A semi-implicit fully exactly well-balanced relaxation scheme for the shallow water system (Q6585314)

The articles studies the numerical discretization of the one-dimensional shallow-water system (also termed Saint-Venant equations), with a topography source term. Stationary (i.e. time-independent) solutions are particularly relevant for applications to straight ideal rivers.\N\NHere, the authors propose the so-called ``well-balanced'' numerical schemes, i.e., schemes that exactly preserve the steadiness of (discrete approximations to smooth) stationary solutions, including in the case of non-zero discharge/velocity (beyond the more usual case of ``water-at-rest'' equilibria).\N\NBuilding on their previous work [\textit{C. Caballero-Cárdenas} et al., Appl. Math. Comput. 443, Article ID 127784, 24 p. (2023; Zbl 1511.76059); \textit{M. J. Castro} and \textit{C. Parés}, J. Sci. Comput. 82, No. 2, Paper No. 48, 48 p. (2020; Zbl 1440.65109)], the authors first propose a semi-discrete (time-continuous) scheme (2.22)--(2.23)--(2.18) that is well balanced using formulas (2.25)--(2.26) for the numerical flux, cf. Theorem 2.1.\N\NThe numerical scheme (2.22)--(2.23)--(2.18) builds on existing discretization approaches of the one-dimensional shallow-water system, using a relaxation of the nonlinear pressure term (similar to Suliciu's relaxation). It consists in a (first-order) splitting of the time evolution (i.e. of the fluxes) into acoustic and then in transport steps -- the transport of the relaxed pressure is thus neglected. The source term is taken into account through terms explicitly written as functions of a (discrete) stationary solution -- i.e., the latter should be known (computed).\N\NFormulas (2.25)--(2.26) for the numerical fluxes reconstruct interface values. They are written as fluctuations around a stationary state, so that the resulting (semi-discrete) scheme is actually well-balanced, cf Theorem 2.1.\N\NLast, explicit and semi-implicit fully-discrete versions are proposed (the acoustics step could be implicited, for a less stringent CFL condition) and numerically tested.