It is possible to create new raster surfaces by selecting New Raster...
from the File
menu, by pressing the
button, or by using the shortcut key
Ctrl-N
. You will be presented with a dialogue window
allowing the bounds and resolution of the new surface to be changed. The default settings create
a raster with 100 rows and 100 columns. The new raster can be either a polynomial expression or a
fractal surface with a user-defined fractal dimension.
New raster window using polynomial expression and metadata.
For modelling and simulation purposes it is sometimes useful to create artificial surfaces with known
characteristics. By selecting Polynomial
from the Create New Raster
window,
you can define surfaces with the following variables and functions.
Expression | Explanation |
x | the x position of each raster cell. x is scaled to be +- nCols/2, where nCols is the number of columns in the raster |
y | the y position of each raster cell. y is scaled to be +- nRows/2, where nRows is the number of rows in the raster |
z1 | the value of any given cell in the primary raster. |
z2 | the value of any given cell in the secondary raster. This will only be meaningful if a secondary raster has been selected; if it does not, z2 will always return 0. |
pi , e | The constants PI (3.1459) and e (2.7183) |
+ , - , * , / , ^ | Addition, subtraction, multiplication, division and power operators. |
cos() , sin() , tan() | trigonometrical functions. Expects angles to be given in radians (to convert from
degrees into radians multiply by pi/180 or 0.01745329). Values of infinity (e.g.
tan(pi/2) ) are returned as 0. |
acos() , asin() , atan() | inverse trigonometrical functions. Returns values in radians (to convert from radians into degrees multiply by 180/pi or 57.2957795). Values outside of the range of +-1 supplied to these functions will return a value of 0. |
sqrt() | Square root. If the value given to the function is negative, the function will return a 0. |
ln() | Natural logarithm (to base e). If the value given to the function is <=0, the function will return a zero. |
() | Brackets to control the order of evaluation. |
Example | Explanation |
x+y | Creates a plane sloping from bottom right to top left. |
0 - (x^2 + y^2) | Creates a convex-up dome. |
z1-z2 | Creates a difference map of the differences between the primary and secondary rasters. |
100 - sqrt((x*y*sin(x*y/200) * cos(y/30))+(x^2+y^2)) | Complex polynomial (peak with 8 radial valleys) |
Expressions that are incorrectly specified (e.g. using unknown functions, or failure to close brackets), are highlighted
in the edit window and preventing it from being closed.
Polynomial and combined polynomial and fractal surfaces
By selecting Fractal
from the Create New Raster
window, more realistic terrain surfaces may be
created for modelling and simulation purposes. The roughness of such surfaces can be controlled by entering a fractal dimension
between 2.0 (smooth) and 3.0 (very rough). Typical landscapes have fractal dimensions of around 2.1.
These may also be combined with smoother polynomial surfaces, as shown in the figure above. This can be achieved by creating a
fractal surface and a separate a polynomial surface and finally creating a new polynomial as
z1+z2
, thus combining the two simulated fields.
It is possible to combine the primary and secondary raster into a single raster layer by selecting the
Edit->Combine rasters...
menu option. This allows rasters that do not have the same spatial
extent to be merged into a single object. This can be particularly useful for assembling 'tiled' datasets.
Assuming a primary and secondary raster have been selected in the thumbnail view, selecting the Combine Rasters
option will bring up a window similar to the one shown below. If the spatial extents of both rasters are
overlapping but not identical, you will have the option of either selecting the intersection of the two
layers, or their union. For cell locations that are present in both rasters, you can select the
output value for those locations to be the primary raster cell value, the secondary raster cell value,
or the average of the two. Additionally, you can choose to override this selection for cells where the
chosen raster value is null but would be a numeric value if selected from the other raster.
Raster combination options.
New vector maps can be created either by selecting New Vector...
from the File
menu or by pressing the
button. You will be presented with a dialogue window allowing the bounds of the new map to be changed.
Initially this vector map will be 'empty', but it is possible to add vector objects through on-screen
digitizing.
To digitize new data, make sure a vector map is selected (either an empty vector map as described above,
or an existing one to which manually digitized objects are to be added), then select either the
Edit->Digitize Mode
menu option or the
button.
Digitizing new vector objects.
You will then be presented with a small window (see the figure above) that provides you with the option
of creating vector points, lines or areas. By clicking in the main LandSerf window with the left mouse
button, coordinates can be added to a new vector object. The numeric attribute to be associated with that
object can also be defined at this stage. To store the digitized points, press the Store object
button.
To make the process of digitizing easier and more precise, any raster data can be displayed on screen while digitizing, and the main display can be placed in zoom mode and panned with the right mouse button.
Both raster and vector data are associated with numeric attribute values. However, this can be limiting for data that have non-numeric or multiple attributes. In order to overcome this, any raster or vector map in LandSerf can be associated with an attribute table that defines the relations between numeric attributes (the 'ID' or 'key') and any other numeric or textual attribute values.
To create an attribute table, select either Edit Raster
or Edit Vector
as
appropriate and then click the Edit
button in the Attributes
area (adjacent
to the minimum/maximum attribute values). This should open a new attribute table editor that allows
you to add rows and columns to the attribute table as well as load or save them from/to disk. Any
attribute can be edited by clicking on the appropriate cell in the table.
All attribute tables have an active attribute associated with them. This attribute is indicated
in bold in the editor window (see figure below for an example). The active attribute is the one that
is displayed when querying a raster or vector. To make any attribute active, simply click one of the
values in the relevant table column.
Attribute table.
All spatial objects (raster and vectors) can be associated with metadata describing the spatial properties of the object (bounds and resolution). They can also be classified into different object types (e.g. elevation, slope, features, fuzzy membership, image, contours, TINs etc.) that determine the type of operations that can be performed on them and the way in which they are displayed.
To edit either a raster or vector object, select it then choose the Edit->Edit Raster
or Edit->Edit Vector
menu item. Bounds, resolution (raster only), type, colour table
and supplementary notes can all be changed from this window. Bounds can be changed either by typing
new coordinates for each of the four edges of the object, or by dragging the mouse in the main display
window to identify a sub-area of the object. To 'cut out' a subset of an object,
make sure the Extract subset
option is selected. To change the resolution of a raster
object, ensure the Interpolate to new resolution
option is selected and appropriate values
are entered in the N-S Res
and E-W Res
fields. This technique can be used to
remove 'steps' in a coarse DEM, or to provide an optimal sampling strategy for DEMs that are too large
for effective processing.
To commit changes made in the edit window, press the OK
button. If there is an inconsistency
between the new raster bounds and the resolution and neither the resampling or subset options are selected,
the relevant fields will be highlighted when you attempt to exit the window.
Rasters and vectors can be removed by selecting the relevant thumbnail view and then selecting either the
Edit->Delete raster
or Edit->Delete vector
menu option. You can also
remove all spatial objects by selecting the Edit->Delete all...
option.
LandSerf allows you to perform three types of transformation. Coordinate transformations
change the locational representation of rasters and vectors and include map projection and
rectification. Model transformations include the conversion of vector to raster models,
and the generation of contours and TINs (Triangulated Irregular Networks) from raster surfaces.
Raster Value Transformations allow the values of individual raster cells to be changed
through scaling, reclassification, translation etc. All are accessible through the Transform
menu.
Both rasters and vectors may be reprojected from global latitude/longitude to either Universal Transverse Mercator
(UTM) or Ordnance Survey National Grid Coordinates. All reprojections are selected from the
Transform->Project
menu item. When reprojecting into UTM, the zone of used by the
transformation is stored in the 'notes' field of the new spatial object (choose Info->Summary Info
to view this field).
Lat/long (left) and OSGB (right) coordinate projections
When rasters are reprojected you are given the option of setting the new resolution and bounds of the projected raster. This allows allows square pixels to be created in UTM and OSGB systems by setting identical x and y resolutions. You are also given the option to control whether or not to interpolate new values. Interpolation is suitable for continuous measurement data such as elevation. For categorical data, make sure this option is not selected so that new raster values are resampled from the original raster.
Many raster object that need to be referenced to some spatial units can be georeferenced simply by editing the bounds of the four edges (see Section 3.4 above). However for some data, more sophisticated transformations are required. Rectification allows a number of 'before' and 'after' points (known as ground control points) to be defined. LandSerf will then attempt to find an appropriate transformation that will convert the georeferenced locations of the 'before' points into appropriate 'after' locations. This transformation can then be applied to the entire raster.
To perform image rectification, display the raster that will be transformed in the main LandSerf window
and then select the Transform->Rectify...
menu option. The a new window similar to that
shown below will appear. Making sure the Add points
option is selected, click on the main
raster display at a point at which you know the true georeferenced location. The untransformed raster
coordinates will appear in the rectification window in columns 2 and 3. Enter the georeferenced coordinates
in columns 4 and 5. Repeat this process for other points on the raster. You should aim to create a spread
of typically 10-20 points over the entire raster. To make this process easier, you may also use zoom mode
to zoom into particular parts of the raster (making sure the Add points
is not selected
when using the mouse to zoom and pan).
Rectification control point editor
You can control the complexity of the transformation to be applied by selecting one of the linear
,
quadratic
or cubic
transformation options. The calculate the transformation,
press the calculate transformation
button. This will display the coefficients of the transform
along with a measure of Root Mean Squared Error (RMSE) which should be as small as possible. The error
associated each point is also displayed, and this can give a clue as to which of the control points should
be changed in order to reduce the error. It is often worth trying different linear/quadratic/cubic options
and examining their effect on the overall RMSE. The figure below shows the degree of flexibility each of
these has on the transformation process. Higher order transformations require progressively more points in
order to model the transformation (linear - 3, quadratic - 6 and cubic - 10).
Linear (left), quadratic (middle) and cubic (right) rectification
When you are happy with an appropriate set of control points and transformation type, these can be saved
as a separate file for later use using the Save points
button. To transform the raster itself,
press the Rectify map
button. You will then be presented with the option of changing the
new bounds and resolution of the new raster before applying the rectification.
A raster DEM can be converted into a vector Triangulated Irregular Network (TIN) be selecting
the Transform->DEM to TIN...
menu item. You will then be presented with a dialogue window
asking for various triangulation options (see figure below). To control the number of triangles produced
by the process, either specify the number explicitly and select the relevant check box, or specify some
error criterion. The error (either average as specified by the Root Mean Squared Error, or the maximum
error) represents the difference in elevation between any point on the original DEM and its elevation in
the TIN. The smaller the error specified, the greater the number of triangles required in the network.
A representation of the spatial pattern of errors can be produced by selecting
Create Error Surface
from the dialogue box.
The triangulation process works by successively adding triangles to the network until either any of
the selected error criteria are met, or the maximum number of triangles is reached (if specified).
It is also possible to convert a given TIN back into a DEM. This is achieved by selecting the
Transform->TIN to DEM...
menu option. LandSerf will apply a planar interpolation of
the triangle network that tends to produce a surface composed of flat facets. If necessary, these
can be smoothed using Quadratic Interpolation (see the Analysis
chapter for details).
TIN over 'TIN to DEM' surface
It is possible to create a contour representation of a given DEM by selecting the
Transform->DEM to contours...
menu option. Once selected, you can control the contour
interval and the elevation of the lowest contour. For DEMs with large flat areas (e.g. sea around an
island), more useful results are produced when a contours do not coincide with the elevation of the
extended flat region. Alternatively, flat areas can be reclassified as null values and be excluded
from the transformation. The Grid width
option controls the resolution at which DEM cells
are sampled in order to thread contour lines. A value of 1 will sample every DEM cell and produce the
most detailed contour lines; larger values will be faster and will generalise the resulting contours.
Contour creation options
By default, generated contour vectors are given the same colour scheme as the DEM from which they were
derived. This can make them difficult to see when overlaid on the same DEM, so it can useful to change
the colour scheme of one of the two models (e.g. black contours shown in the figure below).
DEM with contour overlay
More general vector to raster transformations can be made by selecting the
Transform->Vector to raster...
menu option. This will attempt to rasterize the currently
selected vector map to a given resolution specified. Point data can be rasterized in this way, or they
can be converted rasters of point density by selecting Transform->Point to density surface...
.
On selecting this menu option, a new dialogue window is shown requesting the raster resolution required
and the size of the local window used to calculate density values. This can range from 1 to the size of
the raster, but should be an odd number. The larger the number, the smoother and more generalised
the resulting density surface will be.
A series of simple transformations can be applied to the values of a raster's cells by selecting
the Transform->Raster Values...
menu item. Raster values are scaled by the
value in the Scale
field, translated 'up' or 'down' by the Translate
value
or rounded to the nearest Round
value. The elevation model can also be 'flooded' such that
all values below that in the Flood
field will be assigned that value. Transformations
are only applied if the relevant check box is selected.
Raster value transformation replacing zeros with null values
Additionally, you can replace given raster values by entering appropriate figures in the
Replace ... with
fields. If the two numeric values here are different, then all occurences
of the number in the first field will be replaced with that in the second. This can be useful for simple
reclassifications of categorical information. Rasters can also contain null values that are not
displayed and are removed from processing operations. This is particularly useful for representing missing
data, for water bodies and for excluding raster edges from processing. To replace numeric values with a
null, or a null value with a numeric equivalent, enter an n
in the relevant field (see figure
above).