TransDEM offers a large set of functions to edit or even create Digital Elevation Models (DEM).
Basic functions comprise opening DEM in a number of file formats, cropping, merging and saving. Advanced functions include re-sampling (new raster width), smoothing, filling of holes, and elevation adjustment. Several of those functions utilise a so-called, Triangulated Irregular Network (TIN) with arbitrarily placeable control points. With the help of the TIN algorithm a DEM can even be created from scratch, using contour lines from a topographic map as a guidance.
Note: Internally, TransDEM applies a lot of mathematics. At the surface, however, almost nothing will be revealed of the underlying algorithms. The TransDEM user interface requires minimal knowledge about DEMs, coordinates and correction methods which is provided in the manual and illustrated with tutorials.
TransDEM originally concentrated on DEMs based on space research missions, but can nowadays also read Canadian and USGS DEMs plus all metric and UTM based DEMs in MicroDEM dem format and also save them in this format. ASCII .xyz is another format support for import and export.
Note: For a couple of years ASTER DEMs were available for free. Since early 2006 this is no longer the case.
ASTER DEMs have a horizontal resolution of 30 m and have UTM coordinates. Each DEM (called a granule) has a net size of about 60 x 60 km. Due to the slant of the satellite orbits the actual DEM file has an extension of about 70 x 70 km. TransDEM can read ASTER DEMs in geotiff format.
ASTER DEMs are available for limited areas only and do not have uniform quality. Data acquired by the satellite needs to processed first to create the DEM. To find a DEM search the database for both coarse data and finished DEMs. DEM creation may take weeks or months, but finished DEMs will be made available in a couple of hours (fees are involved in both cases).
ASTER DEMs usually have a couple of defects. Typical defects are:
Relative elevation error, for reference points and for vertical scale (applies to ASTER relative DEMs).
Holes: Water bodies, snow/ice and clouds
Artefacts: Peaks or dips not existing in nature.
TransDEM offers a number of function to treat many of those defects.
Image 1 shows a typical ASTER DEM. This is the Maas Valley in the Belgian/Dutch/German border region. The Mass (La Meuse), Mass lakes and some of the canals are water bodies which partially appear as a DEM hole. UTM grid with in this image is 20 km.
SRTM DEMs have a resolution of 3 arc sec (outside the USA, 1 arc sec within the USA). 3arc sec are approximately 90 m on the north axis and about 60 m on the east axis in Central Europe. Each DEM covers one degree longitude and latitude. TransDEM read the native SRTN hgt format and automatically converts to UTM.
SRTM DEMs are available for most landmasses of the Earth (up to 60° North) and can be downloaded directly.
SRTM DEMs can have defects as well ( some of them were corrected in version 2, available since late 2005):
Holes: Water bodies, snow/ice; to a lesser extent than ASTER
Noise: Uneven water bodies.
SRTM defects can be corrected by TransDEM as for ASTER.
Image 2 depicts an SRTM DEM, already converted to UTM by TransDEM. This is the Thames Valley, west of London with the Cheltern Hills. The UTM grid in this example is 50 km.
ASTER DEMs for Central Europe have the nominally higher resolution of 30 x 30 m, compared to SRTM DEMs with 60 x 90 m. However, ASTER DEMs most often have more defects and require intense post-processing. The following images show ASTER and SRTM side by side, ASTER to the left, SRTM to the right, with an identical resolution of 30 m. The SRTM DEM has been re-sampled for this purpose. ASTER looks grainy and SRTM appears “washed out”.
Image 3: ASTER and SRTM side by side. City of Aachen and surroundings, Germany
TransDEM directly supports Canadian DEMs of type CDED, which have a resolution of 0.75 or 1.5 arc sec (ca 20m and 40m raster width).
Additionally NED type USGS DEMs in geotiff format can be opened directly. TransDEM supports 1 and 1/3 arc sec (ca 30 m and 10 m raster width).
TransDEM renders heights as atlas colours. Shading can be activated in addition for a three-dimensional effect. TransDEM can also compute contour lines from DEM data and draw them on top of the DEM.
Image 4 shows a DEM with shading activated and TransDEM computed DEM contour lines.
The merge function allows to combine individual DEMs to a larger one. The DEMs may overlap, have different resolution and lie in adjacent UTM zones.
Image 5 with the result of a TRANSDEM merge for the north-western part of Germany, 6 x 6 degrees. The Harz mountains or the Rhine valley are clearly identifiable. The DEM has been sampled to 60 resolution, consist of 85 milliard height points and requires 107 MB file space in binary MicroDEM format.
ASTER DEMs in particular tend to have invalid features embedded, peaks and dips not existing in nature and called artefacts in TransDEM
TransDEM offers functions for spatial adjustment of elevations applicable to smaller and larger areas. These functions are also used to remove such artefacts.
With the aid of arbitrarily placeable control points and a triangular network spanned between these control points elevation errors can be compensated by interpolation. The control points placed define the new elevation.
The following sequence of images shows such a correction of artefacts, a peak and a dip. A series of control points encircles the artefacts. Outside, elevations will remain unchanged. Additional control points inside the marked areas define the new desired elevation. In the subsequent re-computation of the DEM the control point network determines new elevations for all interior DEM points and eliminates the artefacts.
Images 6 to 8 show artefact removal. Image 6 shows the two artefacts. Image 6 shows the TIN, defined by a series of control points, encircling the artefacts and defining elevations for the interior.The result of DEM re-computation with the TIN applied is shown in image 8.
There are several options in TransDEM to fill DEM holes. The following series of images illustrates some of these options.
The simplest way would be to drag a rectangular mask around a hole and let the fill algorithm do its job. Here, the control point method is show instead.
The control point network in this example consists of 6 points (image 9), 4 of them forming the boundary with fix elevations. These points ensure current elevations will be retained outside this boundary. The two reddish inner points define a new absolute elevation for each point, the assume elevation of the water body.
One of the more interesting options is the computation of a weighted average with a Gaussian filter (ignoring the inner control points). The result may look acceptable but does not resemble a water surface.
Another option is pure triangular interpolation (image 10). Now the inner control points with new absolute elevation take effect. This comes much closer to a water surface, but transition between old and new is quite abrupt.
For this reason an additional option was implemented, working in two phases. Phase 1 applies triangular interpolation, phase 2 treats the transition range between old and new with a weighted average algorithm to smooth it, as shown in image 12.
For simple holes there is also an option to fill these with a uniform elevation.
The most suitable option should be decided on a case-by-case basis.
Images 9 to 12 illustrate the filling of DEM holes. Image 8 (upper left) shows the control point network, image 9 (upper right) the weighted average algorithm, image 10 (lower left) linear barycentric interpolation and image 12 (lower right) the two-phase combination of the algorithms.
The triangular network can also be used to create an entire DEM from scratch or modify larger areas of a DEM, by tracing contour lines. A topographic map will be needed for this (see raster maps) as a template. Control point network display will be switched to contour lines. The mathematical algorithms remain the same but net rendering changes, now emphasising triangle edges with identical elevation, simplifying the emulation of contour lines.
Image 13: Triangular network, rendered in contour line mode. In this example contour lines were traced with a 20 m spacing. At some places all three triangle edges show the same colour. This is an indication for terracing, and unwanted effect which requires some minor adjustment by adding one or more additional control points.
Image 14 depicts a “home-made” DEM by contour line tracing from a topographic map. Shown in the image is a Zusi2 route terrain created from this DEM for the Rhaetische Bahn in the upper Albula Valley in Switzerland, created by Christian Omlin.
© 2003 - 2007 Roland Ziegler