DOI: 10.1126/science.1239794, 997 (2013);341 Science

et al.Jonathan L. BamberPaleofluvial Mega-Canyon Beneath the Central Greenland Ice Sheet

This copy is for your personal, non-commercial use only.

clicking here.colleagues, clients, or customers by , you can order high-quality copies for yourIf you wish to distribute this article to others

  here.following the guidelines

can be obtained byPermission to republish or repurpose articles or portions of articles

  ): September 1, 2013 www.sciencemag.org (this information is current as of

The following resources related to this article are available online at

http://www.sciencemag.org/content/341/6149/997.full.htmlversion of this article at:

including high-resolution figures, can be found in the onlineUpdated information and services,

http://www.sciencemag.org/content/suppl/2013/08/29/341.6149.997.DC1.html can be found at: Supporting Online Material

http://www.sciencemag.org/content/341/6149/997.full.html#ref-list-1, 4 of which can be accessed free:cites 25 articlesThis article

registered trademark of AAAS. is aScience2013 by the American Association for the Advancement of Science; all rights reserved. The title

CopyrightAmerican Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. (print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by theScience

on

Sep

tem

ber

1, 2

013

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

11. P. S. Lykawka, T. Mukai, Icarus 192, 238–247 (2007).12. L. Dones et al., Icarus 142, 509–524 (1999).13. M. S. Tiscareno, R. Malhotra, Astron. J. 126, 3122–3131

(2003).14. M. H. M. Morais, A. Morbidelli, Icarus 160, 1–9 (2002).15. M. H. M. Morais, A. Morbidelli, Icarus 185, 29–38

(2006).16. J.-M. Petit et al., Astron. J. 142, 131 (2011).17. N. A. Kaib, R. Roskar, T. Quinn, Icarus 215, 491–507

(2011).18. J. Horner, N. Wyn Evans, Mon. Not. R. Astron. Soc. 367,

L20–L23 (2006).19. H. F. Levison, M. J. Duncan, Icarus 127, 13–32 (1997).

20. B. Gladman et al., Astron. J. 144, 23 (2012).21. J. L. Elliot et al., Astron. J. 129, 1117–1162 (2005).22. C. Shankman, B. J. Gladman, N. Kaib, J. J. Kavelaars,

J. M. Petit, Astrophys. J. Lett. 764, L2 (2013).23. A. A. Christou, Icarus 144, 1–20 (2000).

Acknowledgments: The data are available at the IAU MinorPlanet Center’s online database under MPEC K13F19. Wethank N. Kaib for making the scattering object model availableto us. We thank S. Lawler, C. Shankman, and N. Kaib forproofreading and constructive comments. M.A., S.G., andB.G. were supported by the National Sciences and EngineeringResearch Council of Canada. This work is based on observations

obtained at the Canada-France-Hawaii Telescope, operatedby the National Research Council of Canada, the InstitutNational des Sciences de l’Univers of the Centre National de laRecherche Scientifique of France, and the University of Hawaii.

Supplementary Materialswww.sciencemag.org/cgi/content/full/341/6149/994/DC1Supplementary TextFigs. S1 to S4References (24–51)

20 March 2013; accepted 25 July 201310.1126/science.1238072

Paleofluvial Mega-Canyon Beneaththe Central Greenland Ice SheetJonathan L. Bamber,1* Martin J. Siegert,1 Jennifer A. Griggs,1Shawn J. Marshall,2 Giorgio Spada3

Subglacial topography plays an important role in modulating the distribution and flow of basalwater. Where topography predates ice sheet inception, it can also reveal insights into formertectonic and geomorphological processes. Although such associations are known in Antarctica, littleconsideration has been given to them in Greenland, partly because much of the ice sheet bed isthought to be relatively flat and smooth. Here, we present evidence from ice-penetrating radardata for a 750-km-long subglacial canyon in northern Greenland that is likely to have influencedbasal water flow from the ice sheet interior to the margin. We suggest that the mega-canyonpredates ice sheet inception and will have influenced basal hydrology in Greenland over pastglacial cycles.

Greenland is an ancient craton dominatedby crystalline rocks of the Precambrianshield, composed largely ofArchaean [3100

to 2600million years (My) before the present (B.P.)]and Proterozoic (2000 to 1750 My B.P.) forma-tions, withmore limited sedimentary deposits alongparts of the continental margins (1). These rocksare resistant to glacial erosion compared with thesedimentary deposits that are thought to underliesubstantial sectors of Antarctica (2, 3). Althoughthere is evidence for some glaciation dating backto 38 My B.P. (4), the island has only been exten-sively ice-covered for no more than ~3.5 My (5),as compared with the Antarctic ice sheet that firstgrew at ~34My B.P., and has existed persistentlysince ~14MyB.P. There are, therefore, importantgeologic and glacio-morphological differencesbetween the subglacial environments of the twoice sheets.

Ice-penetrating radar (IPR) is capable of imagingthe ice-sheet base and, if collected as a series oftransects, can reveal the landscape beneath theice. In Greenland, such data have been acquiredon numerous occasions over the past 40 years(6, 7). Using these data in combination with ice-surface topography measurements, a quasilinearfeature was previously identified in northernGreenland from a combination of ice-surface

topography and IPR data, but with limited knowl-edge on its extent and configuration (8). By in-specting IPR-derived basal topography from theseand more recent data sources, we have discoveredthis bed feature to be a major subglacial canyonthat extends from almost as far south as Summitin central Greenland to the northern margin ofthe ice sheet, terminating in the fjord that drainsPetermann Gletscher (Fig. 1). The extent of thecanyon is at least ~750 km, but its southern limitmay continue further thanwe are able to identifybecause the density of IPR tracks below 74.3° Nis limited. However, two tracks at 73.8° N, 42°Wcross an over-deepening that appears likely to bethe southern extension of the canyon. There is noobvious evidence of a geological boundary in thispart of Greenland (9). The southern limit of thecanyon, however, does lie close to a minimum inthe free-air gravity anomaly and amaximum in theterrain-correlated gravity anomaly, which indi-cates crust that is isostatically thin and subject tocompressive inflow of crustal material (9), whichcould form an explanation for its genesis.

The IPR data indicate that the canyon has adepth and width of up to ~800 m and ~10 km,respectively, toward the coast and a slightly re-duced width and substantially shallower depth of~200m further south (Fig. 2 and figs. S1 and S2).The valley width-to-height ratio (used as a mea-sure of river channel morphology and formation)for the profile in Fig. 2C is 30, which is indicativeof a broad, shallow channel, subject to limitedstream erosion (10, 11). The ratio for the profilein Fig. 2A is ~0.5, suggesting linear stream in-

cision was important in the development of thechannel at this location, with no evidence of gla-cial erosion being the dominant process (12). In-deed, none of the profiles are typical of glaciallyeroded valleys (fig. S2). The canyon follows a

1Bristol Glaciology Centre, University of Bristol, BS8 1SS Bristol,UK. 2Department of Geography, University of Calgary, T2N 1N4Calgary, Canada. 3Dipartimento di Scienze di Base e Fondamenti,Urbino University, 61029 Urbino, Italy.

*Corresponding author. E-mail: j.bamber@bristol.ac.uk

Fig. 1. Bedelevation [between–900and900mabove sea level (asl)] for northern Greenland.The area plotted is indicated by the red box in theinset. Airborne ice penetrating radar flight lines areshown in gray, and the three bed profiles plotted inFig. 2 are shown by the solid blue (Fig. 2A), black(Fig. 2B), and red (Fig. 2C) lines. The mauve lineshows the approximate location of the groundingline on Petermann Gletscher. Other places discussedin the text are also labeled: NGRIP, the onset areaof the North East Greenland Ice Stream (NEGIS),and the camp near the highest point of the ice sheet,Summit. The red ellipse marks the approximatelocation of a marked minimum in crustal thickness,and hence maximum in geothermal heat flux, iden-tified from gravity data (9).

www.sciencemag.org SCIENCE VOL 341 30 AUGUST 2013 997

REPORTS

on

Sep

tem

ber

1, 2

013

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

meandering path more typical of a large riversystem. The dimensions of the Greenland canyonare comparable with parts of the Grand Canyon

in the United States in terms of width and lengthand, in its deepest region, are about half of theGrand Canyon’s depth.

To assess the canyon’s influence on the flowof subglacial water, and the extent to which it wasa mega-channel for surface water, we calculatedthe hydraulic potential for both the present-day,ice-free, and last glacial maximum (LGM) con-figurations of the Greenland ice sheet (Fig. 3),accounting where necessary for isostatically com-pensated ice-free bedrock. We used the methodsdescribed in Shreve (1972) (13), which have beencommonly applied to determine subglacial hydro-logical pathways (14). Because of its size, thecanyon provides a hydraulic pathway in all casestested, suggesting that it has influenced basalwater flow in northern Greenland since ice sheetinception and, indeed, before this. For the ice-free(preglacial) state (Fig. 3A and fig. S3B), water ispredicted to flow south-to-north along the ma-jority of the canyon. The isostatically compensatedchannel gradient, without the ice sheet load, is0.3mkm−1, which is similar to the southern 500 kmof the Colorado River and around three timeslarger than the average for the Mississippi River.Regionswherewater diverts away from the canyon(such as at 79.3° N, 48° W) are coincident withlimited IPR data coverage (Fig. 1), implying poorresolution of the bed topography in such places.Hence, we believe water will likely have routed

Fig. 2. Ice-penetrating radargramprofiles across the canyon at three locations. The locations areindicated in Fig. 1 by (A) the solid blue line, (B) the black line, and (C) the red line. There is anexaggeration in the vertical by a factor of 26. The depth of the canyon in meters below sea level isindicated in each profile alongside internal reflections within the ice and the bed return. North isroughly into the page.

Fig. 3. Subaerial and subglacial hydraulic potential flowpaths. (A) Anisostatically compensated bed without an ice sheet. (B) An ice sheet con-figuration at LGM. (C) The present-day ice sheet and bed configuration.

H, Helheim; P, Petermann; R, Ryder. The bed elevations shown are all forpresent day. The isostatically compensated bed elevation is illustrated infig. S3B.

30 AUGUST 2013 VOL 341 SCIENCE www.sciencemag.org998

REPORTS

on

Sep

tem

ber

1, 2

013

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

continually along this feature before extensive icesheet formation ~3.5 My B.P. (5).

In full glacial conditions [based on LGM(supplementary materials)], the flow of water isinfluenced by the ice overburden pressure, whichforces water out of the canyon in parts of theupper Petermann catchment and along the direc-tion of ice flow (Fig. 3B). Nonetheless, the canyonremains a conduit for the flow of basal watertoward the coast even under LGM conditions, es-pecially northward of ~79° N. For the present ice-sheet configuration, which is small and short-livedcompared with its expanded full glacial state,steeper ice surface slopes (relative to LGM) forcemore water from the canyon to the west at certainlocations (Fig. 3C and fig. S3A). In all cases,above ~76° N and within the entire length of thePetermann catchment, the canyon exerts a controlon basal water flow. For ~200 km, it provides anuninterrupted hydraulic pathway (Fig. 3 and fig.S3A) that ends at the terminus of PetermannGletscher. However, although extensive parts ofnorthern Greenland have been identified as havinga wet bed (fig. S4) (15), water currently may notbe ubiquitous throughout its length. Under fullglacial conditions, the ice sheet was larger andthicker than at present day (4), and therefore, agreater proportion of the bed was likely melt-ing during the longer lasting LGM ice sheetconfiguration.

The lack of substantial subglacial lakes in themain Greenland ice sheet may be partly explainedby generally steeper ice surface slopes as com-pared with those of Antarctica, but subglacialwater must be evacuated by somemeans or it willpool. The subglacial canyon provides an efficientpathway for this water routing over an otherwiserelatively flat bed (Fig. 3C). Given that waterbeneath former ice sheets in Greenlandwas likelyto have been influenced by the canyon, it seemsprobable that conditions have never beenwell suitedto substantial long-term basal water storage inthis sector of Greenland.

The floating ice shelf in front of PetermannGletscher has been found to possess sub–ice-

shelf channels 1 to 2 km wide and 200 to 400 mdeep incised upward into the ice (16). These chan-nels are associated with high basal melt rates,exceeding 30 m year−1. Their existence has beenattributed entirely to the action of the ocean (16).We believe the efficient routing of basal water viathe canyon to the northern ice sheet margin is alsoimportant in explaining these features. Analysisof IPR echo-strength data has identified extensiveregions of the northern half of the Greenland icesheet that have water at the bed (15). Subglacialmelt rates in the vicinity of the North GreenlandIce Core Project (NGRIP) drill site have beenestimated to exceed 1 cm year−1 (17) and are ashigh as 15 cm year−1 near the onset of the North-east Greenland Ice Stream (18). A study exam-ining crustal thickness and geothermal heat flowin Greenland found a minimum in the former (andhence, a maximum in the latter) close to the north-ern limit of the canyon (Fig. 1) (9). Thus, sub-stantial volumes of subglacial meltwater aregenerated at the bed, and additional surface melt-water may also penetrate to the bed near the coast(19). An effectivemeans bywhich this basal wateris routed to the ice-sheet margin must exist, elsesubglacial lakes would form. Basal water gen-erated in the Petermann catchment is highly likelyto be routed through the Canyon to the ice sheetterminus (Fig. 3C). Thus, basal water is likely tobe delivered at a point source to the ice-shelf cav-ity, which is likely to influence the local subshelfwater circulation (20). Similar sub-shelf channelshave been identified underneath Pine Islandglacier and elsewhere in Antarctica (21), and wesuggest that well-organized subglacial meltwaterflow is likely to play a similar role in the de-velopment of these features.

References and Notes1. N. Henriksen, A. K. Higgins, F. Kalsbeek, T. C. R. Pulvertaft,

Geol. Surv. Denmark Greenland Copenhagen 18, 126 (2009).2. D. J. Drewry, Tectonophysics 36, 301 (1976).3. M. Studinger et al., Geophys. Res. Lett. 28, 3493–3496

(2001).4. R. B. Alley et al., Quat. Sci. Rev. 29, 1728–1756 (2010).5. G. Bartoli et al., Earth Planet. Sci. Lett. 237, 33–44 (2005).

6. J. L. Bamber et al., Cryosphere 7, 499–510 (2013).7. Data set for (6) is available at https://docs.google.com/

file/d/0BylqEEvDu_qtWWdIYTFVcVpkd2s/edit?usp=sharing).

8. S. Ekholm, K. Keller, J. L. Bamber, S. P. Gogineni,Geophys. Res. Lett. 25, 3623–3626 (1998).

9. A. Braun, H. R. Kim, B. Csatho, R. R. B. von Frese, EarthPlanet. Sci. Lett. 262, 138–158 (2007).

10. W. B. Bull, L. D. McFadden, in Geomorphology in AridRegions, D. O. Doehring, Ed. (State University of NewYork, New York, 1977), pp. 115–139.

11. The valley height-to-width ratio is Vf = 2Vfw/[(Eld – Esc) +(Erd – Esc)], where Vfw is the valley floor width, Esc is thevalley floor elevation, Eld is the elevation of the left riverdivide, and Erd is elevation of the right river divide.

12. M. A. Summerfield, Global Geomorphology. AnIntroduction to the Study of Landforms (Longman,Harlow, UK, 1991).

13. R. Shreve, J. Glaciol. 11, 205 (1972).14. G. W. Evatt, A. C. Fowler, C. D. Clark, N. R. J. Hulton,

Phil. Trans. R. Soc. A 364, 1769–1794 (2006).15. G. K. A. Oswald, S. P. Gogineni, IEEE Trans. Geosci.

Rem. Sens. 50, 585–592 (2012).16. E. Rignot, K. Steffen, Geophys. Res. Lett. 35, L02503

(2008).17. D. Dahl-Jensen, N. Gundestrup, S. P. Gogineni,

H. Miller, Ann. Glaciol. 37, 207–212 (2003).18. M. Fahnestock, W. Abdalati, I. Joughin, J. Brozena,

P. Gogineni, Science 294, 2338–2342 (2001).19. I. Joughin, S. Tulaczyk, M. Fahnestock, R. Kwok, Science

274, 228–230 (1996).20. F. Straneo et al., Nat. Geosci. 4, 322–327 (2011).21. D. G. Vaughan et al., J. Geophys. Res. 117, F03012 (2012).

Acknowledgments: J.L.B., J.A.G., and G.S. were supported byfunding from the ice2sea program from the European Union7th Framework Programme, grant 226375. This work is ice2seacontribution number 158. J.L.B. and M.J.S. were supported byNatural Environment Research Council grant NE/H021078/1.We acknowledge the use of data products from CReSISgenerated with support from NSF grant ANT-0424589, NASAgrant NNX10AT68G, and the NASA Operation IceBridgeproject. J.L.B. wrote the main text with substantial contribu-tions from M.J.S. and input from all authors. J.A.G. carried outthe hydraulic potential calculations. G.S. calculated theisostatically adjusted bed, and S.J.M. produced the LGM icesheet configuration.

Supplementary Materialswww.sciencemag.org/cgi/content/full/341/6149/997/DC1Materials and MethodsFigs. S1 to S5References (22–30)

29 April 2013; accepted 31 July 201310.1126/science.1239794

Social Learning of Migratory PerformanceThomas Mueller,1,2,3* Robert B. O’Hara,2 Sarah J. Converse,4 Richard P. Urbanek,5 William F. Fagan1

Successful bird migration can depend on individual learning, social learning, and innate navigationprograms. Using 8 years of data on migrating whooping cranes, we were able to partitiongenetic and socially learned aspects of migration. Specifically, we analyzed data from a reintroducedpopulation wherein all birds were captive bred and artificially trained by ultralight aircraft ontheir first lifetime migration. For subsequent migrations, in which birds fly individually or ingroups but without ultralight escort, we found evidence of long-term social learning, but no effectof genetic relatedness on migratory performance. Social learning from older birds reduceddeviations from a straight-line path, with 7 years of experience yielding a 38% improvementin migratory accuracy.

Mechanisms underlying the complex phe-nomenon of animal migration have beenparticularly well studied in birds (1–5).

In some taxa, individuals may migrate alone, un-aided by conspecifics and relying instead mostlyon endogenous, genetically inherited navigation

programs (6). In other species, innate programsalone are not sufficient, and experiential learningis critical to successful navigation, as adult ani-mals often have markedly better navigational ca-pabilities than juveniles (7–9). Information transferfrom more experienced individuals to inexperi-enced ones can be essential to navigational success,

1Department of Biology, University of Maryland, College Park,MD 20742, USA. 2Biodiversity and Climate Research Centre(BiK-F), Senckenberg Gesellschaft für Naturforschung, and GoetheUniversität Frankfurt, Senckenberganlage 25, 60325 Frankfurt(Main), Germany. 3Smithsonian Conservation Biology Institute,National Zoological Park, 1500 Remount Road, Front Royal, VA22630, USA. 4U.S. Geological Survey, Patuxent Wildlife ResearchCenter, 12100Beech Forest Road, Laurel, MD 20708,USA. 5U.S.Fish and Wildlife Service, Necedah National Wildlife Refuge,N11385 Headquarters Road, Necedah, WI 54646, USA.

*Corresponding author. E-mail: muellert@gmail.com

www.sciencemag.org SCIENCE VOL 341 30 AUGUST 2013 999

REPORTS

on

Sep

tem

ber

1, 2

013

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om