Abstract:A first-order system least squares formulation for the sea-ice dynamics is presented. In addition to the displacement field, the stress tensor is used as a variable. As finite element spaces, standard conforming piecewise polynomials for the displacement approximation are combined with Raviart-Thomas elements for the rows in the stress tensor. Computational results for a test problem illustrate the least-squares approach.
What problem does this paper attempt to address?
This paper attempts to solve the key problems in the dynamic modeling of Antarctic sea ice. Specifically:
1. **Improve simulation accuracy**: Traditional sea - ice dynamic models usually calculate the stress tensor through post - processing of the velocity field, which may lead to non - physical solutions and loss of accuracy. Therefore, this paper proposes a new method, directly incorporating the stress tensor as an independent variable into the variational formula to improve the accuracy and stability of the simulation.
2. **Improve numerical methods**: The paper introduces the Least Squares Finite Element Methods (LS - FEM) to solve the sea - ice dynamic problems. LS - FEM provides a robust and efficient numerical method by minimizing the \(L^2\) residuals of partial differential equations, especially suitable for complex boundary conditions and nonlinear problems.
3. **Consider multi - physical field coupling**: Sea - ice dynamics involves multiple physical processes, including the changes of velocity field \(u\), ice concentration \(A\) and average ice height \(H\). There are complex interactions among these variables. The paper constructs a more comprehensive model by introducing the stress tensor \(\sigma\) and coupling it with the velocity field and other variables.
4. **Verify and optimize the model**: To verify the effectiveness of the proposed method, the author uses a test case to simulate the sea - ice dynamic behavior within 8 days and shows the velocity field results at different time points. In addition, the paper also discusses further research directions, such as the analysis of ellipticity, the application on curved - edge domains and the relationship with other standard or hybrid methods.
### Summary of mathematical formulas
- **Momentum equation**:
\[
\rho_{\text{ice}} H \frac{\partial u}{\partial t}+F(u)-\text{div}\,\sigma(u, A, H) = 0
\]
- **Ice concentration conservation equation**:
\[
\frac{\partial A}{\partial t}+\text{div}(uA) = 0
\]
- **Average ice height conservation equation**:
\[
\frac{\partial H}{\partial t}+\text{div}(uH) = 0
\]
- **Stress - strain relationship**:
\[
\sigma=\frac{P(A, H)}{2}\left(\frac{\text{dev}\,\varepsilon(u)}{\Delta(u)}+\frac{2\text{tr}\,\varepsilon(u)}{\Delta(u)}I - I\right)
\]
where,
\[
P = P^\star H e^{-C(1 - A)}
\]
\[
\Delta(u)=\sqrt{\text{dev}\,\varepsilon(u):\text{dev}\,\varepsilon(u)+4(\text{tr}\,\varepsilon(u))^2+\Delta_{\min}^2}
\]
\(\Delta_{\min}=2\times 10^{-9}\,\text{s}^{-1}\)
- **Least - squares functional**:
\[
F(\sigma, u, H)=F_m(\sigma, u, A, H)+F_c(\sigma, u, A, H)+F_e(\sigma, u, A, H)
\]
where,
\[
F_m(\sigma, u, A, H)=\left\|\rho_{\text{ice}} H \frac{\partial u}{\partial t}+F(u)-\text{div}\,\sigma\right\|^2_0