Você está na página 1de 7

1

Whitebox Geospatial Analysis Tools Tutorial Series

Tutorial 2: Digital Elevation Models and Terrain Analysis

2
Tutorial version 1.0, March, 2010
Written by John Lindsay, Whitebox Geospatial Analysis Tools Project Lead Developer, The Centre for
Hydrogeomatics, The University of Guelph, Guelph, Canada, jlindsay@uoguelph.ca.
This tutorial has been prepared by John Lindsay as part of Government of Canada contract
#AM1043433W.

3
Preamble
This tutorial will cover the following topics:
1. Creating a DEM from LiDAR
2. Post-processing of DEM data
3. Extracting surface derivatives and common terrain indices

1.0 Introduction
The purpose of this tutorial is to familarize users with the creation, analysis, and visualization of digital
elevation models (DEMs) in Whitebox GAT. DEMs are a common data source for many types of GIS
analysis. This tutorial will walk you through the process of creating a DEM from a raw data set of
elevation points, some of the common pre-processing steps that are commonly applied to DEMs, the
extraction of common terrain indices, and effective DEM visualization.

2.0 Creating a DEM


DEMs are created from source data (contour lines or point data) through the process of interpolation.
There are many methods for interpolating data. Some excellent software, such as Golden Software's
Surfer, is entirely dedicated to interpolation and offers a wide range of interpolation methods. Whitebox
GAT, at least at the time of writing, has far fewer available interpolation methods. The choice of which
interpolation method to use is an important decision because it will ultimately affect the quality of the
DEM. Generally, the finer the spatial resolution of the input data and the higher the quality of the data,
the simpler the interpolation routine that is needed to generate a DEM.
This tutorial should have been distributed with a data file called 'sample LiDAR data.las'. This file
contains airborne laser scanner data (LiDAR data) in a standard binary format known as a LAS file.
LiDAR data is one of the best sources of high-quality and very dense topographic data. In its basic
form, these data can be thought of as an unstructured point cloud, where each point is assigned X, Y,

4
and Z positional values. In fact the LAS file specification allows for much more information to be
attached to each of the XYZ points than simply location. For example, each point also possesses
information about the intensity of the laser return, a land-cover classification, and the return number.
The return number can provide valuable additional information. When a laser pulse from a LiDAR
system is directed towards the ground it may hit many off-terrain objects, such as trees, fences, and
buildings, before hitting the ground surface. As a result a single laser pulse may actually yield multiple
returns with several XYZ points within the data set. Generally we think of the last return as being the
one coincident with or closest to the true ground surface.
Open the 'Nearest Neighbour Interpolation' tool from the LiDAR toolbox. The nearest neighbour
interpolation method is an appropriate method for application with LiDAR data because of its very
high point density. Furthermore, this interpolation routine has the advantage that, unlike other methods,
it allows for sudden breaks in slope, which are commonly associated with off terrain objects (e.g. think
of the sudden change in slope that occurs at the side of a building). This interpolation tool is very
similar to the nearest-neighbour interpolation tool within the Raster Creation toolbox, except that it is
designed to work with LAS files and allows the user to take advantage of the additional information
contained within these data. Specify 'sample LiDAR data.las' as the input LAS file and call the output
raster 'DEM first return'. Set the point return to First Return, the maximum search distance to 2 m, and
the grid resolution to 1 m. Do not change any of the other settings. Press OK. When the tool has
completed the resulting DEM should be automatically displayed. Notice that there are several trees and
buildings present in the DEM. The linear features that are apparent are hedge rows adjacent to farmers'
fields.
Now run the 'Nearest Neighbour Interpolation' tool a second time, this time set the point return to Last
Return, the maximum search distance to 2 m, the grid resolution to 1 m, and exclude points with
classifications of low vegetation, medium vegetation, high vegetation, building, power line, and power
line tower (you will need to scroll to the bottom to see these options for excluding points based on their
classification). Call the output raster 'DEM last return' and press OK. When the tool has completed,
remove 'DEM last return' from the active map. run the 'Fill Missing Data Holes' tool, also in the
LiDAR toolbox to fill in grid cells with missing data. Notice that many of the trees and buildings that
were apparent in the first-return DEM are no longer present in this last-return DEM (this might be more
obvious if you make the palette and display minimum and maximum values the same between the two

5
images). Despite the fact that we have selected the last-return points and excluded points that have been
classified as vegetation and buildings, there are still some areas of the DEM that are apparently affected
by the presence of off-terrain objects. Use the 'Remove Non-ground Points' tool to further remove these
points from the DEM. Set the maximum non-ground object size to 150 pixels and call the output raster
'DEM ground'.
Remove all of the displayed images on the active map except for 'DEM ground'. Open the Raster
Calculator and type the following expression into the expression textbox:
[OTOs]=if(([DEM first return]-[DEM ground])>0,([DEM first return]-[DEM ground]),NoData)
Notice that this expression evaluates grid cells for which there is a positive difference in elevation
between the first return and ground DEM. For grid cells where there is no difference between the two
DEMs, the NoData value is assigned. NoData is a special value (-32,768) used in Whitebox GAT to
signify cells that have no valid value and cells that should be displayed transparently. This will create a
map of the off-terrain objects and will display transparently over the 'DEM ground' image.

3.0 Post-processing of DEM Data


If you zoom into any part of 'DEM ground' more closely you will likely find that the surface contains
significant small spatial scale roughness. This is a characteristic of the nearest-neighbour interpolation
method, which does not smooth the data to the extent that most other interpolators typically do. For
many applications, this small-scale roughness is not desirable and should be removed. Use the 'Edge
Preserving Smoothing' filter, found within the 'Image Processing' toolbox, to remove the small-scale
roughness from 'DEM ground' without removing important linear features, such as the roads and
accompanying ditches. Call the output raster 'DEM smoothed', use a standard deviation distance of 1.5
pixels, and a standard deviation intensity of 0.2 m. Notice the improvement in the resulting image with
respect to its smoothness. This will also improve the calculation of surface derivatives, such as slope
gradient, aspect, and curvatures, and other common terrain indices, such as shaded relief images.
There are a number of other common pre-processing steps that are often applied to DEM that are used
for specific applications. For example, DEMs that are used in any hydrological application, such as

6
stream and watershed mapping and surface drainage modelling, typically have their topographic
depressions and flat areas removed prior to analysis. It may also be the case that these DEMs have a
stream layer, derived from a topographic map, 'burned' into the DEM to force flow within these known
channels. Stream burning is not typically done with high quality, fine resolution data such as LiDAR,
where the inherent streams within the DEM are likely of better quality than those off of a coarser scale
map. Nevertheless, these pre-processing steps can be carried out using the tools within the 'DEM Preprocessing' toolbox, itself within the 'Hydrological Tools' toolbox. Notice that there are several
approaches to removing topographic depressions from DEMs, but that at the time of writing, the only
approach available in Whitebox is depression filling. Depression filling is probably the most commonly
used approach to the remove of depressions from DEM, though some work has suggested that it is not
as appropriate as depression breaching. Depression breaching will be available in future versions of the
software.

4.0 Extracting Surface Derivatives and Common Terrain Indices


Whitebox GAT allows you to extract a number of commonly used surface derivatives from a DEM.
The tools for performing this type of analysis are contained within the 'Surface Derivatives' toolbox, a
sub-directory of the 'Terrain Analysis' toolbox. By far the most commonly extracted derivatives of
DEMs include slope gradient and aspect. These data provide valuable information about the steepness
and orientation of the local topographic surface. These attributes are related to the rates of movement of
materials over the surface (water, sediment, nutrients, and contaminants) and the solar energy recieved
by the surface. Use the 'Slope' tool to extract slope gradient from the 'DEM smoothed' raster. Call the
output 'Slope'. Slope gradient is reported here in degrees. To better visualize the slope data, you may
need to clip the upper tail by 1% in the 'Layer Properties' tab; do not rescale the lower tail (i.e. set it to
zero during the clipping). If you display the histogram associated with the layer, you will see why this
rescaling is necessary. Slope is used in many applications, including the secondary terrain attributes,
relative stream power, sediment transport index, and the wetness index, each of which can be derived
from a DEM using Whitebox GAT.
Now use the 'Aspect' tool to derive slope aspect, or orientation, from the 'DEM smoothed' raster. Notice
that it is displayed using a specialized palette used for circular data, such as aspect. If you move your

7
mouse over this image, you will find that the direction that each grid cell is facing is reported as an
azimuth from 0-360 degrees. Grid cells that are flat, i.e. upwards facing, are flagged with a -1 value.
Aspect is most directly related to the direction of surface flow (discussed further in Tutorial 3) and the
amount of insolation, whereby south-facing slopes generally receive more solar radiation than northfacing slopes. In fact, the amount of insolation is also affected by a number of other factors including
the solar azimuth (i.e. the direction of the sun along the horizon) and the solar altitude (i.e. the angle of
the sun above the horizon), which both change with time of year and day. Note: there are many other
factors affecting insolation including the characteristic of the atmosphere.
Open the 'Hillshade' tool from the 'Terrain Analysis' toolbox. Analytical hillshading is a common tool
used to effectively visualize topography. Notice that the user is able to specify the solar azimuth and the
solar altitude. The default values (a sun position of 30o above the north-west horizon) do not make
much sense from the perspective of insolation modelling. That is, in the northern hemisphere, the sun is
never in the northwest. Run the tool using the default settings, specifying 'DEM smoothed' as the input
raster and the output raster name as 'Hillshade'. If you examine the image histogram you will find that
only a small portion of the range of display values (0-1) actually contain data. Clip the upper and lower
tails by 1% each and the image contrast should be improved considerably. Run the 'Hillshade' tool a
second time, this time specifying a solar azimuth of 135 degrees. Call the output raster 'Hillshade 135'
Compare the two hillshade images by toggling the visibility of the top image. Notice that ridges and
valleys appear as you might expect in the original hillshade image but that they are opposite (i.e.
valleys are ridges and ridges are valleys) in the 'Hillshade 135' image. This phenomena is known as the
pseudoscopic effect.
Close any layers that you currently have open on the active map. Now add the 'Hillshade' image, 'DEM
smoothed' and 'OTOs'. If they do not appear in that order, use the Raise Layer and Lower Layer tools on
the toolbar to rearrange the order. Clip the lower tail of the DEM by 0.5% to improve the image
contrast. Set the transparency of the DEM to 15%. This will allow the hillshade image to show through
the DEM, creating an effective display of the topography in the area. You can now save the map and
display it at a future time.

Você também pode gostar