MIKE ZERO: Creating 2D Bathymetries

Bathymetry Editor & Mesh Generator

Scientific Documentation

MIKE by DHI 2014

Agern Allé 5
DK-2970 Hørsholm
Denmark

Tel: +45 4516 9200

Support: +45 4516 9333
Fax: +45 4516 9292

mikebydhi@dhigroup.com
www.mikebydhi.com
PLEASE NOTE
COPYRIGHT This document refers to proprietary computer software,
which is protected by copyright. All rights are reserved.
Copying or other reproduction of this manual or the related
programs is prohibited without prior written consent of DHI.
For details please refer to your 'DHI Software Licence
Agreement'.
LIMITED LIABILITY The liability of DHI is limited as specified in Section III of
your 'DHI Software Licence Agreement':
'IN NO EVENT SHALL DHI OR ITS REPRESENTA-
TIVES (AGENTS AND SUPPLIERS) BE LIABLE FOR
ANY DAMAGES WHATSOEVER INCLUDING,
WITHOUT LIMITATION, SPECIAL, INDIRECT,
INCIDENTAL OR CONSEQUENTIAL DAMAGES OR
DAMAGES FOR LOSS OF BUSINESS PROFITS OR
SAVINGS, BUSINESS INTERRUPTION, LOSS OF
BUSINESS INFORMATION OR OTHER PECUNIARY
LOSS ARISING OUT OF THE USE OF OR THE
INABILITY TO USE THIS DHI SOFTWARE PRODUCT,
EVEN IF DHI HAS BEEN ADVISED OF THE
POSSIBILITY OF SUCH DAMAGES. THIS
LIMITATION SHALL APPLY TO CLAIMS OF
PERSONAL INJURY TO THE EXTENT PERMITTED
BY LAW. SOME COUNTRIES OR STATES DO NOT
ALLOW THE EXCLUSION OR LIMITATION OF
LIABILITY FOR CONSEQUENTIAL, SPECIAL,
INDIRECT, INCIDENTAL DAMAGES AND,
ACCORDINGLY, SOME PORTIONS OF THESE
LIMITATIONS MAY NOT APPLY TO YOU. BY YOUR
OPENING OF THIS SEALED PACKAGE OR
INSTALLING OR USING THE SOFTWARE, YOU
HAVE ACCEPTED THAT THE ABOVE LIMITATIONS
OR THE MAXIMUM LEGALLY APPLICABLE SUBSET
OF THESE LIMITATIONS APPLY TO YOUR
PURCHASE OF THIS SOFTWARE.'

PRINTING HISTORY January 2009 ..................................................... Edition 2009

November 2009 ................................................. Edition 2009
July 2010 ........................................................... Edition 2011
July 2012 ........................................................... Edition 2012
August 2013 ....................................................... Edition 2014

MIKE by DHI 2014

2013-08-26/BATHYMETRYGENERATION_SCIENTIFICDOC.DOCX/JNH/JFB/AJS
CONTENTS

MIKE ZERO: Creating 2D Bathymetries

Bathymetry Editor & Mesh Generator
Scientific Documentation

1 INTRODUCTION .............................................................................................. 1

2 GRID GENERATION ....................................................................................... 3

3 GRID INTERPOLATION ................................................................................. 5

3.1 Land Generation ...................................................................................................................... 5
3.1.1 Polygon Fill ............................................................................................................................. 5
3.1.2 Extensive ................................................................................................................................. 6
3.2 Box Grouping .......................................................................................................................... 6
3.3 Gap Filling .............................................................................................................................. 6
3.3.1 Bilinear Interpolation .............................................................................................................. 7
3.3.2 Triangular Interpolation ........................................................................................................ 10
3.3.3 Inverse Distance Weighted Interpolation .............................................................................. 11
3.3.4 Inverse Squared Distances Weighted Interpolation .............................................................. 11

4 MESH GENERATION .................................................................................... 13

4.1 Triangular Mesh Generation ................................................................................................. 13
4.2 Quadrangular Mesh Generation ............................................................................................ 13
4.2.1 Algebraic Box Method .......................................................................................................... 14
4.2.2 Transfinite Interpolation........................................................................................................ 18
4.3 Generating Combined Mesh ................................................................................................. 20
4.4 Smoothing the Mesh ............................................................................................................. 24
4.5 Refinement by Bisection ....................................................................................................... 25

5 MESH INTERPOLATION .............................................................................. 27

5.1 Interpolation Accommodating the Stream Direction ............................................................ 28
5.2 Interpolation using Prioritization of Scatter Data ................................................................. 31

6 MESH ANALYSIS AND EDITING ............................................................... 33

6.1 Mesh Analysis ....................................................................................................................... 33
6.2 Collapsing Elements.............................................................................................................. 34

7 REFERENCES ................................................................................................. 37

i
ii MIKE ZERO: Creating 2D Bathymetries
1 INTRODUCTION
The Bathymetry Editor and the Mesh Generator provides you an
environment for creating, editing and presenting detailed digital 2D
bathymetries. The Bathymetry Editor generates bathymetries in a
rectangular grid (dfs2) whereas the Mesh Generator generates
bathymetries in a flexible mesh format (mesh).

The program provides you the utilities for importing raw data from
various external source (i.e. xyz soundings, xyz contours, MIKE 21 /
MIKE 3 formatted data), or to manually create data by using the built-
in drawing tools. To aid the process of manually drawing data and for
presentation, graphical background images such as maps can be
imported and overlaid with the bathymetry data.

Figure 1.1 Example of import of raw data into a Bathymetry Editor workspace

Various interpolation options are available to you to provide the best

possible method for your type of data. When the bathymetry has been
prepared, you can use the export utilities to output bathymetry data in
various formats.

The purpose of this document is to provide the user with the scientific
background for the 2D bathymetry generation.

1
 The generation of a bathymetry is generally divided into two phases:

• The generation of the grid/mesh

• The interpolation of bathymetry values onto the grid/mesh

As a consequence this document is divided in similar way: the

generation of a dfs2 bathymetry in the Bathymetry Editor is described
in the sections Grid Generation and Grid Interpolation whereas the
generation of a mesh bathymetry is described in the sections Mesh
Generator and Mesh Interpolation.

2 MIKE ZERO: Creating 2D Bathymetries

2 GRID GENERATION
The bathymetry area to describe the dfs2 bathymetry is defined by a
map projection, the origin of the grid, i.e. the center of cell (0.0), the
orientation of the grid and the extension of the grid, defined by the
number of grid steps and equidistant grid spacing in the x- and y-
direction, respectively.

In the process of creating the grid the limiting bathymetry value for a
cell to be recognized as a land cell must be defined.

The resulting grid will be a rectangle resolved with rectangular

elements (often squares).

3
4 MIKE ZERO: Creating 2D Bathymetries
3 GRID INTERPOLATION
Before the interpolation starts an empty bathymetry is created. This
means that it contains only delete values. The interpolation of a
bathymetry is done in 3 steps:

• Finding the grid cells that have a centre inside the land polygons
identifies all the land points. This is known as Land Generation.

• The grid cells that aren’t land points needs to be assigned a depth.
The grid is used for sorting the data (loose points, contour points
and polygon vertices). All points used for the interpolation are
distributed into optional lookup-tables for each grid cell. This
enables a much more efficient search instead of something which
is proportional to number of horizontal grid cells times number of
vertical grid cells times number of raw data points. This is known
as Box Grouping

• Only grid cells defining land have been assigned an elevation (z-
value). Each of the remaining points needs to be assigned. The raw
data points are used for this interpolation. This is the Gap Filling
process.

3.1 Land Generation

Land generation is based on an input data set that consists of any
number of land polygons.

3.1.1 Polygon Fill

The polygon fill algorithm defines a set of scan lines aligned with the
bottom of the bathymetry. There is a scan line defined for each k-grid
cell. For each scan line the list of polygons is searched and points
inside these polygons are marked as land points according to the
algorithm described on the two figures below. This is by far the most
efficient algorithm for identifying cells inside polygons.

5
 Figure 3.1 Polygon Sweep-Line Algorithm

3.1.2 Extensive

The extensive method for computing land points loops over the total
number of grid points in a bathymetry and check versus inside/outside
boundaries of the polygon. This is significantly more time consuming
than the ‘Polygon fill’-algorithm.

3.2 Box Grouping

Group by Cell Index

All raw data points are attached to the grid cell that they are located in.
By doing this it is very easy to search through the raw data points and
identifying the points located close to the grid point.

3.3 Gap Filling

The gap filling is based on the concept that we have to calculate the
depth in the point (xc, yc). We define this as the function zc = f (xc, yc).
If we place our self in this point, we can divide the world up into four
quadrants Q1 - Q4. From here it’s a matter of finding some points
from the raw data set relatively close to this point. The search radius
for all possible techniques can be entered - in grid cell distance. Points
outside this distance will never be taken into account.

6 MIKE ZERO: Creating 2D Bathymetries

 Figure 3.2 Definition of quadrants

3.3.1 Bilinear Interpolation

This technique finds four points from the raw data set - one in each
quadrant. The search is done in the following way. A mask of relative
indices is created. The cells in this mask are sorted according to the
distance. For the quadrant Q1 the cells are sorted in the following way,
the grid point itself being excluded.

Figure 3.3 Illustration of the neighbouring grid cells being sorted

Note that the grid cells with a crosshatch pattern contain raw data
points. When the closest raw data point in each quadrant is found, we
have four points that form a quadrangle. This quadrangle contains the
centre point, where we want to calculate the z-value. This is illustrated
in Figure 3.4.

7
 Figure 3.4 Illustration of the closest raw data points in each quadrant

Note that each grid cell might contain more raw data points. If this is
the case, the closest of these is chosen. We now have an irregular
quadrangle, where the elevation is defined in each vertex. We need to
compute the elevation in (xc, yc). If we transform our quadrangle into a
square, we can perform bilinear interpolation. This is illustrated in
Figure 3.5.

Figure 3.5 Illustration of bilinear interpolation

First the interpolation requires the transformation from quadrangle to a

normalized square. This is done in the by computing 8 coefficients in
the following way:

8 MIKE ZERO: Creating 2D Bathymetries

 A1  x0
A2  y 0
B1  x1  x0
B2  y1  y 0
(3.1)
C1  x3  x0
C 2  y3  y0
D1  x 2  x1  x 0  x3
D2  y 2  y1  y 0  y 3

Mapping the coordinates (xc, yc) to the normalized square (dx, dy) is
done by solving Equation (3.2).

ax 2  bx  c  0 (3.2)

Where the coefficients are

a  D1 B2  D2 B1
b  D2 xc  D1 y c  D2 A1  D1 A2  C1 B2  C 2 B1 (3.3)

c  C 2 xc  C1 y c  C1 A2  C 2 A1

Solving Equation (3.2) gives us dx

 b  b 2  4ac
dx  (3.4)
2a

Where 0  dx  1 is used to choose the correct root. dy can now be

computed in two ways:

xc  A1  B1 dx
dy  (3.5)
C1  D1 dx

or

xc  A2  B2 dx
dy  (3.6)
C 2  D2 dx

Choosing between Equations (3.5) and (3.6) is done in such a way that
division by zero is avoided. (xc, yc) has been mapped to (dx, dy). The
task was to compute the elevation in the point (xc, yc) and this is done
in the following way using regular bilinear interpolation:

9
 z c  1  dx 1  dy z 2  dx1  dy z3  1  dx dy z1  dxdy z 0
(3.7)

If less than four points are found (if one or more quadrants are empty),
the double linear interpolation is replaced with inverse distance
weighted interpolation. This is done according to the following
scheme:

1
wi 
xi  xc 2   yi  yc 2
(3.8)

N
ws   wi (3.9)
i 1

N
1
zc 
ws
w z
i 1
i i (3.10)

The method works fairly efficiently, but it has one drawback. The
quadrant search is heavily dependent on the orientation of the
bathymetry. If the bathymetry is rotated 45 degrees 4 completely
different points might be used for the interpolation. For this reason
there is also a Triangular interpolation method, which can be used, and
this method should be direction independent.

3.3.2 Triangular Interpolation

As mentioned previously the ‘Bilinear Interpolation’ is dependent on
the orientation of the bathymetry. The ‘Triangular Interpolation’ is
made as an answer to this problem. First the closest point to (xc, yc) is
found, cf. Figure 3.6.

Figure 3.6 Illustration of triangular interpolation

10 MIKE ZERO: Creating 2D Bathymetries

 In this example the point (x0, y0, z0) is the closest point. When this
point is identified, two quadrants are identified – indicated by the light
grey and the dark grey areas. The closest points in these two quadrants
are then found. They can be seen on the figure as (x1, y1, z1) and (x2, y2,
z2). The interpolation is then done in two steps.

First the coefficients describing the plane defined by the 3 found

points are computed:

  y1  y 0 z 2  z 0    y 2  y 0  z1  z 0 
A
x1  x0  y 2  y0   x2  x0  y1  y0 

B
x1  x0 z 2  z 0   x2  x0 z1  z 0  (3.11)
x1  x0  y 2  y0   x2  x0  y1  y0 

C  z 0  Ax 0  By 0

And secondly, the actual interpolation is done:

z c  Ax c  By c  C (3.12)

If less than 3 points are found, inverse distance weighted interpolation

is used. The triangular interpolation is more time consuming due to the
more complex direction independent search, but better end results
should be achieved with this method.

3.3.3 Inverse Distance Weighted Interpolation

The ‘Inverse Distance Weighted Interpolation’ is based on the scheme

described in Equations (3.8) - (3.10).

Please note that the quadrant search is heavily dependent on the

orientation of the bathymetry. Thus, if the bathymetry is rotated 45
degrees four completely different points may be used for the
interpolation.

3.3.4 Inverse Squared Distances Weighted Interpolation

The ‘Inverse Squared Distance Weighted Interpolation’ is based on the

‘Inverse Distance Weighted Interpolation’ but using the squared
distance points in the search criteria. The applied scheme is thus
defined by:

11
 1
wi 
 xi  x c    yi  yc 
2 2 (3.13)

N
ws   wi (3.14)
i 1

N
1
zc 
ws
w z
i 1
i i (3.15)

Please note that the quadrant search is heavily dependent on the

orientation of the bathymetry. Thus, if the bathymetry is rotated 45
degrees four completely different points may be used for the
interpolation.

12 MIKE ZERO: Creating 2D Bathymetries

4 MESH GENERATION
The mesh generator can construct meshes that consist of both
triangular and quadrangular elements. The approach being that the
area of interest is divided up into regions described through polygons.
Each polygon may have a triangular or a quadrangular mesh generated
within. The complementary area is by default populated with a
triangular mesh.

The steps to generate such a mesh are the following:

• Define polygons to be used for quadrangular mesh

• Set properties for each (default values used if local properties are
not supplied)

• Generate the mesh within each polygon

• Use the triangular mesh approach for the area not contained within
any of the polygons.

Thus the overall idea is to generate the quadrangular mesh first and
then to use the triangular mesh to patch the mesh together.

4.1 Triangular Mesh Generation

The generation of triangles is based on the ‘Triangle’ code, developed
by Jonathan Shewchuk /1/.

Additional information regarding the functionality can be found on the

web page http://www.cs.cmu.edu/~quake/triangle.html.

4.2 Quadrangular Mesh Generation

The quadrangular grid generation may be carried out using two
different routines:

• A simple algebraic method using a boxing technique

• An algebraic grid generator using transfinite interpolation

The first of these is robust but may generate skewed grids. The second
is efficient but may not always be successful.

13
4.2.1 Algebraic Box Method

The aim is to generate a mesh for a polygon which aligns itself with
the polylines that make up the polygon. The main idea is to break any
closed polygon into quadrangles. Below is a polygon consisting of two
polylines l1 and l2 with a number of vertices. The number of vertices
on each polyline may be unequal in general. The polylines are joined
by two arcs. These end arcs consist each of only one line segment. The
polygon is considered as an abstraction of a stream tube i.e. the flow
direction is in the direction of the two polylines and the end arcs are
transversal to the flow direction. Ideally the end arcs should be chosen
so that these are perpendicular to the flow direction.

Vertex
Node
l1 End arc

l2

Figure 4.1 The polygon to be meshed. The vertices are displayed along with the
end arcs

The procedure is as follows: The length of the two polylines is

determined and the position of the vertices along these polylines is
given relative to this length i.e. the position of the vertices may thus be
identified by a scalar value in the range 0,…,1. Let this scalar value be
given by sn,m where n = 1,2 (polyline index) and m is the index of the
vertex along the polyline. Each of the vertices along the polyline has a
unique value of sn,m. The union set of sn,m values of the vertices along
l1 and l2 are then considered. For each of the two polylines vertices are
introduced at the values sn,m, if they are not already present along the
polyline. This will give rise to a polygon where the two polylines have
the same number of vertices. Vertices with the same value of s will be
referred to as a complementary pair. By joining the complementary
vertex pairs the polygon is divided in a number of quadrangles. Each
of these quadrangles may then be treated individually.

14 MIKE ZERO: Creating 2D Bathymetries

 Vertex
Node
End arc
Interpolated
complementary vertex

Figure 4.2 The polygon with the interpolated complementary vertices displayed.
The polygon has now been divided into a number of quadrangles

Figure 4.3 A mesh is applied for each of the quadrangles in the polygon. Note that
the number of intervals in the t direction is fixed for all individual boxes
whereas the number in the s direction may vary significantly

To generate the mesh for each of the quadrangles the overall

maximum mesh side length should be considered. It must be possible
to control the element side length transversal to the stream direction as
well as in the stream direction. Let the two maximum element side
lengths be denoted by t and s. These values apply for all the
quadrangles in the polygon. The number of mesh lines in the direction
of the stream needs to be the same for all the quadrangles making up
the polygon. Thus to evaluate the number of mesh lines in the stream
direction requires evaluating the maximum distance between all
complementary vertex pairs. The number of intervals in the t direction
is given by

 vi 
N t  Ceiling  max i 1,...M 
 (4.1)
 t
 

15
 Where v i is the length of the vertex pair i.

S2
(x0,Nt,y0,Nt) (xNs,Nt,yNs,Nt)

(xNs,0,yNs,0)
(x0,0,y0,0) S1

Figure 4.4 Notation used for the individual quadrangles

The notation for the quadrangles follows from the above figure. The
individual quadrangles should be meshed in the following way. First
the number of intervals in the s-direction is determined through

 max( s1 , s 2 ) 
N s  Ceiling   (4.2)
 s 

Where

s1  ( x Ns,0 , y Ns,0 )  ( x0,0 , y0,0 )

(4.3)
s 2  ( x Ns, Nt , y Ns, Nt )  ( x0, Nt , y0, Nt )

Once the number intervals is known the node locations may be found
through

16 MIKE ZERO: Creating 2D Bathymetries

  i  j  i  j 
xi , j  1  1   x0, 0  1   x N s , 0 
 N s  N t  Ns  Nt 
j  i  i j
1   x0, N t  xN ,N
Nt  Ns  N s Nt s t
(4.4)

 i  j  i  j 
y i , j  1  1   y 0, 0  1   y N s , 0 
 N s  N t  Ns  Nt 
j  i  i j
1   y 0, N t  y N ,N
Nt  Ns  N s Nt s t

Where

i  0,, N s and j  0,, N t (4.5)

Problematic sections may arise for some polygons. Such a polygon is

displayed below.

Figure 4.5 A problematic polygon

Problematic sections may be identified as sections with opposing sides

having a scalar product which is non-positive. Using the notation from
the earlier figure the criteria is expressed as

 x Ns,o  x0,0   x Ns, Nt  x0, Nt 

      0 (4.6)
y
 Ns,o  y y
0 , 0   Ns, Nt  y 0 , Nt 

17
4.2.2 Transfinite Interpolation

The second option uses a so-called transfinite interpolation to

determine the grid points. The formula uses relative coordinates along
the four polylines making up the polygon to be meshed.

‘East’
‘North’

‘West’
‘South’



Figure 4.6 The notation used for the transfinite interpolation grid generation

and vary between 0 to 1 along the arcs. The intersection of the

boundaries are identified The intersection of the ‘South’ and the
‘West’ boundary corresponds to (

he intersections are identified by:

Intersecting boundaries (ξ, η)

‘South’ and ‘West’ (0,0)

‘West’ and ‘North’ (0,1)

‘North’ and ‘East’ (1,1)

‘East’ and ‘South’ (1,0)

18 MIKE ZERO: Creating 2D Bathymetries

 The grid is defined through the expressions

x ,   1   x0,   x1,   1   x ,0   x ,1 

1   1   x0,0   1   x1,0 
1   x0,1   x1,1
(4.7)

y  ,   1    y 0,   y 1,   1    y  ,0  y  ,1 

1   1    y0,0   1    y1,0 
1   y0,1   y1,1
The expression requires that the four polylines making up the polygon
have been parameterized using the parameters and respectively.
These parameters are simply the distances along the arcs relative to the
full length of the polylines.

The distribution of the grid points along the polylines

The grid points along the polylines must obey the location of the
original vertices. Thus the concept of vertex pairs for opposing
polylines should be used. Though note that the vertex pairs no-longer
by default are connected by a straight line due to the piecewise linear
nature of the end arcs.

‘East’
‘North’

‘West’
‘South’




Figure 4.7 The polygon with the original vertices along with the interpolated
vertices. Opposing polylines have identical number of total vertices.
Note that the end arcs also have additional interpolated vertices.

19
 The number of grid points along the arcs is controlled by the
maximum grid element side length. More specifically a length is
applied for the  and  directions. This length is non-dimensionalised
by the maximum arc lengths in the respective direction. For example,
the maximum side length in the direction is denoted by s. This
value shall be used to define the location of the mesh nodes along the
‘south’ and ‘north’ polylines.

The maximum length of the ‘south’ and ‘north’ polylines is used to

define the non-dimensional mesh element side length . Let the two
lengths be denoted by lsouth and lnorth. The non-dimensional maximum
side length is defined by

max( lsouth, lnorth)

  (4.8)
s

This value must be satisfied for all line segments along the ‘south’ and
‘north’ polylines. The approach is to determine the number of intervals
for the line segment i through

   
N ,i  Ceiling  i i 1  (4.9)
  
  

Where i corresponds to the i’th vertex.

The procedure for the ‘west’ and ‘east’ arcs follows along the same
lines.

4.3 Generating Combined Mesh

The present section outlines the approach that will be adopted for
merging the quadrangular mesh polygons with the surrounding
triangular mesh.

The triangular mesh generator has the flexibility to insert so-called

Steiner points. Steiner points are additional vertices used to subdivide
line segments into several line segments. Steiner points are necessary
to allow the segments to exist in the mesh while maintaining the
required properties of the mesh such as constraints on the minimum
angle and maximum triangle area.

20 MIKE ZERO: Creating 2D Bathymetries

Figure 4.8 A polygon that has been meshed where additional Steiner points have
been inserted

The above figure illustrates the concept of Steiner points. Steiner

points are useful when the whole area is meshed using triangles, but do
create a problem when the mesh is mixed. To overcome this problem
the triangulisation routine may be used with the restriction that Steiner
points are not inserted. If this is used on the above case the result
would be along the lines as illustrated in Figure 4.9.

Figure 4.9 A triangulisation with the restriction that Steiner points are not allowed

It is especially important that the mesh routine does not insert Steiner
points along arcs which form part of a polygon designated for a
quadrangular mesh. To overcome this problem the idea is to run the
triangulisation twice. The first triangulisation is done to determine the
location of Steiner points along arcs without constraints (i.e. arcs not
part of a polygon for quadrangulisation).

Once the Steiner points have been identified along these ‘free’ arcs the
Steiner points are internally converted to vertices. The second
tringulisation is then done with the constraint that Steiner points are
not allowed. The procedure is outlined on the following pages.

21
 Step 1:

generate the quadrangular mesh

Step 2:

Insert vertices along the mesh points at the

arcs used for quadrangular mesh.

Step 3:

Mark all polygons used for a quadrangular

mesh internally as not included in the mesh.
Then mesh the remaining area.

Step 4:

Convert all mesh points along ‘free’ arcs to

vertices. Notice the Steiner point at the
north arc of the quadrangular mesh polygon.

22 MIKE ZERO: Creating 2D Bathymetries

Step 5:

Remove the triangular mesh.

Step 6:

Regenerate the mesh, but this time do not

allow the triangulation to insert Steiner
points.

Step 7:

Reintroduce the mesh for the quadrangular

polygons. Resulting in the final mesh. Note
that the Steiner point previously introduced
at north arc of the box is no longer present.

23
4.4 Smoothing the Mesh
After the generation of the mesh it is prudent to smooth the mesh to
obtain a better applicability in a simulation. Smoothing a mesh is the
effort to position the nodes in a way such that the angles in each
element and the element area are as large as possible. The smoothing
only affects triangular mesh elements.

Figure 4.10 Definition: Moving node point due to smoothing

The location of the Smoothed point Q' is found through

1  N elements

Q'   N elements Q   2 Pi  (4.10)
3N elements  i 1 

The smoother may be applied iteratively and will converge to a mesh

where the following holds for each point.

N elements
1
Q' 
N elements
P
i 1
i (4.11)

I.e. every point is placed in the centre of mass of the surrounding

polygon.

Please note that though the system is converging towards a unique

solution this may not be a solution which satisfies the mesh constraints
as set up in the Mesh Generation menu (plus local polygons if they
exist). The idea is to do the mesh smoothing one step at a time and
then evaluate the mesh and reset any nodes which give rise to elements
which do not satisfy the constraints.

The measure consists of the element area and the smallest angle in the
element. If one of these is smaller than the allowed values then the
calculated measure evaluates to false.

24 MIKE ZERO: Creating 2D Bathymetries

 Two options can be included when smoothing the mesh:

• Smoothing constrained by mesh criterion

If this is not selected the measure will always evaluate to true

• Leave mesh nodes at arcs untouched

If this is not selected ALL internal nodes may be relocated.
Otherwise, if a node originally was part of a defined arc or is a
single user defined node then the node is not allowed to be moved
during the smoothing.

4.5 Refinement by Bisection

A mesh may be refined by subdividing the faces of the existing
elements. The bisection method is designed to only change the size of
the elements in a given mesh without changing the overall quality of
the mesh i.e. the angles of the refined mesh will not become less than
the original mesh.

The mesh bisection method follows some clear principles:

Take all elements faces of the original mesh and place a node at the
mid points (hence bisection). Consider each element in the original
mesh. Depending on whether the element is triangular or quadrilateral
the approach varies slightly.

For a triangular element the approach is as illustrated in Figure 4.11.

Figure 4.11 Bisection of triangular element

As can be seen from Figure 4.11, the operation will give rise to 4 new
elements per triangular element in the original mesh and 3 new
element faces.

For a quadrilateral element the approach is as illustrated in Figure

4.12.

25
 Figure 4.12 Bisection of quadrangular element

As can be seen from Figure 4.12 also for the quadrangular elements
the process introduces 4 new elements per quadrangular element in the
original mesh and 4 element faces. Note that the quadrangular
elements also generate an additional mesh node on top of the ones
caused by the face bisection.

The refinement may in principle be applied an arbitrary number of

times. The table below summarizes the number of topological
elements (faces, elements, nodes) as a function of the number of
refinements (n).

n’th refinement (2n)

st nd
Number of Original mesh 1 refinement 2 refinement
topological
elements

Faces F 2F+3T+4Q 4F+18T+24Q

∑

Triangular T 4T 16T
elements

Quadrangular Q 4Q 16Q
elements

Nodes N N+F+Q N+3F+3T+9Q

∑ ∑

∑ (∑ )

26 MIKE ZERO: Creating 2D Bathymetries

5 MESH INTERPOLATION
The Mesh Generator gives two possibilities with respect to
interpolation for triangular elements. The two possible interpolation
routines are

• Natural neighbour
• Linear interpolation

These interpolation methods are valid independent of the type of mesh

used. The interpolation requires values only at the mesh nodes and will
base the interpolation solely on the scatter data. Thus whether the
nodes are part of a triangular or quadrangular mesh is of no
significance. These methods may however not be the correct ones to
use for river beds or other water ways in an in land setting. In
particular the interpolated surface generated for a strongly meandering
river may give rise to a highly erroneous river bed (see Figure 5.1).

A B

C D

Figure 5.1 A meandering river with a point (O) at which the topography is to be
interpolated using the natural neighbour method. The crosses indicate
location at which data is located. The grey polygon illustrates the
weights of the values at A-E

As can be seen from the above figure the point B will have a
significant weight and thus impact on the interpolated value at O.
Point B is located on the opposite bank of the interpolation location O
and thus the bed level may be very different from the true value at O.
One would assume that the actual level at O is closer to the values at A
and C in the stream direction. Further the interpolation routine also
takes the values at D and E into account even though the points are
located outside the main river.

27
 To create sound interpolated topographies in river beds an
interpolation routine that emphasises values in the stream direction and
does not interpolate across riverbanks has been developed for
quadrangular elements.

5.1 Interpolation Accommodating the Stream Direction

To accommodate higher weights on points positioned in the stream
direction a modified version of the inverse distance method is applied.
The inverse distance is a simple interpolation routine. A
neighbourhood about the interpolated point is identified and a
weighted average is taken of the observation values within this
neighbourhood. The weights are a decreasing function of distance ri
and are typically given by

1
 N 1  1
wi    p  p (5.1)
 i 1 ri  ri

for point i where p is greater than 0.

The distance r is often given by the Pythagorean distance but may be

any measure. The latter may be utilised to accommodate the weighting
of points in the stream direction. The coordinate system will be
constructed so that at the mesh nodes the coordinates take on integer
values. At intermediate locations the coordinates are found through
bilinear interpolation. The river segment or polygon when pictured in
the mesh based coordinate system is simply a rectangular as illustrated
in Figure 5.2.

A B
A B
E
E

C D
C D

Figure 5.2 The meandering river on the left without the mesh displayed. The right
figure illustrates the river overlaid with the mesh

28 MIKE ZERO: Creating 2D Bathymetries

k B
C
A j

Figure 5.3 The polygon displayed in the mesh coordinate system (j,k)

In the new coordinate system the points A, B and C will have the
following coordinates:

A: (jA,kA)
B: (jB,kB)
C: (jC,kC)

where j and k may take on non-integer values.

The location at which the interpolation is to take place has coordinates

(jo,ko). The z-value at O may then be expressed as
1
 1   zi 
z O    A , B ,C  
 
 ji  jO  ki  kO
p q
  A , B , C
j  j
p
 k  k
q
 (5.2)
   i O i O 

where q > p. The latter is to ensure that the effect of the j coordinate
decays more slowly than the effect of the k-coordinate. Both p and q
should be user defined (advanced option). Some experimentation is
needed to identify proper default values of p and q.

The bilinear interpolation to be used for determining the (j,k)

coordinates for scatter points not placed at a mesh node is found
through

(x01,y01) (x11,y11)

(x,y)

(x10,y10)
(x00,y00)

Figure 5.4 A mesh element with four nodes with mesh coordinates (j,k),
(j+1,k),(j,k+1) and (j+1,k+1). The mesh coordinates of (x,y) are sought

29
 The coordinates of (x,y) satisfy the following equation

 x   x00   x10  x00 

      s  
 y   y 00   y10  y 00 
(5.3)
 x   x  x01   x00   x10  x00 
t  01   s 11      s 
 01   11
y y  y 01   00   10
y y  y 00  

Where s,t lies within the interval [0;1]. This equation may be re-
arranged as

 x  x00   x10  x00   x01  x00 

   s   t   
 y  y 00   y10  y 00   y 01  y 00 
(5.4)
 x  x  x01  x11 
st 00 10 
 y 00  y10  y 01  y11 

which in turn may be expressed as

 d1   a1   b1  c 
   s   t    st 1  (5.5)
 d 2   a2   b2   c2 

This equation may be rewritten as a set of two equations

a2c1  a1c2 s 2  (a2b1  a1b2  d1c2  d 2c1 )s  b2 d1  b1d 2  0

d1  sa1 (5.6)
t
b1  sc1

of which the first is second order algebraic equation in s. Once the two
solutions are found t may be determined from equation number. The
true solution is the one that gives rise to a set of s and t lying within
the interval 0 to 1.

The mesh coordinates of the point (x,y) are thus given by (j+s,k+t).

30 MIKE ZERO: Creating 2D Bathymetries

5.2 Interpolation using Prioritization of Scatter Data
The prioritization is carried out by applying weights to the individual
scatter data sets. These weights may be assigned globally and locally.

The methodology used for prioritization during interpolation is the

following:

Assume n sources of data each being supplied in any of the accepted

data formats (xyz, dfs2, dfsu, mesh). Let these data sets be denoted by
D1 , D2 …, Dn.

• The interpolation is done separately for each data source Di.

Hereby there will be a mesh Mi with values only dependent on
data source Di.

• Each data source has an associated weight wi in the range 0 to 1.

• The resulting mesh is obtained through the sum

∑ (5.7)

Where the weights satisfy

∑ (5.8)

• The concept is also used with localized weights associated with

user specified regions.

The process is illustrated below using a set-up with three different data
sources.

31
 Figure 5.5 Mesh point P in setup with three different data sources. Each scatter
data source has associated weight

After interpolation, the mesh point P will have a z value based on the
three interpolations from data source 1 to 3 weighted according to the
weights:

(5.9)

Where Z(P,Di) is the result of an interpolation only including data

source Di.

To apply local weights prioritization polygons must be defined.

Prioritization polygons are user defined areas where a specific set of
prioritization weights are to be used. Prioritization polygons are
different from the polygons used for specifying mesh resolutions and
mesh type.

32 MIKE ZERO: Creating 2D Bathymetries

6 MESH ANALYSIS AND EDITING
During the process of creating the mesh it is possible to analyse the
mesh for its applicability in a simulation and to edit local mesh
elements.

6.1 Mesh Analysis

It is possible to analyse a mesh for its applicability in a simulation by
evaluating the restricting time step for each mesh element.

The analyses of a mesh consists of considering

• The element area

• The smallest angle within each element
• The CFL stability criteria

Figure 6.1 Definition of node point and position in mesh analysis

The area of a triangular element is given by

A
1
x2  x1  y3  y1    y 2  y1 x3  x1  (6.1)
2

The area of a quadrilateral is calculated as the sum of areas of the two

triangles that matches the quadrilateral.

The smallest angle may be evaluated by calculating the normalized

scalar product:

x2  x1 x0  x1    y 2  y1  y0  y1 
s 1 (6.2)
x2  x1 2   y 2  y1 2 x0  x1 2   y0  y1 2

33
 The quantity S varies between 0 and 2 where 0 refers to the smallest
angle and 2 the largest.

The time step t evaluated on basis of the CFL number is given by

min  facelength 
t  CFL  (6.3)
2 gh

where h is water depth (water level - topography) and g is 9.82 m2/s.

6.2 Collapsing Elements

Effectively the mesh element (triangular or quadrangular) is collapsed
by taking all nodes in the mesh and collapsing them to the centre point
of the element (centre of mass). Note that the procedure may remove
up to four mesh elements.

The collapse of an element may be achieved by subsequent collapses

of cells faces and a final move of the node to the centre of mass of the
original mesh element. The process is illustrated in Figure 6.2 for a
triangular element. The procedure for the quadrangular element
requires four face collapses instead of three.

Figure 6.2 Process for collapsing triangular element

34 MIKE ZERO: Creating 2D Bathymetries

The z value at the new node is obtained through interpolation based on
the user settings. The code value of the mesh node is set to the
maximum of the nodes of the collapsed element. The data used as
scatter data for the interpolation are the values at the mesh nodes.

35
36 MIKE ZERO: Creating 2D Bathymetries
References

7 REFERENCES
/1/ Jonathan Richard Shewchuk, Triangle: Engineering a 2D Quality Mesh Generator
and Delaunay Triangulator, in ``Applied Computational Geometry: Towards
Geometric Engineering'' (Ming C. Lin and Dinesh Manocha, editors), volume 1148
of Lecture Notes in Computer Science, pages 203-222, Springer-Verlag, Berlin, May
1996.

Scientific Documentation 37
 Bathymetry Editor & Mesh Generator

38 MIKE ZERO: Creating 2D Bathymetries