Digital Terrain Data: Scientific Discussion
Table of Contents
Topographic data are vital for many scientific, technical, and other applications. Even when not used directly in a study, topographic data are often used in preparing visualization tools such as perspective or stereoscopic views of terrain.
Topographic data have traditionally been presented:
- as local topographic maps, in series of "rectangular" quadrangle sheets. These maps often have contours representing specific elevation values. The contours are typically spaced in even increments of elevation or bathymetry. This "contour" interval helps to determine the vertical resolution of the contour map. Topographic maps often have point elevations, such as at airports, on mountain peaks, for typical surface elevations of lakes, or at other landmarks. Other implicit topographic information may include graphical sketches of terrain (as mountains are usually assumed to be topographically higher than their surroundings, while streams and lakes are usually considered to be topographically lower than their surroundings. Of course, this may not be true. The lower Mississippi and Yellow Rivers (the latter river is in China) are built-up in natural levees (sometimes supplemented by man-made levees), so that their water level is sometimes higher than that of the surrounding countryside.)
- or as more generalized maps assigning colors to various ranges of elevations.
These maps usually have "minimum mapping units." This means that features smaller than a specified amount are omitted from coverage in the map. In addition, production-grade data are usually slightly stylized, in order to increase production efficiency. Relatively abrupt changes in direction of an elevation "line" on the ground (for example, at a sharp promontory) may be smoothed on a map, contours may be dropped in areas of steep slope, subsidiary contours may add detail in flat areas, etc. We have found apparent cultural differences in topographic maps, and derived digital topographic data, produced by different institutions and countries.
Now that we are entering the digital age, one may ask "what are digital topographic data, and how should they be prepared and presented?" There are surprisingly few good discussions available on this topic.
Even terminology can be confusing:
- Digital Topographic Data (DTD) is occasionally used as a generic term.
- Digital Elevation Data (DED) - same as above.
- Digital Terrain Data (DTD) - same as above.
- Digital Terrain Tape (DTT) an old term for early DTD produced by the U. S. Army Map Service and reformatted for distribution by the U. S. Geological Survey.
- Digital Bathymetric Data (DBM) a generic term for depths in water bodies (rivers, lakes, seas, oceans).
- Digital Relief Data, or Digital Relief Models, generic terms for elevations above and below water levels. That is, a generic term for wet and dry elevations/bathymetry.
- Digital Elevation Model (DEM). Some people associate this with the U. S. Geological Survey's data (and file format[s]). However, we consider this to be the most common generic acronym for digital elevation data/models. Incidentally, the term "model" is used by some folk who consider "data" to imply more accuracy than is currently possible. Thus "model" is considered an attempt to recreate topographic elevations, and recognizes that such "models" may be getting better, but are still hardly perfect. They are certainly not reliable for certain life-critical applications, without appropriate analysis and possible repair.
- Digital Terrain Elevation Data (DTED) is generally considered the specific name for digital terrain data produced by the U. S. National Geospatial-Intelligence Agency and its precursors (primarily the National Imagery and Mapping Agency, the Defense Mapping Agency, and before that the Army Map Service).
So what are DEMs?
- DEMs are traditionally presented as two-dimensional computer arrays (like putting graph paper over the terrain). Data may be presented in ASCII (character text) or 2-byte integer binary formats. Traditional DEMs have origins in the southwestern corner of the grid, with subsequent values to the north of the previous value. After the first north-south row is covered, the next column to the east is traditionally presented in the same way. Some newer DEMs are presented as raster data, in binary format with the origin at the top-left-hand corner of the grid. Such images scan as television images.
- However, elevations are also presented as:
- Points or spots, such as at mountain peaks, lake surface elevations, confluences of streams, and cultural landmarks (e.g. airports, cities, geodetic control marker locations). Points may be determined by cadastral survey, Global Positioning Survey, photo interpretation or other technique.
- Digitized contour line segments as vector lines.
- Sounding depth estimates (for bathymetry).
- Physical line drop measurements (for bathymetry), which are presumed to be vertical, but which often deviate slightly from the vertical because of currents in the water, movement of the ship from which measurements are made, slope of the bottom surface, etc.
How should we describe the Earth's surface via DEMs?
- Most DEMs give single values for each location. These values are often considered to represent the Earth's surface at that "point." However, matters may not be so simple:
- Many elevation values determined from photointerpretation estimate the height of vegetation canopy (that is, the tops of trees with their leaves), rather than the ground surface. Sometimes at attempt is made to "correct" for the typical height of such canopy. However, this "correction" is usually incomplete. For example, trees are often taller in stream valleys than on ridges. Tree heights may also vary by soil type and other factors, let alone by species. "Corrections" for vegetation canopy rarely handle such variations.
- In areas of steep slope, slight errors in horizontal positional accuracy may result in significant vertical errors in the DEM.
- In some areas, overhanging cliffs may actually result in three elevation values at a given map coordinate: the top surface, the elevation of the underhanging cliff face, and the elevation of the ground directly below that overhanging cliff face.
- In some cases, DEMs represent point determinations of elevation at the exact point referenced in the computer file. In other cases, interpolated surfaces are sampled at the location referenced in the computer file. In other cases, additional sampling is done to estimate the value at the referenced location. In yet other cases, the value given for a location might represent a "representative value reasonably near to" the referenced location.
- One should assume that there is systematic error in every DEM. Some of this error is:
- known and documented, some is
- known and undocumented (some DEM developers believe that users don't want to be bored with "all the details"!), some is
- unknown (and thus undocumented).
- Rarely does a DEM attempt to give a more sophisticated view that "gimme a number." Within a 30 meter or 100 meter grid cell for a local quadrangle-based DEM (let alone a 30-arc-second GLOBE DEM grid cell) there can be considerable variety in elevation. The GLOBE DEM Project negotiated an original design for a 30" grid from the Defense Mapping Agency, that included a maximum, minimum, mean and point elevation known for each 30" area, based on higher-resolution source data. This pioneering effort was revised, and released as DTED Level 0 data by the National Imagery and Mapping Agency (now National Geospatial-Intelligence Agency) (which absorbed DMA at NIMA's creation after this prototype was developed). One can imagine more advanced DEMs, containing estimated maximum, minimum, mean, median, standard deviation, and perhaps computed slope and aspect of terrain. The traditional paradigm of "Gimme a number" has been broken.