s_algorithm/text/shap.tex

% $Header: $
% $Name: $

\subsection{Shapiro Filter} 

The Shapiro filter (Shapiro 1970, 1975) is a high order horizontal 
filter that efficiently remove small scale grid noise 
without affecting the physical structures of a field. 
It is applied at the end of the time step %(the\_correction\_step),
on both velocity and tracer fields.

Three different space operators are considered here (S1,S2 and S4).
They differs essentially by the sequence of derivative in
both X and Y directions. Consequently they show different 
damping response function specially in the diagonal directions 
X+Y and X-Y.

Space derivatives can be computed in the real space, 
taken into account the grid spacing. 
Alternatively, a pure computational filter can be defined,
using pure numerical differences and ignoring 
grid spacing. 
This later form is stable whatever the grid is, and therefore
specially useful for highly anisotrope grid such as spherical 
coordinate grid.
A damping time-scale parameter $\tau_{shap}$ 
defines the strength of the filter damping.

The 3 computational filter operators are :
$$
\mathrm{S1c:}\hspace{2cm}
[1 - 1/2 \frac{\Delta t}{\tau_{shap}}
   \{ (\frac{1}{4}\delta_{ii})^n 
    + (\frac{1}{4}\delta_{jj})^n \} ] 
$$

$$
\mathrm{S2c:}\hspace{2cm}
[1 - \frac{\Delta t}{\tau_{shap}} 
\{ \frac{1}{8} (\delta_{ii} + \delta_{jj}) \}^n]
$$

$$
\mathrm{S4c:}\hspace{2cm}
[1 - \frac{\Delta t}{\tau_{shap}} (\frac{1}{4}\delta_{ii})^n]
[1 - \frac{\Delta t}{\tau_{shap}} (\frac{1}{4}\delta_{jj})^n]
$$

In addition, the S2 operator can easly be extended to 
a physical space filter: 
$$
\mathrm{S2g:}\hspace{2cm}
[1 - \frac{\Delta t}{\tau_{shap}} 
\{ \frac{L_{shap}^2}{8} \overline{\nabla}^2 \}^n]
$$

with the Laplacien operator $\overline{\nabla}^2 $
and a length scale parameter $L_{shap}$.
The stability of this S2g filter requires 
$L_{shap} < \mathrm{Min}^{(Global)}(\Delta x,\Delta y)$.

\marginpar{Add Response functions and figures}
1	adcroft	1.1	% $Header: $
2			% $Name: $
3
4			\subsection{Shapiro Filter}
5
6			The Shapiro filter (Shapiro 1970, 1975) is a high order horizontal
7			filter that efficiently remove small scale grid noise
8			without affecting the physical structures of a field.
9			It is applied at the end of the time step %(the\_correction\_step),
10			on both velocity and tracer fields.
11
12			Three different space operators are considered here (S1,S2 and S4).
13			They differs essentially by the sequence of derivative in
14			both X and Y directions. Consequently they show different
15			damping response function specially in the diagonal directions
16			X+Y and X-Y.
17
18			Space derivatives can be computed in the real space,
19			taken into account the grid spacing.
20			Alternatively, a pure computational filter can be defined,
21			using pure numerical differences and ignoring
22			grid spacing.
23			This later form is stable whatever the grid is, and therefore
24			specially useful for highly anisotrope grid such as spherical
25			coordinate grid.
26			A damping time-scale parameter $\tau_{shap}$
27			defines the strength of the filter damping.
28
29			The 3 computational filter operators are :
30			$$
31			\mathrm{S1c:}\hspace{2cm}
32			[1 - 1/2 \frac{\Delta t}{\tau_{shap}}
33			\{ (\frac{1}{4}\delta_{ii})^n
34			+ (\frac{1}{4}\delta_{jj})^n \} ]
35			$$
36
37			$$
38			\mathrm{S2c:}\hspace{2cm}
39			[1 - \frac{\Delta t}{\tau_{shap}}
40			\{ \frac{1}{8} (\delta_{ii} + \delta_{jj}) \}^n]
41			$$
42
43			$$
44			\mathrm{S4c:}\hspace{2cm}
45			[1 - \frac{\Delta t}{\tau_{shap}} (\frac{1}{4}\delta_{ii})^n]
46			[1 - \frac{\Delta t}{\tau_{shap}} (\frac{1}{4}\delta_{jj})^n]
47			$$
48
49			In addition, the S2 operator can easly be extended to
50			a physical space filter:
51			$$
52			\mathrm{S2g:}\hspace{2cm}
53			[1 - \frac{\Delta t}{\tau_{shap}}
54			\{ \frac{L_{shap}^2}{8} \overline{\nabla}^2 \}^n]
55			$$
56
57			with the Laplacien operator $\overline{\nabla}^2 $
58			and a length scale parameter $L_{shap}$.
59			The stability of this S2g filter requires
60			$L_{shap} < \mathrm{Min}^{(Global)}(\Delta x,\Delta y)$.
61
62			\marginpar{Add Response functions and figures}