SmoothBatimNesting

Animation consisting of three frames showing respectively i) the father bathymetry, ii) the son bathym, iii) and the interpolated bathym.

This program creates a new sub-model bathimetry to allow a smooth transition in the boundary between the coarser model and the high resolution model. See also SmoothBatimCoefs.

Quick start

1. Create the father and son bathymetries (you may use Mohid GIS).
2. Create the SmoothCoeficients griddata file (you may use SmoothBatimCoefs).
3. Create and edit the options SmoothBatimNesting.dat file and save it in the same path as the executable.
4. Run the executable.

Options file

Here's what the options file SmoothBatimNesting.dat looks like:

FATHER_BATIM                : ..\..\WestIberiaTide\GridData.dat
SON_BATIM                   : ..\GridData_2.dat
SMOOTH_COEF                 : SmoothCoef.dat
NEW_SON_BATIM               : SmoothData.dat

Smoothing algorithm

Let it be $X \in \mathcal{F}\equiv\left\{F,\,S,\,I\right\}$, as an element of the father, the son and the interpolated bathymetries discrete fields. Let it be $I_X : \mathbb{N}^2 \longrightarrow \mathbb{R}^2$ as the respective index mapping function. Let it be $X^{\prime}: \mathbb{R}^2 \longrightarrow \mathbb{R}$, a continuous map, so that $X \equiv X^\prime\circ I_X,\;\forall X \in \mathcal{F}$. Let $r_X \subset \mathbb{R}^2$ define the work domain of $X^\prime \in \mathcal{F}^\prime$. In particular, $X^\prime$ verifies $X^\prime(\overline{r_X})\equiv 0$. The son grid is the same as the interpolated grid, meaning $I_S = I_I$. However, the father field is completely independent from the other fields, meaning that $r_{FS}\equiv\left(r_F \cap r_S\right) \subseteq \mathbb{R}^2$. The smoothing algorithm is a simple linear interpolation of the father and son fields:

$I^\prime = \left(1 - C^\prime\right) F^\prime + C^\prime\,S^\prime$

where the smoothing coefficient is given by $C^\prime: \mathbb{R}^2 \longrightarrow \left[0\;1\right]$. Thus, the smoothing coefficient gridded counterpart $C$ is defined; and its index mapping function $I_C$ is equal to the son's; meaning $I_C = I_S$. Let us define the grid $G \subseteq \mathbb{N}^2$ such that $I_S(G)\subseteq r_S$ and $I_S(G)\rightarrow r_S$. Thus the smoothing coefficient $C$ needs can be defined in $G$. Concretely speaking, $G \mapsto (I_S(G),C(I_S(G)))$ is the smoothing coefficient's Grid Data file described above, thus it leaves to the modeller the choice of the smoothing coefficient field.

Implementation

Let us define a collection of four-vertices polygons taken from the father grid:

$\begin{matrix} V : & \mathbb{N}^2 \longrightarrow \mathbb{R}^8 \\ & \left(i,\,j\right) \mapsto (I_{F\;i\,j},\,I_{F\;i+1\,j},\,I_{F\;i\,j+1},\,I_{F\;i+1\,j+1})\end{matrix}$

where $D\left(V_{i\,j}\right)\subset \mathbb{R}^2$ is the subdomain defined by the vertices of $V_{i\,j}$. Let us define an auxiliary grid of the son grid given by:

$\forall\; _{ij} \in \mathbb{N}^2,\;\tilde{I}_{S\;i\,j}=\frac{1}{4}\left(I_{S\;i\,j}+I_{S\;i+1\,j}+I_{S\;i\,j+1} +I_{S\;i+1\,j+1}\right)$.

Let us define a new mapping into a subset of $F$'s image:

\$\forall\;_{i\,j} \in \mathbb{N}^2\quad \exists !\;_{u\,v} \in \mathbb{N}^2\; :\quad \tilde{I}_{S\;i\,j} \in D(V_{u\,v}) \wedge \tilde{F}_{i\,j}=F_{u\,v}$.

Here's the implemented smoothing algorithm:

$I_{i\,j} = \left(1 - C_{i\,j}\right) \tilde{F}_{i\,j} + C_{i\,j}\,S_{i\,j}$.

Test image

Animation consisting of three frames showing respectively i) the father bathymetry, ii) the son bathym, iii) and the interpolated bathym.