3D cut-cell modelling for high-resolution atmospheric simulations

TL;DR

This paper introduces a high-resolution 3D atmospheric model using cut-cell terrain representation, cell-merging, and adaptive mesh refinement, enabling accurate simulation of flows over steep terrain with good scalability.

Contribution

It develops a novel 3D Cartesian model with cut-cell terrain, cell-merging, and adaptive mesh refinement for high-resolution atmospheric simulations.

Findings

Successfully simulates flow over steep terrain with high accuracy.

Demonstrates good scalability on locally refined grids.

Reproduces mountain waves without large errors.

Abstract

Owing to the recent, rapid development of computer technology, the resolution of atmospheric numerical models has increased substantially. With the use of next-generation supercomputers, atmospheric simulations using horizontal grid intervals of O(100) m or less will gain popularity. At such high resolution more of the steep gradients in mountainous terrain will be resolved, which may result in large truncation errors in those models using terrain-following coordinates. In this study, a new 3D Cartesian coordinate non-hydrostatic atmospheric model is developed. A cut-cell representation of topography based on finite-volume discretization is combined with a cell-merging approach, in which small cut-cells are merged with neighboring cells either vertically or horizontally. In addition, a block-structured mesh-refinement technique is introduced to achieve a variable resolution on the model grid with the finest resolution occurring close to the terrain surface. The model successfully reproduces a flow over a 3D bell-shaped hill that shows a good agreement with the flow predicted by the linear theory. The ability of the model to simulate flows over steep terrain is demonstrated using a hemisphere-shaped hill where the maximum slope angle is resolved at 71 degrees. The advantage of a locally refined grid around a 3D hill, with cut-cells at the terrain surface, is also demonstrated using the hemisphere-shaped hill. The model reproduces smooth mountain waves propagating over varying grid resolution without introducing large errors associated with the change of mesh resolution. At the same time, the model shows a good scalability on a locally refined grid with the use of OpenMP.