Numerical method to solve the Cauchy problem with previous history (Q2362815): Difference between revisions

The present paper investigates the theoretical and practical aspects of constructing a family of single-stage multi-step methods for solving initial value problems formulated as a classic Cauchy problem. These problems model atmospheric processes which are significant in the context of climate changes and global warming. A successful solution of the Cauchy problem by numerical methods requires special numerical techniques, despite the fact that, theoretically speaking, the problem is uniquely solvable, under mild conditions, relaxing differentiability, for instance, by a Lipschitz condition. The authors consider general facts related to discretization, approximation, convergence, and stability. In the authors' opinion the scheme proposed by them has all the good attributes like numerical stability, high-order accuracy/convergence, cost effectiveness, and easy implementability and is as good as -- if not better -- than the best available schemes in the literature. In designing the scheme, the authors' main purpose is to provide a mathematical framework to overcome the essential difficulties and seek improvement in the existing finite difference based single stage multistep schemes, by analyzing the fundamental structural properties of the solution. To this end, to make their approach systematic and as self-contained as possible, all relevant definitions of mesh with constant and variable step length, characteristic polynomials, generating matrices, division algorithms for polynomials, accuracy, and convergence are provided. First, on the basis of their experience with climate modelling, they claim that the advection operator appearing in the equation causes the greatest amount of problems in the application of numerical methods to solve Cauchy-type problems. Second, they claim that non-central spatial difference operators are more effective than schemes in which spatial derivatives are approximated by central difference operators. Third, the nonlinearity of hydrodynamic equations causes nonlinear instability, eventually leading to an explosion of the solution. Single-step methods, besides being computationally convenient, have a low accuracy. Motivated by that, the paper presents the underlying theory. To establish the versatility and universality of the scheme, the construction of the family of multistep single-stage methods on a uniform mesh is undertaken, followed by an extension to a mesh with variable step length. To choose between explicit or implicit methods, single or multistep, and single or multistage methods, in order to obtain the greatest computational efficiency in different situations, is both challenge and opportunity to the researchers. It is important to admit that no method is universally the best one, and the choice of an effective method depends upon a careful consideration of the features of each one. Any investigation of numerical methods must address theory as well as practice. Here both aspects are well considered.