Earth elevation models have experienced truly dramatic improvements in recent decades. Various technologies and algorithms applied to both satellite and airborne platform sensors have pushed the limits of resolution of the best currently available elevation models by orders of magnitude beyond what was previously available. Today’s capabilities for the production of elevation model data truly define a new era in several respects, with new applications rapidly emerging and many more that are yet to be discovered.
Historical Methods of Collection and Representation of Elevation Models
Prior to the application of modern remote sensing approaches to elevation modeling, techniques used to measure elevation data were manual and laborious. The time required to conduct surveys placed inherent limits on the amount of data that could be collected. More automated techniques, and in particular those involving remote sensing, needed to be fully developed before realization of a significant expansion in the amount of data that could be collected.
Aerial film photography permitted photogrammetric approaches to be applied to the task of elevation modeling over larger areas. Application of advanced photogrammetric techniques to digital imagery today can produce extremely high-resolution digital surface models (DSMs) and digital terrain models (DTMs). Examples of what is possible using state-of-the-art photogrammetric techniques will be shown later in this article. Due to industry-wide investment in digital image sensor technology and its use in commercial high-resolution imaging satellites, photogrammetry currently can provide the highest-resolution elevation models at the greatest distances.
Other sensors and technologies provide elevation model information in forms that are complementary to photogrammetric techniques. The invention of the laser in 1960 enabled the development, beginning in the 1970s, of airborne laser scanning (LiDAR, Light Detection and Ranging), which calculates the distance to the ground more directly through time-of-flight measurement of pulses of emitted laser light. The density of points captured by LiDAR sensors has steadily increased in support of topographic mapping requirements. Use of different laser wavelengths enables such applications as topographic-bathymetric elevation model capture, among others.
The first International LiDAR Mapping Forum (ILMF) convened less than ten years ago in 2001 and has been growing steadily along with advances in the technology. One of the important characteristics of LiDAR is its ability to collect multiple returns from each pulse, which helps improve determination of the true ground surface. Due to its much smaller wavelength, LiDAR is able to detect smaller objects than radar can. Current research in LiDAR technology will result in higher density collection capabilities leading to further advances in mapping our earth and neighboring planets.
Distance measurement using radar has similarly evolved. Resolution of imaging radars has improved, and will continue to improve. Radar of appropriate frequencies can penetrate foliage, supporting better determination of the ground surface, and even can provide penetration below ground. Radar, like LiDAR, is an active self-illuminating system that supports collection day or night. Interferometric radar processing techniques use the phase difference of the returned radar energy from multiple radar images, rather than the amplitude, to compute topographic height more accurately.
Elevation models have been represented by various means according to their intended purposes. They have long been represented by two-dimensional descriptions; typical of these are contour lines, engineering drawings using an “elevation” view or profiles, or even early cartographic products enhanced with artistic oblique renderings of terrain or cultural features. The first three-dimensional elevation models, predating digital computers, were physical models of the terrain. These were produced from stacked layers of material, such as wood, or made as plaster casts. Relief pantographs were used to reproduce and cut models using a contour map as reference.
In the 1940s, the U.S. Army Map Service first used thermoplastic vacuum forming to construct physical models more rapidly. Red-blue anaglyph stereo images have been superseded by holographic printing, which can render full color terrain models in 3D and do not require the viewer to wear special glasses. Using similar technology, dynamic displays that render holographic models in real time have been demonstrated. Today, 3D printing techniques used for rapid prototyping of 3D models can be used to produce more complex and detailed physical elevation models, and to incorporate application of imagery to the surface in the same process.
To support digital processing and visualization, elevation models may be stored in different forms. In traditional GIS applications, elevation data are primarily stored as a grid or TIN (Triangulated Irregular Network). As terrain data sets have grown to very high resolution, representing elevation models in formats previously applied to imagery such as TIFF, GeoTIFF, or NITF has become more common. Specialized formats, such as LAS for LiDAR, have evolved to address the need to capture elevation data and metadata that is important at today’s higher resolutions.
Also, in the same way that image pyramids have been used for efficient display of very large image files in the past, elevation models of higher resolution now must employ similar techniques for efficient processing and visualization. Pyramids, level of detail, and data streaming are applied to terrain data increasingly as the resolution improves. Historical distinctions between elevation models, imagery, features, and attribution intermingle as elevation models reach the fidelity currently achievable. Applications of elevation data in the new era will be progressively more complex and demanding and will drive further innovation in data representations necessary to support them.
From Kilometer to Decimeter “X”treme Elevation (100,000,000 “X”)
Earth elevation models have experienced truly dramatic increases in resolution in recent decades. In the public domain, the only global coverage elevation model just ten years ago was the 30 arc second post spacing data set, GTOPO30. Figure 1 is a shaded relief model covering a portion of Hobart, Australia, approximately 65 sq km in size. Although global in scope, the GTOPO30 elevation data are very low resolution by today’s standards, with the elevation sample spacing of approximately one kilometer. (For comparison, see the same geographic area in Figures 2-5.)
In less than a decade, we now have multiple near-global elevation data sets that are almost one thousand fold denser than GTOPO30, providing post spacings of 1 arc second. The Shuttle Radar Topography Mission (SRTM), flown aboard the Space Shuttle Endeavour in February 2000, generated earth terrain model coverage at 1 arc second from 60 degrees north latitude to 56 degrees south latitude, a feat that was unprecedented in scope and resolution at the time. Figure 2 shows SRTM terrain data, the 3 arc second public release of the same area covered by GTOPO30 in Figure 1. The GTOPO30 data set has been subsequently upgraded (SRTM30) using the SRTM data.
On June 29, 2009, NASA and Japan’s Ministry of Economy, Trade and Industry jointly released GDEM terrain data derived from the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) instrument on the Terra satellite. The GDEM terrain models are 1 arc second resolution. This data set incorporates additional area not covered by SRTM; GDEM data spans from 83 degrees north latitude to 83 degrees south latitude. Figure 3 shows an example of GDEM over the same area as Figure 2 for comparison. Note the significant improvement in resolution of both the SRTM and GDEM data as compared to the GTOPO30 data.
If that resolution were not enough (and it’s clearly not), various sensor and processing technologies applied using both satellite and airborne platforms have pushed the limits of resolution of the best currently available elevation models even further. Figures 4 and 5 show a 5-meter DTM and a 1-meter DSM, respectively, of the type that Harris Corporation routinely produces from stereo satellite imagery using state-of-the-art photogrammetry. Our DSM extraction algorithm has been designed to use multiple stereo pairs viewing the same scene. The major benefit of this is that portions of the scene that are occluded in one stereo pair may be seen in stereo in another pair. In this way, we can produce more complete DSMs with fewer occlusions.
The source used to create the products in Figures 4 and 5 is 0.5-meter GeoEye-1 satellite imagery collected by GeoEye. The 1-meter DSM is a two-dimensional array of points that contains the elevation of the earth’s surface at each point. In this context, it is referred to as a reflective surface model because the height is for the objects that reflect visible light. As such, man-made features such as buildings and bridges are included in the DSM. The 5-meter DTM is similar to the DSM product, except that vegetation and man-made features have been removed and replaced with the bare earth underneath.
See Figure 6 for examples of similar resolution surface models generated using WorldView-1 satellite imagery collected by DigitalGlobe. The color image is a 1-meter DSM. The gray scale figures are detail images (center) and the corresponding detail 1-meter DSMs (bottom). These 1-meter DSMs are approximately one million times the density of GTOPO30.
These imaging satellites, and others, combined with appropriate photogrammetric extraction technology, provide for generation of similar high-fidelity elevation models practically anywhere on the earth. Elevation models generated from these sources do not possess accessibility limitations of airborne LiDAR or IFSAR.
This same state-of-the-art photogrammetric technology is also applied to aerial imagery to yield DSMs with one-decimeter resolution. Decimeter post spacing elevation models produced today have approximately one hundred million times the density of GTOPO30. The source used to create DSMs in Figure 7 is aerial imagery collected by Pictometry International Corp.
Applications in the New Era
In view of the history and the current state and variety of elevation model sensor and production technologies, focusing this discussion on photogrammetric approaches and applications is appropriate. With visible light photogrammetry, the main concern is the effect of clouds and other atmospheric conditions. Fortunately, visible imagery is most often simultaneously the most commonly available source, most current, and highest resolution – particularly with respect to satellite sources.
Areas of high interest are imaged for other purposes at high frequency relative to the timescale of changes to the terrain, allowing terrain models to be built up and improved over time and exploiting largely redundant information. When available, other sensor data (LiDAR, IFSAR, etc.) may be used to augment photogrammetric elevation models; the type of information other sensors provide can be highly complementary to the imagery source.
Elevation is used in numerous applications, and the need is growing fast. It is used in the GIS/Mapping, Military, Oil & Gas, and Wireless Telecom industries and in many more applications. The world’s need for energy and “green” initiatives has put pressure on the geospatial technology community to produce high-resolution and accurate elevation models. High-resolution terrain data provide essential information when planning for corridor right of way, land restoration, damage prevention, risk management and emergency response.
High-resolution elevation models play a major rule in the demanding requirements of the wireless telecom industry. DSMs are used for point-to-point microwave analysis and DTM, along with other geospatial data layers such as 3D building models and clutter data, are used for WiFi and WiMAX applications. Using such datasets minimizes surveying costs and offers efficient ways to expand existing networks.
Height “Z” Matters
Elevation is foundational for many geospatial applications. Without elevation data, features extracted from imagery have relief displacement, and processed data are not accurate. When it comes to elevation models, it might be that even one decimeter spacing is not good enough. In this new era, the challenge is processing and managing so many millions of elevation values efficiently and accurately for the many wide-spread applications.