Volume 44, Issue 6 p. 2810-2816
Research Letter
Free Access

The projected demise of Barnes Ice Cap: Evidence of an unusually warm 21st century Arctic

A. Gilbert

Corresponding Author

A. Gilbert

Department of Earth Sciences, Simon Fraser University, Burnaby, British Columbia, Canada

Correspondence to: A. Gilbert,

[email protected]

Search for more papers by this author
G. E. Flowers

G. E. Flowers

Department of Earth Sciences, Simon Fraser University, Burnaby, British Columbia, Canada

Search for more papers by this author
G. H. Miller

G. H. Miller

INSTAAR and Department of Geological Sciences, University of Colorado Boulder, Boulder, Colorado, USA

Search for more papers by this author
K. A. Refsnider

K. A. Refsnider

INSTAAR and Department of Geological Sciences, University of Colorado Boulder, Boulder, Colorado, USA

Search for more papers by this author
N. E. Young

N. E. Young

Lamont-Doherty Earth Observatory, Columbia University, Palisades, New York, USA

Search for more papers by this author
V. Radić

V. Radić

Department of Earth, Ocean and Atmospheric Sciences, University of British Columbia, Vancouver, British Columbia, Canada

Search for more papers by this author
First published: 20 March 2017
Citations: 11

Abstract

As a remnant of the Laurentide Ice Sheet, Barnes Ice Cap owes its existence and present form in part to the climate of the last glacial period. The ice cap has been sustained in the present interglacial climate by its own topography through the mass balance-elevation feedback. A coupled mass balance and ice-flow model, forced by Coupled Model Intercomparison Project Phase 5 climate model output, projects that the current ice cap will likely disappear in the next 300 years. For greenhouse gas Representative Concentration Pathways of +2.6 to +8.5 Wm−2, the projected ice-cap survival times range from 150 to 530 years. Measured concentrations of cosmogenic radionuclides 10Be, 26Al, and 14C at sites exposed near the ice-cap margin suggest the pending disappearance of Barnes Ice Cap is very unusual in the last million years. The data and models together point to an exceptionally warm 21st century Arctic climate.

Key Points

  • Barnes Ice Cap will probably disappear in the next 300 years
  • The pending disappearance of Barnes Ice Cap is very unusual in the last million years
  • The 21st century Arctic climate is exceptionally warm in the context of the last million years

1 Introduction

At the Plio-Pleistocene transition, 2.6 Ma, global climate cooled and Northern Hemisphere ice sheets grew, expanding and contracting initially at 41 ka periodicities, then shifting to 100 ka periodicities at ~800 ka, during the mid-Pleistocene transition [Shackleton and Opdyke, 1973; Lambeck et al., 2002; Sosdian and Rosenthal, 2009]. In contrast, projections for the next centuries predict atmospheric warmth that may not have been reached since the Pliocene [International Panel on Climate Change (IPCC), 2013]. Arctic ice caps are already experiencing melt rates unprecedented in several millennia [Fisher et al., 2012] with a strong increasing trend since 2005 [Sharp et al., 2011]. Our study aims to put the modern and projected future warming into a longer-term perspective by constraining the Quaternary history of ice sheet change in inland Baffin Island.

The Laurentide Ice Sheet, the largest of the Pleistocene ice sheets, covered much of North America during the last glaciation before it began to retreat after ~21 ka. By ~5 ka its final remnant was an elongate ice cap on Baffin Island, Arctic Canada, that eventually stabilized at ~2 ka as Barnes Ice Cap (Figure 1) [Dyke et al., 2003; Briner et al., 2009]. The current ice cap rests on a gently sloping landscape at elevations too low to allow an ice cap to nucleate under Holocene climates [Gilbert et al., 2016]. The narrow elevation range of its accumulation zone (700–1100 m above sea level (asl)) makes the ice cap more sensitive to climate change than any other glacier on Baffin Island [Gardner et al., 2012], with a slight increase in the equilibrium line altitude causing a marked reduction of the accumulation area. Barnes Ice Cap was probably close to equilibrium at the end of the 19th century but has experienced a negative mass balance trend since [Gilbert et al., 2016], with an acceleration in this trend since the mid-1990s [Abdalati et al., 2004; Sneed et al., 2008; Gardner et al., 2012; Gilbert et al., 2016] causing the disappearance of its accumulation area since about 2010 [Gray et al., 2015; Gilbert et al., 2016], and surface lowering at all elevations (see Figure S5 in the supporting information) on the south dome profile [Holdsworth, 1973; Hooke et al., 1987].

Details are in the caption following the image
Map of Barnes Ice Cap and location on Baffin Island, Canada (inset, blue circle). The labeled red dots show CRN sample locations (Table 1). The labeled green dots show other sample locations (Table 2). Surface elevation is contoured in black; bedrock elevation is contoured in color.

In this study, we use a thermomechanical ice-flow model coupled to a surface mass balance model to investigate the likely disappearance of Barnes Ice Cap under a range of future emissions scenarios. We used an ensemble of general circulation models (GCMs) up to 2300. Cosmogenic radionuclides (CRNs) in exposed bedrock were utilized to evaluate how frequently the ice cap disappeared in previous interglaciations. The future projections compared with past fluctuations of the ice cap aid in evaluating whether the predicted warming has any precedent in the past 1 to 2 Ma.

2 Models and Data

2.1 Ice Cap Model

The ice-flow model is based on the Stokes equations coupled to the energy conservation equation (using an enthalpy method [Aschwanden et al., 2012]) and the equation governing free-surface evolution [Gilbert et al., 2016]. This model is implemented and solved by using the finite element code Elmer/Ice [Gagliardini et al., 2013], which is based on the Elmer open-source multiphysics package (see http://elmerice.elmerfem.org for details). The model setup for Barnes Ice Cap, as well as the strategy to establish an initial state in 1960, is described in detail in Gilbert et al. [2016]. The initial state is obtained by assuming a steady state thermal regime calibrated to measured borehole temperatures and basal drag coefficients computed from inversion of surface velocities. We assume a constant but nonuniform basal drag coefficient. The ice-flow model runs at a biannual time step forced by biannual surface temperatures and mass balance. It also includes evolution and rheology of a basal layer of Pleistocene ice [see Gilbert et al., 2016].

Mass balance is computed from daily air temperatures and a constant precipitation rate, which has been shown to be a valid approximation for simulating the 1960–2010 mass balance [Gilbert et al., 2016]. We also perform simulations with temperature-dependent precipitation rates in future projections. The model employs a degree-day formulation and was calibrated for the 1960–2010 period by using multiple data sets including geodetic and in situ surface mass balance [Gilbert et al., 2016].

2.2 Climate Scenarios

We selected five different GCMs from the Coupled Model Intercomparison Project Phase 5 (CMIP5) intercomparison project [IPCC, 2013] and three different Representative Concentration Pathways (RCPs; see supporting information), from which we extract daily 2 m air temperatures at the grid cell that includes Barnes Ice Cap. We selected RCP 2.6, 4.5, and 8.5, which assume that greenhouse gas emissions will reach a maximum in 2020, 2040, and 2100, respectively. The temperature time series are downscaled to the resolution of the ice-flow model by using a constant lapse rate. The GCMs have been selected according to their skill in reproducing interannual climate variability over the second half of the 20th Century in North America [Radić and Clarke, 2011]. Biases between modeled temperatures and local meteorological observations around Barnes Ice Cap are corrected by using monthly data for a 9 year overlap between 2006 and 2015. The local measurements are derived from meteorological observations at Dewar Lake (inland) and Clyde River (coastal), both located about 200 km from the ice cap [Gilbert et al., 2016].

2.3 Cosmogenic Radionuclides

Glacier ice shields the underlying rock from cosmic rays, whereas deglaciation results in the exposure of bedrock surfaces to cosmic ray bombardment and the in situ production of CRNs [Gosse and Phillips, 2001]. We measured the concentrations of CRNs 10Be, 26Al, and 14C, with half-lives of 1.4 Ma, 0.7 Ma, and 5.7 ka, respectively. The short half-life of 14C ensures that only the most recent deglacial period is represented; any 14C that accumulated during previous interglaciations would have decayed to undetectable concentrations after ~30 ka of ice cover. In contrast, the long half-lives of 26Al and 10Be record the integrated exposure/burial/erosion history of rock surfaces through the Quaternary. The difference between the measured 26Al/10Be ratio and the constant exposure production ratio reflects the duration of sample burial, during which CRN inventories decay at different rates. CRN ages (±1σ) are calculated by using the online CRONUS calculator, “St” scaling [Lal, 1991; Stone, 2000]; for in situ 14C ages we use a globally calibrated production rate that includes calibration sites from Baffin Bay [Young et al., 2014].

We assume that the patterns of previous Laurentide deglaciations were similar to the last deglaciation [Dyke et al., 2003], with the final ice remnant occupying the same space as the current Barnes Ice Cap. With that assumption, the extent of prior deglaciations is recorded by CRN inventories preserved in rock surfaces at and around the current ice-cap margin. If most deglaciations resulted in remnant ice as small or smaller than the contemporary ice cap, then rock surfaces just beyond the ice-cap margin that were not extensively eroded during recent glaciations will have significant accumulations of CRNs, but the 26Al/10Be ratio will be below the production ratio of 6.75 due to prolonged burial during glacial cycles. Conversely, if most deglaciations concluded with remnant ice larger than the contemporary ice cap, then the CRN inventories in our bedrock locations would be much smaller and with far greater 26Al/10Be disequilibrium due to less frequent exposure and CRN production. With these assumptions in mind, we collected the upper 2 cm of flat-lying exposed rock surfaces from six sites situated between 1 and 20 km from Barnes Ice Cap (red dots in Figure 1) for CRN measurements. Quartz was isolated from bulk rock, from which Be and Al were isolated by column chromatography at the University of Colorado, following procedures modified from Kohl and Nishiizumi [1992]; 14C was isolated from quartz separates at Lamont-Doherty Earth Observatory following procedures outlined in Goehring et al. [2014]. Isotope concentrations were measured by accelerator mass spectrometry at the Center for Accelerator Mass Spectrometry (CAMS) at Lawrence Livermore National Laboratory.

To interpret the 26Al/10Be disequilibrium in our samples, we use a simple model in which CRN inventories that had accumulated during the Pliocene subsequently evolve during the Pleistocene through burial, erosion, and exposure corresponding to a 41 ka periodicity until 800 ka and subsequently to a 100 ka periodicity. Preglacial erosion rates, the duration of exposure during interglaciations, and glacial erosion rates are prescribed iteratively; CRN production occurs only during interglaciations. Model parameters specific to cosmogenic nuclide systematics follow those used in the CRONUS-Earth online age calculator with St production rate scaling [Balco et al., 2008; http:// hess.ess.washington.edu/]. Although an infinite number of combinations of exposure, burial, and erosion can explain the measured CRN inventories in many of our samples, the known timing of the cyclical glaciation-interglaciation (burial-exposure) histories of these samples significantly limits the range of geologically reasonable scenarios.

3 Results

3.1 Modeling the Future Evolution of Barnes Ice Cap

All simulations of the evolution of the ice cap begin in 1960, allowing for a 55 year calibration period during which the model is forced by data from the Dewar Lakes and Clyde River meteorological stations and constrained by in situ and geodetic measurements. Mass balance prior to 1960 (Figure 2) was estimated by Gilbert et al. [2016] by using the Climatic Research Unit data set from West Greenland and assuming no change in ice-cap surface elevation. This calibration period contains a significant warming trend (see supporting information), which allows us to validate the model sensitivity to increasing air temperature.

Details are in the caption following the image
Future projections of Barnes Ice Cap. (a) Cumulative and (b) annual mass change modeled for the different RCP scenarios. (c–e) Margin positions for the three RCP scenarios at the years indicated in the legend. The shaded areas show standard deviation from the ensemble of GCMs used in the projections. The red dots (Figure 2a) and lines (Figure 2b) are geodetic measurements of mass change. The dashed lines (Figure 2a) are averaged mass evolution assuming a precipitation increase of 8% K−1 [IPCC, 2013]. The vertical black line is the beginning of the dynamically coupled simulation (1960), before which mass balance is computed assuming constant surface elevation (1960 DEM).

All future projections lead to the disappearance of Barnes Ice Cap in less than ~500 years, even in response to the lowest modeled temperatures (Geophysical Fluid Dynamics Laboratory, RCP2.6) and including a precipitation increase of 8% K−1, the average value for GCM projections in the Arctic [IPCC, 2013]. For the moderate scenario, which assumes that greenhouse gas emissions peak around the year 2040 (RCP 4.5), the ice cap is projected to disappear in less than 300 years. Maximum mass-change rates reach −8 Gt yr−1 in 2060, −10 Gt yr−1 in 2120, and −20 Gt yr−1 in 2100 for RCP 2.6, RCP 4.5, and RCP 8.6, respectively. Sensitivity tests to mass balance model parameters and basal sliding are shown in the supporting information. Adopting the most favorable degree-day parameters, assuming 8% K−1 precipitation increase, and disabling sliding extend the lifetime of the ice cap by ~30, 100, and 400 years, respectively, for RCP 8.5, 4.5, and 2.6. The sensitivity of these results to uncertain model parameters changes markedly with climate forcing. RCP8.5 simulations exhibit very low sensitivity to model parameters.

The surface hydrology of Baffin Island in the vicinity of Barnes Ice Cap is interrupted by the ice itself, with large lakes dammed along the ice cap's eastern margin (Figure 1). Topography dictates that the drainage of the two largest lakes, Conn and Bieler (Figure 1), will occur as the ice dams thin and eventually fail; both lakes will drain, perhaps catastrophically, southwest into Foxe Basin. Thermal erosion at the base of the ice by these drainage events, combined with the possible collapse of a thin ice roof over the flow path, may hasten the disappearance of the ice cap. All climate scenarios employed here lead to a similar retreat of the ice margin until 2100 (Figure 2), after which the retreat patterns diverge. Simulations driven by the RCP 4.5 scenario indicate that Conn Lake will most likely drain around 2150 and Bieler Lake between 2200 and 2300.

3.2 Constraining Past Deglaciations

Although the decaying Laurentide Ice Sheet lost mass rapidly through both calving and melting during the early Holocene, the rate of ice margin retreat slowed after 5 ka as the remnant ice cap became land-based and Northern Hemisphere summer insolation decreased [Dyke et al., 2003]. We constrain the time when Barnes Ice Cap attained its 20th century dimensions by using four dating methods (Tables 1 and 2). Striated bedrock in the valley floor 20 km W of Barnes Ice Cap (B204; Figure 1) has a 10Be exposure age of 3210 ± 335 years and 26Al/10Be (6.6 ± 1.7) that is indistinguishable from the production ratio (Table 1). Striated bedrock on a summit 1.7 km E of the ice cap (B158; Figure 1) has an in situ 14C age of 2757 ± 403 years, and a nearby large (3 × 2 × 2 m) boulder (B164), 40 m lower than B158 and ~300 m closer to the ice cap, has an in situ 14C age of 1014 ± 315 years, a reasonable age difference given the lower elevation and slow rate of ice retreat after 2 ka. The in situ 14C ages are maximum ages due to muogenic 14C production once the ice cover thins below ~30 m and via spallation when ice cover is <6 m, so some of the 14C inventory was formed before deglaciation if ice thinned slowly over our collection sites. We obtained complementary ages from optically stimulated luminescence in quartz sand sampled in the cutbank of a deglacial terrace ~10 km west of Barnes Ice Cap (2760 ± 340 years; Table 2) and lichenometric exposure ages of ~2.5 ka on Rhizocarpon thalli from sites 6 km NW [Andrews and Webber, 1964] and 1.5 km NE of the ice cap (Figure 1 and Table 2). The in situ 14C age of 6.6 ka for bedrock sample B160 (near B158) is spurious, conflicting with all other ages and with the pattern of deglaciation which documents a much larger ice cap ~6 ka [Dyke et al., 2003].

Table 1. Sample Coordinates, Characteristics, and CRN Agesa
Sample ID Latitude Longitude Material Elevation (m asl) 14C Exposure Age (yr ± 1σ) 10Be Apparent Age (ka ± 1σ)b 26Al/10Be ±1σ
M09-B158r 70.41831 −73.67100 Bedrock 665 2757 ± 403 3.3 ± 0.3 ka 4.93 ± 0.67
M09-B164r 70.42320 −73.69382 Erratic 640 1014 ± 315 4.2 ± 0.6 ka 2.65 ± 0.66
M09-B160r 70.41621 −73.67486 Bedrock 676 6629 ± 796 5.5 ± 0.3 ka 5.43 ± 0.48
M09-B204r 70.25797 −75.24011 Bedrock 198 3.1 ± 0.3 ka 6.61 ± 1.70
M10-B019r 70.49137 −74.83885 Bedrock 615 39.7 ± 3.6 ka 4.94 ± 0.30
M10-B015r 70.57971 −74.16063 Bedrock 904 27.1 ± 2.5 ka 5.09 ± 0.56
  • a See Figure 1 for sample locations.
  • b 10Be ages calculated by using the CRONUS calculator version 2.4 and Lal [1991] and Stone [2000] production rates.
Table 2. Sample Coordinates, Characteristics, and Ages for Other Samples Constraining Ice Cap Deglaciationa
Sample ID Latitude Longitude Elev (m asl) Age (yr ± 1σ) Method
M09-B203o 70.27008 −76.17117 197 2760 ± 340 Optically stimulated luminescence (quartz sand)
King Moraines 70.44079 −74.89648 290 2500 ± 500 Lichenometry
NE BIC 70.43108 −73.74570 660 2500 ± 500 Lichenometry
  • a See Figure 1 for sample locations.

The frequency with which prior deglaciations resulted in a residual Laurentide Ice Sheet similar in size to or smaller than Barnes Ice Cap is best constrained by 26Al and 10Be in sample B158 because it is a bedrock and has coherent 10Be and 14C ages. Bedrock exposed continuously results in 26Al/10Be at the production ratio, 6.75, whereas samples buried for more than ~100 ka after having accumulated some CRNs will have lower 26Al/10Be due to the shorter half-life of 26Al. The 26Al/10Be ratio in B158 (4.93) is well below the production ratio, but the 10Be inventory is very low (Table S1 in the supporting information), particularly since the 14C inventory in B158 (Table S2) demands at least 2 ka of 26Al and 10Be formation at the production ratio in the Holocene.

The most complete Quaternary deglaciations likely occurred during two “superinterglacials.” Marine oxygen isotope stage (MIS) 5e (~125 ka) was exceptionally warm because of favorable tilt and precession orbital configurations; all of the Canadian Arctic Archipelago is thought to have been ice-free then [Koerner, 1989]. The other extreme interglacial, MIS 11c (~400 ka), although not quite as warm as MIS 5e, was twice as long, with evidence of persistent warmth in the Arctic [Melles et al., 2012] and elsewhere [Masson-Delmotte et al., 2010].

Given at least 2 ka of exposure in the late Holocene, we postulate similar exposure in the warm interglaciations MIS 5e and 11c. Reconciling the low 10Be inventory coupled with the low 26Al/10Be ratio in B158 and ~2 ka of exposure in MIS 1, 5e, and 11c requires a long preglacial (>2.6 Ma ago) exposure, followed by continuous burial beneath ice until MIS 11c, with average glacial erosion rates between 15 and 30 m Ma−1 (Figure 3). This history of long burial with rare exposure is supported by the similarly low 10Be inventory (5.5 ka) in bedrock sample B160 (26Al/10Be: 5.4) assuming lower erosion rates on average and boulder sample B164 (10Be: 4.2 ka) with exceptionally low 26Al/10Be (2.65), although there is a much wider range of plausible burial and exposure histories for a boulder than for bedrock. Additional support for long burial and only limited erosion comes from two summit bedrock samples collected north of Barnes Ice Cap (B015 and B019; Figure 1) that both have 26Al/10Be (~5.0; Table 1) well below the production ratio and similar to B158. Both of the summits have considerably higher 10Be inventories than B158, consistent with earlier and longer deglaciation throughout the Quaternary; higher summit elevations farther from the margin of Barnes Ice Cap would have deglaciated earlier and more frequently than our lower elevation ice-proximal sample sites.

Details are in the caption following the image
10Be-26Al two-nuclide evolution plot; the ellipses show ±1σ analytical uncertainties for four bedrock and one large erratic collected close to the BIC margin (Figure 1). The y axis values are normalized to the 26Al/10Be surface production ratio, and the x axis values are normalized to sea level-high latitude production rates. The solid black lines show theoretical CRN inventories for no erosion (upper) and steady erosion (lower) scenarios; data falling below those lines have experienced significant burial. The gray dashed lines are minimum burial (horizontal) and exposure (vertical) isochrons. The two jagged curves show the cosmogenic-nuclide evolution for the most reasonable burial-exposure-erosion scenarios for samples M09-B158r (blue) and M09-B164r (gray), each of which includes the onset of glacial-interglacial cycles at 2.7 Ma, 40 ka glacial cycles prior to 1 Ma, and 100 ka glacial cycles subsequently. The open stars show the endpoint for each scenario. These scenarios prescribe glacial erosion rates of 0.4 m and 0.3 m for 41 and 100 ka glacial cycles, respectively (M09-B158r), and 0.2 m for both 41 and 100 ka glacial cycles (M09-B164r) and sufficient ice cover to attenuate all CRN production during all interglacials except MIS 11, 5e, and 1 which are specified to include 700 years of exposure.

4 Conclusions

The Laurentide Ice Sheet retreated throughout the Holocene, until finally reaching an approximate mass balance equilibrium ~2 ka and stabilizing as Barnes Ice Cap. After 2000 years of little change in its dimensions, recent observations show that the ice cap is now losing mass at all elevations, despite the continued decrease in summer insolation [Berger and Loutre, 1991]. Using CMIP5 future climate projections to drive a coupled model of surface mass balance and ice flow, we project that Barnes Ice Cap will disappear within the next 300 years. CRN data (14C, 10Be, and 26Al) in bedrock <2 km from the ice margin indicate that only in the three warmest interglaciations of the past 2.6 Ma has Barnes Ice Cap been as small as it is presently. The combination of CRN evidence suggesting that earlier deglaciations rarely approached the current ice-cap dimensions, and glacier modeling indicating that the ice cap is likely to disappear in the near future despite summer insolation near minimum levels, adds compelling evidence to the argument that the current level of greenhouse gas forcing is exceptional and may result in an Arctic climate not seen since the Pliocene.

Acknowledgments

We are grateful for financial support provided by the Natural Sciences and Engineering Research Council of Canada, the Simon Fraser University, and the U.S. National Science Foundation. This research was enabled in part by WestGrid (www.westgrid.ca) and Compute Canada/Calcul Canada (www.computecanada.ca). We acknowledge the World Climate Research Programme's Working Group on Coupled Modeling, which is responsible for CMIP, and we thank the climate modeling groups for producing and making available their model output (available at http://pcmdi9.llnl.gov/). For CMIP the U.S. Department of Energy's Program for Climate Model Diagnosis and Intercomparison provides coordinating support and led development of software infrastructure in partnership with the Global Organization for Earth System Science Portals. We thank NASA's Operation IceBridge Project that acquired bedrock topography data used in this paper (available at https://nsidc.org/icebridge/portal), with special thanks to W. Abdalati. We thank S. L. Forman at Baylor University for the OSL date, T. Guilderson at LLNL-CAMS for careful 14C measurements, J. Lamp for help with 14C sample preparation, and S. Crump and S. Pendleton for input on the manuscript. CRN sample collections and measurements were supported by NSF grants ARC 1204096 and ARC 0454662.