    Next: 3.2.1 Upwind Scheme Up: 3. Surface Evolution Previous: 3.1.3 The Level Set

# 3.2 Solving the Level Set Equation

In this section numerical schemes for solving the LS equation are discussed. Since the LS equation belongs to the class of Hamilton-Jacobi equations for (3.4)

with denoting the Hamiltonian, many numerical schemes developed for the solution of Hamilton-Jacobi equations can also be applied to the LS equation.

Generally, the solution of a partial differential equation requires a discretization in time as well as in space. In case of the LS equation the simulation domain is usually discretized using regular grids which allow the application of simple finite difference schemes. In the following, denotes the grid spacing. Individual grid points are identified by an index vector , where denotes the set of index vectors of all grid points of the discretized simulation domain.

For regular grids the finite difference method is the most suitable method to solve the LS equation. As the name suggests, it approximates derivates by finite differences. The discretized version of (3.4) using the first-order forward Euler method  writes as (3.5)

Here is an appropriate finite difference approximation to the Hamiltonian of the LS equation evaluated at grid point . depends on the discretized values of the LS function at time and also on the current value of the velocity field at grid point .

In the following discussion different approximations for the Hamiltonian are presented, which result in different time integration schemes. Since these approximations are based on finite differences, it is useful to introduce a notation for finite differences. The central difference operator is represented by  (3.6)

where is the unit vector along the -th grid direction. Furthermore, the forward difference and the backward difference operator are defined as (3.7)

Subsections    Next: 3.2.1 Upwind Scheme Up: 3. Surface Evolution Previous: 3.1.3 The Level Set

Otmar Ertl: Numerical Methods for Topography Simulation