The early rise and late demise of New Zealand’slast glacial maximumHenrik Rothera,1, David Finkb, James Shulmeisterc, Charles Mifsudb, Michael Evansd, and Jeremy Pughe,2

aInstitute for Geography and Geology, University of Greifswald, 17489 Greifswald, Germany; bAustralian Nuclear Science and Technology Organisation,Menai, NSW 2234, Australia; cSchool of Geography, Planning and Environmental Management, The University of Queensland, St. Lucia Campus, Brisbane,QLD 4072, Australia; dMoultrie Geology, Banyo, QLD 4014, Australia; and eDepartment of Geological Sciences, University of Canterbury, Christchurch,New Zealand

Edited by Mark H. Thiemens, University of California, San Diego, La Jolla, CA, and approved June 13, 2014 (received for review February 3, 2014)

Recent debate on records of southern midlatitude glaciation hasfocused on reconstructing glacier dynamics during the last glacialtermination, with different results supporting both in-phase andout-of-phase correlations with Northern Hemisphere glacial sig-nals. A continuing major weakness in this debate is the lack ofrobust data, particularly from the early and maximum phase ofsouthern midlatitude glaciation (∼30–20 ka), to verify the compet-ing models. Here we present a suite of 58 cosmogenic exposureages from 17 last-glacial ice limits in the Rangitata Valley of NewZealand, capturing an extensive record of glacial oscillations be-tween 28–16 ka. The sequence shows that the local last glacialmaximum in this region occurred shortly before 28 ka, followedby several successively less extensive ice readvances between26–19 ka. The onset of Termination 1 and the ensuing glacial retreatis preserved in exceptional detail through numerous recessionalmoraines, indicating that ice retreat between 19–16 ka was verygradual. Extensive valley glaciers survived in the Rangitata catch-ment until at least 15.8 ka. These findings preclude the previouslyinferred rapid climate-driven ice retreat in the Southern Alps afterthe onset of Termination 1. Our record documents an early lastglacial maximum, an overall trend of diminishing ice volume inNew Zealand between 28–20 ka, and gradual deglaciation untilat least 15 ka.

surface exposure dating | global climate linkages

According to the Milankovitch orbital theory of glaciation,variations in northern high-latitude summer insolation are

responsible for glacial–interglacial cycles (1, 2). On this basis, itis commonly assumed that climatic changes in the NorthernHemisphere (NH) constitute the principle forcing mechanismfor glaciation in the Southern Hemisphere (SH) (3). However, inrecent times, this view has been challenged by studies showingthat at least some glacial signals in the SH have no NH correl-ative or precede events in the NH by up to 1.5 ka (4–7). Al-though some of these patterns can be explained by an extendedbipolar seesaw model, other features, including the decreasingexpression of hemispheric antiphasing away from the polarregions, indicate that additional forcing mechanisms must beinvolved (8).A compilation of glacial records from the NH indicates that

midlatitude ice sheets in North America and northern Eurasiabegan a major expansion phase between 33–29 ka (9). From thistime onward, relative sea level reconstructions indicate a steadyincrease in global ice volume, until most ice sheets had reacheda near-maximum extent by 26.5 ka, followed by 6–7 ky of equi-librium conditions until the onset of final retreat around 20–19ka (9). For the midlatitude SH, specifically New Zealand (NZ),paleoecologic data (10), supported by dated glaciofluvial aggra-dation sequences (11), have also suggested an onset of mountainglacier growth around 30–28 ka (12). Contrary to NH ice volumetrends, however, recent records from NZ may indicate thatmaximum glaciation in the Southern Alps occurred before 28 ka,several thousand years before the NH ice maximum (13–15). A

second aspect in the debate centers on determining the preciseonset of last glacial maximum (LGM) retreat (a period referredto as Termination 1 and dated to ∼19–10 ka) and the mode ofthis deglaciation (i.e., models of collapse versus slow recession)in NZ. Together, these issues are responsible for ongoing con-troversy about the evolution of the last glacial-to-interglacialtransition in the SH midlatitudes and the relative importance ofinterhemispheric climate forcing versus regional insolation and/or synoptic forcing of SH glaciation.

Previous WorkNZ represents one of only three areas in the SH midlatitudesthat experienced extensive late Quaternary glaciation. Duringthe LGM, valley glaciers 70–100 km in length reached the alpineforelands of the eastern Southern Alps, whereas shorter andsteeper glaciers in the western Alps typically terminated into theTasman Sea. According to cosmogenic exposure ages from LakePukaki terminal moraines in the central Southern Alps, it hasbeen suggested that large-scale glacial retreat in NZ commencedat 17.4 ka, representing “the start of the termination of theLGM” (16). This appeared to be virtually indistinguishable fromthe onset of the final LGM retreat at similar NH and SH mid-latitude sites, supporting the idea of a global mechanism re-sponsible for triggering a “near-synchronous” ice retreat at thesefar-field sites. Conversely, evidence for millennial-scale climatereversals in NZ at the end of Termination 1, during the AntarcticCold Reversal (14–15 ka) and the Younger Dryas chronozone(11–12 ka), has been interpreted to indicate late-glacial climaticantiphasing with NH events (17, 18).

Significance

We present here a comprehensive record of glaciation froma New Zealand valley glacier system covering the critical15,000-y period from the local last glacial maximum (LGM) tonear the end of the last ice age. This record from a key site inthe midlatitude Southern Hemisphere shows that the largestglacial advance did not coincide with the coldest temperaturesduring this phase. We also show that the regional post-LGM iceretreat was very gradual, contrary to the rapid ice collapsewidely inferred. This demonstrates that glacial records fromNew Zealand are neither synchronous with nor simply lag orlead Northern Hemisphere ice sheet records, which has im-portant implications for the reconstruction of past inter-hemispheric climate linkages and mechanisms.

Author contributions: H.R. and J.S. designed research; H.R., M.E., and J.P. performed re-search; D.F. and C.M. contributed new reagents/analytic tools; H.R. and D.F. analyzeddata; and H.R., D.F., and J.S. wrote the paper.

The authors declare no conflict of interest.1To whom correspondence should be addressed. Email: henrik.rother@gmail.com.2Deceased February 2012.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1401547111/-/DCSupplemental.

11630–11635 | PNAS | August 12, 2014 | vol. 111 | no. 32 www.pnas.org/cgi/doi/10.1073/pnas.1401547111

Dow

nloa

ded

by g

uest

on

Apr

il 10

, 202

0

Compounding these apparently incompatible interpretationsof the NZ glacial history using cosmogenic exposure dating is theintroduction of a local 10Be reference production rate for the NZregion (19). This new production rate is significantly lower (by16%) than the statistically weighted mean global production rateobtained from a collection of geographic calibration data sets(20). As a consequence, any intercomparison of glacial chro-nologies, both relative and absolute, across the globe, as well asfrom within NZ itself, requires great care to ensure a consistentsynthesis of the published data. Specifically, exposure ages fromNZ calculated using the global production rate (20) are nowconsidered to underestimate landform ages by anywhere from1.8–4.5 ky across the period of interest (i.e., 15–30 ka). Thisdraws into question the previously implied synchronicity of theonset of the final LGM retreat in NZ and NH midlatitude sites(16) as the rescaled NZ timing for “glacial collapse” increases(by 2.8 ky) to 20.2 ka. Recent studies by Putnam and colleagues(15, 21) report that at Lake Ohau and the Rakaia Valley in thecentral Southern Alps, fast retreat from LGM moraines com-menced around 17.8 ka, followed by a substantial reduction inoverall glacier length (−50%) over a period of ∼1,000 y. Asimilar rapid post-LGM ice decay has also been proposed foroverall glaciation across NZ, with an estimated 60% reduction inglaciated area soon after the onset of the final retreat (12).However, a different view is provided by Shulmeister and col-leagues (14), who reconstruct a much lower rate of ice retreatduring the same period (10–20% glacier reduction) on the basisof exposure dating of terminal and recessional moraines in theRakaia Valley (renormalized using the lower local 10Be pro-duction rate; see ref. 19).Despite significant progress in reconstructing NZ glacial dy-

namics, a common deficiency in the available glacial chronolo-gies is the relatively short-term nature of the investigatedmoraine sequences. Missing are moraine records spanning thewhole sequence from the ice maximum to the subsequent retreatstages pertaining to the whole LGM cold period of 30–15 ka. InNZ, such records are rare: in the western Alps because most majorLGM glacier systems were marine terminated, and in the easternAlps because many cross-valley terminal and retreat moraineswere buried during ice retreat in deep proglacial troughs ordestroyed by extensive postglacial fluvial processes. Here wepresent an extended glacial reconstruction from the RangitataValley, a major drainage system of the central Southern Alps inwhich an extensive moraine record survived in a stranded posi-tion on the interfluves between the entrenched Rangitata andRakaia trunk valleys. This allows the assessment of glacier dy-namics within a single catchment from the local LGM, throughTermination 1, and into the last glacial–interglacial transition.

Study Site and MethodsThe targeted moraines are located in the Lake Clearwater and Lake Heronareas of the Rangitata Valley, ∼50 km downstream from the modern glaciermargins. At its maximum extent, the Rangitata Glacier comprised two majorlobes: a southern ice tongue that followed the trunk Rangitata Valley(Rangitata lobe), and a northern lobe that advanced onto a higher-lyingarea north of Harper Range (Clearwater lobe). A smaller distributary lobefrom the main Rakaia Glacier entered the Lake Heron area from the northvia the Prospect Hill saddle (Heron lobe; Fig. 1). Ice marginal landformsconsist primarily of lobate laterofrontal dump moraines, lateral moraines,and hummocky terrain that formed during the disintegration of the icelobes. These landforms are present over a considerable down-valley distanceand altitudinal range and have formed the basis for an earlier geomorphicdifferentiation of five late Pleistocene glacial landform associations (22).Until now, no absolute ages have been available to constrain the timings ofthese glacial advances in the area.

Covering a valley distance of ∼12 km in-board from terminal ice positionsof the Heron and Clearwater lobes, we sampled 56 quartzo-feldspathicsandstone (greywacke) erratics from 17 moraine sites and two ice overriddenbedrock surfaces (SI Appendix, Fig. S3). Five samples of this total wererejected as outliers. We calculate exposure ages on the basis of a sea-level

high-latitude spallation 10Be production rate for NZ of 3.87 ± 0.08 atoms/gquartz/a (19) and a 26Al production rate of 26.08 ± 0.55 atoms/gquartz/acorresponding to a 26Al/10Be production ratio of 6.727 (20). All exposureages were calculated using the CRONUS Web-based calculator. Both spall-ation and muon production rates were then scaled to geographic locationand altitude of the individual sample sites, using the scaling methods ofT. Dunai (ref. 23; for age results based on other scaling schemes, seeSI Appendix).

ResultsOf the 73 zero-erosion ages (58 10Be and 15 26Al ages; Fig. 1 andSI Appendix) presented in this study, 48 are from the Clearwaterand 25 from the Heron area. Apart from two sites (site 10, site11 ∼600 m above the valley floor, all in-valley terminal and lateralmoraine weighted average ages fall between 27.7 and 15.8 ka,providing one of the most extensive glacial sequences from NZto track ice extent from the local LGM until well after the onsetof the last termination. We bracket this sequence into four dis-tinct periods consistent with an early local LGM ice extent, aninitial phase of ice volume reduction followed by stabilization(or readvance) of the ice front at 25 ka, further recessionand renewed ice front stabilization (or readvance) at 19 ka, anda gradual final retreat phase from 19 to 15 ka (Fig. 2 and SI Ap-pendix, Fig. S4).Two ice-overridden hills in the Heron Basin yield mean site

ages of 27.7 ± 0.5 ka (site 12) and 27.6 ± 0.8 ka (site 13). On thebasis of the required ice thickness to overflow these sites (site-13surface ∼230 m above valley floor), we infer that during itsmaximum extent, the Heron ice lobe coalesced with the glaciertongue occupying the Clearwater area. The cumulative weightedmean age for site 12, and site 13, representing the minimum localLGM age, is 27.7 ± 0.4 ka (n = 6, SI Appendix, Fig. S4A). Afterthe initiation of ice retreat from this local LGM ice position, thetwo lobes separated again, with an independent terminal mo-raine position established in the Heron area soon thereafter at27.6 ± 0.8 ka (site 14). Evidence for a comparable down-valleyice extent is also indicated by an elevated lateral moraine NW ofLake Heron, giving an age of 25.0 ± 0.6 ka (site 17; SI Appendix,Fig. S4A), suggesting only minor changes in ice volume duringthis time. Interestingly, the next up-valley moraines (site 15, site16), located no further than 3–4 km from the ∼27.6 ka ice po-sition (site 14), yield significantly younger site ages of 19.0 ± 0.4 ka(site 15) and 18.6 ± 0.3 ka (site 16).Ice recession from these younger moraine positions and

through Termination 1 is preserved on the Clearwater interfluve,where nine moraine sites were investigated. Two Clearwatermoraines (site 1, 18.7 ± 0.7 ka, and site 8, 18.5 ± 0.4 ka) and thetwo Heron moraines at site 15 (19.0 ± 0.4 ka) and site 16 (18.6 ±0.3 ka) clearly show that a stable ice front was established by∼19 ka in both areas (site 1, site 8, site 15, and site 16; weighted meanage, 18.7 ± 0.2 ka; n = 17). During the next 3–4 ky, ice recessionfrom these positions was gradual, with the ice front moving atmost 8–10 km up-valley, leaving several nested clusters of closelyspaced moraines. The largest of these recessional moraineclusters (site 2, site 3, and site 4) is located southeast of SpiderLakes, yielding overlapping ages of 17.0 ± 0.6 ka (site 2), 16.6 ±0.5 ka (site 3), and 17.9 ± 0.6 ka (site 4), with a weighted meanage of this moraine group of 17.1 ka ± 0.3 (n = 10, one reject; SIAppendix, Fig. S4C). Coeval retreat of the main Rangitata Valleyglacier during this time is observed at site 9 (17.2 ± 0.4 ka). Thefinal stages of retreat from the Clearwater area are preservedthrough three moraines ∼8 km upstream (site 6, 16.8 ± 0.4 ka;site 7, 15.8 ± 0.4 ka; site 5, 15.8 ± 0.3 ka), with a weighted meanage of 16.1 ± 0.2 (n = 11, one reject; SI Appendix, Fig. S4D).Collectively, the moraines show a broad up-valley decrease in

age consistent with a retreating ice margin, which was punctu-ated by repeated ice still stands. Ice thickness reduction withinthe ∼12 ky of retreat from the local LGM limit to our youngest

Rother et al. PNAS | August 12, 2014 | vol. 111 | no. 32 | 11631

EART

H,A

TMOSP

HER

IC,

ANDPL

ANET

ARY

SCIENCE

S

Dow

nloa

ded

by g

uest

on

Apr

il 10

, 202

0

moraine at about 16 ka amounts to 400–500 m. The oldest ageswere obtained from six glacial boulders from two sites in theDogs Hill Range (∼600 m above valley floor; Fig. 1C), yieldingindividual boulder ages of 113.2 ± 4.0, 104.5 ± 3.5, and 84.6 ± 2.5 ka(site 11) from the higher surface and 54.6 ± 3.4, 48.6 ± 1.9,and 31.4 ± 1.1 ka (site 10) from the lower of these surfaces.Despite their broad age range, these erratics clearly relate toprevious, larger-scale glacier advances in the central SouthernAlps. Removing the youngest boulder age from each site, theirweighted mean site ages (site 11, 108.3 ± 2.4 ka; site 10, 50.1 ±1.7 ka) fall into the last glacial cycle. A more robust interpretation

of the ages requires a reliable measure of long-term surfaceweathering of greywacke rock. An upper limit estimate of grey-wacke erosion rate of 1–2 mm/ky would yield site-11 erosion-corrected exposure ages between 141–114 ka and site-10 agesbetween 53–32 ka, which are effectively coeval with glaciationduring late marine isotope stage (MIS) 6/MIS-5 and MIS-3,respectively. We report erosion-uncorrected ages but note thatin some cases, thin quartz veins were found to be ∼5 mm proud ofthe sampled surface (SI Appendix, Fig. S3D), giving a minimumerosion rate of ∼0.3 mm/ky. This is almost identical to the amountof postemplacement erosion on greywacke surfaces reported from

Fig. 1. Location of the Rangitata catchment in the South Island, NZ (A), study areas (B), and positions of all sampled moraines and resulting cosmogenicexposure ages (C and D). The shaded relief map (B) shows the topographic context and the ice outlet routes of the major Rangitata and Rakaia glaciers. Insetrectangles denote regions of sampled moraines, with their relative locations detailed for the Lake Clearwater (C) and Lake Heron (D) interfluves. Multipleterminal and lateral moraine sequences span the period from maximum ice extent during the LGM (29–28 ka) to subsequent retreat phases in the Clearwaterand Heron areas (19–15 ka). Sites selected for 10Be and 26Al exposure dating are indicated by numbered black circles and comprise from two to five bouldersper site. All ages shown are derived from in situ 10Be with associated analytical errors (for 26Al ages, see SI Appendix). Bold ages represent site-weighted meanages with their standard mean error (ages using scaling schemes calculated via the CRONUS on-line web-calculator are given in SI Appendix, Tables S4 and S5).Five sample ages denoted with an asterisk are interpreted to be outliers (two of these relate to prelocal LGM sites). Sample ages denoted with a b werederived from bedrock.

11632 | www.pnas.org/cgi/doi/10.1073/pnas.1401547111 Rother et al.

Dow

nloa

ded

by g

uest

on

Apr

il 10

, 202

0

NZ by Kaplan and colleagues (17). Assuming a much higher ero-sion rate of 1–2 mm/ky, the ages would increase by 2–4% (i.e., ca.300 y at 15-ka and 1 ky at 30-ka exposure), showing that withinthese limits, the uncertainties resulting from potential surfaceerosion would have no effect on our main conclusions.

Discussion and ConclusionOur detailed chronology of the Rangitata glacial system es-tablishes four distinct phases of the LGM and last deglaciationin NZ (Fig. 2): a maximum local LGM ice extent reachedshortly before 28 ka (phase A), an initial phase of ice volumereduction followed by stabilization or readvance of the ice until25 ka (phase B), further ice volume reduction after 25 ka andrenewed stabilization of the ice at 18.7 ka at moraine positionsonly a few kilometers (∼3–4 km) distant from the local LGMlimit (phase C), and a gradual and final glacier retreat com-mencing shortly after 18.7 ka (and lasting to about 15.8 ka),producing numerous closely spaced recessional moraines rep-resenting the regional onset of Termination 1 (phase D). Co-eval with this last phase, Antarctic ice core records indicatecessation of severe cold conditions that had persisted for theprevious 10 ky and commencement at 18–19 ka of continuouswarming proxied via increasing δ18O and a synchronous rise inatmospheric CO2 (24–26).The exposure age results show that recession of the Rangitata

Glacier from the innermost LGM moraine (site 1) commencedat about 18.7 ± 0.7 ka, which is similar, although slightly earlier,than the 17.8 ± 0.2 ka onset of major LGM retreat foundby Putnam and colleagues (21) in the nearby Rakaia Valley.There, dating of recessional moraines points to extensive localdeglaciation between 17.8–15.6 ka, interpreted to coincide with

Heinrich Stadial 1 in the North Atlantic (21). Considering thetime evolution of the presented Clearwater–Heron glacial se-quence, we also document recession during this time, but sig-nificantly, we find no evidence for a rapid pace of deglaciation inNZ (12). In contrast, we see that retreat of the Rangitata Glacierbetween 18.7–15.8 ka was very gradual, with ∼70–80% of thelocal LGM glacier length surviving until 15.8 ka. This outcomediffers markedly from other NZ deglaciation studies, specificallyfrom Lake Pukaki (16), Lake Ohau (15), and the Rakaia Valley(21), which all report recession or collapse of major ice tonguesover tens of kilometers during a short period. However, we notethat in these areas, the expression of rapid local ice retreat washeavily influenced by the formation of large and deep proglaciallakes during deglaciation (27, 28). These proglacial lakes estab-lished heat advection and wave impact as significant additionalmechanisms of ice loss, probably leading to locally enhancedrates of ice recession. This mode of post-LGM retreat wasa common feature of glacier systems across the Southern Alps, asevidenced by widespread glaciolacustrine deposits, particularly inthe Eastern Alps, that have been documented by numerousglacial workers (e.g., refs. 29–35). Within this type of retreatsetting (i.e., early transition to lake-terminated glaciers after theonset of retreat), ice-loss dynamics frequently decouple fromclimatic influences, implying that these glacier systems may notpresent reliable indicators for inferring overall regional degla-ciation trends and/or reconstructing climatic signals.In this context, we suggest that the Rangitata record offers

a distinct advantage because this glacier retreated predomi-nately via a grounded glacier margin with little to no influencefrom lake calving on ice retreat dynamics. Our results indicatethat the Rangitata Glacier reached its largest LGM ice extent

Fig. 2. Summary of the Rangitata glacial chronology based on 49 10Be exposure ages (excluding three outliers) that characterize four phases of last glacial–interglacial activity (A–D in graph). The width of the bars represents the timing of moraine formation during each glacial phase according to the weighedmean (with standard mean error) for moraines grouped according to valley position, and thus indicating similar ice extent. The y axis denotes relative glacialextent, with 1.0 representing the approximate maximum ice extent during the LGM (total Rangitata glacier length, 65–70 km). Ice recession from the localLGM position commenced at 18.7 ± 0.3 ka. At 15.8 ka, the Rangitata glacier terminated ∼11 km up-valley from the first inbound recessional moraines, in-dicating the survival of more than 80% of the overall glacier length (and nearly 70% of the full LGM glacier length) through to 15.8 ka.

Rother et al. PNAS | August 12, 2014 | vol. 111 | no. 32 | 11633

EART

H,A

TMOSP

HER

IC,

ANDPL

ANET

ARY

SCIENCE

S

Dow

nloa

ded

by g

uest

on

Apr

il 10

, 202

0

about 28 ka ago and maintained these glacial limits until 25 ka,with only a minor reduction in ice volume during this time. Thistiming corresponds well with the first major LGM cooling event(28.8–25.4 ka), as defined by the composite NZ climate eventstratigraphy (36), and is also broadly contemporaneous withglacial maxima recorded at other midlatitude Southern Hemi-sphere sites (37–39) and the onset of major cooling observedin the Talos Dome and European project for ice coring inAntarctica-Dome C ice cores of Antarctica (24, 40, 41). How-ever, we note that within the NZ pollen-based climate stratig-raphy, the largest Rangitata ice advance occurred near thetransition from mild interstadial conditions [New Zealand cli-mate event (NZce)-11 phase ∼30–28.8 ka: pollen flora 81–95%trees and shrubs] to a progressively cooler climate (NZce-10phase 28.8–25.5 ka: pollen flora 49–71% trees and shrubs).Compared with the last glacial coldest period in NZ (NZce-6 datedto 21.7–18.3 ka; ref. 36), we infer that the ice maximum coincidedwith less-severe but probably moister climatic conditions.Recent records from both hemispheres have emphasized that

similar to many continental alpine glaciers, various parts of themajor ice sheets in Eurasia and Antarctica were not in synchronywith the global LGM (42). More important, although most sec-tors of the largest ice sheet, the Laurentide Ice Sheet, alongwith the Cordilleran Ice Sheet, and the southeast sector of theScandinavian Ice Sheet appear to have reached their maximumextent synchronous with the global LGM (43, 44). We find thatcontrary to these NH trends, in which where ice sheets expandedstrongly between 29–26 ka (9) and then remained at near-maximum limits until the end of the LGM (19–18 ka), ourfindings show that in NZ each ice advance subsequent to the 28-ka

maximum was less extensive, suggesting an overall trend of dimin-ishing regional glaciation in NZ between 28–20 ka.The available terrestrial and marine evidence from the NZ

region, supported by regional paleoclimatic modeling (45),indicates that the persistence of relatively high levels of pre-cipitation against a background of moderate cooling during theearly LGM (∼30–28 ka) are responsible for generating thelargest LGM ice advance in NZ; conversely, the subsequent less-extensive ice advances reflect a trend of decreasing precipitationthat limited glaciation during the remainder of the LGM inNZ, despite intensifying cooling. Although global ice volumeremained stable or even slightly increased between 26–20 ka,glacier systems in NZ were in slow but continuous retreat. Thepattern is repeated at ∼19 ka, when our record shows a region-wide hiatus in warming or ice-stillstand that is coeval with theonset of major retreat in the NH. This implies an overall scenarioin which increasing NH summer insolation initiates the slowretreat of NH ice sheets (from 20 ka onward), whereas Antarcticice extent remains stable at this time. It is now also evident thatthe Southern Ocean plays an important role in the further de-mise of global ice volume by invoking major changes in oceanicCO2 discharge. Once CO2 levels rise sharply (from 18 ka on-ward), Antarctic temperatures increase synchronously (26),triggering the onset of retreat in Antarctica and the midlatitudeSH (NZ).

ACKNOWLEDGMENTS. Cosmogenic surface exposure dating was fundedby Australian Institute of Nuclear Science and Engineering Grant 05/219(Australian Nuclear Science and Technology Organisation) and the Austra-lian Nuclear Science and Technology Organisation’s Cosmogenic ClimateArchives of the Southern Hemisphere Project 0203v.

1. Milankovitch M (1941) Canon of Insolation and the Ice-Age Problem, Special Publi-cation 132 (Serbian Academy of Sciences and Arts, Belgrade). German.

2. Imbrie J, Berger A, Shackleton NJ (1993) Role of orbital forcing: A two-million-yearperspective. Global Changes in the Perspective of the Past, eds Eddy JA, Oeschger H(John Wiley, New York), pp 263–277.

3. Lowell TV, et al. (1995) Interhemispheric correlation of late pleistocene glacial events.Science 269(5230):1541–1549.

4. Vandergoes MJ, et al. (2005) Regional insolation forcing of late Quaternary climatechange in the Southern Hemisphere. Nature 436(7048):242–245.

5. Huybers P, Denton G (2008) Antarctic temperature at orbital timescales controlled bylocal summer duration. Nat Geosci 1:787–792.

6. Wolff EW, Fischer H, Röthlisberger R (2009) Glacial terminations as southern warm-ings without northern control. Nat Geosci 2:206–209.

7. Garcia JL, et al. (2012) Glacier expansion in southern Patagonia throughout theAntarctic cold reversal. Geology 40:859–862.

8. Newnham RM, et al. (2012) Does the bipolar seesaw extend to the terrestrial southernmid-latitudes? Quat Sci Rev 36:214–222.

9. Clark PU, et al. (2009) The Last Glacial Maximum. Science 325(5941):710–714.10. Newnham R, Vandergoes M, Hendy C, Lowe D, Preusser F (2007) A terrestrial paly-

nological record for the last two glacial cycles from south-western New Zealand. QuatSci Rev 26:517–535.

11. Suggate P, Almond P (2005) The Last Glacial Maximum (LGM) in western South Island,New Zealand: Implications for the global LGM and MIS 2. Quat Sci Rev 24:1923–1940.

12. Alloway B, et al.; NZ-INTIMATE Members (2007) Towards a climate event stratigraphyfor New Zealand over the past 30,000 years. J Quaternary Sci 22:9–35.

13. Fink D, Williams P, Augustinus P, Shulmeister J (2005) Are glacial chronologies acrossthe southern hemisphere over the past 80ka regionally synchronized? Annual Geo-logical Society of America, October 5–10, 2005, Salt Lake City. Available at https://gsa.confex.com/gsa/2005AM/finalprogram/abstract_95488.htm. Accessed March 11, 2014.

14. Shulmeister J, Fink D, Hyatt D, Thackray G, Rother H (2010a) Cosmogenic 10Be and26Al exposure ages of moraines in the Rakaia Valley, New Zealand and the nature ofthe last termination in New Zealand glacial systems. Earth Planet Sci Lett 297:558–566.

15. Putnam A, et al. (2013a) The Last Glacial Maximum at 44°S documented by a 10Bemoraine chronology at Lake Ohau, Southern Alps of New Zealand. Quat Sci Rev 62:114–141.

16. Schaefer JM, et al. (2006) Near-synchronous interhemispheric termination of the lastglacial maximum in mid-latitudes. Science 312(5779):1510–1513.

17. Kaplan MR, et al. (2010) Glacier retreat in New Zealand during the Younger Dryasstadial. Nature 467(7312):194–197.

18. Putnam A, et al. (2010a) Glacier advance in southern middle-latitudes during theAntarctic Cold Reversal. Nat Geosci 3:700–704.

19. Putnam A, et al. (2010b) In situ cosmogenic 10Be production-rate calibration from theSouthern Alps, New Zealand. Quat Geochronol 5:392–409.

20. Balco G, Stone J, Lifton N, Dunai T (2008) A simple, internally consistent, and easily

accessible means of calculating surface exposure ages and erosion rates from Be-10

and Al-26 measurements. Quat Res 69:242–249.21. Putnam AE, et al. (2013b) Warming and glacier recession in the Rakaia valley,

Southern Alps of New Zealand, during Heinrich Stadial 1. Earth Planet Sci Lett 382:

98–110.22. Mabin M (1980) The glacial sequences in the Rangitata and Ashburton Valleys, South

Island, New Zealand. PhD thesis (Univ of Canterbury, Christchurch, New Zealand).23. Dunai T (2001) Influence of secular variation of the geomagnetic field on production

rates of in situ produced cosmogenic nuclides. Earth Planet Sci Lett 193:197–212.24. Stenni B, Buiron D, Frezzotti M, Albani S, Barbante C, et al. (2011) Expression of the

bipolar see-saw in Antarctic climate records during the last deglaciation. Nat Geosci

4:46–49.25. Veres D, et al. (2013) The Antarctic ice core chronology (AIC2012): An optimized

multi-parameter and multi-site dating approach for the last 120 thousand years.

Climate of the Past 9:1733–1748.26. Parrenin F, et al. (2013) Synchronous change of atmospheric CO2 and Antarctic

temperature during the last deglacial warming. Science 339(6123):1060–1063.27. Shulmeister J, et al. (2010b) The stratigraphy, timing and climatic implications of pre-

LGM glaciolacustrine deposits in the middle Rakaia Valley, South Island, New Zealand.

Quat Sci Rev 29:2362–2381.28. Evans DJA, Rother H, Hyatt O, Shulmeister J (2013) The glacial sedimentology and

geomorphic evolution of an outwash head/moraine-dammed lake, South Island, New

Zealand. Sediment Geol 284-285:45–75.29. Speight R (1926) Varved glacial silts from the Rakaia Valley. Records of the Canterbury

Museum 3:55–82.30. Gage M (1958) Late Pleistocene glaciations of the Waimakariri Valley, Canterbury,

New Zealand. NZ J Geol Geophys 1:123–155.31. Clayton J (1968) Late Pleistocene glaciations in the Waiau Valleys, North Canterbury.

NZ J Geol Geophys 11:753–767.32. Suggate RP (1990) Late Pleistocene and Quaternary glaciations of New Zealand. Quat

Sci Rev 9:175–197.33. Hart JK (1996) Proglacial glaciotectonic deformation associated with glaciolacustrine

sedimentation, Lake Pukaki, New Zealand. J Quaternary Sci 11:149–160.34. Mager S, Fitzsimons J (2007) Formation of glaciolacustrine Late Pleistocene end

moraines in the Tasman Valley, New Zealand. Quat Sci Rev 26:743–753.35. Rother H, Shulmeister J, Rieser U (2010) Stratigraphy, optical dating chronology (IRSL)

and depositional model of pre-LGM glacial deposits in the Hope Valley, New Zealand.

Quat Sci Rev 29:576–592.36. Barrell DJA, Almond PC, Vandergoes MJ, Lowe DJ, Newnham RM, INTIMATE members

(2013) A composite pollen-based stratotype for inter-regional evaluation of climatic

events in New Zealand over the past 30,000 years (NZ-INTIMATE project). Quat Sci Rev

74:4–20.

11634 | www.pnas.org/cgi/doi/10.1073/pnas.1401547111 Rother et al.

Dow

nloa

ded

by g

uest

on

Apr

il 10

, 202

0

37. Harrison S, Glasser NF (2011) The Pleistocene glaciations of Chile. Quaternary Glaci-ations—Extent and Chronology, eds Ehlers J, Gibbard P, Hughes PD (Elsevier, Amsterdam),Vol 15, pp 739–756.

38. Colhoun EA, Barrows T (2011) The glaciation of Australia. Quaternary Glaciations–Extent and Chronology, eds Ehlers J, Gibbard P, Hughes PD (Elsevier, Amsterdam),Vol 15, pp 1037–1045.

39. Barrows T, et al. (2013) Late Pleistocene glacial stratigraphy of the Kumara-Moanaregion, West Coast of South Island, New Zealand. Quat Sci Rev 74:139–159.

40. EPICA Community Members (2006) One-to-one coupling of glacial climate variabilityin Greenland and Antarctica. Nature 444(7116):195–198.

41. Lemieux-Dudon B, et al. (2010) Consistent dating for Antarctic and Greenland icecores. Quat Sci Rev 29:8–20.

42. Hughes PD, Gibbard PL, Ehlers J (2013) Timing of glaciation during the last glacial cycle:Evaluating the concept of a global ‘Last Glacial Maximum’. Earth Sci Rev 125:171–198.

43. Dyke A (2004) An outline of North American deglaciation with emphasis on centraland northern Canada. Quaternary Glaciations—Extent and Chronology, Part II: NorthAmerica. Developments in Quaternary Science, 2b, eds Ehlers J, Gibbard P, Hughes PD(Elsevier, Amsterdam), Vol 2, Part B, pp 1–7.

44. Briner JP, Gosse JC, Bierman PR (2006) Applications of cosmogenic nuclides to LaurentideIce Sheet history and dynamics. Application of Cosmogenic Nuclides to the Study of EarthSurface Processes: The Practice and the Potential, eds Siame LL, Bourlès DL, Brown ET(Geological Society of America Special Papers, Boulder, CO), Vol 415, pp 29–41.

45. Lorrey A, et al. (2012) Palaeocirculation across New Zealand during the last glacialmaximum at ∼21 ka. Quat Sci Rev 36:189–213.

Rother et al. PNAS | August 12, 2014 | vol. 111 | no. 32 | 11635

EART

H,A

TMOSP

HER

IC,

ANDPL

ANET

ARY

SCIENCE

S

Dow

nloa

ded

by g

uest

on

Apr

il 10

, 202

0