Description

Sea ice forms at the interface between the ocean and the atmosphere at high latitudes. It regulates mass, heat, and momentum exchange between the ocean and the atmosphere in these polar regions. Its formation and melting drive global thermohaline circulation, and changes in its extent significantly affect reflection of solar radiation. Strong interactions between planetary systems occur across the ocean-ice-atmosphere interface that are challenges for Earth System Models to simulate accurately. Inter-annual variability in ice characteristics are also difficult to capture in numerical simulations of climate. The objective of this project is to improve the physical description and numerical computation of sea ice in the sea-ice component of the Department of Energy's Energy Exascale Earth System Model (E3SM) to address these challenges.

The current sea-ice model in E3SM treats ice as a continuous fluid with variable viscosity.  This project will replace the current model with a solid model for ice that includes explicit representation of fracture, accounting for lead opening and, for example, concomitant ice and dense water formation. The improved physical description of ice is expected to improve modeling of ice characteristics and improve accuracy of coupled climate simulations.

The improved constitutive model for ice will be implemented in the material-point method (MPM), a numerical method that combines Lagrangian material points with a background mesh to discretize equations of motion. The Lagrangian aspect of the method facilitates the use of the proposed constitutive model and provides a natural way to handle horizontal advection of sea ice and any number of tracer variables, without artificial numerical diffusion, as is inherent in Eulerian remapping algorithms currently in use. The background grid in MPM is under user control, allowing use of the Model for Prediction Across Scales (MPAS) framework incorporated in E3SM. MPAS meshes are spherical centroidal Voronoi tessellations. This unstructured mesh technology allows meshes that concentrate mesh points in regions of interest and importance, providing flexibility and improving computational efficiency.  Moreover, the algorithms we develop are designed to make use of current and next-generation GPU-enabled, high-performance computing platforms.

Model evaluation is an integral part of this proposal. Evaluation will focus on how well the model simulates overall sea ice behavior, as well as the spatial and temporal pattern of sea ice characteristics and their variability. Analysis of the model will include standard pointwise error estimation. However, new methods are required to account for misaligned or misshapen simulated features such as fractures. We propose novel metrics and related calibration techniques that are central to our model evaluation but are also generally applicable to many models with spatiotemporal output.

Code and model performance will be demonstrated within a capstone simulation with the new model fully coupled to global atmosphere, land surface, ocean, and river components. A simulated 50 year duration will allow a comprehensive assessment of predicted sea ice characteristics and their effect on other climate components. The main question to be answered is 'Does the new model adapt better to the changing conditions in the Arctic and Antarctic, allowing the simulated Earth system to evolve more accurately through coupling of the sea ice to the other Earth system components?'