Basal Drainage System Response to Increasing Surface Melt on the Greenland Ice Sheet

See allHide authors and affiliations

Science  16 Aug 2013:
Vol. 341, Issue 6147, pp. 777-779
DOI: 10.1126/science.1235905

Draining Through Ice

Water formed by surface melting of the Greenland Ice Sheet is transferred rapidly to the underlying bedrock, but how the water is then dispersed is less clear. This question is important because how the ice-rock interface is lubricated affects how fast the ice sheet moves. Existing conceptual models are based on observations of mountain glaciers, but Meierbachtol et al. (p. 777; see the Perspective by Lüthi) now show that those ideas may not be applicable to the Greenland Ice Sheet. Measuring water pressures in a transect of 23 boreholes revealed that drainage structures differ between the edge, where large melt channels form, and further inland, where more distributed pathways are found.


Surface meltwater reaching the bed of the Greenland ice sheet imparts a fundamental control on basal motion. Sliding speed depends on ice/bed coupling, dictated by the configuration and pressure of the hydrologic drainage system. In situ observations in a four-site transect containing 23 boreholes drilled to Greenland’s bed reveal basal water pressures unfavorable to water-draining conduit development extending inland beneath deep ice. This finding is supported by numerical analysis based on realistic ice sheet geometry. Slow meltback of ice walls limits conduit growth, inhibiting their capacity to transport increased discharge. Key aspects of current conceptual models for Greenland basal hydrology, derived primarily from the study of mountain glaciers, appear to be limited to a portion of the ablation zone near the ice sheet margin.

Measurements on the Greenland ice sheet (GIS) show widespread meltwater forcing on velocity, affecting marine (1, 2) and terrestrially terminating regions (210) many tens of kilometers inland from the margin (4, 5, 10). During the summer melt season, velocities commonly increase up to 100% or more above winter averages (2, 3, 510) and can exceed 300% (4, 10). However, the future response of the ice sheet to enhanced surface melt intensity, longer melt seasons, extended melt zones, and increasing high-melt or -rainfall events is unclear, with apparently conflicting possibilities. Positive feedbacks may be limited. Observations show that velocity peaks in the melt season and diminishes later in the summer as ablation increases (3, 10). Further, years with high ablation display smaller average seasonal speedups near the ice sheet margin (6). Alternatively, longer surface melt seasons in a warmer climate enhance ice sheet motion due to more short-term accelerations from melt pulses repeatedly overwhelming the subglacial system (4, 11). The nature of changes in sliding motion affects sea level by dictating the extent to which ice is drawn to lower elevations where melt rates are high and ice is discharged through calving termini.

The degree to which meltwater input influences sliding dynamics is driven by the disparity between the rate of change of water input and the capacity of the subglacial drainage network (4, 11, 12). Observations on the GIS show the basal network evolves through the melt season, with enhanced efficiency extending upward of 50 km inland from the ice sheet margin (13, 14). The mechanism driving evolution is commonly interpreted to follow the smaller mountain glacier conceptual model, whereby an inefficient network of linked cavities switches to a system of efficient channels by melting of the overlying ice roof. The sensitivity to sustained meltwater input is reduced as the inefficient network evolves to a high-capacity channelized system, but short-term perturbations still overwhelm even these efficient drainage features. Basal pressure measurements, which are restricted in time and space (1518), and dye-tracing experiments along the ablation zone (14) provide the only direct measurements of subglacial hydrologic conditions to date. Whereas basal pressure observations have proven indispensable for elucidating hydrologic processes in the mountain glacier setting, scarce observational data sets from the GIS so far have limited the extension of these processes to the ice sheet scale with confidence. Using in situ water pressure measurements in boreholes drilled along a 34--km transect in the western GIS and modeling, we tested the hypothesis that basal conduit melt processes drive the formation of a channelized network across the drainage regime.

During the summers of 2010–2012, we used hot-water methods to drill 23 boreholes to the ice sheet bed at sites along an E-W transect in western Greenland (Fig. 1). Borehole sites extend inland from the margin of terrestrially terminating Isunnguata Sermia and represent a range of settings, from shallow marginal conditions with ice depth of 100 to 150 m (hereafter referred to as S1) to 34 km deep in the interior, where ice thickness is >800 m (S4) (19). Sites also sample a bedrock trough (S3) and an adjacent shallow area located 17 km inland (S2). Before borehole refreezing, we performed hydrological impulse tests (19) and instrumented boreholes with basal water pressure sensors for long-term (up to >1 year) monitoring.

Fig. 1 Site setting with bedrock topography from ICEBRIDGE airborne radar (29), extending east from Isunnguata Sermia.

Drill sites are denoted by red dots, with site number and ice thickness for reference.

We measured distinctive spatial variability in borehole water pressure behavior at the ice sheet margin (S1) as compared to deeper in the interior (S2 to S4). Of the 13 holes drilled at S1, 6 immediately connected to the active subglacial system as evidenced by rapid water level drops when the drill intersected the bed (fig. S1). Basal water pressures in these holes suggest a mix of distributed and channelized drainage during the summer melt season. Daily water pressure swings exceeding 70% of ice overburden (OB) (Fig. 2) indicate borehole connection to a highly efficient, channelized system (20). In other holes, small-amplitude variations superimposed on steady high pressure suggest limited drainage capacity (Fig. 2), which is corroborated by the slow accommodation of water introduced to holes during impulse testing. Similar variability over high mean pressures has been interpreted to reflect a distributed drainage configuration (21, 22).

Fig. 2 (A to D) Characteristic long-term water pressure records at S1 to S4 as a fraction of OB.

S1 (D) showed significant spatial variability, with some holes showing large diurnal variations (black line), whereas others showed small-amplitude variations at high pressure (red line). The slow increase in water level during days of year 205 to 210 at S1 (black line) may represent the slow closure of an efficient connection or advection of the borehole away from such a hydrologic feature. Boreholes were drilled over the course of three field seasons, thus presented pressure records do not all span the same time period (the drill year is shown in the boxes). Head as fraction of OB pressure (Embedded Image) is calculated as Embedded Image, where Embedded Imageis head equivalent of water pressure measured in the borehole, Embedded Imageis ice thickness, and Embedded Imageand Embedded Imageare water and ice densities, respectively.

All 10 boreholes located 17 to 34 km from the margin experienced 50- to 125-m water level drops [12 to 17% OB equivalent (fig. S1)] when the drill intersected the bed, indicating connection to an active drainage system. At S2 and S4, pressure records are similar to those at boreholes near the ice sheet edge interpreted to be connected to a linked cavity network. Water pressures 34 km inland at S4 remained between 88 and 94% OB during the late melt season (days of year 200 to 240), with diurnal variations limited to 4% OB or less. Pressures at S2 were steadily between 82 and 92% OB with small variations of similar magnitude as those at S4 (Fig. 2). In the adjacent trough (S3), hydraulic connections between boreholes were evident over 20-m length scales, but long-term connectivity to a broader drainage system appears to be transient. Water level behavior at S3 periodically switched from steadily near or above 100% OB, to exhibiting diurnal variations up to 14% OB. We interpret this behavior to reflect temporary establishment and loss of connections along the bed.

Consistent with existing observations to the north near Jakobshavn Isbræ (17, 18), we see no direct evidence of high-capacity basal melt channels in our inland (17 and 34 km from the ice sheet margin) boreholes, as manifested by reduced mean pressures and large-amplitude diurnal pressure variations. However, limited sample locations do not preclude the existence of such basal pathways, motivating comparison of basal pressure conditions through numerical analysis. Following previous theoretical development of steady-state conduit dynamics (23), we modeled conduit conditions on the ice sheet domain under a range of constant discharge values (19). We find that steady conduit pressures less than 70% OB are limited to <10 km from the margin and increase toward OB further inland (Fig. 3). Near the ice sheet margin, expected conduit pressures lower than our distributed measurements reinforce the previous theory that water is driven into these low-pressure features, enhancing their stability (11). The reverse is true away from the margin. Pressure differentials between our inland measurements and conduit theory are diminished, and measured pressures at S4 are lower than expected in conduits. In this interior setting, water driven away from high-pressure conduits would feed the distributed network represented by our measurements, resulting in conduit instability and collapse.

Fig. 3 Simulation results from steady-state conduit analysis (A) and transient experiments (B to D).

Steady-state conduit simulations were performed for an envelope of discharges (Q) ranging from a low of 1 m3 s−1 [solid red line in (A)] to an upper limit of 300 m3 s−1 (dashed red line), guided by proglacial measurements (4). Mean pressures encompassing days of year 200 to 240 from distributed network measurements are shown by black dots; vertical bars denote maximum and minimum pressures during the time period. Transient experiments were forced with diurnally varying input [red line in (B)]. Conduit discharge (B), cross-sectional area (C), and head as fraction of OB (D) are displayed at 3 and 40 km for the margin (dashed line) and inland (solid black line) scenarios, respectively. Margin and inland scenarios are assumed to be representative of conditions expected near S1 and S4.

Steady-state analysis neglects variable input as surface meltwater routed to the bed through discrete moulins undergoes seasonal (24) and diurnal (25) change. We therefore extended our analysis to include transient forcing by implementing a conduit melt-closure model (19). Creep closure from ice deformation and opening from meltback of conduit walls determined cross-sectional area, and turbulent flow was simulated through the semicircular conduit. We focus here on two test cases representing a conduit near the margin (with a length of 3 km) and extending to the drainage interior (40 km) to illustrate geometric effects on conduit growth via melting. We assume a preexisting conduit and force the system with diurnally varying discharge. Meltback of the ice walls near the ice sheet margin promotes basal conduit growth (Fig. 3), so that changes in input are rapidly accommodated, thereby preventing elevated pressures. In contrast, interior conduit melt rates are >95% slower than creep closure. Stunted conduit geometry from slow growth can only accommodate a fraction of the imposed input and at pressures largely constrained to 100% OB. The imposed input range (1 to 3 m3 s−1) requires a period of variation that is unreasonably long (90 days) to be fully accommodated by the conduit (supplementary text).

Enhanced conduit closure under thicker ice has been highlighted as a driving difference between alpine and ice sheet subglacial hydraulics (10, 12). However, our pressure records imply that maximum closure rates are similar at marginal sites and in the interior. This is supported by our numerical test cases, despite ice thickness at the inland setting that is three times that at the margin. Instead, low basal conduit melt rates limit channel evolution at inland areas of the ice sheet. The melting of conduit walls is a function of discharge and hydraulic gradient. Steep surface slopes are common in both alpine glaciers and near the ice sheet margin, leading to large hydraulic gradients at the glacier bed and hence to high basal conduit melt rates. In contrast, much of the ice sheet ablation zone away from the margin is characterized by muted surface slopes, which limit conduit growth through dissipative heating effects.

Asymmetry between conduit meltback and closure rates dictates that basal conduits can only be sustained by generally steady discharge conditions. Despite variable surface input, it is possible that the modulation of discharge through basal or englacial water storage and release could facilitate near-steady basal flow. Other work suggests that reversing hydraulic gradients enable an adjacent network to temporarily store water driven out of conduits during periods of high flux and return it during times of low flow (20, 26). However, that model assumes preexisting basal conduits. Our results suggest basal melt rates are too slow to fully establish a channelized network during a single summer season (supplementary text). Further, our measurements of the basal pressure regime question interior conduit stability under steady flow conditions.

Our data and numerical experiments do not support the widespread growth of a conduit network in the drainage interior by ice-wall melt processes alone. Nevertheless, velocity interpretations (10), proglacial stream measurements (13), and dye-tracing experiments (14) provide evidence for seasonally increasing drainage efficiency away from the margin. This implies that other physical processes at the bed are more important than previously recognized in regions where gradient-driven basal melting is muted. We surmise that, toward the ice sheet interior, a network of efficient distributed pathways develops in contrast to large melt channels. Accelerated sliding through the melt season may increase pathway discharge by enlarging space on the lee sides of bed asperities, allowing such a network to transport significant quantities of water (supplementary text). Regardless, any working model of GIS subglacial hydrology must reconcile observations of both pressure and drainage efficiency under relaxed gradient regimes atypical of mountain glaciers.

Accurate representation of processes driving the behavior of ice/water interactions at the ice sheet bed is crucial to understanding drainage system evolution and our ability to quantify the impact of surface meltwater input on ice sheet acceleration. Our results caution against the direct transfer of processes from the alpine setting to the ice sheet interior, where geometric differences are accentuated. Future efforts to assess drainage dynamics away from the ice sheet margin should focus on developing additional mechanisms for growing drainage system efficiency, such as sliding over bedrock. The importance of processes in this interior regime is paramount, considering that surface melt extent and intensity have increased over the GIS during the observational era (27, 28), and future warming is projected to expand melt intensity toward the ice sheet interior.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S6

Tables S1 to S3

References (3043)

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: We thank J. Johnson for his thoughtful review and comments that improved this manuscript. Many thanks to all those who provided field assistance. This work is funded by SKB-Posiva-NWMO through the Greenland Analogue Project and NSF (Office of Polar Programs–Arctic Natural Sciences grant no. 0909495).
View Abstract

Stay Connected to Science

Navigate This Article