Spatial Technologies Working in a 3D World

31 Mar, 2005 By: James L. Sipes

Understanding Terrain Models

The World Of Gis (geographic information systems) is a 3D world. Much of the geospatial data is not only georeferenced to fit an actual location on the earth, but also includes z-coordinates that define the elevation and height of both natural and manmade objects. A critical part of any GIS that takes advantage of 3D data is the modeling of the terrain (figure 1).

 Figure 1. Spot 5 2.5M resolution images of the Dead Sea in Israel and Jordan. Copyright CNES 2005, provided courtesy of SPOT Image Corp. (<a
Figure 1. Spot 5 2.5M resolution images of the Dead Sea in Israel and Jordan. Copyright CNES 2005, provided courtesy of SPOT Image Corp. (

Most architects, engineers and landscape architects are familiar with using primarily CAD and digital survey data to help create 3D landform models for a particular site. Site surveying involves measuring the location of objects on the Earth's surface. Monuments or benchmarks that have predefined x, y and z coordinates serve as reference points for determining the location of objects. I learned to do site surveying years ago using a Philadelphia rod, plumb-bob and dumpy level. Fortunately, surveying equipment has improved dramatically since then, and today much of the equipment uses lasers and GPS to supplement traditional approaches.

The major limitation of site surveying is that it isn't effective for creating 3D landforms at a large scale. DTMs (digital terrain models) and DEMs (digital elevation models) used for GIS models are available from federal, state, local and private sources. Most digital elevation data is calculated using stereo pairs, digitized from existing maps or generated from laser pulses, and is available in many different formats and resolutions. Users need to know where to access DTMs and DEMs, how they were created, the level of accuracy and detail they can expect from each, and appropriate uses for different digital elevation data.


Both DTMs and DEMs contain elevation data for landforms. A DTM represents the theoretical surface on the ground, and a DEM represents the theoretical surface of high points, such as treetops. DEMs consist of a series of elevation points arrayed in a rectangular grid. Each grid point has a unique x and y coordinate. The spacing of these grid points determines the level of resolution of the resulting DEMs. The finest quality DEMs you can get from USGS (U.S. Geological Survey), for example, are based on a 10-meter grid.

DEM Development

The USGS has produced and distributed DEM data since the 1970s. USGS takes two primary approaches to creating DEMs: remote sensing and photogrammetric techniques, and scanning contour map information.

Remote sensing technology acquires knowledge about the Earth's surface from airborne or satellite platforms. The technology has been used for decades to collect geospatial data, and new instruments and processes are continuously being updated and developed. The USGS originally used photogrammetric techniques in the late 1970s and early 1980s to produce orthophotos and magnetic tape recordings of DEMs. These photogrammetric techniques typically focused on stereo aerial photography and stereo satellite imagery. Today, photogrammetry involves the digital representation of stereo pairs (figure 2).

 Figure 2. IKONOS satellite image and DEM provided by Satellite Imaging Corp. (<a
). Copyright 2005, Space Imaging, all rights reserved.Corp. ("/>
Figure 2. IKONOS satellite image and DEM provided by Satellite Imaging Corp. ( Copyright 2005, Space Imaging, all rights reserved.Corp. (

Scanning contour maps. In this approach, the contour map is scanned at a fairly high resolution (500-1000dpi) to create a raster-based file that is then converted to vectors by some type of graphics translation software. In the past, the USGS used four different methods to collect DEM data, but the only one it currently uses is the interpolation from vectors or DLGs (digital line graphs).

The instruments used to collect data to generate DEM files can be impaired by weather conditions, and some landscapes, such as very mountainous terrains, have a greater degree of error. The resulting gaps in data are typically filled by interpolation or with other source data.

Available USGS Data

The USGS produces five types of elevation data: 7.5-Minute DEMs, 7.5-Minute Alaska DEMs, 15-Minute DEMs, 30-Minute DEMs and 1-Degree DEMs.

7.5-Minute DEMs correspond to the USGS 1:24,000- and 1:25,000-scale topographic quadrangle maps. They are available in 30m or 10m grids, depending on the areas.

7.5-Minute Alaska DEMs also correspond to the USGS 1:24,000- and 1:25,000-scale topographic map series for Alaska. Grid spacing for these DEMs is 1 arc-second of latitude by 2 arc-seconds of longitude.

15-Minute Alaska DEMscorrespond to the USGS 1:63,360-scale topographic map series for Alaska, and have a grid spacing of 2 arc-seconds latitude by 3 arc-seconds longitude.

30-Minute DEMs correspond to the east or west half of the USGS 30-minute by 60-minute topographic quadrangle map. Grid spacing is 2 arc-seconds.

1-Degree DEMs correspond to the 1:250,000-scale USGS topographic map series and have a grid spacing of 3 arc-seconds.

One way to break down this DEM data is based on scale. Large-scale DEM data includes 7.5-minute units for the continental United States and Alaska that correspond to the USGS 1:24,000- and 1:25,000-scale topographic quadrangle map series, and 15-minute DEM data that corresponds to the USGS 1:63,360-scale topographic quadrangle map series. The 7.5-minute DEMs are generated from digitized contour maps or from scanned National Aerial Photography Program photographs. 15-minute DEM data is produced by combining digitized hypsographic and hydrographic data from 1:63,360-scale graphics.

At the Intermediate scale is 30-minute DEM data that corresponds to USGS 30- by 60-minute topographic quadrangle map series. Intermediate-scale DEMs are produced from DLG contours of any map series 7.5-minutes to 30- by 60- minutes (1:24,000-scale to 1:100,000-scale), or from resampling other DEM models.

Small-scale DEMs include the 1-degree DEM for all of the contiguous United States, Hawaii and most of Alaska. They are typically produced from cartographic and photographic sources.

DEM data from USGS is available in a variety of formats, depending on the location and scale of the original data (see "USGS Data Formats" below).

Light Detection and Ranging

A relatively new type of remote sensing application that is currently very popular for creating DEMs is LiDAR (light detection and ranging). Many think of LiDAR as a new technology, but it was first paired with digital imaging in the mid-1990s. A LiDAR system uses powerful laser transmitters and receivers, a GPS receiver and an inertial flight management system that is used to calculate the altitude of the plane. With LiDAR, a laser beam fired from an airplane to the ground helps determine elevation points. The time it takes the laser beam to hit the ground and return is used to calculate the precise distance to the ground.

The level of detail obtainable from LiDAR is staggering. For example, on a LiDAR terrain model that shows downtown Seattle, you can actually pick out the layout of all the streets because of the elevation changes of the curbs that separate the streets from sidewalks. LiDAR, however, is not necessary for every project. For many planning projects, that level of detail is just not needed. Because LiDAR generates so many points, a LiDAR file can be monstrous in size.

In general, LiDAR systems provide a horizontal accuracy of better than 18" and a vertical accuracy of better than 6". Commercially available LiDAR systems are typically advertised to have a vertical accuracy of around 15 centimeters if ground control points or stereo superposition is used to verify the data. A network of ground control points can help provide an even greater level of accuracy. The digital data from aircraft and multiple ground stations is processed together using postprocessing software.

The biggest advantage of LiDAR technology is that it produces high-quality DTMs faster and less expensively than traditional stereo-compilation methods. As an example of the difference, a traditional processing using stereo pairs can collect 1,500 to 2,000 points per hour. LiDAR can collect about 10,000 points per second! The logical question, then, is how to handle all of that data.

In discussions of laser technologies such as LiDAR, "bare earth" modeling is often discussed. LiDAR filters can be used to eliminate all features that are above the ground, such as vegetation and buildings. What's left is a digital terrain model that shows only what's on the bare ground.

Like any technology, LiDAR has limitations. One potential problem is that it can pick up points that show trees, buildings, rock outcroppings and other vertical features. LiDAR can be confused by smooth water bodies and other surfaces that absorb or reflect the laser pulses, and very dark surfaces can also be difficult to model. Bridges, highway ramps and other structures with voids underneath typically need to be adjusted by editors. One way to clear up potential discrepancies as well as reduce the size of a LiDAR dataset is to use filters that eliminate information that isn't needed. A number of different algorithms are used to filter and thin LiDAR data.


One assumption that too many users make about DEM data is that the information is accurate and precise. Nothing could be further from the truth. All DEM data has some variation in it—the question is just how much. For example, a 1978 study by USGS indicated that there can be as much variation as 15' to 40' for steep hillsides when using 1:48,000-scale profiled stereo aerial photography.

The accuracy of a DEM is based on many factors, including the source of the data and its scale and spatial resolution. The horizontal accuracy of DEM (spatial resolution) data is based on the horizontal spacing of the elevation matrix used. Vertical accuracy is based on the spatial resolution (horizontal grid spacing), quality of the source data, collection and processing procedures and digitizing systems.

USGS Data Formats
USGS Data Formats

The method of determining DEM accuracy involves computing what is called RMSE (root-mean-square error). USGS quality control groups collect test point data and compare the DEM to the quadrangle hypsography. Standards for accuracy are established by ASPRS (American Society for Photogrammetry and Remote Sensing). For example, for Class 1 Large-scale maps with 1' contours, the ASPRS standards determine that the RMS error can be up to one third of the contour interval, or, ±0.33 feet. Most LiDAR systems meet ASPRS Class 1 accuracy for 2' contours.

Grid and TIN

Most of the digital terrain models available are in either a grid or TIN (triangulated irregular network) form. Many large data providers, including the USGS, use the grid format for terrain data because it takes up less space and is easier to store. With a grid DTM, the smaller the cell size, the higher the spatial resolution of the data.

A TIN generates a surface from a series of x, y and z points and is particularly good at picking up elevation changes associated with ridges, drains and other breaklines. Producing a TIN without using breaklines can result in terrain models that don't capture detailed landform features.

A TIN is a terrain model that uses a sheet of continuous, connected triangular faces. TIN is typically used where a greater level of accuracy is needed. TIN files are not used to store large amounts of terrain data because the file size is considered too cumbersome.

Remember that the level of resolution of a DTM can't be enhanced once it's established. Users can down-sample a grid DTM, which means that they increase the cell size and lower the spatial resolution. A TIN is vector based, so it doesn't suffer a loss of resolution.

Selecting the Right DEM Data

Selecting the right DEM data depends in large part on the location of the area in question, the spatial resolution required and the intended use of the data. Some DEM data is available only for selected parts of the world, so obviously that has a significant affect on whether a specific type of DEM data is used.

Spatial resolution is an important part of selecting the most appropriate DEM. For example, GTOPO30 and DTED0 1km both have a horizontal grid spacing of 30 arc-seconds, which is about 1 kilometer. This means that there is roughly 1,000 meters (1 kilometer) between elevation points. If you are standing in central Kansas, there may not be much change in 1,000 meters, but here in Seattle, significant elevation changes take place in that distance. Any data format with such an extremely low resolution is useful primarily to show general elevation changes.

A couple of years ago I worked on a computer animation project for PBS that focused on a series of catastrophic floods that occurred during the last great Ice Age. The floods originated when a glacial lake in Montana collapsed and flooded parts of Idaho, Washington and Oregon. We needed a 3D terrain model that would encompass much of the four states, yet provide the level of detail necessary to create animated flyovers through mountain canyons and valleys. We used GTOPO30 data to create a 3D landform of the four-state area for broader animations and NED 30m DEM data for more detailed landforms for flying through valleys and canyons (figure 3).

Figure 3. To create a 3D model of the Wenatchee Watershed in the Cascade mountain range in Washington, 10M DEM data from USGS was downloaded and combined with thematic layers that show vegetation patterns. Image by J. Sipes.
Figure 3. To create a 3D model of the Wenatchee Watershed in the Cascade mountain range in Washington, 10M DEM data from USGS was downloaded and combined with thematic layers that show vegetation patterns. Image by J. Sipes.

For an urban design project in Seattle, on the other hand, we needed the resolution and detail that LiDAR provides.

Potential Uses

Potential uses for 3D DEM models are limited only by the imagination. Terrain models are frequently used to generate altitude matrices that can be used to calculate contours, slope angles and aspects, hill shading, watershed definitions and other related types of surface analysis.

Shaded relief maps show the impact the sun has about the landscape. They visually enhance GIS maps and make it easier to see changes in landform patterns. 3D landform models are also used for hydrological analysis and for volumetric calculations to determine how much cut or fill may be required.

A home builder can generate aspect maps that show desirable south-facing slopes that provide the best solar characteristics. Geologists can use curvature maps to measure the rate that slopes are changing. We built 3D landform models of all the major transportation corridors in the state of Nevada for a recent project to determine what you could see, or not see, from the highway.

In This Article
In This Article

The FAA requires that all passenger aircrafts above a certain size be equipped with TAWS (Terrain Awareness Warning Systems), which uses DTMs to help pilots determine where their plane is in relationship to the earth's surface. The U.S. Army is exploring how to interactively incorporate digital terrain data into their modern battlefield weapons.

How will you be using 3D DEM models?

James L. Sipes is the founding principal of Sand County Studios in Seattle, Washington.

More News and Resources from Cadalyst Partners

For Mold Designers! Cadalyst has an area of our site focused on technologies and resources specific to the mold design professional. Sponsored by Siemens NX.  Visit the Equipped Mold Designer here!

For Architects! Cadalyst has an area of our site focused on technologies and resources specific to the building design professional. Sponsored by HP.  Visit the Equipped Architect here!