FESOM 2.0 AWI-CM3 version 3.0

• Main Authors: Scholz, Patrick, Sidorenko, Dmitry, Gurses, Ozgur, Danilov, Sergey, Koldunov, Nikolay, Wang, Qiang, Sein, Dmitry, Smolentseva, Margarita, Rakowsky, Natalja, Jung, Thomas
• Format: Article in Journal/Newspaper
• Language: unknown
• Published: Zenodo 2022
• Subjects: Petersen; Sea ice
• Online Access: https://dx.doi.org/10.5281/zenodo.6335382; https://zenodo.org/record/6335382

FESOM2.0 builds on the framework of its predecessor, FESOM1.4, using its sea ice component Finite-Element Sea Ice Model (FESIM; Danilov et al., 2015), general user interface and code structure. Both model versions work on unstructured triangular meshes, although the horizontal location of quantities and vertical discretization are different. FESOM2.0 uses a B-grid-like horizontal discretization, with scalar quantities at triangle vertices and horizontal velocities at triangle centroids, while in FESOM1.4 all quantities were located at the vertices. In the vertical, FESOM2.0 uses a prismatic discretization where all the variables, except the vertical velocity, are located at mid-depth levels, while in FESOM1.4 each triangular prism is split into three tetrahedral elements and variables are located at full depth levels. In addition, in FESOM2.0, the interfaces for data input and output are further modularized and generalized to facilitate massively parallel applications. The new numerical core of FESOM2.0 is based on the finite-volume method (Danilov et al., 2017). Its boost in numerical efficiency comes largely from the more efficient data structure, that is, the use of two-dimensional storage for three-dimensional variables. Due to the use of prismatic elements and vertical mesh alignment, the horizontal neighborhood pattern is preserved in the vertical (see Fig. S4 in the Supplement). In FESOM1.4, three-dimensional variables are stored as one-dimensional arrays, which requires more fetching time. More importantly, the vertices of tetrahedral elements and derivatives on these elements need to be assessed for each tetrahedron separately, thus resulting in lower model efficiency. Other major advantages of using finite volume are the clearly defined fluxes through the faces of the control volume and the availability of various transport algorithms, whose choice was very limited for the continuous Galerkin linear discretization of FESOM1.4 (Danilov et al., 2017). Arbitrary Lagrangian–Eulerian (ALE; Petersen et al., 2015; Ringler et al., 2013; White et al., 2008; Danilov et al., 2017) vertical coordinates became an essential part of the numerical core of FESOM2.0. In principle, ALE allows a choice of different vertical discretizations such as geopotential, terrain-following and hybrid coordinates, as well as the usage of a linear free-surface or full free-surface and generalized vertical layer displacement within the same code.