some scattered superficial peaty deposits but the only basin situated in the dry NE part of the island, the Central Basin of Prosperous Bay Plain, is filled in by pre-Quaternary lake sediments. The volcanism is also old without any traces of crater lakes. The Tristan da Cunha island group (Fig. 1) is comprised of Tristan da Cunha, Inaccessible Island and Nightingale Island (Fig. 1B) found at 37
S, and Gough Island slightly further south (40
S), just at, or north of some important oceanographic and atmospheric frontal systems, namely the Subtropical Front (STF) and the Southern Hemisphere Westerlies (SHW). All these islands are of volcanic origin of varying age, and have been known to hold peat and lake deposits (Hafsten, 1951, 1960; Wace and Dickson, 1965; Bennet et al., 1989), with indications of pre-Holocene age on scattered peat finds. From an expedition to the Tristan da Cunha group in 2003 it was clear for us that young volcanic activity and tephra fallouts on the main island, Tristan da Cunha (TDC), prevented attempts to reach down to sediments older than 2.3 cal ka BP (Ljung et al., 2006). However, on the near-by Nightingale Island (NI) it was possible to obtain an almost complete Holocene sequence in one of the four “Ponds” (2nd Pond) on the island (Fig. 1B) without € rck, 2007). In 2010 reaching the bottom of the basin (Ljung and Bjo

three of the ponds - which are made up of basins between lava ridges possibly formed during MIS3 (Bjørk et al., 2011) - were cored with stronger coring equipment. It was possible to reach down into the Last Termination in all three of them (1st, 2nd and 3rd). One of the records, 1st Pond, spans 37 cal ka, and here we present two of the records (1st and 2nd Pond), focusing on the period 11e16.2 cal ka BP. The concept of a bipolar seesaw climate mechanism (Broecker, 1998) during the Last Deglaciation has been convincingly shown to exist between the two polar areas as recorded in their ice core records (EPICA, 2006). The most distinct, well-dated and striking anti-phase shifts between the polar areas are the onset and warming of GIS-1e in Greenland, corresponding almost perfectly to the start of the Antarctic Cold Reversal (ACR), and the onset of the present interglacial warming in the very south during the early part of the Younger Dryas (YD) cool event in the NH. It has also been shown that the climate in some southerly regions outside Antarctica have responded in the same pattern as many of the Antarctic ice cores do, but they are in general situated south of 40
S in the Pacific (Newnham et al., 2012). Such records are, e.g. from glacial advances in southern New Zeeland (Turney et al., 2007; Putnam et al., 2010) and southerly shifts of the SHW in Tierra del € rck et al., 2012), and thereby also latitudinal changes in Fuego (Bjo sea ice extent, during the YD. The geographic extent of the effects of the bipolar seesaw in the SH is, however, still not resolved (Newnham et al., 2012). The close coupling between the atmospheric and oceanic circulation in the SH means that latitudinal shifts of the main atmospheric and oceanic fronts play a key climate role. Theoretical studies show that a cooling of the Northern Hemisphere (e.g. cooling of the North Atlantic) leads to a southwards shift of the ITCZ (Chiang and Bitz, 2005; Kang et al., 2008;

K. Ljung et al. / Quaternary Science Reviews 123 (2015) 193e214

Cvijanovic and Chiang, 2013) as the thermal equator also shifts due to the anomalous inter-hemispheric temperature gradient. These shifts have in modeling exercises been shown to propagate to midlatitudes and strengthen SHW in response to NH cooling (Lee et al., 2011). The NH cooling and shifts of ITCZ and SHW are interconnected with the northward ocean heat transport. As the NH Hadley circulation intensifies to compensate for the cooling, the northward ocean heat transport in the Atlantic decreases due to weakened AMOC, leading to a warming of the South Atlantic (Kageyama et al., 2013). It has also been shown that a shifting STF is a powerful climate modulator (Bard and Rickaby, 2009), and areas situated south of the STF in the southern oceans have displayed a more Antarctic climate signature during the Last Termination (Pahnke and Sachs, 2006). Mapping the temporal and spatial shifts of this front is therefore essential for a better understanding of the mechanisms behind the bipolar seesaw effects. We here present paleoclimatic records from two over-grown lake basins on NI in the central South Atlantic (Fig. 1), 1st Pond (1P) and 2nd Pond (2P) situated in the centre of the island at 210 and 200 m a.s.l., respectively. The Tristan da Cunha archipelago is strategically situated at the northern boundary of the SHW, and close to the southern limit of the STF, where salinities and temperatures decrease, 0.3‰ and 3e4
C, respectively, in a few degrees € rck, 2007; of latitude. From our previous studies (Ljung and Bjo Ljung et al., 2008; Lindvall et al., 2011; Holmgren et al., 2013) it has been shown that the area has experienced recurrent phases of increased precipitation during the Holocene, in many cases coinciding with phases of North Atlantic ice-rafting (Bond et al., 2001). This precipitation signal in the central South Atlantic is either a direct response to changes in the intensity of the SHW or an effect of rising sea surface temperatures (SST) of the ambient ocean (Renssen et al., 2001; Ljung et al., 2008). Since no unambiguous temperature proxy could be analyzed it has not been possible to determine if a bipolar mechanism influenced this fairly northerly situated area of the South Atlantic in the Holocene. However, we know that the preceding deglacial period, the Last Termination, was characterized by large-scale climatic changes and with clear bipolar climate effects between the poles (Blunier and Brook, 2001), but Last Termination paleoclimatic records from this central South Atlantic region have so far not been available. To document largescale changes of atmospheric and ocean circulation shifts is therefore the key aim of this study.

195

sequence at the depth of 618.8 cm (described below), where a previously continuous sedimentation seems to have halted between c. 18.5e16.2 cal ka BP, and the fact that we only have proxy data down to c. 13.7 cal ka BP in 2P, made it logical to first focus on the Last Termination. At 2P coring reached 2 m deeper than the 9.7 m managed in 2003. In addition, with a screw corer we reached c. 2 m further, where it was a clear stop. Due to the lack of continuous cores of these lowermost sediments it is impossible to give exact depths. However, between c. 0.5e1 m above the deepest part it was possible to obtain samples of clean sediment material from the screw, while the sediment on the deepest screw depth was not clean enough for any type of sampling. At an approximate depth of 13.17 m a screw sample was dated to 13,935 ± 85 14C yrs BP without SH correction. This age will not be included in the age model, since we did not trust detailed analyses of the “screw samples”, but the 1P sequence covers this time interval. 2.2. Radiocarbon dating In order to obtain reliable chronologies for the two sequences a number of radiocarbon (14C) measurements were carried out. All dated samples were pre-treated and measured at the Lund University Radiocarbon Dating Laboratory. Radiocarbon ages were calibrated using a stratigraphically constrained Bayesian model (PSequence) implemented in OxCal (Bronk Ramsey and Lee, 2013). The SHCal13 calibration curve was used for all calibrations (Hogg et al., 2013). In 1P we dated 19 levels between 510.5 and 618 cm, spanning 14 C ages between 10 and 13.5 kyr BP, corresponding to c. 11.5e16.2 cal ka BP (Table 1, Fig. 2). The material dated consists of 1 cm thick bulk sediment or peat (upper levels), but at two levels (504.5 and 587.5 cm) wood remains were dated, in the former case possibly from Phylica trees. In addition we show two of the dated, much older, levels just below the assumed hiatus (Fig. 2). In 2P 17 levels between 968 and 1172.5 cm were measured for 14 C, spanning 14C ages between 8.8 and 12 ka BP, corresponding to c. 10e13.8 cal ka BP (Table 2, Fig. 3). All samples consisted of 0.4e1 cm thick bulk sediment or peat (upper levels). In addition, one 4 cm thick sample from the end of the screw (1317 cm) was dated to 13.9 14C yr BP, but no other sediments between 1175 and 1317 could be sampled.

2. Material and methods 2.1. Field work, core correlations and sampling Field work on NI was carried out in February 2010. The ketch Ocean Tramp provided the transport from the Falkland Islands and back to Uruguay. The coring was carried out with 1 m long Russian chamber samplers (Ø 5 or 7.5 cm), with 15e50 cm overlap between each section. In order to penetrate as deep as possible into the often very stiff sediments a chain-hoist was used for coring the deeper parts of the sequences. The sediments were described immediately in the field before being wrapped in plastic film and put into PVC tubes. The cores were transported to the Department of Geology in Lund where they have been stored in a cold room. Before subsampling, the field-based lithostratigraphy and correlations between individual core sections were adjusted in the laboratory. This was aided by magnetic susceptibility (k) measurements to confirm and adjust the visual correlation between overlapping core segments. Due to difficulties in obtaining large volumes of sediments from the deeper parts of the sequences sampling resolution varies between different proxies. The discovery of a hiatus in the 1P

2.3. Elemental analyzes (C and N) Total carbon and nitrogen content were measured on dried and homogenized samples at 1e2 cm intervals with a Costech Instruments ECS 4010 elemental analyzer. The accuracy of the measurements is better than ±5% of the reported values based on replicated standard samples. Total carbon is used as a proxy for the organic content of the sediments. C/N atomic ratios were obtained by multiplying by 1.167 and are used to discriminate between terrestrial and aquatic organic matter sources (Meyers and Teranes, 2001). 2.4. Stable isotope analyzes (13C and

15

N)

Bulk stable carbon and nitrogen isotope compositions were measured on dried and homogenized samples using a ThermoFisher DeltaV ion ratio mass spectrometer. The isotopic composition of samples is reported as conventional d-values in parts per thousand relative to the Pee Dee Belemnite (13C) and atmospheric N (15N):

196

K. Ljung et al. / Quaternary Science Reviews 123 (2015) 193e214

Table 1 List of the radiocarbon dated levels in Fig. 2 from 1st Pond (1P). Calibrations with OxCal 4.2 based on SHCal13 (Reimer et al., 2013). All calibrated ages are in the 2s range. *regarded as outlier and hence excluded from the age-depth model. All samples were measured at the Lund University Radiocarbon Dating Laboratory. Lab no. all LuS dates

Dated material

Sample depth, (±0.5) cm

14

10,864 10,683 9487 10,315 9143 9142 9486 10,316* 9998 9485 9484 9483* 10,317 9717* 10,318 9482 10,319 10,320 10,321 9718 9716

Wood of Phylica Gyttja Peat Gyttja Peat Gyttja Gyttja Gyttja Gyttja Gyttja Wood Gyttja Gyttja Gyttja Gyttja Gyttja Gyttja Gyttja Gyttja Gyttja Gyttja

510.5 541 548.5 555 561 563 565.5 568 571 577.5 587.5 589.5 595 600.5 605 610.5 613 616 618 620.5 628.5

9985 10,060 10,120 10,390 10,515 10,620 10,670 10315* 11,185 11,500 12,130 13,090* 12570 11,230* 13,100 13,120 13,385 13,490 13,305 15,450 15,920

C age BP

h . i dsample ð‰Þ ¼ ðRsample  Rstandard Þ ðRstandard Þ  1000 where R is the abundance ratio of standard.

13

1s error

Age range, 2s error, cal yr BP

Mean calibrated age, cal yr BP

65 60 70 65 60 65 70 70 70 65 70 75 70 75 80 75 75 75 75 90 90

11,701e11,224 11,766e11265 11,959e11,324 12,428e11,849 12,635e12,059 12,691e12,417 12,704e12,432 12,402e11,762 13,132e12,805 13,443e13,147 14,125e13,760 15,899e15,313 15,120e14,314 13,206e12,825 15,928e15,317 15,951e15,346 16,278e15,794 16,458e15,922 16,196e15,705 18,886e18,472 19,435e18,912

11,463 11,516 11,642 12,139 12,347 12,554 12568 12,082 12,969 13,295 13,943 15,606 14,717 13,016 15,623 15,649 16,036 16,190 15951 18,669 19,174

clay minerals. We say that we use a modified CIA in that we have no NaO data.

C/12C in the sample or in the

2.5. Measurements for magnetic susceptibility Magnetic susceptibility (k) was measured using a Bartington MS2E1 high-resolution surface scanning sensor coupled to a TAMISCAN automatic logging conveyor. The measurements were carried out on non-sampled half cores and with a resolution of 5 mm. The magnetic susceptibility gives a relative estimate of the ability of the material to be magnetized, i.e. the magnetic mineral concentration. 2.6. XRF analyzes XRF analyzes were made on 66 freeze-dried sediment samples from 1P using a handheld Thermo Scientific portable XRF analyzer (h-XRF) Niton XL3t 970 GOLDDþ set in the Cu/Zn mining calibration mode. All analyses were performed by using an 8 mm radius spot size in order to get a representative analysis. The elemental detection depends partly on the duration of the analysis at each point and for this reason the measurement time was set to 6 min. Although a larger suite of elements were acquired, we have chosen to work with Al, Si, P, S, K, Ca, Ti, Mn, Fe, Rb, Sr and Zr. These elements were selected based on their analytical quality (i.e. level above the detection limit) and with the help of Principal Component Analysis (PCA). PCA was made using JMP 10.0.0 software in correlation mode using a Varimax rotation. Before analysis all data were converted to Z-scores calculated as (Xi  Xavg)/Xstd, where Xi is the normalized elemental peak areas and Xavg and Xstd are the series average and standard deviation, respectively, of the variable Xi. The number of factors extracted was determined using a scree test. To further support the interpretation of our principal components (PC) we use both Zr/Ti ratios and a modified Chemical Index of Alteration (CIA) as defined by Nesbitt and Young (1982):

CIA ¼ ½Al2 O3 =ðAl2 O3 þ CaO þ NaO þ K2 OÞ  100 This index expresses the relative proportion of Al2O3 to the more labile oxides and is an expression of the degradation of feldspars to

2.7. Diatom analyzes For preparation of diatom slides ~200 mg freeze-dried sediment was oxidized with 15% H2O2 for 24 h, then 30% H2O2 for a minimum of 24 h, and finally heated at 90
C for several hours. A known quantity of DVB (divinylbenzene) microspheres was added to 200 mL aliquots of the digested and cleaned slurries in order to estimate diatom concentrations (Battarbee and Keen, 1982; Wolfe, 1997). The diatoms were mounted in Naphrax® medium (refractive index ¼ 1.65). 68 of the 77 samples in 1st Pond and 36 of 63 samples in 2nd Pond contained enough diatom valves to be counted. The rest of the samples either had too few specimens, only small fragments of diatoms or contained no diatoms at all. In all samples containing diatoms, at least 300 valves per sample were counted and identified largely using Van de Vijver et al. (2002), Lange-Bertalot (1995), Krammer and Lange-Bertalot (1986e1991), Le Cohu and Maillard (1983, 1986) and Moser et al. (1995). Diatom results are expressed as relative abundances for each taxon, and as total concentrations of valves per g dry sediment. 2.8. Biogenic silica analyzes Seventy samples were analyzed for biogenic silica (BSi) in 1P using the wet-alkaline digestion technique described by Conley and Schelske (2001). The samples were freeze-dried and gently ground prior to analysis. Approximately 30 mg of sample was digested in 40 ml of a weak base (0.47 M Na2CO3) at 85
C for a total duration of 3 h. Subsamples of 1 ml were removed after 3 h and neutralized with 9 ml of 0.021 M HCl. Dissolved Si concentrations were measured with a continuous flow analyzer applying the automated Molybdate Blue Method (Grasshoff et al., 1983). The analytical results are considered to be within ±10% of the measured value. 2.9. Pollen analyzes Pollen samples of 1 cm3 were processed following standard method A as described by Berglund and Ralska-Jasiewiczowa (1986) with added Lycopodium spores for determination of pollen

K. Ljung et al. / Quaternary Science Reviews 123 (2015) 193e214

197

Fig. 2. Depth-age model for 1st Pond, including the calibrated 14C dates with 2s errors and the modeled dating uncertainties, related to the lithostratigraphic units and magnetic susceptibility. Note the two lowermost anomalous dates just below the hiatus and the three outliers not included in the modeling.

concentration values. The counting was made under a light microscope at magnifications of 400 and 1000. Identification of pollen grains and spores was facilitated by published photos (Hafsten, 1960), standard pollen keys (Moore et al., 1991), and a small collection of type slides. Pollen diagrams were plotted in C2 (Juggins, 2003) and divided into local pollen zones based on the frequencies of the major taxa. Concentration and influx calculations of the major taxa are also presented. 2.10. Climate simulations The main purpose of the climate simulations was to test our hypothesis that the hiatus in 1P could have been caused by arid (and warm?) conditions in the Tristan da Cunha region during the

Heinrich 1 event (H1). For the climate modeling we use the IPSL_CM4 coupled ocean-atmosphere model (Marti et al., 2006, 2010). The atmospheric component is the LMDZ.3.3 model with a horizontal resolution of ~3.75
 2.5
in a regular grid and 19 vertical hybrid levels. The ocean component is the ORCA2 model with an irregular horizontal resolution of ~2
and 31 vertical levels. The model is run for two simulations with constant forcings, a preindustrial control (CTRL) and a 17 ka BP run (17k), as well as a 17 ka BP simulation with freshwater hosing in the North Atlantic (17k_FW). The CTRL is a standard PMIP2 pre-industrial control run, while the 17k run uses the PMIP2 Last Glacial Maximum boundary conditions (Braconnot et al., 2007 and http://pmip2.lsce.ipsl.fr) except for the astronomical parameters which are the 17 ka BP values according to Berger (1978). The simulation is initialized from

198

K. Ljung et al. / Quaternary Science Reviews 123 (2015) 193e214

Table 2 List of radiocarbon dated samples from 2nd Pond (2P). Calibrations with OxCal 4.2 based on SHCal13 (Reimer et al., 2013). All calibrated ages are in the 2s range and all samples were used in the age-depth model. All samples were measured at the Lund University Radiocarbon Dating Laboratory. Lab no. all LuS dates

Dated material

Sample depth, cm (±0.2e0.5)

14

1s error

Age range, 2s error, cal yr BP

Mean calibrated age, cal yr BP

9224 9225 9226 9227 9228 9229 9230 9231 9232 9665 9233 9666 9667 9670 9669 9668 9063

Peat Peat Peat Peat Peat Peat Peat Gyttja Gyttja Gyttja Gyttja Gyttja Gyttja Gyttja Gyttja Gyttja Gyttja

968 1010 1040 1055 1070 1075 1095 1100 1121 1131 1146 1149 1151 1155 1161 1165 1172.5

8815 9170 9595 9770 9990 10,075 10075 10,160 10,200 10,270 10,560 10,850 10,830 10,845 11,340 11,260 11,995

60 60 65 60 65 70 65 70 70 70 75 70 70 70 70 70 80

10,136e9554 10,489e10,194 11,158e10,680 11,253e10,794 11,704e11,227 11,916e11,258 11,820e11,265 12,006e11,393 12,041e11,405 12,385e11,616 12,664e12,084 12,830e12,570 12,808e12,564 12,824e12,568 13,297e13,046 13,260e12,896 14,022e13,570

9845 10,342 10,919 11,024 11,466 11,587 11,543 11,700 11,723 12,001 12,374 12,700 12,686 12,696 13,172 13,078 13,796

C age BP

the corresponding PMIP2 run and run for 200 years. For the 17k_FW a freshwater flux of 0.1 Sv is imposed north of 40
N in the Atlantic and the Arctic Oceans in order to slow down the Atlantic Meridional Overturning Circulation (AMOC), hence emulating the climate system behavior during the H1 event. The specifications of the model experiments are summarized in Table 3. We analyze the climatology of the last 100 years of the CTRL and 17k simulations, while the mean of year 80e100 after the initiation of the freshwater hosing of 17k_FW is analyzed. 3. Results and proxy interpretation 3.1. Lithostratigraphy The deposits in the cored basins on NI have some general and common features: they are very rich in organic material, are often dark colored (gray-brown to blackish brown) and are very stiff and compacted only a few meters below the surface. As such it is difficult to identify vague changes within units and to optically draw lithologic boundaries between, e.g. a highly organic lacustrine gyttja and highly humified swamp/fen peat. In 1P we reached down to 937 cm, but here we will only describe the deposits from slightly below the supposed hiatus to the onset of the Holocene, i.e. 620e510 cm. The sediment unit between 620 and 551 cm in 1P consists of a dark brown silty algal-rich fine detritus gyttja with higher minerogenic content below 615 cm. Rather elusive color changes appear throughout the unit. A 1 mm thin sand/silt layer occurs at 618.8 cm (Fig. 2), slightly undulating and with some 3e4 mm large ball-like accumulations of sand connected to this layer. The upper boundary to the following 8 cm thick unit, 551e543 cm, a low humified swamp peat with coarser plant remains, is gradual. Much of the coarser plant remains consist of wood from Phylica arborea (Island Tree). With a fairly sharp upper boundary, this peat is overlain by a 6 cm thick unit of dark brown compact fine detritus gyttja, very rich in organic material, between 543 and 537 cm. The upper boundary of this lacustrine unit is very gradual, shifting into a dark brown highly humified swamp peat, which is 19 cm thick. At 518 cm it gradually turns into a dark brown relatively low humified swamp peat. A large piece of Phylica wood is found at 511e510 cm. This type of peat, but with variable humification degree, occurs up to the top of the sequence. The sediments in 2P found on the screw below the deepest cored level are very similar to the bottom sediments found in the

deepest cored sequence. This first unit, between 1175 and 1097 cm, consists of an extremely compact blackish-brown clayey-silty gyttja without any distinct lithologic changes. However, the upper part between 1110 and 1097 shifts between more peat-like (1110e1104 cm) and gyttja (1104e1097 cm) deposits (Fig. 3), indicating that the transition from lake sediments to peat was not completely gradual. The upper boundary at 1097 cm is fairly sharp where it turns into a black, highly humified swamp peat, and this unit is 8.5 cm thick. At 1088.5 cm it changes rather abruptly to a brown, less humified peat, possibly a swamp peat. This peat, with a thickness of 17.5 cm, is significantly different from the under- and overlying peat. At 1071 cm it turns rather rapidly into a very compact, black, highly humified swamp peat. This unit continues upwards until at least 950 cm, overlapping with the sequence € rck (2007). Because of this overlap the described by Ljung and Bjo present study from 2P will describe the development below c. 990 cm. To sum up, both sequences are characterized by highly organic lake sediments (Fig. 4) ending up with overgrowth of the basins. The very minor lithologic changes of the gyttjas can hardly be used for climatic interpretations, but one important aspect with the overgrowth is if it occurred simultaneously or not. 3.2. Chronology The Bayesian age modeling (Bronk Ramsey and Lee, 2013) of the two sequences resulted in the two age-depth models (Figs. 2e3). In 1P the accumulation rate is low before peat deposition starts, with a mean rate of 0.15 mm/yr, varying between 0.27 and 0.11 mm/yr, and without any obvious abrupt changes but rather gradual ones. In contrast to the low sedimentation rates in the gyttja, the peat between 550 and 500 cm displays a mean accumulation rate of 1.4 mm/yr. The sedimentation rates of the gyttja in 2P are more than twice as high compared to 1P, with a mean rate of 0.36 mm/yr, varying between 0.24 and 0.88 mm/yr. The latter value is found between 1110 and 1097 cm, where shifts between gyttja and more peaty deposits occur, but in the section below the highest value is 0.5 mm/yr. The mean accumulation rate in the peat is 0.73 mm/yr with only minor variations according to the age model. 3.3. Magnetic susceptibility Magnetic susceptibility is a rough proxy for amount of mineral matter in the sediments, and the 1P susceptibility record shows an

K. Ljung et al. / Quaternary Science Reviews 123 (2015) 193e214

199

Fig. 3. Depth-age model for 2nd Pond, showing calibrated 14C dates with 2s errors and the modeled dating uncertainties, related to the lithostratigraphic units and magnetic susceptibility. Note the 14C plateau at 11.5 cal ka BP and that all dated levels lie within the modeled uncertainties.

Table 3 Summary of specifications for model runs with orbital parameters (eccentricity, obliquity, precession) according to Berger (1978), atmospheric greenhouse gas concentrations (CO2, CH4, N2O), ice sheet configuration and sea level reconstruction and imposed freshwater flux in the North Atlantic. See also text. Experiment name

Ecc.

Obl.

Pre.

CO2 (ppm)

CH4 (ppb)

N2O (ppb)

Ice sheets/sea level

Fresh-water hosing

CTRL 17k 17k_FW

0.016715 0.019486 0.019486

23.441
23.612
23.612


102.7
180.7
180.7


286 185 185

791 350 350

275 200 200

Present day PMIP2 21 ka BP PMIP2 21 ka BP

None None 0.1 Sv

upwards slightly declining trend (Fig. 2), possibly as a function of gradually less mineral particles, but interrupted by several more or less distinct peaks. Such peaks occur at 15.75, 14.9e14.8, 13.85, 13.7, 13 and 12.8 cal ka BP, and are most likely the result of increased inwash of mineral particles as an effect of higher precipitation. The

low values after 12 cal ka BP reflect the transition into swamp peat, while the higher values at 11.6 cal ka BP with its gyttja sediment, and a minor peak at 11.4 cal ka BP, show that the depositional environment shifted between a peaty wetland and a very shallow limnic setting.

200

K. Ljung et al. / Quaternary Science Reviews 123 (2015) 193e214

Fig. 4. Graphs showing percentage changes in total organic carbon (TOC) and nitrogen (N) of the sediments in 1st Pond, left, and 2nd Pond, centre. To the right the C/N ratios are shown from both sites.

In 2P we also see an upwards declining and oscillating trend interrupted by several more distinct peaks, e.g. at 13.05, 12.5, 12.3 and 11.65 cal ka BP (Fig. 3). The values are fairly similar to 1P, with the distinction that the older peaks are more pronounced in 1P; the 2P record starts when the 1P peaks become less prominent. This may be related to gradually denser vegetation, reducing the effect of surface run-off/soil erosion during periods of higher precipitation. After 11.6 cal ka BP the values reach a minimum as a consequence of the transition from lake sediments (gyttja) to peat, coinciding in time with the lowest values in 1P as well as with the onset of the Holocene (Walker et al., 2009). This implies that the earliest centuries of the Holocene were the driest period since the hiatus was developed in 1P. The low values continue for 300e400 yrs whereafter they increase and stay relatively high until 10.75 cal ka BP. This is followed by 450 yrs of low values, suddenly shifting into 300 yrs of higher values until 10 cal ka BP. This pattern, possibly related to wetness changes, is similar to the 1P record and shows that both sites experienced parallel hydrologic changes during the early Holocene.

3.4. Total carbon (TOC) and nitrogen (TN) values and C/N ratios The amount of total carbon (TOC) in an environment devoid of carbonates, such as a volcanic island like Nightingale Island, is a measure of organic carbon in the sediments. Organic matter in lake sediments is either produced in the lake itself or originates from the terrestrial environment around the basin, while most of the measured nitrogen (TN) in sediments has been produced in the aquatic environment. Therefore increased C/N ratios are good proxies for higher flux of terrestrial organic matter into the lake.

1P shows an upwards-rising trend in TOC (Fig. 4) reflected in the minimum values (5.7%) in the bottommost sample and the maximum values (41%) in the top. The low values in the bottom are related to the fact that the hiatus in the bottom is developed as a thin silty layer. After the low values in the bottom the upwardsincreasing trend shows an oscillating pattern, with high values at 15.9e15.6, 15.3e15.2, 14.1e13.7, 13.4, 12.7, 12.25, 11.65 and at the very top at 11.5 cal ka BP, with some distinct minima in between these peaks. The most conspicuous of these occurs at 14.9 cal ka BP, but others are found at 15.4, 13.5, 13, 12.5, 12.1, and at 11.6 cal ka BP. This pattern is fairly similar to the TN pattern, implying that it is partly a good proxy for aquatic productivity. However, the pattern is far from perfect; some TOC peaks occur without TN peaks indicating more influence of terrestrial organic matter, such as at 12.7, 12.4 and 12 cal ka BP. This can be deciphered with the C/N ratios, which show a general rising trend, and the increasing influence of terrestrial organic matter (Fig. 4). Five C/N peaks occur before the gyttja grades into peat: at 14.8, 14, 12.7, 12.4e12.3 and 12 cal ka BP. All these peaks are interpreted as being related to increased inwash of terrestrial organic matter due to high precipitation and surface run-off. The transition into peat at 11.7 cal ka BP is revealed as a maximum in C/N ratio (43) followed by lower values during the deposition of the gyttja at 11.6 cal ka BP, followed again by higher values in the overlying peat. The upwards-increasing trends of the TOC and C/N records in 1P are also found in 2P (Fig. 4). In fact, parts of the curves show striking similarities, in spite of the fact that they are based on separate chronologies, while other parts differ rather significantly. The 2P TOC record displays a series of peaks, at 13.3, 12.9, 12.4, 12.1, 11.8, 11.65, and finally at 11.5 cal ka BP when the peat begins to form, and between these some clear minima occur at 13.1, 12.7, 12.35e12.2,

K. Ljung et al. / Quaternary Science Reviews 123 (2015) 193e214

11.75 and 11.6 cal ka BP. It is noteworthy that the gyttjaepeat transition is characterized by large TOC and C/N shifts, simultaneous with the shifts we see in the 1P TOC and C/N records (Fig. 4), implying that the two basins responded in the same way at the PleistoceneeHolocene transition. The 2P TN record displays fairly subtle changes between 13.65 and 12.05 cal ka BP, followed by three peaks at 12e11.8, 11.6 and 11.5 cal ka BP, coinciding with TOC peaks. The C/N ratios show an oscillating and rising trend, as in 1P, implying increased influence of terrestrial organic matter, with peaks coinciding with the most clear TOC peaks at 12.9, 12.4, 12.1 and 11.65 cal ka BP. In contrast to these periods we see a more aquatic organic production before 13 cal ka BP and at 12.8e12.7, 12.4e12.2 and 12e11.7 cal ka BP when TN values are high and C/N ratios low. Short-lived phases like that can also be seen at 11.6 and just before 11.5 cal ka BP, suggesting that the transition into a more or less peat-covered wetland was characterized by the occurrence of larger or smaller pools in the wetland. The following peat displays three distinct maxima at 10.7, 10.3 and 10.15 cal ka BP with values of 42, 62 and 52, respectively, coinciding with TN minima (Fig. 4). These periods were probably characterized by more arid conditions. The partly coinciding trends between the two records mixed with separate responses during certain time periods can give us clues about the hydrologic changes that the basins have undergone. While 1P is the highest situated and exposed basin of the drainage area, 2P is most likely the deepest (Bjørk et al., 2011) and most sheltered. Therefore one would expect a higher sedimentation rate in 2P due to its larger drainage area, but its greater depth could partly compensate for that. It would, however, be unlikely that the general in-filling and over-growth of the two basins would occur simultaneously if it would not be forced by climatic/hydrologic factors. The partly coinciding trends we see in the two C and N records are possibly linked to the general in-filling, while the anomalies between them can probably be explained by how their different settings influence how they respond to hydrologic changes. For example, periods with higher precipitation and rising water table would most likely have larger effect on 2P, while arid periods would affect 1P more. 3.5. Stable isotopes (13C and

15

201

a substantial input of nitrogen from sea-birds nesting in the catchment. It has been shown on sub-Antarctic islands that seabird populations can contribute significantly to the deposition of nitrogen, shifting 15N ratios to higher values (Erskine et al., 1998; Bokhorst et al., 2007). Today 2e3 million pairs of different sea birds breed on the island. These birds bring enormous amounts of nutrients from the sea to the island. Sea-bird excrement is heavily enriched in 15N resulting in highly enriched soils and terrestrial plants in breeding areas (Erskine et al., 1998). The highly enriched d15N values (Fig. 5) clearly show that there has been substantial input of nitrogen-enriched in 15N by sea birds. There are two principal transport ways for nitrogen into the basin, one via direct deposition of excrement and detrital material and, another by leaching of ammonia and nitrate into the aquatic systems. The shifts in d15N therefore reflect the abundance of sea birds in the catchment and changes in the nutrient transport. 3.6. Biogenic silica (BSi) In general, BSi values are rather low in 1P varying between 0.7% and 10.9%. However, substantial centennial-scale variability is present in the data (Fig. 5). From 16.2 cal ka BP, a gradual increase in BSi occurs reaching a maximum of 10.9% at 15.6 cal ka. BSi remains high for c. 100 yr followed by a stepwise drop, first to 5e6 % between 15.5 and 15 cal ka BP, followed by a rapid decrease to <2% and remaining low until 14.6 ka. Thereafter an increase to almost 4% occurs, lasting between 14.5 and 14.1 cal ka BP after which BSi drops to 1% until 13.8 cal ka BP. From 13.7 cal ka BP and onwards, BSi increases and fluctuates between 3 and 5% until 13.1 cal ka BP after which the values gradually decrease to 2.5%. A sudden rise to 7% occurs at 12.7 cal ka BP and values remain high till 12.5 cal ka BP, followed by lower and oscillating values (3e5%) until 12.2 cal ka BP. Thereafter there is a general decreasing trend, with the lowest values (0.7%) in the topmost analyzed sample at 11.7 cal ka BP. The BSi values show some remarkable shifts, implying rather high but oscillating lake productivity between 16.2 and 14.9 cal ka BP (Fig. 5) as well as between 13.6 and 12.1 cal ka BP. In between, 14.8e13.7 cal ka BP, aquatic productivity was lower and with significantly smaller fluctuations.

N) 3.7. XRF data

Stable isotope analyses were only carried out on 1st Pond (Fig. 5), and the bulk sediment d13C values vary between 28.4 and 25.8‰ with an average of 27.2‰. Such values are typical for C3 land plants. The d13C values display a general decline from 25.5‰ in the lowermost part, to values lower than 28‰ at the start of the peat in the top. This general decline is interrupted by periods of higher values at 15.2e14.7 cal ka BP and 13.7e13 cal ka BP. The upwards declining d13C values coincide with increasing TOC and C/N, indicating a higher proportion of terrestrial organic matter in the sediments. There is also a sharp decline in d13C at 12.7 cal ka BP when the transition from lake to wetland conditions starts and C/N ratios increase. The general trend of declining d13C values is therefore explained by increased deposition of terrestrial organic matter. Changes in the catchment vegetation can also have influenced the d13C of the sediments. The major species in the tussock grass vegetation, Spartina arundinacea, is a C4 plant with heavily enriched d13C. Declining grass coverage in the catchment of 1st Pond is indicated by the pollen results, and the lowermost pollen zone with highest grass pollen frequencies and influx values coincides with the highest d13C values (25.5 permil). Bulk sediment 15N values vary between 14.3 and 11.4‰ with an average of 12.8‰ if the two outliers with significantly lower 15N, 2.1 and 7.9 are not included. The generally high values of 15N indicate

Principal component analyses (PCA) of the XRF data from 1st Pond extracted several components but here we focus on the first two (Fig. 1 in Supplementary material). The first principal component (PC1), accounting for 38% of the variance in the data, extracted Rb, Sr and Zr on the positive axis and Ca and S on the negative axis. It is possible to see how PC1 changes over time by plotting factor scores versus age. Significant positive and negative factor scores in PC1 are found between 16.13 and 12.25 cal ka BP (Fig. 6). From 12.14 cal ka BP to the top of the record values go from moderately positive to the most extreme negative PC1 factor scores of the profile at 11.68 cal ka BP. Zirconium does not play a role in biotic processes and is mainly found in the weathering resistant silicate mineral zircon. Rubidium is generally found in K-rich minerals such as potassium feldspar and displays inert behavior during weathering while Sr is associated with the Ca-bearing minerals like plagioclase feldspar (Brady, 1990). Zircon is normally enriched in medium to coarse silts while feldspars are enriched in coarse silts and sands (Taboada et al., 2006; Taylor and McLennan, 1985). Changes in PC1 are thus partly driven by variations in allochthonous silicate mineral inputs to the lake basin. Such variations can be the result of transportdepositional processes affecting the size and properties of the deposited materials and/or changes in the dominant sediment

202

K. Ljung et al. / Quaternary Science Reviews 123 (2015) 193e214

Fig. 5. Graphs of proxies from 1st Pond sediments show from left to right: the content of biogenic silica (%), total diatom concentrations and concentrations of S. venter (a diatom species often dominant in larger pools/water bodies), TOC (%), d13C (‰) and d15N (‰).

PC1 -3

-2 -1

0

Zr/Ti 1

2

3 0

0.1

0.2

PC2 0.3

-3

-2 -1

0

Modified CIA 1

2

3 50 55

60

65

70

75

11.5

12.0

12.5

Age (cal ka BP)

13.0

13.5

14.0

14.5

15.0

15.5 16.0

16.5 Increasing grain size and/or change in source

Increasing clay content/ chemical weathering

Fig. 6. Plots of PCA analysis of the XRF data set with, from left to right, PC1 factor scores (38%), Zr/Ti ratios, PC2 factor scores (27%) and a modified CIA against age at 1st Pond.

K. Ljung et al. / Quaternary Science Reviews 123 (2015) 193e214

source. Given the small size of the catchment it is likely that this is a grain size controlled response. This cannot be confirmed without grain size analyses, which was not possible to carry out because of too small sample sizes. This interpretation is confirmed by the high correlation of PC1 to the Zr/Ti ratio (r ¼ 0.85) (Fig. 7). The Zr/Ti ratio has been used previously to track changes in silicate inputs to a lake (Kylander et al., 2013) and is based on the fact that Zr is hosted in coarser fractions while Ti is associated with finer fractions (Taboada et al., 2006). As such an increase in the Zr/Ti ratio indicates less clay/ more silt in the sediments and/or a change in the silicate source. The negative association of Ca and S with PC1, both found bound to organic matter in lake sediments, would suggest that productivity and vegetation cover in the catchment are important controls on sediment grain size and PC1. PC2 explains 27% of the variance in the data and highlights the association between Al, K, Ti and, to a lesser extent, Si. PC2 factor scores show a number of extreme positive and negative values (Fig. 7). Positive values for PC2 are found at 15.86, 15.53, 14.83, 14.59e14.21, 13.49, 12.91, 12.61 and 12.53 cal ka BP while negative peaks are found at 15.32, 14.028, 13.24, 12.72, 12.44 and 11.92 cal ka BP. The bedrock of Nightingale Island is mainly comprised of trachybasalts (Baker et al., 1964). The dominant minerals in trachybasalts are plagioclase and alkali feldspars, pyroxene and olivine. Chemical weathering of these mineral groups produces secondary clay minerals, which readily host Al, Si, Ti and K. As such increases in PC2 factor scores indicate greater inputs of clay minerals, whose formation is driven by the chemical weathering of feldspars. Another way to examine clay minerals and chemical weathering is to look at the CIA. It should be noted that we use the CIA here as a supporting indicator only since Na data are not available and we have not corrected for post depositional processes (Xiao et al., 2010). Nonetheless, the correlation between PC2 and CIA is rather good (r ¼ 0.40) considering that these two indicators are built on different suites of elements (Al, Ca, K versus Al, K, Si, Ti) and Na data are missing; examination of the profiles

203

finds that most major peaks are captured in both PC2 and the CIA (Fig. 6). The CIA expresses the ratio of feldspars to clay minerals where higher values are indicative of higher clay content and therefore greater rates of chemical weathering (Nesbitt and Young, 1982). Chemical weathering of silicates is generally thought to be controlled by temperature and precipitation but other interlinked factors include physical erosion rates, presence/depth of soil cover, mineralogy, runoff and the presence of acids (e.g., Anderson, 2005; West et al., 2005; White and Blum, 1995). 3.8. Diatom assemblage changes The diatom records from the two basins are presented as relative abundances and as total diatom valve concentrations (Figs. 7e8) related to age. In order to simplify the description of the assemblage variations in 1st Pond and 2nd Pond the diatom stratigraphies are divided into 3 and 6 zones respectively (Tables 1e2 in Supplementary material). Below follow short accounts of what type of environment the diatom assemblages imply. 3.8.1. 1st pond Between 16.15 and 15.69 cal ka BP (Zone 1) the high abundances of Staurosira venter, Stauroforma exiguiformis, Aulacoseira spp. with high total diatom valve concentrations and relatively high concentrations of both terrestrial and meroplanktonic taxa show that 1st Pond was a rather big lake/pool with a high inwash of terrestrial diatoms from the catchment, most likely due to enhanced precipitation. The meroplanktonic Aulacoseira distans included in the Aulacoseira spp. is a heavy silicified species and needs turbulent water to stay in suspension, either from shallow waters or high wind exposure (Saunders et al., 2009). The dominating taxa S. venter prefers rather large pools (Van de Vijver et al., 2002), and it is therefore likely that Aulacoseira spp. thrived due to high wind exposure on the lake.

Fig. 7. Diatom percentage diagram from 1st Pond with the most common types, grouped into meroplanktonic and benthic (and epiphytic) taxa, and also a curve for terrestrial/ aerophytic taxa. To the right concentration curves of some of the most common types are shown as well as the zonation.

204

K. Ljung et al. / Quaternary Science Reviews 123 (2015) 193e214

Fig. 8. Diatom percentage diagram from 2nd Pond with the most common types, grouped into benthic (and epiphytic) and terrestrial (and aerophytic) taxa. To the right total concentrations and concentrations of terrestrial taxa are shown as well as the zonation.

The high abundances of Pinnularia sp and Frustulia cf. rhomboides in Zone II, 15.63e15.39 cal ka BP, suggest rather acidic waters. Very low abundances of S. exiguiformis and S. venter with low total diatom concentrations indicate that 1st Pond was a smaller lake or a shallow pool with turbulent water. The flora between 15.31 and 13.95 cal ka BP (Zone III) suggests that 1st Pond was a relatively large water body at the start of this zone, with the lower part being dominated by S. venter whereas the upper part is dominated by S. exiguiformis. This suggests a transition from a larger to a smaller pool, possibly with more acidic water as shown by increasing abundances of Pinnularia spp and Eunotias, and possibly ending up with a more or less dried out phase. The mix of high total diatom concentrations, disappearance of diatoms, the low concentration of terrestrial taxa and the dominating S. exiguiformis between 13.88 and 13.29 cal ka BP (Zone IV) suggest a shallow, occasionally dried out pool with restricted catchment erosion. The disappearance of S. exiguiformis and S. venter between 13.24 and 12.2 cal ka BP (Zone V) and the dominance of Frustulia cf. rhomboides, Pinnularia viridis and Eunotia sp. imply that 1st Pond was a very shallow and small pool with acidic waters. The complete dominance of Pinnularia viridis at 12.14e11.6 cal ka BP (Zone VI) and the lack of benthic diatoms suggest that 1st Pond was a wetland during this period. The sudden and short dominance of S. exiguiformis, S. venter, Staurosira leptostauron, Staurosirella pinnata and Gomphonema spp. at the very onset of the Holocene suggests a very short phase of more waterlogged conditions; a small water-body might have developed in the wetland before the basin was totally over-grown. 3.8.2. 2nd pond In general the diatom assemblage in 2nd Pond is surprisingly different from the assemblage in 1st Pond, but with a few common types such as Frustulia cf. rhomboides and Pinnularia viridis (Figs. 7e8).

The diatom assemblages between 13.69 and 12.38 cal ka BP (zones I and II in Table 2, Supplementary information) have the character of a slightly acidic, oligotrophic lake or pool. Frustulia cf. rhomboides, Achnanthes saxonica, Chaemopinnularia spp and Naviculadicta spp are common species in oligotrophic lakes and pools in the sub-Antarctic area (Van de Vijver et al., 2002) indicating rather acidic waters. The lower abundances of Frustulia cf. rhomboides, A. saxonica, Chaemopinnularia spp, Naviculadicta spp, Eunotia paludosa var. paludosa and the dominance of Eunotia serra tetraodon and Pinnularia viridis in Zone III, 12.73e11.71 cal ka BP, suggest that the previously aquatic environment turned into a wetland. The terrestrial diatoms were likely transported to the site from the catchment by surface run-off during periods of increased erosion, i.e. during periods of increased precipitation. While zones IeII contain terrestrial diatom valves in relatively high concentrations, zone III hardly contains any at all. This lends further support for the hypothesis of drier conditions in zone III, ending up with a bog after 11.7 cal ka BP.

3.9. Pollen zonation e vegetation changes The pollen results from 1st Pond are presented as percentages (Fig. 9), and pollen concentration and influx (Fig. 10) for the major taxa. The pollen percentage diagram has been divided into six pollen assemblage zones based on diagnostic pollen types. In total 16 taxa of flowering plants and seven taxa of Pteridophytes have been identified. One unidentified taxa occurs throughout the sequence with frequencies up to 40%. The pollen grain is tri-colpate, and it has not been possible to unambiguously identify it. In the lowermost zone 1P-1 (16.18e15.74 cal ka BP) (see Table 3, Supplementary material) the relatively high Poaceae and low P. arborea and herb pollen frequencies and influx values indicate that the vegetation around the pond was dominated by tussock grass vegetation. The high influx values of Empetrum rubrum and open

K. Ljung et al. / Quaternary Science Reviews 123 (2015) 193e214

205

)

e in

n le ol lp

To

ta

lyp Po

flu S x (c edi m m /y en ea ta r) tio n ra t

on (c e ea

ia c od

od lyp Po

c.

x) flu e

ea ia c

e ric h al lit C

al lit C

(in

te ris ch

ris ch e ric h

ae Po

ac e

ae ac e

te

) (c

(in

on

flu

c.

x)

on (c ea e Po

ac

ns

en ns

) c.

x) flu yp er C

en

ii

ii

(c

(in

) on (in ea e ac yp er

C

Em pe

tru m

tru m

ru b

ru b

ru m

ru m

(c

(in

on (c re a Em pe

ca yli Ph

c.

x) flu

) c.

x) flu (in ar bo

re a ar bo ca yli Ph

on

flu

c.

x)

)

Fig. 9. A pollen and spore percentage diagram from 1st Pond, including charcoal, grouped into different taxa types. The pollen sum and the zonation are shown to the right.

11.0 1P-6

11.5

1P-5

12.0

1P-4

12.5

Cal ka BP

13.0 13.5 1P-3

14.0 14.5 15.0 1P-2

15.5 16.0

1P-1

16.5 0

24000 0

90000

0

2000 0

3

16000

0

8000

0 200000

2

0

3200

0

150000 0

900

0

16000

0

48000

0

600000

0 16000

0 0.04 0.12 0.20

Fig. 10. A combined concentration (grains/cm ) and influx (grains/cm /year) diagram of the most common pollen grains and spores as well as the total pollen influx and sedimentation rate.

206

K. Ljung et al. / Quaternary Science Reviews 123 (2015) 193e214

ground species indicate that areas on the island were relatively open. Between 15.69 and 14.76 cal ka BP the high Cyperaceae pollen frequencies (Fig. 9) is most likely caused by floating mats of Scirpus, which is a common feature in pools with open water on the island today. The gradual increase in P. arborea pollen indicates an expansion of a local tree cover. Few finds of pollen from taxa favored by open areas, such as Chenopodiaceae or Asteraceae, indicate that the vegetation surrounding the site was dense without much disturbance. Histipoteris incisa often forms the under-storey in dense Phylica stands, limiting other under-storey vegetation. The low percentages and influx of Empetrum rubrum pollen also support a denser vegetation cover. The most marked change between 14.69 and 12.91 cal ka BP is the low values of P. arborea. There are also a relatively high number of taxa associated with open or disturbed ground, such as Chenopodiaceae, Asteraceae, and Aster-type. This indicates more open vegetation, perhaps with disturbance from nesting birds. The high Cyperaceae pollen frequencies indicate that the pond had areas of floating Scirpus mats. Callitriche occurs in low frequencies and is completely absent in some samples. It is possible that Callitriche was outcompeted by Scirpus and that most of the shallow water along the shores of the site was covered by floating Scirpus mats. The high Phylica frequencies in Zone 1P-4 (12.56e11.58) indicate a locally substantial tree growth. This is the start of a gradual overgrowth of the site and Phylica probably grew on the peat surface together with Blechnum palmiforme as indicated by the high influx of Polypodiaceae, which includes Blechnum species. The declining Cyperaceae frequencies are probably a consequence of diminishing open water with floating Scirpus mats. Presence of taxa indicating open or disturbed ground, such as Rumex frutescens and Asteraceae, indicates at least partly open vegetation in the catchment. The increase in Empetrum rubrum is probably a consequence of growth on peat lands around the pond. The next period, 11.54e11.47 cal ka BP, is characterized by the complete dominance of Cyperaceae and indicates a wetland covered with mats of Scirpus, possibly as a consequence of a brief increase in water table, flooding the surface. We also find taxa indicating open or disturbed ground, such as R. frutescens and Asteraceae, indicating at least partly open vegetation in the catchment. The high Phylica frequencies (Fig. 9) and influx values (Fig. 10) in Zone 1P-6, 11.44e11.33 cal ka BP; indicate dense local stands, probably on a relatively dry peat surface. The following increase in Poaceae and Apium australe indicate slightly less dense Phylica vegetation at the site. 3.9.1. Summary of the pollen assemblage changes/vegetation development The pollen diagram from 1st Pond contains similar species as previous pollen diagrams from Nightingale Island and Tristan da €rck, 2007; Ljung et al., 2006). Cunha (Hafsten, 1960; Ljung and Bjo All pollen species present in the diagram, except possibly one yet undetermined pollen taxa are present in the modern day flora. There appears to have been no major changes in the species present on the Tristan da Cunha islands since the Late Glacial period. The top part of the pollen diagram from 1st Pond is very similar € rck, to the lower part of the diagram from 2nd Pond (Ljung and Bjo 2007): Phylica frequencies are high, about 10e20%, in both diagrams, and Cyperaceae and Poaceae are present with fairly high frequencies. The pollen spectra from these periods in 1st and 2nd Pond represent wetland vegetation with dry peat surfaces where Phylica could grow. Phylica is almost completely absent from the pollen spectra before 15.6 cal ka BP, pollen zone 1P-1. The scattered finds of Phylica

pollen in the 1P-1 indicate that Phylica trees were present on the island, but in low numbers. Between 15.6 cal ka BP and 14.6 cal ka BP Phylica pollen grains are more common indicating more abundant tree growth in the catchment of 1st Pond. Between 14.6 and 12.7 cal ka BP Phylica frequencies decline sharply and only occur scattered. Above 12.7 ka, Phylica increases rapidly and is one of the major pollen taxa. Phylica growth around 1st Pond was probably limited by low temperatures before 15.6 cal ka BP and between 14.7 and 12.7 ka. Phylica is sensitive to frost and only grows in regions with mild winters without extended periods of frost. On Tristan da Cunha it has been shown that Phylica seeds germinate poorly in frost conditions, even if the trees survive (Milton et al., 1993). On Tristan da Cunha Phylica grows up to an altitude of about 800 m.a.sl., and on Gough Island up to about 500 m.a.sl., where it is limited by temperature. On Tristan da Cunha, mean winter temperature is 8.7
C close to sea level (26 m a.s.l.), and on Gough Island 6.7
C (54 m a.s.l.). With a lapse rate of 2
C/300 m, this corresponds to a mean winter temperature of 3.5
C (8.7-(2*774/300)) and 3.7
C (6.7(2*446/300)), respectively, for the highest elevation of Phylica growth. The single coldest month on Tristan da Cunha and Gough Island today is 6.5
C and 4.8
C, respectively, and including lapse rates this corresponds to 1.3
C (6.5e5.2) and 1.8
C (4.8e3.0) for the single coldest month at the upper limit of where Phylica grows today on the two islands. We use a mean value of 1.5
C in the following discussion. The calculated approximate temperatures limiting Phylica growth today can be used to estimate the temperature changes during the Late Glacial, assuming that the changes in the Phylica pollen frequencies are indeed temperature dependent. The almost complete absence of Phylica pollen 16.2 to 15.6 and 14.6 to 12.7 cal ka BP implies cold conditions with frost limiting tree growth. 1st Pond is situated at an altitude of 210 m.a.sl., and if we use the Tristan data this corresponds to a present day mean winter temperature of around 7.4
C and coldest month temperature of 5.2
C at 1P. We also need to take into account a sea-level lowering of around 100 m (0.7
C cooling) during the Late Glacial. Comparing today's situation (5.2
C) with the boundary conditions for Phylica during the period studied (0.7
C and 1.5
C) it translates to temperatures at least 3
C (5.2e0.7e1.5
C) lower than present for Phylica to be severely frost limited at 1st Pond. Since Phylica survived on the island the minimum winter month temperatures at sea level of that time cannot have been below 1.5
C for extended periods. The rapid increase in tree growth implied by the rapid increase in Phylica pollen frequencies after 12.7 cal ka BP, was probably the effect of a combination of rapidly increasing temperatures and increasing areas of dry peat land on the overgrown Pond colonized by Phylica. 3.10. Results of the climate simulations To provide a context for the analysis we first describe the general climate conditions in the subtropical region of the South Atlantic for the simulated climate of the 17 cal ka BP (17k) simulation. In this simulation the annual mean temperature in the SA is 2e3
C lower than in CTRL. During summer 17k has slightly more precipitation due to a northward displacement of the rain belt on the northern flank of the SHW (not shown). For the 17k_FW simulation increased freshwater flux to the North Atlantic weakens the AMOC leading to a general cooling of the North Atlantic region, which propagates throughout the hemisphere to cause widespread cooling of NH. Kageyama et al. (2009) provides a general analysis of North Atlantic freshwater perturbation experiments under glacial conditions using the IPSL_CM4 model. The results here are broadly equivalent except for our particular focus on the South Atlantic

K. Ljung et al. / Quaternary Science Reviews 123 (2015) 193e214

region during the period around 17 cal ka BP. In accordance with the results of Kageyama et al. (2009), as well as other modeling studies with a range of different models (e.g. Kageyama et al., 2013), we find a warming of the South Atlantic from the tropics to the marine polar front of about 1
C (see Fig. 11a). This is due to a reduced northward ocean heat transport in the Atlantic caused by the slow down of the AMOC. The South Atlantic warming, combined with prevailing dry glacial conditions, result in lowered relative humidity and increased evaporation, which explain the aridity of the entire region around Tristan da Cunha (Fig. 11). This can potentially explain the hiatus in 1st Pond as a drying-out effect. The Northern Hemisphere (NH) cooling and accompanying Southern Hemisphere (SH) warming due to decreased northward ocean heat transport results in intensification of the NH Hadley cell and weakening of the SH Hadley cell to compensate for the anomalous inter hemispheric temperature gradient and heat transport (e.g. Swingedouw et al., 2009; Kageyama et al., 2009). In connection with the weakened SH Hadley cell a weakened subtropical high pressure over the South Atlantic can also be seen in Fig. 11b. Additionally, the NH cooling and SH warming causes a shift of the thermal equator, which moves the ITCZ south, as seen in the shift in tropical precipitation in Fig. 11c. 4. Discussion 4.1. Interpretation of the local environmental-climatic history We can divide the development, in and around the two basins, into eight time periods (Fig. 12, Table 4). Before 16.2 cal ka BP: the 2.4 ka long hiatus. i.e. absence of sediments, in 1P, in combination with the thin sand-silt layer at 618.8 cm, indicates that the basin dried out some time after 18.6 cal ka BP (Fig. 2) and sediments may have also been oxidized when they were exposed to the atmosphere. Furthermore, it is probably not a coincidence that the sediment on the bottommost part of the helical auger in 2P, which could not be screwed further down (1317 cm), was dated to 17 cal ka BP (Bjørk et al., 2011). The stiffness of the sediment could indicate a dried-out horizon, but that 2P with its lower position in the catchment was earlier filled-in with water and sediments when ground-water began to rise as conditions became more humid. It should also be noted that it was

207

impossible to core further than 700 cm in the lowest situated basin, 3rd Pond, and that its bottom sediments have also been dated to ca 17 cal ka BP. We therefore think that today's “Pond area” was a dried-out system with exposed lake sediments, perhaps turning into stiff organic soils. The timing of the extreme part of this driedout phase is likely to have been ca 18e17 cal ka BP. Above the hiatus a period of sediment in-filling of 1P starts at 16.2 cal ka BP with a minimum in TOC, and thereafter rising values (Fig. 4). The falling d13C values and the rising BSi content (Fig. 5) show that a lake was developing as precipitation increased, supported by the diatom assemblages and concentrations (Fig. 7). Precipitation also brought terrestrial diatoms to the basin as well as (older) weathered material from the catchment (Fig. 8). The presence of the heavy Aulacoseira diatoms (Fig. 7) imply good oxygenation and mixing possibly due to windy conditions. Decreasing magnetic susceptibility (Fig. 2) and lower Zr/Ti ratios (Fig. 6) imply less surface run-off, possibly as a consequence of denser groundcovering vegetation (Figs. 9e10). Temperatures were, however, not high enough for the presence of Phylica trees e they were possibly approaching the site - but the steadily rising TOC and N values imply a gradual temperature rise. At around 15.7 cal ka BP the arrival of Phylica trees in the catchment of 1P implies that the mean temperature of the coldest month was 3.5
C at 1P, a temperature that today restricts Phylica growth on Tristan da Cunha, corresponding to 5.5
C at sea level of that time, and compared to the temperature of 6.5
C at today's sea level. The slow and gradual rise may be a sign that the trees were gradually established in the surroundings of 1P. The falling BSi values and diatom concentrations, together with the sudden dominance of Aulacoseira and Frustulia cf. rhomboides diatoms, the latter preferring humic-acidic water, indicate that 1P was exposed to windy conditions and began to be surrounded by a wetland. Occasional heavy rains brought in material from the catchment, cf. Zr/Ti ratios (Fig. 6) and magnetic susceptibility, in the beginning of the period. Between 15.3 and 14.95 cal ka BP conditions calmed down and a larger and calmer lake existed. This is shown by high diatom concentrations and the dominance of the S. venter species preferring larger water bodies (Van de Vijver et al., 2002). The previously fairly warm and wet period ended with a seemingly sudden event 14.9e14.75 cal ka BP, shown by an extreme susceptibility peak (Fig. 2), a TOC minimum, a C/N ratio peak, a d15N

Fig. 11. a) Simulated 2 m temperature (t2m) [
C] JJA anomalies of the freshwater forced experiment (17k_FW) compared to the 17k experiment, without freshwater hosing, b) same as a), but for sea level pressure (slp) [hPa] and 850 mb wind, c) same as a), but for evaporation minus precipitation (EeP) [mm/month]. The star indicates the position of Tristan da Cunha. In figure a) and c) the 2
C and 10
C SST isotherms are plotted as proxies for the marine polar front and southern flank of the marine subtropical front with full lines for CTRL (gray) and stippled lines for 17k (black) and 17k_FW (red). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

i( w t% Bi oS

C /N

al ka BP Ph yl ic a ar bo

C

) M (S ag I u ne ni tic ts s 10 u s -5 c ) ptib In ilit fe y rre d hu m id In ity fe rre (a d ) te m pe % ra Po tu la re rs s (b pe ) ci % es W ( c) ar m sp e ci S es (g tac (d ra ke ) in d s/ IR cm D /2y ea δ 18 r) O (e ) iso to pe (‰ )E N G DM R IP L (f) δ 18 O (‰ )( g)

K. Ljung et al. / Quaternary Science Reviews 123 (2015) 193e214

re a (g Au ra la in co s/ se cm ira /2y al St ea pi au r) ge ro na si ra (% ve ) nt T er (d ota (% ia l d ) to ia m to s/ m cm c TC )3 onc % en tra tio n

208

21000

11.5 12.0

12.5

13.0

13.5

14.0

14.5

15.0

15.5 16.0 0

2000 4000 0

32 64 0

40 80 0

2*1012 5 15

35 55

20

36 0

4.8 9.6

38

86 0

2

4 6 0 2 4 6 8 0 5 10 20 0 2.4 4.8 0.00 0.08 0.16 -48.5

-44.5

-43.6 -38.8 -34.0

Fig. 12. Summary diagram of some of the data presented and discussed from 1P, with red dashed lines delimiting the eight periods discussed in the text. While the Phylica curve represents winter temperatures at Tristan da Cunha (see text), the interpreted 1P and 2P records have qualitatively been estimated into (a) relative humidity during the period of study. Further to the right other relevant records are shown, such as percentage of (b) polar foraminiferal species, and (c) warm foraminiferal species in core TNO57-12 (Barker et al., 2009) in the South Atlantic, (d) a stacked IRD record from the Scotia Sea (Weber et al., 2014) southeast of Tierra del Fuego, (e) the 18O record from the EDML ice core (EPICA Community Members, 2006) in the Antarctic South Atlantic sector with the AICC2012 chronology (Veres et al., 2013), and (f) the 18O record from the NGRIP ice core in Greenland with the GICC05 time scale (Rasmussen et al., 2006; Andersen et al., 2006). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

peak, minima in BSi and diatom concentrations, including S. venter. We also note a peak in input of chemically weathered material (PC2 in Fig. 6), a peak in coarser mineral grains (PC1 in Fig. 6) and a small peak in Phylica pollen grains followed by its total disappearance.

We believe that a change in the climate regime took place with heavy precipitation, increased surface run-off with increased transport of coarser grains and Phylica pollen, and increased lake levels possibly promoting erosion of the surrounding soils with

Table 4 Climatic conditions at the Tristan da Cunha (TdC) archipelago (37
S), and its inferred geographic position in relation to some of the main circulation belts, during different time periods between 16.3 and 11 cal ka BP, and the present situation. MAT ¼ Mean annual temp; MDST ¼ Mean daily summer (NoveApr) temp; MDWT ¼ Mean daily winter (MayeOct) temp; MAP ¼ Mean annual precip; MSP ¼ Mean summer precip; MWP ¼ Mean winter precip; STF ¼ Subtropical Front; SSST ¼ Summer sea surface temp; MSZW ¼ Mean summer zonal wind strength; MWSW ¼ Mean winter zonal wind strength; SHW ¼ Southern Hemisphere westerlies; ITCZ ¼ Intertropical Convergence Zone; STHC ¼ Subtropical high pressure cell. Note that temperature and precipitation were measured 1960-87, with only data from 88% of the 336 months, and the station was situated at .23 m a.s.l. on Tristan da Cunha compared to the altitudes of 1P and 2P at 210 and 200 m, respectively. Temperatures are therefore >1
C colder at 1P and 2P and have higher precipitation due to the orographic effect. Time periods, cal. ka BP

Climatic conditions

Position of TdC in relation to main circulation belts and climatic zones

Today

MAT ¼ 14.3
C; MDST ¼ 16.1
C, MDWT ¼ 12.5
C, MAP ¼ 1514 mm; MSP ¼ 691 mm; MWP ¼ 823 mm; SSST ¼ 18
C; MSZW ¼ 10 m/s; MWZW ¼ 12 m/s Fairly dry but with several short-lasting wet events, warm

TdC is north of STF, and at the northern boundary of SHW, with a clear impact in austral winter Retractions and expansions of the SHW belt, TdC clearly north of STF Expansion of the SHW belt impacting TdC, TdC north of STF Rapid southward shifts of STF and SHW, both ending up at a position clearly south of TdC. Increased impact of STHC Shifting latitudes of the circulation belts, with STF slightly north of, or at TdC, and SHW at or south of TdC TdC south of STF, TdC at southern limit of SHW with only minor impact in winter STF moves to or slightly north of TdC, large impact of SHW TdC north of STF, expansion of SHW belt with varying influence on TdC, ITCZ and STHC moves north TdC clearly north of STF, no influence of SHW, impact from SHC, ITCZ displaced far south

11.6e10.0 11.7e11.6 12.7e11.7 14.1e12.7 14.75e14.1 14.9e14.75 16.2e14.9 Before 16.2

Increased wetness and warm Initially a short humidity peak followed by stable and gradually drier conditions, gradual warming with a rapid change at 12.1 culminating at the end of the period Slightly warmer with oscillating humidity with short-lasting drier and slightly warmer events Stable conditions with wet and cool climate, >3
C cooler in winter than today Humidity maximum, slightly cooler conditions High but variable humidity and windiness, slightly warmer with winter temps <3
C cooler than today Aridy and fairly warm conditions

K. Ljung et al. / Quaternary Science Reviews 123 (2015) 193e214

ultrafine magnetic particles. For a while this created a limnic environment not very suitable for aquatic productivity, as seen by e.g. the minima in BSi and diatom concentrations (Fig. 5), while the suddenly rising d15N values indicate soil erosion of areas with bird colonies, or that the new climate situation could have been favorable for an “invasion” of marine birds, possibly nesting around 1P, a situation similar to today. However, it is of course not impossible that this “14.9 ka event” only represents a short-lasting storm event on the island. The 14.75e14.1 cal ka BP period is characterized by a transition into a colder and progressively drier climate. Phylica pollen grains are absent or occur at very low numbers throughout the period (Figs. 9e10), which shows that climate conditions were too cold for tree growth at the site, and that winter temperatures were at least 3
C colder than today. The declining magnetic susceptibility and PC1 values indicate lower erosion and surface run-off, probably as a consequence of less precipitation. The deposition of chemically weathered material/clay minerals is high, which may seem as a contradiction to the lower deposition of minerogenic material indicated by the declining magnetic susceptibility and PC1 values (Figs. 2 and 6). This can be explained by more exposed soils in the catchment due to the harsher conditions, delivering clay minerals, but precipitation and surface run-off were too low for transporting coarser minerogenic material. The reduced deposition of minerogenic matter (declining magnetic susceptibility and PC1) is followed by decreasing frequencies of S. venter, which shows a gradual change into a smaller lake/pond; a lake level lowering that was probably the result of less precipitation. The next stage starts at 14.1 cal ka BP with a sudden dry-phase with possibly milder conditions. The high TC values and C/N ratios together with low magnetic susceptibility indicate lower lake level and more wet-land like conditions with increased peat growth. The disappearance of lake/pond diatom species (S. venter and Alucoseira), the very low BSi values and the disappearance of diatoms at 13.95 cal ka BP indicate that 1P was almost dried out, but probably with some restricted open water, with very limited diatom production, until it became totally dried out. Polypodiaceae increases rapidly during this period, probably as a consequence of ferns being established on dry peat surfaces around the pond. The absence of Phylica pollen indicates that temperatures were still too low for Phylica trees to establish on the peat lands around 1P, or on disturbed soils around the basin where it is today fairly common. The dry conditions during this period were probably caused by lower precipitation, possibly together with higher temperatures and higher evaporation. However, since Phylica trees were not established around the pond, the temperatures were lower than before 14.75 ka BP. At 13.9 cal ka BP we have evidence of increased surface run-off and catchment erosion (magnetic susceptibility) indicating increased precipitation. The re-appearance of the diatom S. venter in 1P (Fig. 7) also indicates a lake level rise, interrupted by a short dried out phase at 13.75 cal ka BP when diatoms disappear in 1P, culminating around 13.6 cal ka BP. Precipitation and catchment erosion were fairly variable throughout the period, also indicated by the variable carbon content, grain size/silicate material (PC1) and deposition of clay minerals (PC2) (Fig. 6). There is a small but steady increase in C/N, which indicates an increase in catchment vegetation and stabilization of soils. After 13.2 cal ka BP a large change in the diatom flora occurs when Pinnularia, Frustulia and Eunotia replace Stauroforma, which indicates a transition to a wetland pool with acidic waters. Around the same time Phylica pollen grains start to be present continuously, although in very low numbers, indicating expanding tree growth on the island as sea level was rising. Phylica trees were probably not established in the immediate surroundings of 1P, but may have formed denser stands at lower altitudes. In general, this 1.4 ka long period is characterized

209

by highly oscillating conditions with regard to humidity and possibly also temperature, seen in both the 1P and 2P records. This ended up with an initial fairly steep rise in Phylica percentages and concentrations (Figs. 9e10), implying that the last oscillation of this period was characterized by warmer conditions. The onset of the next period, 12.7e11.7 cal ka BP, is marked by the re-appearance of abundant Phylica pollen grains. Simultaneously in both records we note suddenly falling magnetic susceptibility with rising TOC values and C/N ratios. In 1P we also note increased input of weathered material (PC2), but less coarser grains (PC1). Put together these changes imply a fairly large temperature rise with less precipitation. Such a change would result in falling water levels in the two basins, and the transformation of the basins into wetlands, which is also documented by e.g. the changes in magnetic susceptibility, TOC and C/N and the falling d13C values in 1P. However, we also see a simultaneous peak in BSi as well as minor peaks in diatom concentrations at the two sites (Fig. 5), which could be interpreted as increased aquatic productivity. The diatom types that increase are, however, not of open water types but rather characteristic for small acidic pools and wetlands. This lends strong support for the fact that some parts of the diatom communities were favored by the warming, in spite of the fact that the basins were gradually becoming more grown-over. The period ends with a 300 yr long transitional phase, characterized by overgrowth when the basins became more or less fully covered by peat. This is witnessed by the low magnetic susceptibility, the low d13C and BSi values in 1P, the steadily rising TOC values and C/N ratios and the disappearance of diatoms in 2P (Fig. 8), followed by a slightly later disappearance in 1P (Fig. 7). The previously warm and rather dry climate possibly continued, and as a result of that the basins were more or less fully filled-in by peat. Between 11.7 and 11.6 cal ka BP the preceding dry conditions were replaced by increased precipitation. This is shown by a short peak in lake-type diatoms and their concentrations in 1P, minima in TOC values and C/N ratios at both sites and increased magnetic susceptibility. Altogether this points to a sudden and short hydrologic shift, possibly causing temporary flooding of the wetlands. From 11.6 to 10.0 cal ka BP both sites were characterized by subaerial/terrestrial conditions shown by e.g. the disappearance of diatoms, but also by the peat deposits themselves and their high TOC values and C/N ratios (Fig. 4) as well as the low magnetic susceptibility values. In the light of the previous study from 2P €rck, 2007), which reached back to 10.7 cal ka BP, it is (Ljung and Bjo clear that the very first part of the Holocene was characterized by what one would call a warm optimum. In 1P this is, e.g. shown by a maximum in Phylica pollen grains at 11.4 cal ka BP, implying dry and warm conditions making substantial tree growth on the bog possible. The further development does, however, also show changes in hydrology, with ample shifts in the C/N ratios in 2P with large excursions at 10.7, 10.3 and 10.15 cal ka BP (Fig. 4), indicating very dry bog conditions. In this context it is worthwhile noting that in spite of the gradual in-filling of the 2P basin the early Holocene bog conditions later changed into an environment with more or less open water conditions until it turned into today's quag-mire € rck, 2007). This shows that the isabout 1 ka ago (Ljung and Bjo land has undergone significant hydrologic changes during the Holocene, possibly triggered by shifts in the intensity and position of the SHW, and where the earliest Holocene stands out as the driest and warmest period. 4.2. Regional correlations, shifts of main circulation belts and bipolar seesaw effects On a small island like NI the climatic conditions are influenced by a complex interaction of atmospheric and marine processes.

210

K. Ljung et al. / Quaternary Science Reviews 123 (2015) 193e214

Additionally, the climatically sensitive position of the Tristan da Cunha archipelago means that even subtle shifts in the atmospheric and marine frontal systems can have large impact. To attain a better understanding of the processes behind climate shifts on NI it is therefore important to relate and correlate these island records to other records in the vast marine, but also terrestrial, region of the Southern Ocean (Figs. 12e13). The more than 2 cal ka long hiatus in 1P that underlies the sequence presented here is interpreted as being caused by a climate that dried out the basins, and coincides with the first part of the H1 in the NH. It has been shown that the trade winds changed considerably in Central America (Hodell et al., 2008; Escobar, 2012) during the first part of H1 and that the Afro-Asian monsoon region experienced a mega drought during H1 (Stager et al., 2011), most likely as a consequence of a southerly displacement of the ITCZ. Modeling studies (e.g. Menviel et al., 2008) also show that the ITCZ might have migrated southwards during Heinrich events due to a weaker AMOC and effects of the bipolar seesaw mechanism. This should have resulted in a slight southward displacement of a weaker subtropical high-pressure cell and the STF: this is confirmed by our simulations (Fig. 11), with warmer and more arid conditions at NI. Such a H1 scenario is consistent with southern, both terrestrial and marine paleorecords. For example, a rapid and early melt of the Cordillera Darwin ice sheet (55
S) in southernmost Chile has been postulated by Hall et al. (2013) and rising temperatures and melting glaciers in New Zealand (44
S) (Putnam et al., 2013), while marine records show onset of massive IRD deposition at 17 cal ka BP in the Scotia Sea (Weber et al., 2014) at 57e60
S, peaks in STF influence 17e16 cal ka BP at 37
S in the eastern Pacific (De Decker et al., 2012), while South Atlantic records show increased upwelling and ventilation throughout H1 at 44
S (Skinner et al., 2010), and warm waters reach 41
S already at 18.5 cal ka BP (Barker et al., 2009) with peak influence of warm water in the middle of H1. The latter core is from the closest situated site to NI and will therefore also be compared with in the following discussion. Our conclusion is that increased aridity of the subtropical high pressure cell reached the Tristan da Cunha archipelago during H1 in connection with the large-scale NHeSH teleconnection pattern, also resulting in a southerly displacement of the ITCZ. The water level of the lake fell and its floor and surface sediments were exposed to atmospheric conditions. It is also possible that the uppermost previously laid down sediments might have been oxidized; the youngest age of the sediments below the hiatus is ca 18.6 cal ka BP. The wetter and windier conditions after 16.2 cal ka BP show that, possibly as an effect of a slight equator-ward shift of the main circulation belts, or an expansion and strengthening of the SHW wind belt as the eddy-driven jet strengthened (Lee et al., 2011), the island was now more influenced by the SHW but still situated north of the STF, as also implied by Barker et al.'s (2009) record. A gradual warming on NI implies that the surrounding waters were gradually warming up, which is also consistent with the gradual increase of warm water species at 41
S (Barker et al., 2009) and a 1 ka long IRD minimum in the Scotia Sea (Weber et al., 2014). Climate on NI was, however, slightly variable up to ca 14.9 cal ka BP, possibly as a result of changing intensity of the SHW, and perhaps also of smaller latitudinal shifts of the STF. The precipitation maximum at 14.9e14.75 cal ka BP implies a frontal shift (Fig. 13) and coincides (within age uncertainties) with the largest IRD peak (Fig. 12) in the iceberg alley of the Scotia Sea (Weber et al., 2014). The preceding warm conditions might have led to substantial iceberg calving resulting in an armada of icebergs, transported by, and within, the Antarctic Circumpolar Current (ACC). The Atlantic sector of the SO would have been most influenced by the cold and freshwater from the melting icebergs, and

possibly caused a northwards shift of the SHW. Since this wet period at NI was followed by a cooling after 14.7 cal ka BP, which can also be seen in the TNO57-21 core at 41
S (Barker et al., 2009), we assume that the STF moved north, placing NI south of the STF. In most Antarctic ice cores this is also the timing of the onset of the Antarctic cold reversal (ACR), and in the SO the previously strong upwelling/ventilation weakened considerably (Skinner et al., 2010). The tight coupling between Antarctic temperatures and atmospheric CO2 (Pedro et al., 2012; Parrenin et al., 2013; Martinez-Boti et al., 2015) and the synchronous frontal shifts at 41
S (Barker et al., 2009) and at NI, strongly suggests that the circulation belts around Antarctica moved northwards as the Antarctic sea ice belt expanded as a consequence of the bipolar seesaw during the Bølling/G1-1e warming in the north. The frontal shifts resulted in cooler waters around NI, with less precipitation as the SHW were displaced further northwards. NI might have been situated at the southern limit of the SHW, and perhaps with minor SHW impact during winters when precipitation today reaches a maximum. From a climatic point of view the period 14.75e14.1 cal ka BP was characterized by fairly cool and stable conditions with decreased humidity as both the warm waters of the subtropics and the SHW had moved north. According to the AICC2012 time scale of Veres et al. (2013), the EDML ice core in the South Atlantic sector, closest to NI, shows a distinct 18O minimum at 14.1 kyr BP, implying that this could have been the culmination of the ACR in the Atlantic sector. The above described most stable period at NI ended up with a humidity minimum centered around 14.0 cal ka BP, coinciding with the Older Dryas stadial/GI-1d event (Rasmussen et al., 2006; Lowe et al., 2008) in the northern Hemisphere. In the EDML ice core an 18 O maximum is seen at 14.05 kyr BP, just after the above described minimum (Veres et al., 2013). If this is the expression of the bipolar seesaw mechanism the SHW ought to have moved south of NI to create dry enough conditions, implying a bipolar seesaw mechanism also for minor events (Fig. 13). The period 13.9e12.7 cal ka BP is characterized by more high frequency variability than during the earlier and later periods studied here, possibly as an effect of varying influence from the SHW. This oscillatory pattern is also seen in, e.g. the EDML and NGRIP 18O records, the varying percentages of polar and cold species in marine core TNO57-21 at 41
S (Barker et al., 2009), and in the IRD record of the Scotia Sea (Weber et al., 2014). From the North Atlantic region we know that this time period, formerly named the Allerød period, comprises the Greenland Interstadial 1 events GI-1c to GI-1a, but contains even more climatic oscillations, often found in European lake records (e.g. Lotter et al., 2012). These coolings are often related to a weakening of the strength of the AMOC (e.g. €rck et al., 1996) and at least 2e3 coolings occur in the NGRIP ice Bjo core between 13.8 and 13.2 kyr BP (Rasmussen et al., 2006). If the oscillations we see at NI and in the SO are bipolar responses to AMOC shifts, one would expect southerly shifts of the circulation belts and decreased precipitation with warming on NI as the SHW and STF moved south during cold events in the north. On NI we note an oscillating pattern, which can also be seen in the amount of marine polar species in core TNO57-21, but the chronology is not well-constrained enough to date the events in detail on NI. However, we note that at 13.2 cal ka BP when Phylica trees start to become continuously present around the site on NI, a distinct cooling occurs in the NGRIP ice core. The same tendency of a bipolar pattern is seen when compared with the EDML ice core (Veres et al., 2013), and it is noteworthy that these events are followed by IRD peaks in the Scotia Sea (Weber et al., 2014) indicating that southward shifts of the circulation belts increased iceberg calving in the Weddell Sea. This implies that the bipolar mechanism, and its related processes, was active also on a small scale in the Atlantic

K. Ljung et al. / Quaternary Science Reviews 123 (2015) 193e214

211

Fig. 13. Six paleoclimatic scenarios in the South Atlantic and surrounding land areas between 17 cal ka BP and 11.6e11.4 cal ka BP, showing the inferred positions of the Southern Hemisphere Westerlies (light blue), Subtropical Front (dashed), the Intertropical Convergence Zone (blue quadrats) and the subtropical high pressure cell (“H” surrounded by dashed line), based on our interpretations. The sites referred to in the text are also shown. The color coding of the sites indicate the relative climate conditions at each site. € rck et al. (2012), 6: Weber et al. References to the numbered sites in the 17 ka map: 1: Hodell et al. (2008), 2: Stager et al. (2011), 3: Barker et al. (2009), 4: Hall et al. (2013), 5: Bjo (2014), 7: Veres et al. (2013). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

212

K. Ljung et al. / Quaternary Science Reviews 123 (2015) 193e214

region and may have influenced latitudes up to at least 37
S, the latitude of the Tristan da Cunha archipelago. After 12.7 cal ka BP the oscillations of the previous period changed rather rapidly into a new and different climatic scenario with suddenly drier and warmer conditions; both the SHW and the STF must have moved south of NI. The only indication of increased influence from the westerlies as the core of the SHW passed by is a short period of increased precipitation at 12.7 cal ka BP in 1P with peaks in C and C/N ratio, rapid decline in d13C, followed by a BioSi peak, temporary maxima in diatom concentrations in 1P and 2P as well as a temporary maximum in terrestrial diatom taxa in 2P. From 12.7 cal ka BP and onwards we see a gradual warming with drier conditions at NI, including a gradual overgrowth of the two basins studied. We interpret this as the result of increased influence of the subtropic high pressure cell, with the STF and SHW displaced far south. This change coincides with the rather sudden decrease in cold and polar foraminifera species at 41
S (Barker et al., 2009), showing that the STF moved even further south, and with dry conditions in Tierra del Fuego indicating that the SHW could even € rck et al., 2012). The have been situated south of South America (Bjo warming coincides with the onset of the GS-1/Younger Dryas cooling in the north and results in rapid spread of Phylica trees on NI, the ponds/lakes turning into wetlands, the more or less disappearance of cold and polar species at 41
S and an IRD peak in the Scotia Sea (Weber et al., 2014) at 12 cal ka BP. It is very likely that this corresponds to a period when the circulation belts began to reach their southernmost positions since perhaps the penultimate deglaciation. The warming after 12.7 cal ka BP implies that the surface water of the SO warmed, but possibly fairly gradually. This is shown by the gradual increase in Phylica pollen, but also by the gradual decline of Barker et al.'s (2009) polar and cold species. However, at around 12.1 cal ka BP a further warming seems to have taken place, shown by the rapid increase of, e.g. Phylica pollen and TOC values. This is also seen in the EDML record (Veres et al., 2013), while the marine core TNO57-21 shows a 300 yr delay in peak warming, i.e. when cold/polar species more or less disappear, relative to NI and EDML. One explanation may be that the two latter records are more directly influenced by the atmosphere, with its considerably more rapid response, and their effects on climate archives, during changes of the circulation belts. We note that the IRD record from the Scotia Sea seems to have slightly shorter delay time, ca 200 yr, than Barker et al.'s (2009) foraminifera record. The NI record implies peak warming and evaporation sometime after 11.8 cal ka BP, which approximately coincides with the onset of the Holocene (Walker et al., 2009) and the rise into an 18O maximum in the EDML ice core (Veres et al., 2013). It is, however, interrupted by a very distinct humidity event starting slightly before 11.6 cal ka BP shown by a multitude of proxies at both sites. Interestingly a clear 18O minimum appears at 11.6e11.5 cal ka BP in the EDML ice core (Veres et al., 2013), which we think may be an indication that the circulation belts temporarily moved north. It is not likely that the STF moved north of NI, but an expansion of the SHW belt, in combination with the increased SSTs, may have resulted in a short phase of more precipitation on NI, as a bipolar seesaw response to the sudden onset of the Holocene warming in the north. Approximately a century of these humid conditions was followed by a drier climate, but the combined data from 1P and 2P indicate that the island experienced some further wetness shifts in the early Holocene. Such early Holocene short-term climatic coolings are also seen in the Greenland ice core record (Lowe et al., € rck et al., 1997) and 2008), such as the Preboreal Oscillation (Bjo € rck et al., 2001) and 9.7 the 10.7 (Rasmussen et al., 2006), 10.3 (Bjo events (Lowe et al., 2008). Warm oscillations at most of these time periods can also be seen in the EDML ice core (Veres et al., 2013),

but whether these are bipolar seesaw effects or not is difficult to conclude since they are more subtle than the older ones. In contrast to the recent study by WAIS Divide Project Members (2015) our record does not show an obvious time-lag between the supposed northern push for the bipolar see-saw and the response in the south. This can be explained by the fact that our site is situated far north of West Antarctica, since a more rapid reaction to the AMOC induced changes in the north would be expected at our subtropical latitudes. Our chronology is also limited by analytical precision in 14C measurements and uncertainties in the 14C calibration curve. It may, however, become possible in the future to better synchronize methane correlated ice cores with 14C dated records when the calibration curve has been extended and refined, which is necessary to detect any differences in time-lags of a century or two between ice cores and other proxy records. First then can these problems be resolved.

5. Conclusions The two studied sites at NI are believed to reflect large-scale circulation shifts in the South Atlantic, which, as shown in the multi-proxy records, caused variations in temperature, hydrology (through changes in precipitation-evaporation ratios), degree of weathering and wind strength. In Table 4 we summarize the local climatic conditions at the Tristan da Cunha archipelago during different time periods, including today's situation, and the possible large-scale drivers for the changes. We conclude that:  Dry conditions at NI before 16.2 cal ka BP were caused by arid conditions within a weakened high-pressure cell linked to a southerly displacement of the ITCZ and as a consequence of this the STF and SHW were situated rather far south of 37
S.  At 16.2 cal ka BP the STF and ITCZ move north, while the SHW expand north with increasing but oscillating humidity at NI. STF is still rather far south of 37
S and SSTs increase slightly. This variable influence from SHW lasts until 14.9 cal ka BP.  Between 14.9 and 14.75 cal ka BP the humidity peak with slightly cooler conditions at NI is interpreted as a consequence of increased influence from the SHW and the STF moved to or slightly north of the island.  Fairly stable and cool conditions 14.75e14.1 cal ka BP shows that the STF had a steady position north of 37
S. The island was within influence of the SHW, but possibly less in winter since precipitation was much less than during the previous period.  The slightly warmer period 14.1e12.7 cal ka BP began with a distinct arid period centered around 14.0 cal ka BP. It is characterized by oscillating humidity and is the least stable phase of our records. STF was just north of or at the Tristan da Cunha archipelago, while the SHW belt was situated over the islands with varying impact/strength.  The gradual onset of warming at 12.7 cal ka BP is slightly delayed in relation to the onset of the Younger Dryas/GS-1 stadial in the Northern Hemisphere (Lowe et al., 2008). This warm and dry period, lasting between 12.7 and 11.6 cal ka BP, was caused by a fairly large southward shift of both the STF and SHW, and increased influence from the subtropical high-pressure cell.  A short period of increased wetness, 11.6e11.5 cal ka BP, was most likely caused by an expansion of the SHW belt.  In the beginning of the Holocene, 11.5e10 cal ka BP, the warm conditions continued but with several phases of increased humidity. This shows that the STF was still a good distance from NI, but that the SHW expanded and retracted, perhaps as a consequence of AMOC variations in the north as the large continental ice sheets were melting.

K. Ljung et al. / Quaternary Science Reviews 123 (2015) 193e214

 Finally, based on the development described above, we can conclude that our records show that the bipolar seesaw was in action up to at least the Southern Hemisphere subtropical latitudes during the Last Termination. Acknowledgments The study was partly financed by a grant to SB by the Swedish Research Council (VR 621-2008-2894). It is also part of WP3 in us centers, from which it also LUCCI, one of the VR funded Linne received support. The Crafoord Foundation financed parts of the expedition costs. VR also financed much of the salary costs for SH, JS and NvdP and some of the salary for KL. For this generous support we are very grateful. We appreciate many good comments from € rck two conscientious but anonymous reviewers. Martin Bjo (Uppsala), Anders Anker Bjørk (Copenhagen), Anders Cronholm (Lund) and James Haile (Oxford) are thanked for taking part in the expedition and in the fieldwork. The helpfulness and kindness of the Tristan islanders were invaluable, and we especially thank James and Felicity Glass for all their support and Warren Glass and Donny Green for their help with the fieldwork on Nightingale Island. We also thank the ship cook Matthieu Grignon. Special thoughts go to Charlie Porter, the skipper of the ketch Ocean Tramp, who managed to safely sail us from Falkland Islands/Malvinas to Tristan da Cunha (13 days) and back to Punta del Este in Uruguay (33 days), and in all types of weather. He sadly died in February 2014 and we dedicate this paper to the memory of him: a true enthusiast in climate science, but also a true and very special friend, not only in troubled waters! Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.quascirev.2015.07.003. References Andersen, K.K., Svensson, A., Johnsen, S.J., Rasmussen, S.O., Bigler, M., Rothlisberger, R., Ruth, U., Siggaard-Andersen, M.L., Steffensen, J.P., DahlJensen, D., Vinther, B.M., Clausen, H.B., 2006. The greenland ice core chronology 2005, 15e42 ka. Part 1: constructing the time scale. Quat. Sci. Rev. 25, 3246e3257. Anderson, S.P., 2005. Glaciers show direct linkage between erosion rate and chemical weathering fluxes. Geomorphology 67, 147e157. Bard, E., Rickaby, R.E.M., 2009. Migration of the subtropical front as modulator of glacial climate. Nature 460, 380e384. Baker, P.E., Gass, I.G., Harris, P.G., Le Maitre, R.W., 1964. The volcanological report of the Royal Society Expedition to Tristan da Cunha, 1962. Philos. Trans. R. Soc. A Math. Phys. Eng. Sci. 256, 439e575. Barker, S., Diz, P., Vautravers, M.J., Pike, J., Knorr, G., Hall, I.R., Broecker, W.S., 2009. Interhemispheric Atlantic seesaw response during the last deglaciation. Nature 457, 1097e1102. Battarbee, R.W., Keen, M.J., 1982. The use of electronically counted microspheres in absolute diatom analysis. Limnol. Oceanogr 27, 184e188. Bennet, K.D., Gribnitz, K.-H., Kent, L.E., 1989. Pollen analysis of a quaternary peat sequence on Gough Island, South Atlantic. New Phytol. 113, 417e422. Berger, A.L., 1978. Long-term variations of caloric insolation resulting from the Earth's orbital elements. Quat. Res. 9, 139e167. Berglund, B.E., Ralska-Jasiewiczowa, M., 1986. Pollen analysis and pollen diagrams. In: Berglund, B.E. (Ed.), Handbook of Palaeoecology and Palaeohydrology. John Wiley and Sons, Chichester, pp. 455e484. € rck, S., Kromer, B., Johnsen, S., Bennike, O., Hammarlund, D., Lemdahl, G., Bjo Possnert, G., Rasmussen, T.L., Wohlfarth, B., Hammer, C.U., Spurk, M., 1996. Synchronized terrestrial-atmospheric deglacial records around the North Atlantic. Science 274, 1155e1160.  Funder, S., 1997. The Preboreal oscillation € rck, S., Rundgren, M., Ingo  lfsson, O., Bjo around the Nordic Seas: terrestrial and lacustrine responses. J. Quat. Sci. 12, 455e466. € rck, S., Muscheler, R., Kromer, B., Andresen, C.S., Heinemeier, J., Johnsen, S.J., Bjo Conley, D., Koç, N., Spurk, M., Veski, S., 2001. High-resolution analyses of an early Holocene climate event may imply decreased solar forcing as an important climate trigger. Geology 29, 1107e1110.

213

€rck, S., Rittenour, T., Rose n, P., França, Z., Mo € ller, P., Snowball, I., Wastegård, S., Bjo Bennike, O., Kromer, B., 2006. A Holocene lacustrine record in the central North Atlantic: proxies for volcanic activity, short-term NAO mode variability and long-term precipitation changes. Quat. Sci. Rev. 25, 9e32. €rck, S., Rundgren, M., Ljung, K., Unkel, U., Wallin, Å., 2012. Multi-proxy analyses Bjo of a peat bog on Isla de los Estados, easternmost Tierra del Fuego: a unique record of the variable Southern Hemisphere Westerlies since the last deglaciation. Quat. Sci. Rev. 42, 1e14. €rck, S., Cronholm, A., Haile, J., Ljung, K., Porter, C., 2011. Possible Late Bjørk, A.A., Bjo Pleistocene volcanic activity on Nightingale Island, South Atlantic Ocean, based on geoelectrical resistivity measurements, sediment corings and 14C dating. GFF 133, 141e147. Blunier, T., Brook, E.J., 2001. Timing of millennial-scale climate change in Antarctica and Greenland during the last glacial period. Science 291, 109e112. Bokhorst, S., et al., 2007. External nutrient inputs into terrestrial ecosystems of the Falkland Islands and the Maritime Antarctic region. Polar Biol. 30, 1315e1321. Bond, G., Kromer, B., Beer, J., Muscheler, R., Evans, M.N., Showers, W., Hoffman, S., Lotti-Bond, R., Hajdas, I., Bonani, G., 2001. Persistent solar influence on North Atlantic climate during the Holocene. Science 294, 2130e2136. Braconnot, P., Otto-Bliesner, B., Harrison, S., Joussaume, S., Peterchmitt, J.-Y., AbeOuchi, A., Crucifix, M., Driesschaert, E., Fichefet, T., Hewitt, C.D., Kageyama, M., Kitoh, A., Lan, A., Loutre, M.-F., Marti, O., Merkel, U., Ramstein, G., Valdes, P., Weber, S.L., Yu, Y., Zhao, Y., 2007. Results of PMIP2 coupled simulations of the Mid-Holocene and Last Glacial Maximum e part 1: experiments and large-scale features. Clim. Past 3, 261e277. Brady, N.C., 1990. The Nature and Properties of Soil. MacMillan, New York. Broecker, W.S., 1998. Paleocean circulation during the last deglaciation: a bipolar seesaw? Paleoceanography 13, 119e121. Bronk Ramsey, C., Lee, S., 2013. Recent and planned developments of the program OxCal. Radiocarbon 55, 720e730. Chiang, J.C.H., Bitz, C.M., 2005. Influence of high latitude ice cover on the marine intertropical convergence zone. Clim. Dyn. 25, 477e496. Cvijanovic, I., Chiang, J.C.H., 2013. Global energy budget changes to high latitude North Atlantic cooling and the tropical ITCZ response. Clim. Dyn. 40, 1435e1452. Conley, D.J., Schelske, C.L., 2001. Biogenic silica. In: Smol, J.P., Birks, H.J.B., Last, W.M. (Eds.), Tracking Environmental Change Using Lake Sediments, Terrestrial, Algal, and Siliceous Indicators, vol. 3. Kluwer Academic Publishers, Dordrecht, pp. 281e293. De Decker, P., Moros, M., Perner, K., Jansen, E., 2012. Influence of the tropics and southern westerlies on glacial interhemispheric asymmetry. Nat. Geosci. 5, 266e269. EPICA Community Members, 2006. One-to-one coupling of glacial climate variability in Greenland and Antarctica. Nature 444, 195e198. Erskine, P.D., et al., 1998. Subantarctic Macquarie Island e a model ecosystem for studying animal-derived nitrogen sources using 15N natural abundance. Oecologia 117, 187e193. Escobar, J., Hodell, D.A., Brenner, M., Curtis, J.H., Gilli, A., Mueller, A.D., rez, L., Schwalb, A., Anselmetti, F.S., Ariztegui, D., Grzesik, D.A., Pe Guilderson, T.P., 2012. A ~43-ka record of paleoenvironmental change in the Central American lowlands inferred from stable isotopes from lacustrine ostracodes. Quat. Sci. Rev. 37, 92e104. Grasshoff, K., Ehrhardt, M., Kremling, K. (Eds.), 1983. Methods of Sea Water Analysis. Verlag Chemie, Veinheim, p. 314. Hafsten, U., 1951. A Pollen Analytical Investigation of Two Peat Deposits from Tristan de Cunha. Research of Norwegian Scientific Expedition to Tristan de Cunha 1937-1938, No 22, p. 42. Hafsten, U., 1960. Pleistocene development of vegetation and climate in Tristan de Cunha and Gough Island. In: Årbok Univ. Bergen, Mat.-Naturv. Serie, vol. 20, p. 45. Hall, B.L., Porter, C.T., Denton, G.H., Lowell, T.V., Bromley, G.R.M., 2013. Extensive recession of Cordillera Darwin glaciers in southernmost South America during Heinrich stadial 1. Quat. Sci. Rev. 62, 49e55. Hodell, D.A., Anselmetti, F.S., Ariztegui, D., Brenner, M., Curtis, J.H., Gilli, A., Grzesik, D.A., Guilderson, T.J., Müller, A.D., Bush, M.B., Correa-Metrio, A., Escobar, J., Kutterolf, S., 2008. An 85-ka record of climate change in lowland Central America. Quat. Sci. Rev. 27, 1152e1165. Hogg, A.G., Hua, Q., Blackwell, P.G., Niu, M., Buck, C.E., Guilderson, T., Heaton, T.J., Palmer, P.J., Reimer, P.J., Reimer, R.W., Turney, C.S., Zimmerman, S., 2013. SHCal13 southern hemisphere calibration, 0e50,000 years cal BP. Radiocarbon 55, 1889e1903. € rck, S., 2013. Holocene environmental changes on Holmgren, S., Ljung, K., Bjo Nightingale Island, South Atlantic, based on diatom floristic changes in an infilled pond. Palaeogeogr. Palaeoclimatol. Palaeoecol. 378, 45e51. Juggins, S., 2003. User Guide. Software for Ecological and Palaeoecological Data Analysis and Visualization. University of Newcastle, Newcastle upon Tyne, UK, p. 69. Kageyama, M., Mignot, J., Swingedouw, D., Marzin, C., Alkama, R., Marti, O., 2009. Glacial climate sensitivity to different states of the Atlantic Meridional overturning circulation: results from the IPSL model. Clim. Past 5, 551e570. Kageyama, M., Merkel, U., Otto-Bliesner, B., Prange, M., Abe-Ouchi, A., Lohmann, G., Roche, D.M., Singarayer, J., Swingedouw, D., Zhang, X., 2013. Climatic impacts of fresh water hosing under Last Glacial Maximum conditions: a multi-model study. Clim. Past 9, 935e953. Kang, S.M., Frierson, D.M.W., Zhao, M., 2008. The response of the ITCZ to extratropical thermal forcing: idealized slab-ocean experiments with a GCM. J. Clim. 21, 3521e3532.

214

K. Ljung et al. / Quaternary Science Reviews 123 (2015) 193e214

Krammer, K., Lange-Bertalot, H., 1986-91. Süsswasserflora von Mitteleuropa Bacillariophyceae Teil 1-4. Gustav Fisher, Stuttgart, p. 2486. € wemark, L., 2013. Geochemical reKylander, M.E., Klaminder, J., Wohlfarth, B., Lo sponses to paleoclimatic changes in southern Sweden since the late glacial: the €sseldala Port lake sediment record. J. Paleolimnol. 50, 57e70. Ha Lange-Bertalot, H., 1995. Diatomeen der Anden/Diatoms of the Andes. In: Iconographia Diatomologia, vol. 9, p. 671. es monoraphide es des îles Kerguelen. Le Cohu, R., Maillard, R., 1983. Les diatome Ann. Limnol. 19, 143e167. es d’eau douce des îles Kerguelen ( Le Cohu, R., Maillard, R., 1986. Diatome a es). Ann. Limnol. 22, 99e118. l'exclusiondes Monoraphide Lee, S.-Y., Chiang, J.C.H., Matsumoto, K., Tokos, K.S., 2011. Southern Ocean wind response to North Atlantic cooling and the rise in atmospheric CO2: modeling perspective and paleoceanographic implications. Paleoceanography 26, PA1214,. http://dx.doi.org/10.1029/2010PA002004. €rck, S., Holmgren, S., Ljung, K., Van der Putten, N., Porter, C., 2011. Lindvall, H., Bjo A Holocene peat record in the central South Atlantic: an archive of precipitation changes. GFF 195e206. €rck, S., Hammarlund, D., Barnekow, L., 2006. Late Holocene multiLjung, K., Bjo proxy records of environmental change on the South Atlantic island Tristan da Cunha. Paleogeogr. Paleoclimatol. Paleoecol. 241, 539e560. € rck, S., 2007. Holocene climate and vegetation dynamics on NightinLjung, K., Bjo gale Island, South Atlantic e an apparent interglacial bipolar seesaw in action? Quat. Sci. Rev. 26, 3150e3166. €rck, S., Renssen, H., Hammarlund, D., 2008. South Atlantic Island record Ljung, K., Bjo reveals a South Atlantic response to the 8.2 kyr event. Clim. Past 4, 35e45. Lotter, A.F., Heiri, O., Brooks, S., Leeuwen, J.F.N., Eicher, U., Ammann, B., 2012. Rapid summer temperature changes during termination 1a: high-resolution multiproxy climate reconstructions from Gerzensee (Switzerland). Quat. Sci. Rev. 36, 103e113. € rck, S., Hoek, W.Z., Steffensen, J.P., Walker, M.J.C., Lowe, J.J., Rasmussen, S.O., Bjo Yu, Z., the INTIMATE group, 2008. Synchronisation of palaeoenvironmental events in the North Atlantic region during the last termination: a revised protocol recommended by the INTIMATE group. Quat. Sci. Rev. 28, 6e17. Marti, O., Braconnot, P., Bellier, J., Benshila, R., Bony, S., Brockmann, P., Cadule, P., Caubel, A., Denvil, S., Dufresne, J.-L., Fairhead, L., Filiberti, M.-A., Foujols, M.-A., Fichefet, T., Friedlingstein, P., Goosse, H., Grandpeix, J.-Y., Hourdin, F., Krinner, G., vy, C., Madec, G., Musat, I., de Noblet, N., Polcher, J., Talandier, C., 2006. The Le ^ le de New IPSL Climate System Model: IPSL-CM4, Tech. Rep. 26, IPSL, Note du Po lisation 84 iSSN 1288, 1619. Mode Marti, O., Braconnot, P., Dufresne, J.-L., Hourdin, F., Denvil, S., Friedlingstein, P., Swingedouw, D., Mignot, J., Goosse, H., Fichefet, T., Codron, F., Guilyardi, E., Bellier, J., Benshila, R., Bony, S., Brockmann, P., Cadule, P., Caubel, A., Fairhead, L., vy, C., Foujols, M.-A., Grandpeix, J.-Y., Hourdin, F., Kageyama, M., Krinner, G., Le Madec, G., Musat, I., de Noblet, N., Talandier, C., 2010. Key features of the IPSL ocean atmosphere model and its sensitivity to atmospheric resolution. Clim. Dyn. 34, 1e26. Martinez-Boti, M.A., Marino, D., Foster, G.L., Ziveri, P., Henehan, M.J., Rae, J.W.B., Mortyn, P.G., Vance, D., 2015. Boron isotope evidence for oceanic carbon dioxide leakage during the last deglaciation. Nature 518, 219e222. Menviel, L., Timmermann, A., Mouchet, A., Timm, O., 2008. Meridional reorganizations of marine and terrestrial productivity during Heinrich events. Paleoceanography 23, PA1203,. http://dx.doi.org/10.1029/2007PA001445. Meyers, P.A., Teranes, J.L., 2001. Sediment organic matter. In: Last, W.M., Smol, J.P. (Eds.), Tracking Environmental Changes Using Lake Sediments, Physical and Chemical Techniques, vol. II. Kluwer, Dordrecht, pp. 239e269. Milton, S.J., et al., 1993. Disturbance and demography of Phylica arborea (Rhamnaceae) on the Tristan-Gough group of islands. Bot. J. Linn. Soc. 111, 55e70. Moore, P.D., Webb, J.A., Collinson, M.E., 1991. Pollen Analysis. Blackwell Scientific, Oxford. Moser, G., Steindorf, A., Lange-Bertalot, H., 1995. Neuekaledonien. Diatomeenflora einer Tropeninsel. Revision der Collection Maillard und Untersuchung neuen Materials. Bibl. Diatomol. 32, 1e340. Nesbitt, H.W., Young, G.M., 1982. Early Proterozoic climate and plate motions inferred from major element chemistry of lutites. Nature 299, 715e717. Newnham, R.N., Vandergoes, M.J., Sikes, E., Carter, L., Wilmshurst, J.M., Lowe, D.J., McGlone, M.S., Sandiford, A., 2012. Does the bipolar seesaw extend to the terrestrial southern mid-latitudes? Quat. Sci. Rev. 36, 214e222. Pahnke, K., Sachs, J.P., 2006. Sea surface temperatures of southern midlatitudes 0e160 kyr B.P. Paleoceanography 21 (2), 001117. http://dx.doi.org/10.1029/ 2005PA001191. €hler, P., Raynaud, D., Paillard, D., Schwander, J., Parrenin, F., Masson-Delmotte, V., Ko Barbante, C., Landais, A., Wegner, A., Jouzel, J., 2013. Synchronous change of atmospheric CO2 and Antarctic temperature during the last deglacial warming. Science 339, 1060e1063. Pedro, J.B., Rasmussen, S.O., van Ommen, T.D., 2012. Tightened constraints on the time-lag between Antarctic temperature and CO2 during the last deglaciation. Clim. Past 8, 1213e1221.

Putnam, A.E., Denton, G.H., Schaefer, J.M., Barrell, D.J.A., Andersen, B.G., Finkel, R.C., Schwartz, R., Doughty, A.M., Kaplan, M.R., Schluchter, C., 2010. Glacier advance in southern middle-latitudes during the Antarctic Cold Reversal. Nat. Geosci. 3, 700e704. Putnam, A.E., Schaefer, J.M., Denton, G.H., Barrell, D.J.A., Andersen, B.G., Koffman, T.N.B., Rowan, A.V., Finkel, R.C., Rood, D.H., Schwartz, R., Vandergoes, M.J., Plummerk, M.A., Brocklehurst, S.H., Kelleym, S.E., Ladig, K.L., 2013. Warming and glacier recession in the Rakaia valley, Southern Alps of New Zealand, during Heinrich Stadial 1. Earth Planet. Sci. Lett. 382, 98e110. Rasmussen, S.O., Andersen, K.K., Svensson, A.M., Steffensen, J.P., Vinther, B.M., Clausen, H.B., Siggaard, M.-L., Johnsen, S.J., Larsen, L.B., Dahl-Jensen, D., € thlisberger, R., Fischer, H., Goto-Azuma, K., Hansson, M.E., Ruth, U., Bigler, M., Ro 2006. A new ice core chronology for the last glacial termination. J. Geophys. Res. 111, D06102. http://dx.doi.org/10.1029/2005JD006079. Reimer, P.J., Bard, E., Bayliss, A., Beck, J.W., Blackwell, P.G., Bronk Ramsey, C., Buck, C.E., Cheng, H., Edwards, R.L., Friedrich, M., Grootes, P.M., Guilderson, T.P., , C., Heaton, T.J., Hoffmann, D.L., Hogg, A.G., Haflidason, H., Hajdas, I., Hatte Hughen, K.A., Kaiser, K.F., Kromer, B., Manning, S.W., Niu, M., Reimer, R.W., Richards, D.A., Scott, E.M., Southon, J.R., Staff, R.A., Turney, C.S.M., van der Plicht, J., 2013. IntCal13 and Marine13 radiocarbon age calibration curves 0e50,000 years cal BP. Radiocarbon 55, 1869e1887. Renssen, H., Goosse, H., Fichefet, T., Camp, J.-M., 2001. The 8.2 kyr event simulated by a global atmosphere-sea-ice-ocean model. Geophys. Res. Lett. 28, 1567e1570. Saunders, K.M., Hodgson, D.A., McMinn, A., 2009. Quantitative relationships betweenbenthic diatom assemblages and water chemistry in Macquarie Island lakes and their potential for reconstructing past environmental changes. Antarct. Sci. 2135e2149. Skinner, L.C., Fallon, S., Waelbrook, C., Michel, E., Barker, S., 2010. Ventilation of the deep Southern Ocean and deglacial CO2 rise. Science 328, 1147e1151. Stager, J.C., Ryves, D.B., Chase, B.M., Pausata, F.S.R., 2011. Catastrophic drought in the Afro-Asian monsoon region during Heinrich event 1. Science 331, 1299e1302. Swingedouw, D., Mignot, J., Braconnot, P., Mosquet, E., Kageyama, M., Alkama, R., 2009. Impact of freshwater release in the North Atlantic under different climate conditions in an OAGCM. J. Clim. 22, 6377e6403. Taboada, T., Cortizas, A.M., Garcia, C., Garcia-Rodeja, E., 2006. Particle-size fractionation of titanium and zirconium during weathering and pedogenesis of granitic rocks in NW Spain. Geoderma 131, 218e236. Taylor, S.R., McLennan, S.M., 1985. The Continental Crust: Its Composition and Evolution. Blackwell Scientific Publications, USA, p. 328. Turney, C.S.M., Roberts, R.G., de Jonge, N., Prior, C., Wilmshurst, J.M., McGlone, M.S., Cooper, J., 2007. Redating the advance of the New Zealand Franz Josef glacier during the last termination: evidence for asynchronous climate change. J. Quat. Sci. 26, 3037e3042. Van de Vijver, B., Frenot, Y., Beyens, L., 2002. Freshwater diatoms from Ile de la Possession (Crozet Archipelago, Subantarctica). Bibl. Diatomol. 46, 1e412.  Mahamadou Kele, H., Lemieux-Dudon, B., Veres, D., Bazin, L., Landais, A., Toye Parrenin, F., Martinerie, P., Blayo, E., Blunier, T., Capron, E., Chappellaz, J., Rasmussen, S.O., Severi, M., Svensson, A., Vinther, B., Wolff, E.W., 2013. The Antarctic ice core chronology (AICC2012): an optimized multi-parameter and multi-site dating approach for the last 120 thousand years. Clim. Past Discuss. 8, 6011e6049. Wace, N.M., Dickson, J.H., 1965. The terrestrial botany of the Tristan de Cunha islands. Philos. Trans. R. Soc. Lond. B 249, 273e360. WAIS Divide Project Members, 2015. Precise interpolar phasing of abrupt climate change during the last ice age. Nature 520, 661e665. Walker, M., Johnsen, S., Olander Rasmussen, S., Popp, T., Steffensen, J.-P., Gibbard, P., €rck, S., Cwynar, L.C., Hughen, K., Kershaw, P., Hoek, W., Lowe, J., Andrews, J., Bjo Kromer, B., Litt, T., Lowe, D.E., Nakagawa, T., Newnham, R., Schwander, J., 2009. Formal definition and dating of the GSSP (global stratotype section and point) for the base of the Holocene using the Greenland NGRIP ice core, and selected auxiliary records. J. Quat. Sci. 24, 3e17. Weber, M.E., Clark, P.U., Kuhn, G., Timmermann, A., Sprenk, D., Gladstone, R., Zhang, X., Lohmann, G., Menviel, L., Chikamoto, M.O., Friedrich, T., Ohlwein, C., 2014. Millennial-scale variability in Antarctic ice-sheet discharge during the last deglaciation. Nature 510, 134e138. West, A.J., Galy, A., Bickle, M., 2005. Tectonic and climatic controls on silicate weathering. Earth Planet. Sci. Lett. 235, 211e228. White, A.F., Blum, A.E., 1995. Effects of climate on chemical weathering in watersheds. Geochim. Cosmoschim. Acta 59, 1729e1747. Wolfe, A.P., 1997. On diatom concentrations in lake sediments: results from an interlaboratory comparison and other tests performed on a uniform sample. J. Paleolimnol. 3, 261e268. Xiao, S., Liu, W., Li, A., Yang, S., Lia, Z., 2010. Pervasive autocorrelation of the chemical index of alteration in sedimentary profiles and its palaeoenvironmental implications. Sedimentology 57, 670e676.