Advances in sensors, imaging techniques, and software propel remote sensing to the top of its class for environmental monitoring.
The world is in constant flux: Everyday there are changes in land use, deforestation, erosion, and the earth’s natural resources. Monitoring such alterations throughout the world would be nearly impossible without remote-sensing technology. In the 1990s, the dimensions of a new generation of container ships required the expansion of the river Elbe, the outlet leading from the harbor of Hamburg, Germany, to the North Sea. The long-term effects of the project on the river’s ecosystem were uncertain, however, so the German Federal Waterways and Shipping Administration (WSV) asked Manfred Ehlers to establish a continuous monitoring program using remote sensing.
Ehlers, director of the Institute for Environmental Sciences (IES) at the University of Vechta, Germany, has been pioneering remote-sensing technologies since the 1970s. “This is the only technique that can get you current information on large geographical areas with accurate data recorded in a very short period of time,” he says.
Seeing in color
In the past, the study requested by the WSV would have been nearly impossible. Researchers, supported by conventional aerial photographs, would have laboriously generated vegetation maps from extensive field work. “This expensive task requires much manpower and often lacks the necessary geometric accuracy,” says Ehlers. In addition, inconsistencies occur whenever the person analyzing the photographs changes.
Modern remote-sensing techniques provided the solution. In cooperation with the German Aerospace Center (DLR; Berlin, Germany), Ehlers’s group developed a three- to five-year monitoring program for the river that is based on the High Resolution Stereo Camera-Airborne (HRSC-A). Originally developed by the German space center for a Mars mission, this camera supplies high-resolution multispectral scanner data, including 8-bit panchromatic imagery (585765 nm) and multispectral imagery (395 to 485 mn, 484 to 576 nm, 729 to
771 nm, 920 to 1020 nm), and 16-bit digital surface model (DSM) data. With 15-cm ground pixel resolution at 3750 m, the camera has an accuracy of ±10 cm along the x- and y-axes and ±15 cm along the z-axis.
The HRSC is a pushbroom-style scanner composed of a trio of 5184-pixel linear charge-coupled-device (CCD) detectors that operate simultaneously. One detector points forward, one to the nadir, and one to the back of the flight line. By combining the information from all three detectors in a technique similar to photogrammetry, one can obtain a “triplet” stereo image that yields the three-dimensional information necessary to develop a DSM. The unit incorporates global-positioning-satellite (GPS) measurements and inertial navigation systems to monitor its movements, providing geometrical corrections that allow an almost perfect overlay on a map or a geographic information system (GIS). “The HRSC is actually the best aircraft remote-sensing system I’ve ever seen because of its high geometric fidelity,” says Ehlers.
The Geoinformatics group of the IES developed a context-based hierarchical image analysis and classification scheme that incorporates a priori GIS information, digital surface models, and multispectral image data. This image analysis process operates with advanced masking techniques, dividing the images into several semantically meaningful layers (for example, nonvegetation/tree vegetation/herbaceous vegetation) by using indices and threshold values derived from the original image data.
The biggest challenge in remote sensing is making sense of the prodigious amounts of data. While the aircraft takes images of the scene from above, Ehlers sends in students to perform “ground truth” studies in which they record what is actually on the ground. Once these points are located by a GPS, Ehlers’s group ties them in with a geographic information monitoring system, a software system for storing, retrieving, manipulating, analyzing, and presenting geographic information.
Using the ground-truth data and the remote-sensing images, Ehlers and associates “train” the image analysis software. Known as supervised maximum likelihood classification, this is a statistical procedure that classifies the whole image based on a set of true ground samples.
Ehlers and his group predominately analyze the images after they are taken. They perform classification, geometric rectification so that data fits with a map or GIS, and image enhancement such as filtering and contrast improvement.
“But this is not an easy task,” he says. Because of the high resolution, a single image may be as large as 80 Gb. Handling such large data files requires significant changes in the commercial imaging software employed by the group. Ehlers begins with commercial packages such as Erdas, Earth Resources Mapper, and E-cognition, then writes add-on programs to fit his needs. “We use the format of the programs and write our own data processing and exchange algorithms. No one writes image-processing software from scratch anymore,” he says.
Although the Elbe monitoring program will continue for another four or five years, short-term results show the tremendous advantages of the new method in terms of the richness of detail and geometric accuracy, especially when contrasted with older, conventional imaging maps of the same area. For comparison, resolutions for conventional satellite imagery range from 30 m for Landsat to 5 m to 20 m for the Indian Remote Sensing (IRS) satellite.
In the future, Ehlers’s group hopes to develop an imaging spectrometer that can produce images over as many as 200 spectral bands, each 20 nm wide. Although such instruments do not provide stereo viewing capability, their narrow channel widths offer high spectral resolution that is useful for remote sensing. Imaging spectrometers can register subtle effects, for example, stress effects on vegetation indicated by changes at the chlorophyl absorption line. Standard wide-band remote sensors cannot detect these minute alterations.
In a separate project, the group is using 15-cm-resolution stereo-scanning data to produce a digital surface model through correlation techniques. This model enhances image interpretation by adding a third dimension, which is particularly useful in urban remote sensing for tracking the elevation of buildings or trees. “We have many areas in Germany where the ground is covered in concrete, buildings, bricks, parking lots, and streets,” says Ehlers. In these places, the water runs off directly into the rivers without treatment. A renaturalization movement is afoot in the country to replace these covered places with lawns or partially open stone that would allow the water to percolate through to the water table, undergoing natural filtration.
There is also a practical reason behind the movement: water taxes. In Germany, property owners are billed on how much water runs off into wastewater treatment systems versus sinking into the ground. The government wants to enforce this tax, but because of the small lots in Germany, researchers have to visit each property to assess the situation. With high-resolution spectral and spatial images from aircraft for analysis of water run-off, coupled with groundwater models, the government will have an accurate assessment quickly and efficiently. “It is wonderful to see how people are trying to improve the environment and how remote sensing can help them,” says Ehlers. oe
The man behind the technology
Manfred Ehlers is considered one of the pioneers of remote sensing, having spent his career focusing on the advancement of this sophisticated science that helps researchers monitor the earth’s changing environment.
After receiving his degree in mathematics in Germany, Ehlers took up oceanography at the Institute for Marine Research in Kiel. When the German Ministry of Science and Technology started a program promoting a new technology called remote sensing, Ehlers and his associates applied it to their studies. “We used scanner data and aerial photography to monitor artificial dye patches in the Baltic Sea and to fit our observations to diffusion models,” Ehlers says. The group tracked and simulated oil spills, and monitored eddies and ocean currents. “Remote sensing has been a part of my professional life ever since,” he adds.
In 1977 he joined a special coastal and marine remote unit at the University of Hannover. “There was no commercial image-processing software for remote sensing available at the time,” he remembers. So he developed software for the analysis of remotely sensed images.
After earning his Ph.D. in surveying engineering, he went to the University of Georgia (Athens, GA), where he switched disciplines to geography in 1984. His intended one-year stay in the United States stretched to six once he moved to Maine to help establish a remote-sensing program at the state university. He also became a research scientist at the National Center for Geographic Information and Analysis, as well as a member of the scientific policy committee.
In 1990 he returned to Europe to join the International Institute for Aerospace Survey and Earth Sciences in the Netherlands as a professor for aerospace data acquisition and photogrammetry, then subsequently headed the newly formed Department of Geoinformatics. Two years later, he returned to his native country as a professor for geographic information systems (GIS) and remote sensing at the University of Vechta. There he started programs in environmental sciences and founded the Institute for Environmental Sciences and the Research Center for Geoinformatics and Remote Sensing, both of which he heads.
Although Ehlers works primarily on the application side, he cooperates closely with the scientific community to develop the equipment required for his work. In Europe, for example, remote sensing needs to be suitable for small field sizes and many buildings. In the United States one can get away with 30-m resolution because there is so much space. “[In Europe,] we need
1-m, 4-m, or 8-m resolution, which we now have,” Ehlers says. He also requested imaging at narrow channels at red and near-infrared spectral bands to enable better image interpretation.
Ehlers now focuses on integrating the three major areas of his academic career: remote sensing, GIS, and image processing. “I would like to see GIS and remote sensing closer together on the analysis side,” he says. Presently there is no true software package that is fully integrated with image analysis and vector GIS capability. Ehlers and associates are working to develop the requirements and prototypes for such a system, while he continues to teach and monitor the earth’s environment.
(By Laurie Ann Toupin, SPIE, OEmagazine, April, 2001)