Develop and apply a fully coupled ice-sheet model to estimate near-term changes to the West Antarctic ice sheet.
OVERALL PERFORMANCE MEASURES
1ST QUARTER METRIC – COMPLETED
Approximately one third of the present day rate of sea-level rise (SLR) is from ice sheet mass loss and that fraction is expected to increase significantly in the coming decades and centuries (Church et al., 2013). Of concern is the potential for increased rates of SLR due to the future initiation of dynamic instabilities within the Antarctic ice sheet (c.f., Schoof, 2007). Indeed, some recent research speculates that such an instability may have already begun (Joughin et al., 2014; Rignot et al., 2014) and uncertainties in the future evolution of Antarctica have been cited as the single biggest uncertainty with respect to projecting rates of future SLR (Church et al., 2013).
Accurate simulations of marine ice sheet dynamics require an ice sheet model with very specific features and capabilities. Two features of utmost importance are that the model must employ a “higher-order,” more accurate treatment of the governing Stokes flow equations than used in the previous generation of simpler models, and the model must be capable of very-high grid resolution (~1 km or less) within isolated portions of the domain.
In accordance with these and other necessary features, new marine ice sheet models have been designed according to the following requirements: (1) numerically and computationally robust solutions of the relevant partial differential equations, representing the conservation of momentum, mass, and energy; (2) the ability to employ extremely high spatial resolution within limited, dynamically complex areas of the computational domain; (3) solvers that allow the models to run efficiently on thousands to tens of thousands of processor cores on current and future Leadership Class computers.
Under the DOE SciDAC Predicting Ice Sheet and Climate Evolution at Extreme Scales (PISCEES) project, two new ice sheet model dynamical cores, BISICLES and FELIX, have been developed in response to the requirements listed above. The application of high-fidelity, high-performance ice sheet models to large-scale simulations of marine ice sheet evolution is new territory for ice sheet modeling and, for this reason, PISCEES has pursued two different, but complementary, modeling approaches.
BISICLES includes a depth-integrated, higher-order approximation to the Stokes flow equations, using a (computationally less expensive) two-dimensional solution with local corrections to recover a fully three-dimensional solution. While a regular, structured mesh is used, BISICLES can refine that mesh locally so that successively finer resolution “blocks” are nested within coarser resolution ones. This block-structured refinement is done adaptively during the course of a solution, based on solution error metrics, to ensure adequate solution accuracy in regions of dynamic complexity (Figure 1). FELIX uses a fully three-dimensional, higher-order approximation to the Stokes flow equations but, through advanced numerical and computational methods, “flattens” the problem so that the bulk of the computational work effectively occurs in two dimensions. Adequate mesh resolution in regions of dynamic complexity is achieved through the use of fully unstructured, variable resolution meshes (Figure 3). Both models have proven to be robust and computationally efficient on a range of idealized and realistic, large-scale (whole ice sheet) problems (see Figures 1-4).
2ND QUARTER METRIC – COMPLETED
To simulate changes in sea levels, dynamic ice sheet models need to be developed and incorporated into climate models. Models also need to accurately capture various processes by which the surrounding climate model components cause the ice sheet to lose or gain mass (e.g., through melting or freezing). The primary couplings that need to be accounted for include (1) heat and precipitation fluxes between the atmosphere and the ice sheets upper-surface, (2) heat and freshwater fluxes between the lower surface of floating ice sheets and the ocean, (3) liquid (meltwater runoff) and solid (icebergs) freshwater fluxes from the lateral margins of ice sheets into the oceans; and (4) changes in surface albedo resulting from snowpack evolution and ice sheet advances or retreats (snow versus ice versus bare land).
Changes in sea level occur when there is an imbalance between the amount of mass gained and lost over a given time period; if net melting, sublimation, and calving exceed net accumulation, ice sheets loose mass and sea level increases (and vice versa). At present, surface melting and iceberg calving are estimated to contribute approximately equally to the annual sea-level rise from Greenland, with melting beneath floating ice contributing a small fraction of the total. For Antarctica, iceberg calving and basal melting of ice shelves are estimated to contribute approximately equally to the annual sea-level rise, with surface melting being a small fraction of the total.
The U.S. Department of Energy (DOE) has funded efforts to include the coupling between ice sheets and other climate components in advanced Earth System Models (ESMs), including the Community Earth System Model and, more recently, DOE’s Accelerated Climate Model for Energy. Coupling between the land ice and the atmosphere models is responsible for passing heat and moisture fluxes between the two components and for evolving the land ice surface brightness. Coupling between the land ice model and the ocean model is responsible for the passing of heat and freshwater fluxes between the two components, both at the base of floating ice shelves and at its margins. Gain or loss of mass from the ice sheet model component via these couplings dictates the rate of ice sheet growth or decay and, in turn the rate at which that ice sheet detracts from, or contributes to, global sea-level rise. These new couplings, described in further detail in the report, have all been developed, integrated, and tested within DOE ESMs. Ongoing testing as part of fully coupled ESM simulations is currently active as part of Accelerated Climate Model for Energy, version 1.0, development and tuning.
3RD QUARTER METRIC - COMPLETED
Potential future sea-level rise from the Antarctic ice sheet is expected to be a strong function of the integrity of Antarctica’s fringing ice shelves, which limit the flux of ice from the continent to the surrounding oceans (Fürst et al. 2016). This hypothesis is supported by both observations (e.g., Scambos et al. 2014) and modeling (e.g., DeConto and Pollard 2016). To better simulate and anticipate such changes, climate-modeling centers need to couple dynamic ice sheet models with climate models and account for the relevant processes by which ice shelves melt, thin, and degrade to the point of collapse. For Antarctica, the potential for large increases in submarine melting of ice shelves due to changes in ocean circulation – and subsequent ocean heat delivery to ice-shelf cavities – is a key concern (e.g., Hellmer et al. 2012). Biases in coupled model simulations that impact the ocean state and circulation around Antarctica strongly impact modeled submarine melting, and in turn ice-sheet evolution in coupled climate-and-ice-sheet simulations. Identifying and understanding the causes for such biases is an important first step towards reducing them and improving projections of ice sheet evolution in coupled climate models.
By comparing model output with observations, we have identified several important coupled model biases that lead to unrealistic sub-ice-shelf melt rates. From a partially coupled ice-sheet-and-ocean modeling framework, we have identified ocean mixed-layer biases that lead to either too little or too much vertical mixing at the ocean surface, with the result that either too much or too little warm, intermediate-depth water gains access to the Antarctic continental shelf. These biases, which cause modeled sub-ice-shelf melt rates to rapidly depart from the range expected from recent observations, leading to either far too much or far too little melting, are attributable to deficiencies in the partially coupled modeling framework. While both of these deficiencies can likely be remedied through full coupling of ocean, atmosphere, sea-ice, and land-ice model components, they point to a clear need for great care in addressing initially small biases in sea-ice formation, transport, and coupling to the ocean model, which can eventually have a strong impact on sea-surface salinity and mixed-layer depth.
From initial simulations using a fully coupled Earth System Modeling framework, we have examined modeled and observed sub-ice-shelf melt rates for Antarctica’s two largest ice-shelf systems. For the Ross Ice Shelf, modeled submarine melt rates are slightly too large, but are generally within the range of observations and we find relatively small ocean temperature and salinity biases in the Ross Sea region. These are likely due to underestimates in sea-ice formation and/or export, which have a strong impact on ocean salinity. For the Filchner-Ronne Ice Shelf, modeled submarine melt rates are far too large, a bias that we attribute to too much warm, salty, intermediate-depth water making its way into the Filchner Trough, which is a key pathway connecting the open ocean and the sub-ice shelf cavity. This bias in the modeled ocean circulation is likely due to inadequate resolution of the Antarctic Slope Front, an ocean density structure responsible for blocking the flow of warm intermediate waters onto the Antarctic continental shelf. A poorly defined Antarctic Slope Front may be the result of a poorly defined Antarctica Coastal Current, the primary feature responsible for the ocean density structure that defines the front. In turn, inadequate model resolution may be responsible for the poorly defined coastal current. Thus, ocean model resolution around the Antarctic continent must be high enough to resolve fine-scale (tens of km or less) features (like the coastal current) in order to reduce the too warm ocean (and too large melt rates) biases currently found in the Filchner-Ronne Ice Shelf region.
We conclude that biases in modeled sub-ice-shelf melt rates, which will lead to biases in modeled discharge of ice to the oceans (and in turn, biases in modeled sea-level rise), can be minimized by correcting or minimizing Southern Ocean temperature, salinity, and mixed-layer biases and by ensuring that coupled model simulations are conducted with adequate spatial resolution.
4TH QUARTER METRIC - COMPLETED
Future sea-level rise from the Antarctic ice sheet is expected to be a strong function of the integrity of Antarctica’s fringing ice shelves, which limit the flux of ice from the continent to the surrounding oceans (Fürst et al., 2016). This hypothesis is supported by both observations (e.g., Scambos et al., 2014) and modeling (e.g., Gudmundsson, 2013; DeConto and Pollard, 2016). One of the primary processes by which Antarctica’s ice shelves are vulnerable to degradation is increases in submarine melting, which occur when warm, intermediate-depth ocean water is brought into contact with ice shelves through changes in circulation and upwelling (e.g., Hellmer et al., 2012). At present, a key oceanographic feature limiting such upwelling onto the Antarctic continental shelf is the “Antarctic Slope Front”, a narrow region of steeply tilted, constant-density surfaces near coastal Antarctica. This density structure is affected by the large-scale structure of the Southern Ocean westerly wind field. Anticipated changes in this wind field as a result of anthropogenic climate change (Fyfe et al., 2007; Zheng et al., 2013) are projected to weaken the Antarctic Slope Front, with concomitant increases in Antarctic coastal warm water upwelling (Spence et al., 2014). Using a coupled climate model and a dynamic ice sheet model, we estimate the impact of these changes on the submarine melting of Antarctic ice shelves and, in turn, on the dynamics of grounded ice currently buttressed by these same ice shelves; over 100 years, we estimate ~30 mm of sea-level rise as a result of these processes.
The ACME climate model allows for coupled, global simulations that include ocean circulation within sub-ice-shelf cavities, allowing for the calculation of sub-ice-shelf melting and freezing rates. Changes in these melting rates occur as a function of changes in ocean circulation, which impact melting rates through the delivery of warm, salty, intermediate-depth waters to ice-shelf cavities (Jacobs et al., 2011). A critical oceanographic feature that limits such changes is the Antarctic Slope Front – a region of steeply inclined isopycnals near the Antarctic coast – which prevents warm intermediate waters from accessing the continental shelves. Degradation of the Antarctic Slope Front is anticipated under future climate change as a result of far field changes in the Southern Hemisphere Westerly winds, which drive the Antarctic Circumpolar Current, flowing West to East. Under future climate warming scenarios, the Westerlies are projected to increase in strength and shift farther south, resulting in a decrease in the strength of Antarctica coastal Easterly winds that drive the Antarctic Coastal Countercurrent (flowing West to East). In turn, this change reduces Ekman transport towards the coast and the downwelling of cold, fresh water, which helps maintain the Antarctic Slope Front (Spence et al., 2014).
We use the ACME climate model, climate perturbation experiments, and a dynamic ice-sheet model to simulate the impacts of these future changes on mass loss and sea-level rise from the Antarctic ice sheet. Following a similar set of experiments to those conducted by Spence et al. (2014), we conduct two simulations: (1) a control simulation with atmospheric forcing taken from the standard “normal year” (Yeager and Large, 2009), and (2) a perturbation simulation in which normal-year forcing includes winds perturbed to simulate the anticipated poleward shift and strengthening of the Westerlies. Unlike Spence et al. (2014), we include sub-ice-shelf cavities in our simulations, allowing us to calculate sub-ice-shelf melt rates, which we then use to force evolution of a dynamic ice-sheet model. We show that the applied wind perturbations result in changes in ocean circulation (a reduction in the speed of the coastal countercurrent) and sub-ice-shelf melt rates (an increase in warm water upwelling and sub-ice-shelf melting) as anticipated by previous results of Spence et al. (2014). Differences between our control simulation and our perturbed-winds simulation indicate that ice-sheet volume loss is greater in the perturbation simulation. We estimate a potential mass loss from the Antarctic ice sheet equivalent to ~90 mm of sea-level over 100 yrs. Because some of this mass loss occurs within portions of the ice sheet already displacing global sea level, ~30 mm of this mass loss would contribute directly to global sea-level rise.