A flux-corrected RBF-FD method for convection dominated problems in domains and on manifolds (Q2307649)

Partial differential equations are a vast field of applications for numerical approximations using kernel-based methods such as radial basis function approximations. The numerical approximations of these solutions of PDEs with radial basis functions can be combined with finite difference schemes to end up in so-called RBF-FD algorithms. In this article, these methods are applied to parabolic partial differential equations, transport problems. They are the basis of a flux-corrected transport method which is described and applied in the paper. One of the advantages of employing radial basis function methods is the availability of interpolants or other with scattered centres that is a major plus whenever difficult PDEs are to be solved in a numerical scheme (because they can, for example, be assembled close to the boundary of a difficult domain or around singularities). Another advantage is the possibility to develop the methods in general dimensions, not restricted to two or three as usually, for instance, the finite element methods are. Finally, collocation (interpolation) schemes are readily available since many of the radial basis function kernels are positive definite or at least conditionally positive definite. This also helps with the so-called positivity preservation which is important in practical applications. The authors make good use of these properties in order to develop a new algorithm. As is usual, the authors apply different discretisations in space and time and then -- after some numerical stabilisation -- apply the schemes to their various numerical examples.