Large-Ensemble Modeling of Past Variations in West Antarctic Embayments

Abstract:

Recent observations of thinning and retreat of the Pine Island and Thwaites
Glaciers identify this sector of West Antarctica as particularly vulnerable to
future climate change. To date, most future modeling of these glaciers has been
validated using recent and modern observations. As an alternate approach,
we apply a hybrid 3-D ice sheet-shelf model to the last deglacial retreat in
this sector, making use of geologic data of ice extents from ~20,000 years BP
to present, both for the Amundsen Sea sector and also for the Ross and Weddell
embayments.

Following recent ice-sheet studies, we use Large-Ensemble statistical
techniques, performing sets of ~500 to 1000 runs with varying model parameters.
The model is run for the last 20 kyrs on 5 to 20-km grids spanning West
Antarctica, with lateral boundary conditions from a prior continental-scale
simulation. An objective score for each run is calculated using reconstructed
past grounding lines, shelf extents, relative sea levels, and modern
conditions. Runs are extended into the future (few millennia) with simple
atmospheric and oceanic forcing. The goal is to produce calibrated
probabilistic envelopes of model parameter ranges and simulated ice retreat.

Preliminary results are presented for Large Ensembles with (i) Latin HyperCube
sampling in high-dimensional parameter space, and (ii) dense sampling with a
lower number of parameters. We focus on optimal parameter differences
between the 3 embayments, validation with other paleo data, contribution
to meltwater pulses ~14 to 12 ka, and future projections. Most reasonable
parameter combinations produce drastic future retreat into the interior
Pine Island and Thwaites basins within ~2000 years, adding ~2 m to global
sea-level rise.