skip to main content
US FlagAn official website of the United States government
dot gov icon
Official websites use .gov
A .gov website belongs to an official government organization in the United States.
https lock icon
Secure .gov websites use HTTPS
A lock ( lock ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites.


Title: Automatic delineation of glacier grounding lines in differential interferometric synthetic-aperture radar data using deep learning
Abstract Delineating the grounding line of marine-terminating glaciers—where ice starts to become afloat in ocean waters—is crucial for measuring and understanding ice sheet mass balance, glacier dynamics, and their contributions to sea level rise. This task has been previously done using time-consuming, mostly-manual digitizations of differential interferometric synthetic-aperture radar interferograms by human experts. This approach is no longer viable with a fast-growing set of satellite observations and the need to establish time series over entire continents with quantified uncertainties. We present a fully-convolutional neural network with parallel atrous convolutional layers and asymmetric encoder/decoder components that automatically delineates grounding lines at a large scale, efficiently, and accompanied by uncertainty estimates. Our procedure detects grounding lines within 232 m in 100-m posting interferograms, which is comparable to the performance achieved by human experts. We also find value in the machine learning approach in situations that even challenge human experts. We use this approach to map the tidal-induced variability in grounding line position around Antarctica in 22,935 interferograms from year 2018. Along the Getz Ice Shelf, in West Antarctica, we demonstrate that grounding zones are one order magnitude (13.3 ± 3.9) wider than expected from hydrostatic equilibrium, which justifies the need to map grounding lines repeatedly and comprehensively to inform numerical models.  more » « less
Award ID(s):
1739003
PAR ID:
10321191
Author(s) / Creator(s):
; ; ; ; ;
Date Published:
Journal Name:
Scientific Reports
Volume:
11
Issue:
1
ISSN:
2045-2322
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Abstract. Knowledge of past ice sheet configurations is useful for informing projections of future ice sheet dynamics and for calibrating ice sheet models. The topology of grounding line retreat in the Ross Sea sector of Antarctica has been much debated, but it has generally been assumed that the modern ice sheet is as small as it has been for more than 100 000 years (Conway et al., 1999; Lee et al., 2017; Lowry et al., 2019; McKay et al., 2016; Scherer et al., 1998). Recent findings suggest that the West Antarctic Ice Sheet (WAIS) grounding line retreated beyond its current location earlier in the Holocene and subsequently readvanced to reach its modern position (Bradley et al., 2015; Kingslake et al., 2018). Here, we further constrain the post-LGM (Last Glacial Maximum) grounding line retreat and readvance in the Ross Sea sector using a two-phase model of radiocarbon input and decay in subglacial sediments from six sub-ice sampling locations. In addition, we reinterpret high basal temperature gradients, measured previously at three sites in this region (Engelhardt, 2004), which we explain as resulting from recent ice shelf re-grounding accompanying grounding line readvance. Atone location – Whillans Subglacial Lake (SLW) – for which a sedimentporewater chemistry profile is known, we estimate the grounding linereadvance by simulating ionic diffusion. Collectively, our analyses indicate that the grounding line retreated over SLW 4300-2500+1500 years ago, and over sites on Whillans Ice Stream (WIS), Kamb Ice Stream (KIS), and Bindschadler Ice Stream (BIS) 4700-2300+1500, 1800-700+2700, and 1700-600+2800 years ago, respectively. The grounding line only recently readvanced back over those sites 1100-100+200, 1500-200+500, 1000-300+200, and 800±100 years ago for SLW, WIS, KIS, and BIS, respectively. The timing of grounding line retreat coincided with a warm period in the mid-Holocene to late Holocene. Conversely, grounding line readvance is coincident with cooling climate in the last 1000–2000 years. Our estimates for the timing of grounding line retreat and readvance are also consistent with relatively low carbon-to-nitrogen ratios measured in our subglacial sediment samples (suggesting a marine source of organic matter) and with the lack of grounding zone wedges in front of modern grounding lines. Based on these results, we propose that the Siple Coast grounding line motions in the mid-Holocene to late Holocene were primarily driven by relatively modest changes in regional climate, rather than by ice sheet dynamics and glacioisostatic rebound, as hypothesized previously (Kingslake et al., 2018). 
    more » « less
  2. Abstract Pine Island, Thwaites, Smith, and Kohler glaciers in the Amundsen Sea Embayment (ASE) sector of West Antarctica experience rapid mass loss and grounding line retreat due to enhanced ocean thermal forcing from Circumpolar Deep Water (CDW) reaching the grounding lines. We use simulated Lagrangian particles advected with a looping 1 year output from the Southern Ocean high‐resolution model to backtrack the transport and cooling of CDW to these glaciers. For the simulated year 2005–2006, we find that the median time needed to reach the grounding lines from the edge of the ASE is 3 years. In addition, the Antarctic Coastal Current contributes an equal number of particles as off‐shelf sources to the grounding lines of Pine Island and Thwaites. For CDW coming from off‐shelf, results from SOhi indicate that 25%–66% of the cooling occurs within ice shelf cavities. 
    more » « less
  3. Abstract. We report cosmogenic-nuclide measurements from two isolated groups of nunataks in West Antarctica: the Pirrit Hills, located midway between the grounding line and the divide in the Weddell Sea sector, and the Whitmore Mountains, located along the Ross–Weddell divide. At the Pirrit Hills, evidence of glacial-stage ice cover extends ∼320 m above the present ice surface. Subsequent thinning mostly occurred after ∼14 kyr BP, and modern ice levels were established some time after ∼4 kyr BP. We infer that, like at other flank sites, these changes were primarily controlled by the position of the grounding line downstream. At the Whitmore Mountains, cosmogenic 14C concentrations in bedrock surfaces demonstrate that ice there was no more than ∼190 m thicker than present during the past ∼30 kyr. Combined with other constraints from West Antarctica, the 14C data imply that the divide was thicker than present for a period of less than ∼8 kyr within the past ∼15 kyr. These results are consistent with the hypothesis that the divide initially thickened due to the deglacial rise in snowfall and subsequently thinned in response to retreat of the ice-sheet margin. We use these data to evaluate several recently published ice-sheet models at the Pirrit Hills and Whitmore Mountains. Most of the models we consider do not match the observed timing and/or magnitude of thickness change at these sites. However, one model performs relatively well at both sites, which may, in part, be due to the fact that it was calibrated with geological observations of ice-thickness change from other sites in Antarctica. 
    more » « less
  4. null (Ed.)
    The use of a boundary-layer parameterization of buttressing and ice flux across grounding lines in a two dimensional ice-sheet model is improved by allowing general orientations of the grounding line. This and another modification to the model’s grounding-line parameterization are assessed in three settings: rectangular fjord-like domains – the third Marine Ice Sheet Model Intercomparison Project (MISMIPC) and Marine Ice Sheet Model Intercomparison Project for plan view models (MISMIP3d) – and future simulations of West Antarctic ice retreat under Representative Concentration Pathway (RCP) 8.5-based climates. The new modifications are found to have significant effects on the fjord-like results, which are now within the envelopes of other models in the MISMIP+ and MISMIP3d intercomparisons. In contrast, the modifications have little effect on West Antarctic retreat, presumably because dynamics in the wider major Antarctic basins are adequately represented by the model’s previous simpler one-dimensional formulation. As future grounding lines retreat across very deep bedrock topography in the West Antarctic simulations, buttressing is weak and deviatoric stress measures exceed the ice yield stress, implying that structural failure at these grounding lines would occur. We suggest that these grounding-line quantities should be examined in similar projections by other ice models to better assess the potential for future structural failure. 
    more » « less
  5. Investigations of the time-dependent behavior of marine ice sheets and their sensitivity to basal conditions require numerical models because existing theoretical analyses focus only on steady-state configurations primarily with a power-law basal shear stress. Numerical results indicate that the choice of the sliding law strongly affects ice-sheet dynamic behavior. Although observed or simulated grounding-line retreat is typically interpreted as an indication of marine ice sheet instability introduced by Weertman (1974), this (in)stability is a characteristic of the ice sheet's steady states – not time-variant behavior. To bridge the gap between theoretical and numerical results, we develop a framework to investigate grounding line dynamics with generalized basal and lateral stresses (i.e. the functional dependencies are not specified). Motivated by observations of internal variability of the Southern Ocean conditions we explore the grounding-line response to stochastic variability. We find that adding stochastic variability to submarine melt rates that produced stable steady-state configurations leads to intermittently advancing and retreating grounding lines. They can also retreat in an unstoppable manner on time-scales significantly longer than the stochastic correlation time-scales. These results suggest that at any given time of their evolution, the transient behavior of marine ice sheets cannot be described in terms of ‘stable’ or ‘unstable’. 
    more » « less