Parameter Optimization in a Sea Ice Model with Elasto-Visco-Plastic (EVP) Rheology
Gleb Panteleev, Naval Research Laboratory, Stennis Space center, MS, United States, Max Yaremchuk, Naval Research Lab, Stennis Space Center, MS, United States, Jacob Stroh, University of Alaska Fairbanks, Fairbanks, United States, Oceana Puananilei Francis, University of Hawaii, Honolulu, United States and Richard Arthur Allard, US Naval Research Laboratory, Ocean Sciences Division, Stennis Space Center, MS, United States
Abstract:
Sea-ice models which allow for deformation are primarily based on rheological formulations originally developed by Hibler (1979). In current versions of the EVP and VP sea ice models (e.g. CICE6) the sea ice rheology and landfast grounding/arching effects are modeled as a functions of variable sea ice thickness and concentration with a set of fixed empirical parameters (e.g. P*, e, k
1 , k
2 , k
T). This work considers a spatially-variable extension of those parameters in the two-dimensional EVP sea ice formulation of (Hunke and Dukowicz, 1997, Lemeaux et al, 2015). Feasibility of optimization of the rheological and landfast sea ice parameters was assessed by applying 4Dvar data assimilation (DA) to the synthetic sea ice observations (concentrations, thickness and velocity).
We found that EVP Tangent Linear and Adjoint models are unstable but can be stabilized through the including Newtonian friction term into the equation for the component of the internal stress. The set of the OSSE’s show that landfast parameters k2, kT can be relatively easy reconstructed after 5-10 iteration of the minimization procedure. Optimization of the spatially varying P*, e and sea ice initial conditions takes more time (~40-50 iteration) but results for better hindcast of the sea ice state and components of the internal stress tensor. Analysis of the inaccuracy in the wind forcing and errors in the sea ice thickness observations show the reasonable robustness of the 4Dvar DA approach and feasibility of it’s application to the available (e.g. BGEP) and incoming (e.g.MOSAiC) observation.
