| Custom DEMs Provide Value for Appalachian Mountain Club
Detailed trail maps are an important companion piece to wilderness guidebooks
By Douglas E. Lierle Over the past 15 years or so, the United States Geological Survey (USGS) has endeavored to produce a nationwide database of 7.5 minute DEMs. A 7.5 minute DEM can be regarded as a regularly-spaced matrix of points, generally 30-meters (m) apart in both horizontal and vertical directions. Using appropriate contouring software, contour lines can be created at a user defined interval. However, different software programs and algorithms within those programs will yield varying results.
In a project with the Appalachian Mountain Club (AMC) to create detailed trail maps of the White Mountains for publication in wilderness guidebooks, the WhiteStar Corporation has ultimately used the ARC/INFO TIN (Triangular Irregular Network) module running on Windows¨ NTª, primarily because of the program's ability to handle large, merged data sets. Other programs used in the mapping project included AutoCAD¨, QuickSurf¨, and a few custom routines developed in-house. Obsticles
The USGS DEMs database encompasses approximately 13,000 DEMs scattered across the country, all of which will suffer from the pitfalls of the manual profiling technique, including a "stripping effect" that renders the DEM unsuitable for many applications, including contour mapping. Normally, the approach to eliminating stripping was to use a low-pass filter that also has the unfortunate side-effect of degrading the quality of the real elevations in the DEM. A technique to suppress this striping artifact, based on the estimation of power spectrums for each column in a DEM, was recently developed by Russell, Kumler, and Ochis.
The USGS DEMs used for the AMC project were derived using a variety of photogrammetric methods that varied over time. In addition, some of the DEMs use feet, while others use meters. Tocreate a seamless database of elevation data for the White Mountains, the DEMs had to be reprocessed using the Russell, Kumler, and Ochis' Power Spectrum technique, then subsequently convert all DEMs to feet, and edge match each DEM to eliminate discontinuities at the boundaries.
Since the spacing between the DEM points is 30m, the resulting contour map only bares a resemblance to the published 7.5 minute hard copy map. In rugged or rolling terrain, the contours more closely approximate the landscape than in broad valleys or flat areas where contour lines are less tightly constrained by the terrain. In extremely flat areas, such as southern Louisiana, the granularity of the DEM, i.e., measurements to the nearest foot or meter, results in the generation of unattractive "stair stepped" contour lines.
On a hiking map, correct contour elevations are extremely important, as outdoor enthusiasts are likely to count the number of contour lines remaining until reaching a summit. This presents a problem with respect to the USGS DEM data because the 30m spacing was coarse enough to miss the tops of craggy mountain peaks, as well as isolated low areas.
As an experiment, the AMC/WhiteStar mapping team digitized key high points and low points for one of the DEMs from the 7.5 minute maps, in addition to points along ridges in an attempt to correct ridge lines. Using a technique pioneered by Ed Russell of Computer Terrain Mapping, these points were triangulated and put on a grid to form a correction surface, which was then added back to the old DEM to generate an entirely new DEM enhanced to better honor the AMC control point file.
Following the initial pass, the team noticed that occasionally the new DEM would show a mountain peak with an anomalous "extra" summit or depression that didn't exist in reality. While AMC accurately digitized control points from source maps, these did not integrate as well as desired at the same x and y position in the DEM.
The USGS is silent on the subject of x and y positional accuracy of the 7.5 minute DEMs. It was suspected from this experiment that the x and y accuracy of a 7.5 minute DEM is in the neighborhood of plus-or-minus 100 feet, based on the horizontal adjustment of the control points. To put this into perspective; this degree of error is, at most, one or two cells (a 30m square) away on the average.
It was also difficult integrating stream data with the DEM-derived contours. In some cases, the stream data derived from other sources did not integrate correctly with the contours derived from the DEM data. This is a classic, complex GIS dilemma, and one can only speculate on the possible sources of error when integrating different data sets, including the DEM itself, the stream data, differing coordinate systems, map projections, datums, the GPS equipment, the stretch of paper, systematic error in various data collection processes, computers, software, and digitization. Given the diversity of sources for cartographic error, it is nearly impossible to locate a single culprit, much less an entire group of culprits.
To address some of these concerns, Ed Russell of Computer Terrain Mapping defined a new process to examine a local region within the vicinity of each control point for a corresponding low or high point so that the DEM could be corrected at the corresponding coordinates. This step has the effect of warping the DEM to fit the accurately digitized AMC control points.
To avoid problems at the edge of the DEM, the warping was constrained to drop to zero at some distance from the control points. In addition, there was no way to warp the DEM using control points that had neither a local high or a local low in the vicinity. Such points were ultimately discarded because they did not enhance the quality of the data.
To check the z values of the AMC control points for typographical errors, Russell created an error or displacement surface by taking the z value of the AMC control point file and substracting it from the corresponding z value of the DEM. The residual file was then turned into a point file for inspection and correction inside of PC ARC/INFO. After some corrections, the mapping team recalculated a new error surface and added it to the original DEM to adjust the key digitized points to their proper elevations. A triangular irregular network was used to interpolate the corrected elevations between the digitized control points. Appearance of Contours
A few USGS DEMs were available (only in terms of meters) in the White Mountain Project area. In a metric DEM, any given elevation is specified to the nearest meter (the overall vertical accuracy of a DEM as stated by the USGS is either plus-or-minus 7m or plus-or-minus 15m). The granularity of the DEMs causes an undesirable stair-stepping effect in the appearance of contour lines in flat areas. Our recourse was to use the spline function in the ArcEdit software module of ARC/INFO. Applied injudiciously, this can cause contour lines to overlap.
Rethinking the process, the ArcEdit densify command was first used to add points at regular intervals along each contour before using the Spline command to smooth the appearance of the contour line. The exact values used for this reprocessing were derived via a tedious process of elimination procedure, but eventually yielded better results. Even then, cartographers at AMC still spent an additional six weeks manipulating and reformatting the data to suit their particular needs.
The final maps also showed some strange anomalies such as contours going the wrong way, peculiar cross-shaped contours which are attributed to the algorithms used in the contouring software, and the realization that the mathematics in most contouring programs have no way of knowing which contour shapes are likely to occur in the real world and which are unlikely to occur.
The team also noticed a tendency of the software to generate very small closed contour polygon artifacts that were undesirable on the finished map. After building polygon topology on the coverage, all polygons with an area smaller than a predetermined threshold value were removed. Manageability of Contour Files
It was then recognized that contouring each DEM as a separate project would create seams among adjacent DEMs, requiring unreasonable amounts of human intervention to repair. Because contouring software only "looks" at the immediate data set it is given, and cannot take into account data in an adjacent DEM not currently loaded onto the system, discontinuities can occur. The solution was to merge as many together as possible into a larger DEM, then contour the resulting file.
The mapping team made an attempt to put together DEMs that matched the geographical size of the desired map sheet. Initially, the team wanted to contour the data at 50-foot intervals, but the resulting data sets bogged down both the PC version of ARC/INFO, as well as the illustration package used to generate publication quality maps. The manageability problem was solved by opting for 100-foot contours. Summary
USGS DEMs, if of a high enough quality and when used under certain conditions and parameters, can provide a more cost-effective solution for elevation mapping of large areas of the country. Prospective users need to think carefully about the accuracy of project requirements, with particular attention to suitability as to the final output scale of the map or engineering requirements. In the case of AMC, the various steps ultimately produced a result that was acceptable for the publication of their Guidebook Series at the 1:95,000 scale. Given an unlimited budget and the man-hours necessary, one alternative would be to acquire digitized and scanned topo quads. However, because of the switch to the metric system, the sizes of the files, and the diversity of contour intervals found on the maps in any given project area, many of these issues would certainly be encountered again. From a long range usability perspective, the USGS DEMs at least need to be edge matched and processed using a technique such as that developed by Russell, Kumler, and Ochis to provide the seamless nationwide database that would be of greatest utility to GIS users, scientists, and cartographers. In addition, some information needs to be provided with respect to the horizontal accuracy of the DEM product.
"A USGS DEM is not a substitution for scanning and vectoring a USGS 7.5 minute map," says Robert White of the WhiteStar Corporation. "However, if you understand your project requirements sufficiently, and look at the limitations of the data sets carefully, you can use the USGS DEMs as a valuable resource to achieve complex contour mapping without busting the budget and also achieve some really good results."
The overall AMC project accomplished several things, including the addition of summits to the DEM and creating a better contour layer, which adds significant value to the end product. The overall opinion of the completed maps is good, and the project is a quantum leap forward for AMC and the organization's guidebooks. AMC realizes that the published Guidebooks are only as accurate and reliable as the sources and data used for their production. Errors are likely, and Guidebook users should utilize the information accordingly and in conjunction with other sources when hiking the back country. More Info
Appalachian Mountain Club, 5 Joy Street, Boston, MA 02108; Tel: 617-523-0636; Fax: 603-466-2721, Web: www.outdoors.org WhiteStar Corporation; Tel: 303-781-5182; Toll Free: 1-800-736-6277; E-mail: [email protected]; Web: www.whitestar.com Back |
