The raster terrain models can be constructed by using the following input data:
Original field leveling data
Stereo pairs of aerial photos
Radar-based elevation of interferometry data
Lased-based elevation or range (LIDAR) data.
The original field levelling data is a three-dimensional point set, which was surveyed by field measurements in some vertical datum (vertical network). The points, of course, are in irregular network in the horizontal plane. The contour lines of topographic maps were drawn using these data, prior to the widespread use of the stereo photogrammetry. In the practice, this kind of data is rarely used. After design and print of the final contour maps, their working material, such as the original field protocols and the derived point lists were often lost.
The map contours (the lines connecting the terrain points with equal elevation) were designed and drawn using the mentioned field surveys, or later by the procedures of the below discussed stereo photogrammetry. Their information content is less than the one of the original field data. The manual or automatic digitalization of these contours provides again a three-dimensional point list: for the horizontal position of the digitized points we couple the elevation value of the contour line. This point set can be interpreted as a model of the original field leveling data. A raster elevation model can be constructed similarly, by TIN→GRID conversion. However, the contour-based elevation models are distorted by three kinds of errors:
In the sharp curves of the contours, there are one or more triangles in the irregular network, whose vertices lie on the same contour. The modeled elevations of all points of these triangles are in horizontal planes. Therefore, along the ridge lines, a ’virtual plateau’ occurs, which is not existing in the real terrain. Thus, in the histogram of the terrain model has peaks connected to every contour elevation.
If we don’t digitize enough vertices along the long, straight sections of the contours (the point interval is less than the distance of the neighboring contour), and it is not even densified later by automatic methods, then the edges of the irregular triangulation network does intersect the contours in some places. The result is a ’fishbone pattern’ at the top or the bottom of the displayed slope (Fig. 47).
In very flat terrains it is a frequent situation as only one contour is crisscrossing through an extent area. Even we digitize thoroughly this line, following the complex structure of oxbows and point bars of a floodplain, the result will be a single, horizontal plane. The original fine relief can be attenuated by virtual auxiliary contours, following the small ridges and valleys.
The above errors can be handled by entering auxiliary data of other pieces of information into the system. At the ridges, we can define the ridge lines themselves. At the valley lines, digitizing the streamlines means such relief information, which decreases or even fully eliminates the above false effects. The most parts of the Earth’s surface is formed by stream erosion. According to this, there are algorithms providing ’hydrologically correct’ terrain models. These algorithms – assuming that the water runs off from all surface points – correct most of the above problems. If our assumption for the surface runoff is true, this terrain model will be the closest one to the real relief and the result in correct also in hydrological applications.
However, in the territories, where this assumption not true, or not even almost true, the resulted model can be of varying quality, sometimes even very bad. Karstic regions with gullies, dolinas, underwater creeks, or an area with many outlets (wind-formed sandy regions, or floodplains with oxbows) are killing the quality of the ’hydrologically correct’ models.
In the case or stereo photo pairs, the elevation information of the area is represented by the different distortion of the two aerial photos, taken from different positions and angles (Fig. 48). These distortions are primarily realized by the mutual image position of the base points of the horizontal geodetic and/or elevation networks (whose coordinates are well-known both in horizontal and vertical sense). Other terrain objects with unknown geodetic position can also be identified in both images, which gives auxiliary information for the above pieces of information. The result of the aerial triangulation was mostly a contour map – and we can derive the grid model according to the above mentioned procedure. With the advance of the computer technologies, however, it is sometimes possible to make a raster terrain model directly from the photo pairs and the detected points on them. For this procedure, not only aerial photos but also satellite images (better than mid-resolution, e.g. the ASTER data) can be processed. It should be mentioned that the identified and paired image elements correspond not necessarily to the terrain but they can be in elevated positions (vegetation, buildings). These ones can be omitted, or if we use them all, we shall recognize the resulted dataset not as terrain but elevation model, containing these elements, too.
The radar technology is connected to the terrain modeling in two different manners. The here only mentioned but not discussed radar-interferometry primarily detects the vertical movements of the surface. However, the radar-based altitude measurement is capable to determine directly the distance of the terrain or terrain objects from radar source and detector, which should be in known position with respect to the Earth. This is the technology, which revolutionized the availability of the digital elevation models in the early 2000s, giving a huge push to any research that needs these data sources. The radar was invented to follow the position of aircrafts from the surface by electromagnetic rays propagating through the relatively dense atmosphere the Earth. However, the reverse way is also possible to detect the surface from board of the air- and spacecrafts with localized radar beams. This ability was clearly shown by Zoltán Bay in 1946, who measured the distance of the Moon by radar experiments. In 2003, using a source and detector pair placed on board of the Space Shuttle Endeavour, the majority of the Earth’s surface was surveyed, resulted in the Shuttle Radar Topography Mission (SRTM) dataset. This model became the most used elevation dataset worldwide, primarily because of its free availability and globally unified characteristics. According to the used technology, the partial effect of the built environment and the vegetation is in the data. Nowadays, the cca. 100 meter spatial resolution is considered to be quite low, however at the time of the publication of the dataset, it brought a real breakthrough for a wide spectra of the sciences.
To improve the resolution, the newest of the discussed technologies, the LIDAR should be applied. The laser range measurement became a part of the toolbox of geodesy in the last decades, after the invention of the portable lasers. The most up-to-date application, the laser scanning is based on the scanner that is capable to alter the direction of the laser beam in a pre-set range and to record the backscattered signal. First because of the huge amount of this data, this application is widespread used only in the very last decade. The laser scanner can be applied in the field and can be also mounted onboard of aircrafts. Its satellite application is limited because of the atmospheric scattering. The laser signals are reflected back from the buildings. Because of the high density of the measurement points (up to several points per square meter), there are soil-reflections even while surveying of vegetated area. The high resolution of the method is ideal for surveying the micro-topography of near-flat terrains.
The quality characteristics of the raster-based terrain and elevation models are:
The horizontal resolution (pixel size);
The numerical representation of the elevations;
The vertical accuracy.
If the source is a contour map, we shall also give:
The scale of the original map
The regular contour interval of the original map, as well as the smallest contour interval (halving and/or auxiliary contours).
It should be underlined that the numerical representation and the accuracy of the elevations are not the same. The representation (e.g. „integer”) shows the smallest elevation difference (e.g. one meter) that can be represented in the model. This is not the same to the accuracy of the elevation estimation (e.g. 3 or 5 meters) that is based on the whole technology chain led to the elevation model. Of course, the representation should be finer than the accuracy, otherwise the representation itself mars the accuracy. The raster-based elevation models are images, whose pixel lines and columns are parallel to the axes of some geodetic or projected coordinate system. This coordinate system, and the place of our image in this system are also very important pieces of meta-data of the terrain or elevation model.