A parallel unstructured mesh model for simulations of stratified flows

The present work extends the computational capabilities of semi-implicit finite volume (FV) non-oscillatory forward-in-time (NFT) solvers for simulating a range of stratified flows past blunt bodies.The numerical model is based on the Multidimensional Positive Definite Advection Transport Algorithm (MPDATA) and employs a non-symmetric Krylov-subspace elliptic solver.A parallel version of the scheme working on fully unstructured meshes has been developed that enables numerical studies of stably stratified flows past isolated and closely positioned objects, exploiting the flexibility and adaptivity attributes of the employed meshes.Flow structures induced in flows past spheres and hills are discussed especially for strong stratification, with computations of flows past a single sphere providing qualitative and quantitative means for validating the numerical approximations.Investigations of flows past two spheres reveal a range of flow patterns also induced in flows past hills, illustrating the unstructured mesh NFT-FV scheme's potential for simulations of atmospheric flows.Developments in elliptic solver preconditioning techniques and immersed boundary method applied to the NFT-FV scheme are documented for strongly stratified flows past hills.In particular, simulations of orographic flows with critical layers are a computationally demanding benchmark where computations' accuracy improves by applying tailored preconditioning techniques.Furthermore, the benchmark of strongly stratified flow past an isolated hill is introduced to provide a validation of the immersed boundary method implementation which opens new perspectives for applications of the unstructured-mesh NFT-FV scheme.