An unconditionally stable, high-order and flux-conservative fluid-fluid coupling method (Q2122035)

Occurrence of extreme weather conditions due to climate change and/or global warming and their attribution to natural variability or human activity or combination of both, has been attracting major attention of climate scientists globally. It is a frontline research area in climatology, due to its implications for protecting lives, livelihoods and critical infrastructure, to securing sustainable supplies of food, water and energy. Climate and Earth system modelling (ESM) is the use of mathematical representations of key components and their properties of the atmosphere, ocean and biosphere to construct computer models that simulate the Earth system are quintessentially based on fundamental physical principles. These models and associated codes (algorithms) are needed to address various issues ranging significantly from achieving stability, flux-conservation and high order accuracy with adequate flexibility in spatial and temporal stepping (resolution), to enhance accuracy in their predictability of the climate behaviour and indicate how they may change in future, especially at regional and local levels. The current article, as claimed by the authors, focuses on one such issue with a desirable and necessary combination of stability, conservation and accuracy not found together with the existing methods. The research work is motivated by a brief literature review revealing some gaps found in the topic under study together with authors recently published work [the authors, Adv. Appl. Math. Mech. 11, No. 6, 1287--1338 (2019; Zbl 1488.65415)] suggesting that there is room for further extension, exploration and improvement. A major shortcoming of the concurrent method presented in [loc. cit.], identified by the authors is freedom to choose longer time intervals without loss of stability (i.e. unconditional stability), a crucial requirement in climate models, despite having all the aforesaid properties. The key highlights of the paper are \begin{itemize} \item[(i)] it is proved rigorously that under the same assumptions as the previous methods, these properties hold. \item[(ii)] It is demonstrated via examples that high order accuracy is obtainable when the analogous method presented in [loc. cit.] is much less accurate or ending in failure of the method (unbounded solution). \item[(iii)] Bulk fluid computations are done advantageously using Crank-Nicolson method a clear deviation from the previous approaches in the literature. \end{itemize} Theoretical and computational overview: As a theoretical background for coupling approaches, authors consider an abstract model of two coupled systems of partial differential equations with appropriate initial and boundary conditions (made precise in the subsequent sections). Simulations are coupled across certain time intervals referred to as coupling windows. Approximations at intermediate times are also considered referred to as sub-stepping. The present theoretical framework allows for sub-stepping and representation of the fluxes using high order, least-square data fitting in time. In an attempt to overcome the limitations of the method proposed in [loc. cit.] requiring coupling windows to be sufficiently small, authors explore a new approach to get flux-conservation, second-order accuracy and unconditional stability simultaneously, keeping in view of the importance of these properties to yield physical fidelity. Unlike methods proposed in the previously published works, Crank-Nicolson as the base time stepping method for the bulk fluid computations is employed to achieve the desired effect. This, according to the authors, is filling up a theoretical gap that has existed for quite some time in the literature for fluid-fluid coupling, which hopefully trigger future research in this area. The second section is devoted to presentation of a model of coupled fluids with natural heat convection, serving the purpose of analysis and evaluation of the proposed method. Mathematical definitions and tools are listed in Section 3. Section 4 defines rigorously the time stepping algorithm (1), after introducing conforming finite element framework in space. In Section 5 rigorous proofs for existence of the discrete solution, the unconditional (independent of step size) convergence of the iterations and the unconditional stability under the same conditions used to analyse its counterpart in [loc. cit.]. Section 6 discusses computational aspects of the method illustrating improved accuracy of results obtained compared with analogous results using the concurrent approach in [loc. cit.]. The model problem for experimenting numerically is already studied in [loc. cit.] and algorithms are implemented using the finite element software. The alternative second-order concurrent method in [loc. cit.] defines algorithm (2). Numerical tests in this section serve two purposes. The first is to provide an illustration of the stability and convergence behaviour of algorithm (1). The second is to demonstrate the advantage of algorithm (1) over algorithm (2). By way of summarizing the main points, the authors seek to highlight the critical innovation of algorithm (1) by enlarging coupling windows. The reference data kinetic energy plots for the air-sea test problem (Fig.1) are shown indicating that air has a faster response to the forcing, but the total kinetic energy is eventually larger for the sea water. This has an important dynamical similarity with an atmosphere-ocean system. The relative errors in computing the kinetic energies of the fluids for algorithm (1) in three cases are plotted in Fig.2 showing that computations are stable in all the cases. Finally, comparison between algorithms 1 and 2, using relative errors in computing kinetic energies of the fluids on sufficiently large coupling windows are plotted in Fig.3. In a nutshell, the main claim of the paper is demonstrated by showing a clear improvement for both accuracy and stability compared to the method [loc. cit.] when computing on large coupling windows. Observations and comments: 1) Climate and Earth System models are based on physical principles, and they reproduce many important aspects of observed climate. Flux (mass, momentum, heat, energy)-conservation property is one such property which complicates the design of the method proposed. 2) Despite great progress in improving the representation of these processes in climate models, the lack of capacity to simulate them in fine detail accounts for the most significant uncertainties in future climate, especially at regional and local levels. As pointed out by the authors, stability has been an elusive issue when it comes to high-order time integration of two fluids (air and sea water in the present case) across an interface. 3) The availability of computing resources has dictated the level of sophistication and the type of simulations that can therefore be performed. Climate models are fundamental to understanding climate change and predicting potential risks. Algorithms developed are implemented using FreeFem++ which is a partial differential equation solver using FEM. It has its own language. freefem scripts can solve multiphysics nonlinear systems of PDEs in 2D and 3D. 4) Multirate methods lower computational cost by using small time steps only for fast evolving components and larger steps elsewhere. The paper presents an iterative algorithm for flux-conservative, fluid-fluid coupling that allows for multirate time stepping. Authors apply the partitioned time stepping to decouple the system. 5) Higher order time integration schemes have a potential gain in efficiency and accuracy in time. To improve computational efficiency, Crank-Nicolson 2nd order implicit time integration schemes as opposed to the traditionally used 1st and 2nd order implicit schemes are investigated. 6) Air-sea coupling is one of the most important processes that can affect weather, climate, and atmospheric chemical composition. The authors have shown that the iterative process of the method ensures consistency of the coupled solution across air-sea interface. This may be viewed as a simplified version of the atmosphere-ocean coupling.