A High-Resolution Granular Sea Ice Model

The Arctic ice pack is composed of parcels of first-year and multiyear ice divided by leads and pressure ridges. In the past, with interest focused on basin-scale processes, sea ice models have used continuum descriptions of sea ice rheology. With the current focus on resolution approaching the scale of individual floes and failure processes that define large-scale rheology, it makes sense to develop a granular model that incorporates this level of detail. The CRREL granular sea ice model is based on a discrete element approach in which individual parcels of first-year and multiyear ice are explicitly modeled. In a nutshell, the granular model takes the continuum assumption from the pack scale down to the floe scale. The yield surface and flow rule that are of central importance for large-scale models are not needed in a granular model. They are replaced by floe-scale failure processes. The individual ice parcels may freeze together, overlap to raft or form pressure ridges or split or separate to form leads, depending on the dynamic conditions. Coulomb frictional forces resist sliding between ice parcels. The model ice pack is driven by boundary deformations, winds, and currents. Because the model explicitly simulates an assembly of thousands of ice parcels with a distribution of sizes, shapes, and thicknesses, it is inherently anisotropic.

The development at CRREL of a granular sea ice model has accelerated in the past year with support from NSF and NASA Offices of Polar Programs. This year we have added a multi-layer thermodynamic model to the granular sea ice model and we have configured the model to be driven by RGPS ice velocity fields. We have performed simulations of a 100x100 km section of the ice pack surrounding the SHEBA field experiment. Model wind forcing is supplied by hourly wind stresses at the SHEBA site provided by the SHEBA Atmospheric Surface Flux Group (Andreas, Fairall, Guest, and Persson). Thermodynamic growth and melt is modeled using the multi-level approach of Ebert and Curry(1993).

The computer code for the thermodynamic model was written and developed by Greg Flato at the Canadian Center for Climate Modelling and Analysis. The boundary forcing is supplied by RGPS velocity grids derived from SAR imagery by Ron Kwok at JPL. The RGPS velocity grids are implemented by superimposing the velocity grids on a background of the Arctic Basin, placing the 100x100 km model ice pack configuration within the velocity grid, and interpolating the grid velocities to the location of the boundary elements surrounding the model ice pack. Once the model pack is configured it is used to simulate the motions of the ice pack within the confines of the model for the time interval covered by the RGPS velocity data, typically 2-3 days. To date we have used RGPS deformations from several periods during winter 1997-1998. In the coming year we will simulate the entire SHEBA year.

This year the explicit floe model was fundamentally changed. The previous approach [Hopkins, 1996] was to model every ice parcel within the model domain as a distinct floe with a single thickness. Not only was this approach extremely slow computationaly, but it suffered from a major problem of a cascading increase in complexity as successive deformation events created, altered, and destroyed ice parcels. One has only to consider the complexity of the real ice pack to appreciate this problem. In the new approach we preserve the granularity of the pack while dealing with the cascading complexity in a consistent manner. This is accomplished by 1) treating the interior of each floe as a continuum with its own thickness distribution and 2) periodically reinitializing the model ice pack by placing a new undeformed configuration over the old deformed configuration. Each floe in the new configuration incorporates the thickness distribution of the floes and leads in the old configuration that it overlaps. This allows the new configuration to incorporate detail such as lead orientation and extent and changes to the thickness distribution occuring from deformation and thermodynamics. During the summer, after break-up, there is no need for the periodic reinitialization. This new approach will allow us to simulate deformation and thermodynamic evolution of a 100x100 km section of the ice pack surrounding the SHEBA field experiment for an entire year.

USES OF A HIGH RESOLUTION GRANULAR SEA ICE MODEL:

1) Because the granular model explicitly treats ridging, rafting and lead formation, changes to the thickness distribution can be directly related to the deformation events that created them. This will lead to improved parameterization of the ice thickness redistribution function.

2) Instead of using RGPS velocity fields to supply boundary forcing, the granular sea ice model could be inserted in place of one or more cells in a large-scale ice model. The large-scale model would furnish the boundary conditions on ice transport and boundary motions needed to drive the granular model. The granular model would allow high-resolution simulation of areas of interest in the central Arctic or in coastal regions.

3) The granular sea ice model could form the basis for a new generation of more powerful single-column models that explicitly incorporate the spatial organization and dynamics of the ice pack.

ADVANTAGES OF THE GRANULAR MODELING APPROACH:

• Spatial resolution at the scale of individual ice parcels.
• Surface stress resolution at the scale of individual ice parcels.
• Internal stress resolution with smart finite element modeling floes.
• Direct coupling of floes with wind and water drag.
• Explicit location and energetics of ridge formation that allows integration with ambient noise models.
• Unlimited scalability: As computer power increases so can scope and detail.
• The ability to directly incorporate SAR data into simulations. The ability to integrate parameterizations of smaller-scale behavior such as freeze and melt, ridging, rafting, and floe splitting.
• A natural transition between a fully consolidated pack and the loose conditions of the MIZ and the late summer pack.

Ebert and Curry (J. Geophys. Res.,98, 10085-10109, 1993)

Hopkins, M.A., (J. Geophys. Res., 101, 18315-18326, 1996)

  Figure 1: A snapshot from a simulation in which a 200 km x 200 km model ice pack is undergoing simple shear with dilatation. Large-scale lead development is evident. The strain rates are du/dy=1.0x10^-6/s and dv/dy = 0.5x10^-6/s.

Figure 2: A snapshot from the same simulation showing the lines of force propagating through the model ice pack. The black lines represent compressive forces and the red lines represent tensile forces. The line width is proportional to the magnitude of the force.

Short animation of the model.

Figure 3. A snapshot from a simulation of the ice pack surrounding the SHEBA field experiment. The simulation began on November 11, 1997 and ended on November 13, 1997. The yellow line shows the path of the SHEBA camp.

Short animation of above model.

For more information, contact us at hopkins@crrel.usace.army.mil or 603-646-4249