Approximation of the Lévy-Feller advection-dispersion process by random walk and finite difference method (Q870598)

Consider the non-random Lévy-Feller advection-dispersion equation (LFADE) \[ \frac{\partial u(x,t)}{\partial t} = a D^\alpha_\theta u(x,t) - b \frac{\partial u(x,t)}{\partial x} \] where \(a > 0\), \(b \geq 0\), \(x \in \mathbb{R}\) (or \(0 < x < L\)), \(t > 0\), and \(D^\alpha_\theta\) is the Riesz-Feller fractional derivative (in space) of order \(\alpha\) (\(1 < \alpha \leq 2\)) and skewness \(\theta\) (\(| \theta| \leq 2 - \alpha\)), subject to initial condition \(u(x,0)=\varphi (x)\). A random walk model for approximating the solution \(u\) governed by (LFADE) is presented. This random walk model converges to model (LFADE) by use of a properly scaled transition to vanishing equidistant space and time steps. An explicit finite difference approximation (EFDA) for (LFADE), resulting from the Grünwald-Letnikov discretization of fractional derivatives, is proposed. As a result of the interpretation of the random walk model, the stability and convergence of (EFDA) for (LFADE) in a bounded domain are discussed. Finally, some numerical examples show the application of the presented techniques.