SWAN is a third-generation wave model for obtaining realistic estimates of wave parameters in coastal areas, lakes and estuaries from given wind, bottom and current conditions. However, SWAN can be used on any scale relevant for wind-generated surface gravity waves. The model is based on the wave action balance equation with sources and sinks.

An important question addressed is how to choose various grids in SWAN (resolution, orientation, etc.) including nesting. In general, we consider two types of grids: structured and unstructured. Structured grids may be rectilinear and uniform or curvilinear. They always consist of quadrilaterals in which the number of grid cells that meet each other in an internal grid point is 4. In unstructured grids, this number can be arbitrarily (usually between 4 and 10). For this reason, the level of flexibility with respect to the grid point distribution of unstructured grids is far more optimal compared to structured grids. Unstructured grids may contain triangles or a combination of triangles and quadrilaterals (so-called hybrid grids). In the current version of SWAN, however, only triangular meshes can be employed.

Often, the characteristic spatial scales of the wind waves propagating from deep to shallow waters are very diverse and would required to allow local refinement of the mesh near the coast without incurring overhead associated with grid adaptation at some distance offshore. Traditionally, this can be achieved by employing a nesting approach.

The idea of nesting is to first compute the waves on a coarse grid for a larger region and then on a finer grid for a smaller region. The computation on the fine grid uses boundary conditions that are generated by the computation on the coarse grid. Nesting can be repeated on ever decreasing scales using the same type of coordinates for the coarse computations and the nested computations (Cartesian or spherical). Note that curvilinear grids can be used for nested computations but the boundaries should always be rectangular.

The use of unstructured grids in SWAN offers a good alternative to nested models not only because of the ease of optimal adaption of mesh resolution but also the modest effort needed to generate grids about complicated geometries, e.g. islands and irregular shorelines. This type of flexible meshes is particularly useful in coastal regions where the water depth varies greatly. As a result, this variable spatial meshing gives the highest resolution where it is most needed. The use of unstructured grids facilitates to resolve the model area with a relative high accuracy but with a much fewer grid points than with regular grids.

It must be pointed out that the application of SWAN on ocean scales is not recommended from an efficiency point of view. The WAM model and the WAVEWATCH III model, which have been designed specifically for ocean applications, are probably one order of magnitude more efficient than SWAN. SWAN can be run on large scales (much larger than coastal scales) but this option is mainly intended for the

A general suggestion is: