A well-balanced operator splitting based stochastic Galerkin method for the one-dimensional Saint-Venant system with uncertainty

Alina Chertock,Shi Jin,Alexander Kurganov

2015-01-01

Abstract:We introduce an operating splitting based stochastic Galerkin method for the SaintVenant system of shallow water equations with random inputs. The method uses a generalized polynomial chaos approximation in the stochastic Galerkin framework (referred to as the gPC-SG method). It is well-known that such approximations for nonlinear hyperbolic systems do not necessarily yield globally hyperbolic systems: their Jacobians may contain complex eigenvalues and thus trigger instabilities and ill-posedness. In this paper, we overcome this difficulty by extending the gPC-SG method from [A. Chertock, S. Jin and A. Kurganov, An operator splitting based stochastic Galerkin method for the one-dimensional compressible Euler equations with uncertainty, Submitted], where it was developed for the Euler equations of gas dynamics. The main idea is to split the underlying system into a linear hyperbolic system and a nonlinear degenerated hyperbolic system, which can be successively solved as scalar conservation laws with variable coefficients and source terms. The gPC-SG method, when applied to each of these subsystems, results in globally hyperbolic systems. We also demonstrate that when these systems are numerically solved by the second-order semi-discrete central-upwind scheme, the resulting method is well-balanced in the sense that it is capable of exactly preserving “lake at rest” steady states. This is a very important feature of the proposed method since lack of balance between the flux and source terms may cause large artificial waves to emerge in the case the method is used on a coarse grid, while using sufficiently fine mesh may be impractical. The performance of the designed gPC-SG method is illustrated on a number of numerical examples.