GEOLOGY | Volume 45 | Number 6 | www.gsapubs.org 1

Recent retreat of Columbia Glacier, Alaska: Millennial contextAnders E. Carlson1*, Zoe Kilmer1, Leah B. Ziegler1, Joseph S. Stoner1, Greg C. Wiles2, Kaitlin Starr2, Maureen H. Walczak1, William Colgan3, Alberto V. Reyes4, David J. Leydet1,5, and Robert G. Hatfield1

1College of Earth, Ocean, and Atmospheric Sciences, Oregon State University, Corvallis, Oregon 97331 USA2Department of Geology, College of Wooster, Wooster, Ohio 44691 USA3Department of Earth and Space Science and Engineering, York University, Toronto, Ontario M3J 1P3, Canada4Department of Earth and Atmospheric Sciences, University of Alberta, Edmonton, Alberta T6G 2R3, Canada5Department of Geography & Environmental Engineering, U.S. Military Academy, West Point, New York 10996, USA

ABSTRACTColumbia Glacier in Prince William Sound, Alaska, has retreated

~20 km in the past three decades. We use marine sediment records to document the Columbia Glacier advance and retreat history over the past 1.6 k.y. in an effort to place its recent retreat in the context of the Common Era (C.E.). A change in magnetic mineralogy coincided with a shift in sediment geochemistry ca. 0.9 ka. This provenance change documents the advance of Columbia Glacier across a fault, resulting in glacial erosion of mafic rocks near the coast; this agrees with the tim-ing of ice advance reconstructed using dendrochronology. Our marine provenance records show that Columbia Glacier remained advanced south of this fault into the 21st century. Columbia Glacier has now retreated north of this fault, making its recent retreat unprecedented since before ca. 0.9 ka. Southern Alaska temperatures have now warmed to pre–0.9 ka levels, based on tree-ring and reanalysis data. We show with glacier model simulations that the warming between C.E. 1910 and 1980, that includes anthropogenic forcing, was sufficient to trigger the recent retreat of Columbia Glacier from its extended position of the past 0.9 k.y., consistent with our data-driven assessment of the relationship between regional climate change and glacier extent. We conclude that the recent retreat of Columbia Glacier is a response to climate change rather than part of a natural internal tidewater-glacier oscillation.

INTRODUCTIONMany Alaskan tidewater glaciers are currently retreating, and are the

largest contributors to global mean sea-level rise from the cryosphere other than ice masses in Greenland and Antarctica (Berthier et al., 2010). However, some Alaskan tidewater glaciers are advancing or remaining at extended positions (Meier, 1987; Post et al., 2011; McNabb and Hock, 2014) despite a warming southern Alaskan climate over the past century (Hartmann et al., 2013). Other Alaskan tidewater glaciers began their retreat well before the beginning of the last century (Barclay et al., 2009), when southern Alaska climate had yet to begin to warm significantly (Wiles et al., 2014). Tidewater glaciers are known to undergo cyclic behav-ior of slow advance and rapid retreat in decades after ice-margin destabili-zation (Post, 1975; Meier, 1987; Post et al., 2011). Post et al. (2011) noted that understanding the cause of this tidewater stability-instability transition is of critical importance for predicting future tidewater-glacier behavior.

Columbia Glacier in Prince William Sound (southern Alaska) is under-going one of the most dramatic and best-documented retreats of a tidewater glacier in the past three decades (Fig. 1) (Meier, 1987; McNabb and Hock, 2014), and is the largest Alaskan glacier contributor to sea-level rise over this time interval (Berthier et al., 2010). Placing the recent behavior of Columbia Glacier in the context of past fluctuations in its extent during periods of minimal anthropogenic influence on climate could shed light

on the trigger of its rapid retreat. Although numerous tree-ring and 14C chronologies exist for glaciers of southern Alaska (e.g., Barclay et al., 2009), relatively little is known about the prehistoric record of Columbia Glacier and its relationship to climate change (Barclay et al., 2009). Here we present the first continuous sediment discharge record for Columbia Glacier that documents its late-Holocene extent, placing its current retreat in the perspective of the past ~1.6 k.y.

METHODSTo track the extent of Columbia Glacier, we use changes in bulk sedi-

ment geochemistry and magnetic grain size in marine jumbo piston core EW0408–95JC/TC and multicore EW0408–94MC (60.66°N, 147.71°W, 744 m water depth; RV Maurice Ewing cruise EW0408) in Prince William Sound that are dated by 14C and 210Pb, respectively (Fig. 1B; Table DR1 in the GSA Data Repository1). The summer through fall regional ocean circu-lation pattern (Musgrave et al., 2013) transports sediment from Columbia Glacier to the core site during the summer ablation season (Fig. 1B) over the course of a few years to a few decades (Jaeger and Koppes, 2016).

Anhysteretic remanent magnetization (ARM), expressed as susceptibil-ity of ARM (kARM) and magnetic susceptibility (k), and bulk sediment geochemistry were measured. We relate the change in bulk sediment, kARM, k, the ratio of the two (a well-known magnetic grain-size proxy), and geochemistry in the cores to changes in sediment source, most nota-bly, the Contact fault, which extends along the north side of Prince Wil-liam Sound (Figs. 1B and 1C). South of the fault, the bedrock contains Paleogene–Miocene mafic, mainly extrusive, crystalline rocks in addition to marine sedimentary rocks; north of the fault, the bedrock is mainly uplifted Cretaceous marine sedimentary rocks (Wilson and Hults, 2012). Therefore, glacially eroded sediments from north of the fault should have magnetic and geochemical signatures distinctly different from those of glacially eroded sediments from south of the fault.

To determine sediment provenance, we collected glacial sediment samples from the historical moraines of Columbia and Shoup Glaciers (Fig. 1B; Table DR2). The Columbia Glacier moraine contains the con-flated geochemistry and magnetic properties of sediment eroded from both sides of the Contact fault, whereas the Shoup Glacier moraine is sediment eroded only from north of the Contact fault.

We used 3020 glacier model simulations (Colgan et al., 2012) that forced Columbia Glacier with different rates and magnitude of climate warming. We binned the simulation results that destabilized the glacier to compare rate of warming versus magnitude of warming against our data-driven observations. (See the Data Repository for methods, dendro-chronology, and glacier modeling.)

1 GSA Data Repository item 2017175, methods, dendrochronology, and glacier modeling, is available online at http://www.geosociety.org/datarepository /2017/, or on request from editing@geosociety.org.*E-mail: acarlson@coas.oregonstate.edu

GEOLOGY, June 2017; v. 45; no. 6; p. 1–4 | Data Repository item 2017175 | doi:10.1130/G38479.1 | Published online XX Month 2017© 2017 Geological Society of America. For permission to copy, contact editing@geosociety.org.

2 www.gsapubs.org | Volume 45 | Number 6 | GEOLOGY

RESULTSForaminiferal 14C ages indicate that the 17.5 m EW0408–95JC core

extends back to ca. 1.6 ka, with sedimentation rates of >1 cm a–1 (Fig. 2F). The 14C ages are in stratigraphic order with the exception of age rever-sals between 554 cm and 1052 cm below the seafloor (Fig. DR1), where foraminifera dates are anomalously old (Table DR1); physical sediment structure in the core (Fig. DR1) supports interpretation of this interval as a gravity flow containing reworked foraminifera (Jaeger and Koppes, 2016). The gravity flow deposit separates a bioturbated section above from a laminated section below (Fig. DR1); the laminated section prob-ably reflects hypoxic or anoxic conditions in Prince William Sound. The magnetic records have two distinct periods: the earlier is characterized by high variability, and low k, kARM, and kARM/k before ca. 1.0 ka, shifting to reduced variability, higher k, kARM, and kARM/k ca. 0.9 ka (Figs. 2A–2C). Sediment geochemical changes mimic magnetic grain-size changes, with a decrease in Si relative to Fe and Ca concentration ca. 0.9 ka (Figs. 2D and 2E).

Bulk elemental analysis of the glacial sediments shows a distinct change in geochemistry; Columbia Glacier sediment has higher Fe and Ca concentrations and lower Si/Ca and Si/Fe than Shoup Glacier sedi-ment (Table DR2). We find a significant difference in magnetic properties between these two terranes. Sediments from south of the Contact fault have higher k, kARM, and kARM/k (a higher concentration of finer mag-netic grain sizes) than sediments north of the fault (Table DR2).

The glacier model simulations show that faster rates of warming require a shorter duration of warming to destabilize Columbia Glacier (Fig. 3).

DISCUSSIONThe magnetic and geochemical data show a change in sediment prov-

enance ca. 0.9 ka. The shift in kARM and k to higher values concurrent (Figs. 2B and 2C) with increased kARM/k (Fig. 2A) is consistent with a source shift to a lithology that is substantially more magnetic with a consistently fine-grained magnetic mineralogy after ca. 0.9 ka, such as the extrusive mafic rocks from south of the Contact fault (Wilson and Hults, 2012). The decrease in Si/Fe and Si/Ca ca. 0.9 ka (Figs. 2D and 2E) also indicates a new sediment source enriched in elements consistent with mafic rocks that are found south of the Contact fault (Wilson and Hults, 2012). Because of the counterclockwise summer to fall ocean circulation pattern in Prince William Sound (Musgrave et al., 2013), three tidewater glaciers, Columbia, Meares, and Shoup, could be major contributors of sediment to the core site during the ablation season (Fig. 1B). However, only Columbia Glacier advanced south of the Contact fault in the late Holocene (Post, 1975; Post and Viens, 2000; Molnia, 2008) (Fig. 1B), and thus would be the only marine-terminating glacier to overlie and erode highly magnetic mafic extrusive rocks with a homogeneous fine-grained magnetic mineralogy, as identified by the glacial sediment provenance data.

We compared the marine sediment record with calendar tree-ring dates on overrun mountain hemlock forests (Tsuga mertensiana) killed during late-Holocene Columbia Glacier advance (Table DR3). The dendrochro-nology shows that Columbia Glacier advanced south of the Contact fault ca. 0.9 ka (Figs. 1C and 2G). This timing is in agreement with the observed marine sedimentary shifts in magnetic properties and geochemical prov-enance ca. 0.9 ka. This advance is consistent with other southern Alaska

1986

1980

1984

1989

1992

2004

20082010

2012

1748

1767

5 km

Contact fault

1808

14501413

1383,1389

1318

1198

1072

10401010

1060,1067

1994

1996

2000

2002

2006

147 ˚W

61 ˚N

C

Gulf ofAlaska

ALASKA

CANADAPrince William

Sound

A

65º N

60º N 140º

W150º

W

500 km

SG

MG 1750 CE

1980CE

age?

CG

EW0408-95JC/TC

20 km

B C

147.0 ˚W

60.5 ˚N

N NAlaska C

oastal C

urrent

Figure 1. A: Prince William Sound, southern Alaska. B: Location of core EW0408–95JC/TC (RV Maurice Ewing cruise EW0408) and the late spring through fall circulation in Prince William Sound (Musgrave et al., 2013). Dotted white lines show the late-Holocene extents of Columbia Glacier (CG), Shoup Glacier (SG), and Meares Glacier (MG) (Post and Viens, 2000; Molnia, 2008); white squares denote the glacier sediment sample locations on moraines. Thick dashed line marks the Contact fault (Wilson and Hults, 2012). C: False-color image of Columbia Glacier on 2 July 2014. Contact fault location is after Wilson and Hults (2012). Solid white lines indicate the retreating Columbia Glacier ice margin since C.E. 1980 (McNabb and Hock, 2014); red circles are sites of tree-ring kill dates (C.E. ages in italics) that map the down-fjord advance of Columbia Glacier.

GEOLOGY | Volume 45 | Number 6 | www.gsapubs.org 3

and Coast Range glaciers that began to expand toward their late Holocene maxima ca. 1 ka (Barclay et al., 2009). We attribute increased variability in magnetic parameters and geochemistry before ca. 1.0 ka to more active glaciers in Prince William Sound eroding heterogeneous sedimentary source material from behind the Contact fault (Wilson and Hults, 2012).

The dendrochronologic record, based on calendar-dated trees over-run at the glacier margin, documents the general advance of Columbia Glacier to its maximum extent at C.E. 1808. The dense tree-ring network dates ice advance, but does not constrain possible intervals of ice-margin retreat, as could be implied by the gap in kill ages of 0.6–0.3 ka (Figs. 1C and 2G). However, given the high sedimentation rates (Figs. 2F and 4D) and a sediment transport time from glacier margin to the outer fjord and

1600

1200

800

400

0

Dep

th b

elow

surfa

ce (c

m)

1234567

kAR

M/k

0

0.0004

0.0008

0.0012

0.0016

kAR

M

0.0001

0.0002

0.0003

0.0004

k

0.16

0.12

0.08

0.04

Si/C

a (in

verte

d)0.008

0.006

0.004

Si/F

e (in

verte

d)

-20

-15

-10

-5

0

Col

umbi

a G

laci

erle

ngth

(km

)

-0.500.51.01.52.0

PWS

∆SAT

(˚C

)

1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0Age (ka)

Contact fault

A

B

D

C

E

G

F

H

40

6080

100

120

140

160

War

min

g du

ratio

n (a

)

0.005 0.010 0.015 0.020 0.025Warming rate (˚C a-1)

-20

-15

-10

-5

0

1900 1920 1940 1960 1980 2000Year C.E.

0.07

0.06

0.05

0.0040

0.0036

0.0032

0.0028

0.1

1

10

210 P

b se

dim

enta

tion

rate

(cm

a-1)

-0.5

0

0.5

1.0

1.5

2.0

Si/C

a (in

verte

d)

Si/F

e (in

verte

d)C

olum

bia

Gla

cier

leng

th (k

m)

PWS

∆SAT

(˚C

)

Contact fault

A

B

DC

E

Figure 2. Columbia Glacier (southern Alaska) change and core EW0408–95JC/TC data (RV Maurice Ewing cruise EW0408) (in A–F, red is trigger core; blue is jumbo piston core). A–C: Ratio of susceptibility of anhysteretic remanent magnetization (kARM) to magnetic suscep-tibility (k) kARM/k, kARM, and k. D, E: Si/Ca and Si/Fe. F: Age-depth relationship with 95% confidence interval. G: Columbia Glacier length from dendrochronology (squares; lines indicate life span of glacier-killed trees) and observations (circles) (Post, 1975; McNabb and Hock, 2014). H: Decadal change in tree-ring February–August surface air temperature (SAT, brown line; annual shown in tan) for southern Alaska (Wiles et al., 2014) and summer SAT from reanalysis data (purple line) at the Columbia Glacier equilibrium line (Colgan et al., 2012), both rela-tive to the C.E. 1870–1890 mean. Dashed line is the summer SAT at the time of Columbia Glacier retreat onset in mid–C.E. 1980. PWS—Prince William Sound. Vertical yellow bar is the gravity flow.

Figure 3. The rate and duration of air temperature warming that triggered Columbia Glacier (southern Alaska) retreat in an ensemble of model simulations (n = 3020) (Colgan et al., 2012). Simulations are divided into 10 bins with a 1σ range (gray bars). The diamond is rate and duration of warming that occurred until Columbia Glacier destabiliza-tion in mid–C.E. 1980.

Figure 4. Columbia Glacier (southern Alaska) change of the past cen-tury. A: Observations of Columbia Glacier length (Post, 1975; McNabb and Hock, 2014); the Contact fault is indicated. B: Si/Ca from core EW0408–94MC (RV Maurice Ewing cruise EW0408). C: Si/Fe. D: 210Pb sedimentation rate in EW0408–94MC with 95% confidence interval on a log scale. E: Decadal change in surface air temperature (SAT) from Figure 2H. Dashed line is the summer SAT at the time of onset of Columbia Glacier retreat. PWS—Prince William Sound. Vertical yellow bar is onset of Columbia Glacier retreat.

4 www.gsapubs.org | Volume 45 | Number 6 | GEOLOGY

continental shelf of a few years to decades (Jaeger and Koppes, 2016), it is very likely that any retreat of Columbia Glacier north of the Contact fault would be recorded as a decrease in kARM/k and an increase in Si/Fe-Si/Ca to pre–0.9 ka levels. The input of high kARM/k and low Si/Fe and Si/Ca sediments to the site through to the core sediment-water interface, reflecting the collection year of C.E. 2004 (Figs. 2 and 4), indicates the continual presence of Columbia Glacier south of the Contact fault (Fig. 1C) since ca. 0.9 ka. The records from EW0408–94MC show increased sediment accumulation with increased input of Fe and Ca relative to Si (Figs. 4B–4D) when Columbia Glacier began northward retreat in mid-1980, consistent with increased sediment discharge during tidewater-glacier retreat (Jaeger and Koppes, 2016). The observational record has Columbia Glacier south of the Contact fault for all of the 20th century, with the glacier margin retreating north of this geologic boundary in C.E. 2002 (Fig. 4A) (Post, 1975; McNabb and Hock, 2014), two years before EW0408 core collection. Based on the marine sedimentary record, this retracted position has not occurred since before ca. 0.9 ka, meaning that the current retreat of Columbia Glacier is unprecedented with respect to the last millennium.

Because the marine sediment record indicates that Columbia Glacier has not retreated to its modern position at any point in the last millen-nium, we argue that an external forcing may be the trigger of the current instability, consistent with tidewater glacier theory (Post, 1975; Meier, 1987; Post et al., 2011). Southern Alaska tree-ring surface air temperature (SAT) reconstructions indicate that February–August SAT was colder than the late 20th century for the past ~0.9 k.y. (Wiles et al., 2014) (Fig. 2H). The last time southern Alaska SAT approached late 20th century levels was prior to ca. 1.0 ka, when southern Alaska SAT was +1.0 °C relative to the late C.E. 1800s (Fig. 2H), during which time the marine sediment record indicates that Columbia Glacier terminated north of the Contact fault. The subsequent Columbia Glacier advance corresponds with cooling below this +1.0 °C threshold. Southern Alaska winter precipitation has increased in the past few centuries (Osterberg et al., 2014); this should support glacier expansion. We therefore suggest that climate warming over the past century thinned the lower Columbia Glacier to the point of destabilization in the decade of C.E. 1980–1990, when SAT warmed above +1.0 °C (Fig. 2H), which then led to the observed catastrophic terminus retreat.

The observed April–September warming of +1.2 °C over 70 yr (0.017 °C a–1) at the Columbia Glacier equilibrium line (Fig. 4E), attributed largely to anthropogenic greenhouse gas emissions (Hartmann et al., 2013), agrees well with the glacier model simulated rate and duration of warming nec-essary to trigger Columbia Glacier instability (Colgan et al., 2012) (Fig. 3). Furthermore, the simulated cumulative warming to trigger Columbia Glacier instability of 1.1 ± 0.2 °C also agrees with the +1.0 °C of warming indicated by the paleodata as necessary to trigger Columbia Glacier retreat.

CONCLUSIONSWe conclude that atmospheric warming of southern Alaska over the

20th century triggered the Columbia Glacier retreat of the last few decades. Columbia Glacier retreat is consequently a response to climate change that includes anthropogenic inputs, rather than a natural tidewater-glacier cyclic oscillation. More broadly, our diagnostic analysis supports the emerging notion, largely derived from prognostic modeling, that air tem-perature increases of <2 °C can result in the loss of ice masses in a fashion that is irreversible on millennial time scales (e.g., Robinson et al., 2012).

ACKNOWLEDGMENTSResearch was funded by U.S. National Science Foundation grants EAR-0711584 (Stoner), EAR-1428421 (Carlson and Stoner), EAR-1049579 (Ziegler), and AGS-1502186/AGS-1202218 (Wiles), National Geographic Society grant NGS9489 (Wiles), and Oregon State University (Kilmer). We thank Maziet Cheseby for help with foraminifera analyses. Four anonymous reviewers, Jason Amundson, and input from editor James Spotila improved this manuscript.

REFERENCES CITEDBarclay, D.J., Wiles, G.C., and Calkin, P.E., 2009, Holocene glacier fluctuations

in Alaska: Quaternary Science Reviews, v. 28, p. 2034–2048, doi: 10 .1016 /j .quascirev .2009 .01 .016.

Berthier, E., Schiefer, E., Clarke, G.K.C., Menounos, B., and Rémy, F., 2010, Con-tribution of Alaskan glaciers to sea-level rise derived from satellite imagery: Nature Geoscience, v. 3, p. 92–95, doi: 10 .1038 /ngeo737.

Colgan, W., Pfeffer, W.T., Hajaram, H., Abdalati, W., and Balog, J., 2012, Monte Carlo ice flow modeling projects a new stable configuration for Columbia Glacier, Alaska, c. 2020: The Cryosphere, v. 6, p. 1395–1409, doi: 10 .5194 /tc -6 -1395 -2012.

Hartmann, D.L., et al., 2013, Observations: Atmosphere and surface, in Stocker, T.F., et al., eds., Climate change 2013: The physical science basis: Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmen-tal Panel on Climate Change: Cambridge, UK, Cambridge University Press, p. 159–254, doi: 10 .7892 /boris .41386.

Jaeger, J.M., and Koppes, M.N., 2016, The role of the cryosphere in source-to-sink systems: Earth-Science Reviews, v. 153, p. 43–76, doi: 10 .1016 /j .earscirev .2015 .09 .011.

McNabb, R.W., and Hock, R., 2014, Alaska tidewater glacier terminus positions, 1948–2012: Journal of Geophysical Research, v. 119, p. 153–167, doi: 10.1002 /2013JF002915.

Meier, M.F., 1987, Fast tidewater glaciers: Journal of Geophysical Research, v. 92, p. 9051–9058, doi: 10 .1029 /JB092iB09p09051.

Molnia, B.F., 2008, Glaciers of North America—Glaciers of Alaska, in Williams, R.S., Jr., and Ferrigno, J.G., eds., Satellite image atlas of glaciers of the World: U.S. Geological Survey Professional Paper 1386-K, 525 p.

Musgrave, D.L., Halverson, M.J., and Pegau, W.S., 2013, Seasonal surface circula-tion, temperature, and salinity in Prince William Sound, Alaska: Continental Shelf Research, v. 53, p. 20–29, doi: 10 .1016 /j .csr .2012 .12 .001.

Osterberg, E.C., Mayewski, P.A., Fisher, D.A., Kreutz, K.J., Maasch, K.A., Sneed, S.B., and Kelsey, E., 2014, Mount Logan ice core record of tropical and solar influences on Aleutian Low variability: 500–1998 A.D: Journal of Geophysical Research, v. 119, p. 11,189–11,204, doi: 10 .1002 /2014JD021847.

Post, A., 1975, Preliminary hydrography and historical terminal changes of Co-lumbia Glacier, Alaska: U.S. Geological Survey Hydrologic Investigations Atlas HA-559, 3 plates.

Post, A., and Viens, R.J., 2000, Preliminary bathymetry of Shoup Basin and late Holocene changes of Shoup Glacier, Alaska: U.S. Geological Survey Water-Resources Investigation Report 94-4093, 11 p.

Post, A., O’Neel, S., Motyka, R.J., and Streveler, G., 2011, A complex relationship between calving glaciers and climate: Eos (Transactions, American Geophysi-cal Union), v. 92, p. 305–312, doi: 10 .1029 /2011EO370001.

Robinson, A., Calov, R., and Ganopolski, A., 2012, Multistability and critical thresholds of the Greenland ice sheet: Nature Climate Change, v. 2, p. 429–432, doi: 10 .1038 /nclimate1449.

Wiles, G.C., D’Arrigo, R.D., Barclay, D., Wilson, R.S., Jarvis, S.K., Vargo, L., and Frank, D., 2014, Surface air temperature variability reconstructed with tree rings for the Gulf of Alaska over the past 1200 years: The Holocene, v. 24, p. 198–208, doi: 10 .1177 /0959683613516815.

Wilson, F.H., and Hults, C.P., compilers, 2012, Geology of the Prince William Sound and Kenai Peninsula region, Alaska: U.S. Geological Survey Scientific Investigation Map 3110, 38 p., scale 1:350,000.

Manuscript received 12 August 2016 Revised manuscript received 17 February 2017 Manuscript accepted 1 March 2017

Printed in USA

GSA Data Repository 2017175 Carlson et al., 2017, Recent retreat of Columbia Glacier, Alaska: Millennial context: Geology, doi:10.1130/G38479.1.

METHODS 1

Marine Core Sedimentology and Chronology 2

Marine jumbo piston core EW0408-95JC, along with its trigger core EW0408-95TC and site 3

multi-core EW0408-94MC (60.66°N, 147.71°W, 744 m water depth), were retrieved in Prince 4

William Sound in the summer of 2004 C.E. aboard the RV Ewing. The composite cm below 5

seafloor (cmbsf) depth scale used to align the jumbo piston, trigger, and multi-cores was 6

determined via correlation of physical properties and geochemical data. Multi-core EW0408-7

94MC captures the sediment-water interface, while EW0408-95TC lost the uppermost 6 cm of 8

the sediment column and EW0408-95JC failed to sample the uppermost 77 cm of the sediment 9

column. Fig. DR1 shows the CT scan and gamma-ray attenuation bulk sediment density of 10

EW0408-95JC. The deepest portion of the core is characterized by laminated sediments 11

indicative of local bottom water hypoxia and/or anoxia, with low-density layers enriched in 12

biogenic silica reflecting elevated primary productivity. This portion of the core is overlain with 13

an erosive contact at 1052 cmbsf. This unit extends to 554 cmbsf and displays sedimentary 14

structure indicative of a geologically instantaneous deposit associated with a submarine gravity 15

flow, an interpretation supported by anomalously aged foraminiferal 14C dates consistent with 16

reworked sediment. 17

We develop a chronology using 18 14C ages (Table DR1) on mixed benthic foraminifera 18

tests in EW0408-95TC/JC. Samples were measured at University of California-Irvine and 19

Australian National University, consistent with the methods outlined in Davies-Walczak et al. 20

(2014). Although planktic foraminiferal preservation in the core was limited, we were able to 21

METHODS

2

evaluate the benthic-planktic age offset at 780 cmbsf in core EW0408-95JC. This benthic-22

planktic age offset of 340±60 14C years, when combined with the regional surface ocean 'R of 23

470±80 years (McNeely et al., 2006), indicates a benthic 'R of 810±100 years. The benthic 24

foraminiferal 14C ages were calibrated using this 'R value on the Marine13 curve (Reimer et al., 25

2013), and integrated with 210Pb and 137Cs data on EW0408-94MC (Walinsky et al., 2009) (Fig. 26

3d) to create an age model using the program BChron (Haslett and Parnell, 2008). The gravity 27

flow deposit is excluded from the age model construction. 28

Magnetic and Geochemical Analyses 29

The magnetic parameters kARM and k were measured at 1 cm intervals on u-channels at 30

Oregon State University, the ratio of which provides a common magnetic grain-size proxy 31

(Hatfield and Stoner, 2013). Bulk sediment geochemistry was measured by X-ray fluorescence 32

at 1 mm resolution on EW0408-95JC and 0.2 mm resolution on EW0408-95TC and EW0408-33

94MC (records shown are smoothed to a decadal average) at Oregon State University following 34

Carlson et al. (2008). Given the >1 cm a-1 sedimentation rate of our cores, these records reflect 35

annual to sub-annual resolution prior to smoothing. 36

Columbia and Shoup Glaciers’ sediments were collected from their terminal late-Holocene 37

moraines to determine sediment provenance source. These locations were chosen as they 38

represent two end members of sediment source to the core site from either side of the Contact 39

Fault that extends along the north side of Prince William Sound (Figs. 1B, C) (Wilson et al., 40

2012). Bulk glacial sediment geochemistry was determined by inductively coupled plasma 41

atomic emission spectrometry (ICP-AES) at the University of Colorado. Bulk glacial sediment 42

magnetic properties were measured in packed plastic cubes at Oregon State University 43

following Hatfield et al. (2013). Results are shown in Table DR2. 44

Dendrochronology 45

METHODS

3

Dendrochronology samples were collected by increment borer and chainsaw from trees 46

exposed (sub-fossil wood) by the recent retreat of Columbia Glacier. The calendar tree-ring 47

dates are from the outer rings of mountain hemlock trees (Tsuga mertensiana) that have been 48

killed by the advancing glacier, pushed against the surrounding bedrock fiord walls, and 49

although they have been transported, most are associated with primary soils and forest debris. 50

We determined the calendar kill dates for the sub-fossil logs by cross-dating the mountain 51

hemlock trees to a master calendar dated tree-ring chronology for southern Alaska (Barclay et 52

al., 2009; Wiles et al., 2014) using standard dendrochronological techniques (Stokes and Smiley, 53

1996). These data have been used in dendro-climatic studies (Wilson et al., 2007; Wiles et al., 54

2014), updated by Starr et al. (2015), and previously unpublished Columbia Glacier tree-ring 55

dating by Kennedy (2003) appeared in Nick et al. (2007) and Barclay et al. (2009). Information 56

on the dendrochronology samples is provided in Table DR3. 57

Glacier Modeling 58

Simulations of Columbia Glacier response to ambient temperature come from Colgan et al. 59

(2012). Briefly, this flow-line model, which includes a first order approximation for longitudinal 60

coupling stresses, was run 20,000 times in a Monte Carlo forward selection approach, whereby 61

simulations that matched the 1980 C.E. extent and thickness of Columbia Glacier (Meier et al., 62

1985) at the end of spin-up to transient equilibrium were retained (n=3020). The selected 63

simulations were then forced with different rates and durations of climate warming that changed 64

the equilibrium line altitude until triggering Columbia Glacier retreat. The selected simulations 65

induced retreat under a diverse range of rates and durations of climate warming (Fig. 4). We 66

divided the selected simulations into ten bins, based on warming rate, with 302 simulations per 67

bin. We then calculated the mean and standard deviation of the rates and durations of warming 68

in each bin (gray bars in Fig. 4) to assess the relation between intensity of climate change (rate 69

and duration of warming) and the instability and retreat of Columbia Glacier. We compare this 70

against the actual rate and duration of warming in April-September air temperature, as inferred 71

METHODS

4

locally over the 1871 to 2008 period by NOAA/OAR/ESRL PSD Twentieth Century Reanalysis 72

V2, which preceded the destabilization and retreat of Columbia Glacier in the mid 1980’s C.E. 73

(Compo et al., 2011; Colgan et al., 2012). 74

METHODS

5

References Cited 75

Barclay, D.J., Wiles, G.C., and Calkin, P.E., 2009, Holocene glacier fluctuations in Alaska: 76

Quaternary Science Reviews, v. 28, p. 2034-2048. 77

Carlson, A.E., Stoner, J.S., Donnelly, J.P., and Hillaire-Marcel, C., 2008, Response of the 78

southern Greenland Ice Sheet during the last two deglaciations: Geology, v. 36, p. 359-362. 79

Colgan, W., Pfeffer, W.T., Hajaram, H., Abdalati, W., and Balog, J., 2012, Monte Carlo ice flow 80

modeling projects a new stable configuration for Columbia Glacier, Alaska, c. 2020: The 81

Cryosphere, v. 6, p. 1395-1409. 82

Compo, G.P., and 26 others, 2011, The Twentieth Century Reanalysis Project: Quarterly 83

Journal of the Royal Meteorological Society, v. 137, p. 1-28. 84

Davies-Walczak, M., Mix, A.C., Stoner, J.S., Southon, J.R., Cheseby, M., and Xuan, C., 2014, 85

Late Glacial to Holocene radiocarbon constraints on North Pacific Intermediate Water 86

ventilation and deglacial atmospheric CO2 sources: Earth and Planetary Science Letters, v. 87

397, p. 57-66. 88

Haslett, J., and Parnell, A., 2008, A simple monotone process with application to radiocarbon-89

dated depth chronologies: Journal of the Royal Statistical Society, v. 57, p. 399-418. 90

Hatfield, R.G., and Stoner, J.S., 2013, Magnetic properties and susceptibility: Encyclopedia of 91

Quaternary Science, v. 2, p. 884-898. 92

Hatfield, R.G., Stoner, J.S., Carlson, A.E., Reyes, A.V., and Housen, B.A., 2013, Source as a 93

controlling factor on the quality and interpretation of sediment magnetic records from the 94

northern North Atlantic: Earth and Planetary Science Letters, v. 368, p. 69-77. 95

Kennedy, M.G., 2003, Advance rates of Columbia Glacier during the last 1000 years, Prince 96

William Sound, Alaska: College of Wooster Senior Thesis, Wooster Ohio, 45 pp. 97

McNeely, R., Dyke, A.S., and Southon, J.R., 2006, Canadian marine reservoir ages, preliminary 98

data assessment: Geological Survey of Canada Open File 5049. 99

METHODS

6

Meier, M., Rasmussen, L., Krimmel, R., Olsen, R., and Frank, D., 1985, Photogrammetric 100

Determination of Surface Altitude, Terminus Position, and Ice Velocity of Columbia Glacier, 101

Alaska: U.S. Geological Survey Professional Paper 1258-F, 40 pp. 102

Nick, F.M., van der Veen, C.J., and Oerlemans, J., 2007, Controls on advance of tidewater 103

glaciers: Results from numerical modeling applied to Columbia Glacier: Journal of 104

Geophysical Research, v. 112, doi: 10.1029/2006JF000551. 105

Reimer, P.J., and 29 others, 2013, IntCal13 and Marine13 radiocarbon age calibration curves 0-106

50,000 years cal BP: Radiocarbon, v. 54, p. 1869-1887. 107

Starr, K., Happ, M., Wiles, G., and Wiesenberg, N., 2015, Reconstructing ice dynamics from 108

forests preserved in the wake of the catastrophic retreating Columbia Glacier, Alaska: 109

Geological Society of America Abstracts with Programs, v. 47, p. 788. 110

Stokes, M.A., and Smiley, T.L., 1996, An Introduction to Tree-ring Dating: The University of 111

Arizona Press, Tucson. 73 p. 112

Walinsky, S.E., Prahl, F.G., Mix, A.C., Finney, B.P., Jaeger, J.M., and Rosen, G.P., 2009, 113

Distribution and composition of organic matter in surface sediments of coastal Southeast 114

Alaska: Continental Shelf Research, v. 29, p. 1565-1579. 115

Wiles, G.C., D’Arrigo, R.D., Barclay, D., Wilson, R.S., Jarvis, S.K., Vargo, L., and Frank, D., 116

2014, Surface air temperature variability reconstructed with tree rings for the Gulf of Alaska 117

over the past 1200 years: The Holocene, v. 24, p. 198-208. 118

Wilson, F.H., and Hults, C.P., 2012, Geology of the Prince William Sound and Kenai Peninsula 119

Region, Alaska: U.S.G.S. Scientific Investigation Map 3110. 120

Wilson, R., Wiles, G., D’Arrigo, R., and Zweck, C., 2007, Cycles and shifts: 1300-years of 121

multidecadal temperature variability in the Gulf of Alaska: Climate Dynamics, v. 28, p. 425-122

440. 123

METHODS

7

124

125

126

Table DR1. 14C ages; italics ages are out of stratigraphic order.

CoreCAMS/ANU

Sample Number

Material Depth in core (cm)

Depth below surface (cm)

14C age 1 sigma Calibrated age 1 sigma

EW0408-95TC 117052 Benthic 5 11 650 60 Post-Bomb Post-BombEW0408-95JC 117046 Benthic 1 77 1535 30 360 90EW0408-95TC 122482 Benthic 91 97 1830 50 600 80EW0408-95JC 117047 Benthic 101 178 1705 50 510 90EW0408-95TC 122483 Benthic 212 218 2110 45 850 110EW0408-95JC 117048 Benthic 201 278 2025 15 770 90EW0408-95JC 107732 Benthic 301 378 1935 20 690 90EW0408-95JC 117049 Benthic 401 478 2265 15 1010 110EW0408-95JC 117050 Benthic 751 828 4255 45 3310 140EW0408-95JC 42435 Planktonic 780 857 1980 55 1070 100EW0408-95JC 42433 Benthic 780 857 2320 30 1070 110EW0408-95JC 107733 Benthic 802 879 3195 20 2020 130EW0408-95JC 42436 Benthic 828 905 2375 35 1120 110EW0408-95JC 42437 Benthic 880 957 2370 35 1110 110EW0408-95JC 117639 Benthic 899 976 2355 20 1100 110EW0408-95JC 117051 Benthic 901 978 2625 15 1370 100EW0408-95JC 42438 Benthic 1298 1375 2665 40 1410 110EW0408-95JC 122481 Benthic 1720 1797 2880 20 1650 120

Table DR2. Geochemistry and magnetic properties of glacial sediments.

Glacier Latitude LongitudeFe

(ppm)Si

(ppm)Ca

(ppm) Si/Fe Si/Ca k kARM kARM/k

Shoup 61.12 -146.59 6412 454 164 0.07 2.8 0.000051 0.000055 0.866Columbia 61.00 -147.02 11707 471 362 0.04 1.3 0.000626 0.002623 3.35

Table DR3. Dendrochronology data with distance from the late Holocene moraineTree Ring Series C.E. Growth ka Kill Latitude Longitude Advance (km)12109 1602-1808 0.192 61.00859 -147.0981 0250401 1520-1767 0.233 61.01153 -146.993 1.6231001 1565-1748 0.252 61.02253 -146.9903 4.8CG05B1 954-1450 0.55 61.0306 -147.1139 6.2TC0526 1260-1413 0.587 61.0375 -146.9882 6.212709 1232-1383 0.617 61.06336 -147.004 8.7100402 1232- 1318 0.682 61.05743 -147.1003 10.201085A 950-1198 0.802 61.07634 -147.016 11.890301 797-1072 0.928 61.08682 -147.0281 12.240101 813- 1067 0.933 61.10511 -147.1103 14.442201 813-1060 0.94 61.10526 -147.1096 15CG14-48 588-1040 0.96 61.11788 -147.0485 16.3EB02-14 814-1010 0.99 61.12605 -147.0108 17.9

METHODS

8

Figure DR1. CT scan false color on the left and bulk sediment density (ρ) on the right versus 127

cmbsf in EW0408-95JC. The red box denotes the gravity-flow deposit. 128

129 Depth bsf

(cm)CT Scan - false color (g/cm3)