source: production/onslow_2006/report/computational_setup.tex @ 3160

To initiate the modelling, a computational triangular mesh is constructed to
cover the study regions which has an area of around 6300 km$^2$. 
The cell size is chosen to balance
computational time and desired resolution in areas of interest,
particularly in the interface between the on and offshore. 
Figure \ref{fig:onslow_area} illustrates the data extent for the 
scenario, the study area and where further mesh refinement has been made. 
The choice
of the refinement is based around the important inter-tidal zones and
other important features such as islands and rivers. The most northern
boundary of the study area is placed approximately around the 100m contour
line. 

\begin{figure}[hbt]

  \centerline{ \includegraphics[width=100mm, height=75mm]
             {../report_figures/onslow_data_poly.png}}

  \caption{Study area for Onslow scenario highlighting areas of increased 
refinement.
The underlying data is as in Figure \ref{fig:onslowdataarea}.}
  \label{fig:onslow_area}
\end{figure}

For the simulations, we have chosen a cell area of 500 m$^2$ per triangle 
for the 
region surrounding the Onslow town centre. It is worth noting here that 
the cell 
area will be the maximum cell area within the defined region and that each 
cell in
the region does not necessarily have the same area. The cell area is increased 
to 2500 m$^2$ for the region surrounding the coast and further increased
to 20000 m$^2$ for the region reaching approximately the 50m contour line.
The remainder of the study area has a cell area of 100000 m$^2$.
The resultant computational mesh is then seen in Figure \ref{fig:mesh_onslow}.

With these cell areas in place, the study area consists of 440150 triangles
in which water levels and momentums are tracked through time.
The associated lateral accuracy 
for these cell areas is approximatly 30m, 70m, 200m and 445m for the respective
areas. This means
that we can only be confident in the calculated inundation extent to 
approximately 30m lateral accuracy within the Onslow town centre. 
Referring to the discussion in Section \ref{sec:anuga}, it is important 
to refine the mesh to be commensurate with the underlying data especially in 
those regions where complex behaviour will occur, such as the inter-tidal
zone and estuaries. Our choice of cell area for the region surrounding the
Onslow town centre is commensurate with the onshore data used for this study 
(see Section \ref{sec:data}). In contrast to the onshore data, the offshore
data is a series of survey points which is typically not supplied on a fixed 
grid which complicates the issue of determining an appropriate cell area. 
If we refer to the discussion in Section \ref{sec:data} 
on modelling a tsunami wave in deep water, we can determine an appropriate 
cell area for the deeper water. Here,
the wavelength of the tsunami wave is approximately 20km
near the boundary, which indicates that our cell area is more than adequate
to propagate the tsunami wave.

\begin{figure}[hbt]

  \centerline{ \includegraphics[width=100mm, height=75mm]
              {../report_figures/mesh.jpg}}

  \caption{Computational mesh for Onslow study area.}
  \label{fig:mesh_onslow}
\end{figure}

To complete the model setup, we describe the form of the
tsunami wave from the earthquake source described
in Section \ref{sec:tsunamiscenario}. The Method of Splitting Tsunami (MOST)
was used to initiate the event and propagate the wave in deep water. The
waves amplitude and velocity is then determined at the 
boundary of the study area (see Figure \ref{fig:onslow_area}) whereupon
ANUGA continues to propagate the tsunami wave in shallow water and onshore.
To illustrate the form of the tsunami wave, we select a number
of point locations outside, on and within the boundary (shown in 
Figure \ref{fig:boundarypoints}) and show the moving
wave amplitude and speed. 

\input{MOST_input_onslow}