KSi Knowledge Sciences, Inc.
go to KSi home page Terrain Analysis System - Integrated elevation and spectral data

Why Use Spectral Data?

Digital elevation data provides three dimensional surface configuration information which is the critical identifier of landform type. The use of knowledge based, physiographic landform models and the application of various morphometric operators to the elevation data can lead to the inference of landform type. Accurate identification of landform type leads to the prediction of probable composition and properties. Thus, knowing landform type should provide key information regarding terrain characteristics for military and civil applications.

Spectral data by itself, are of little use in landform identification because spectral characteristics relate to surface composition rather than shape. However, spectral information in conjunction with surface configuration can help to identify some landform types. In addition, the use of landform and spectral information together can provide information on surface composition that can then be used to infer soil condition factors. In these situations the interpretation of spectral signatures is significantly more constrained and thus should be more accurate.

Elevation Data

Several IFSAR data sets were used that consisted of elevation data and the associated radar image collected using the ERIM airborne sensor. The resolution of these data sets are 2.5 meter horizontal post spacing with an .86 meter rms vertical error. The following IFSAR flight lines were used: m37p23s1, m37p19s1 which cover a portion of the Avawatz mountains in the Silurian Valley . Figure 1 shows flight line m37p23s1 which is just north of m37p19s1 shown in figure 5. These figures are shaded relief models and contour plots using the 2.5 meter data undersampled by a factor of 16 (i.e., 40 meter post spacing). All elevation data processing used this undersampled data.

Spectral Data

The Landsat scene containing the area of interest was scanned such that it would pixel register with the corresponding IFSAR elevation data set. Using the Landsat and our knowledge of the study area, spectral signatures were assigned to the appropriate pixels (not all pixels were assigned a signature). These signatures represent the following materials: varnish; fine sand/silt; sand/sand sheet; and grus/sand/silt. Figures 3 and 7 show the resulting classifications (refer to figure 13 for the legend). All spectral signatures used were from either in-situ ground measurements or laboratory measurements of samples running from 400 to 2500 nm with a spectral resolution of 1.5 to 4.5 nm. These signatures were taken from the Spectral Information Technology Applications Center (SITAC) Spectral Signatures Library, and the Hypermedia Terrain Database.

Results - Valley Floors and Playas

Using the elevation data the mountain regions were first delineated. This was done by the Basin and Range landform knowledge base using the slope, elevation and curvature characteristics representative of the Basin and Range physiography. If mountains are found, then the possible presence of valleys are inferred. Areas of relatively low slope, and low surface variation that are large and adjacent to mountain regions are delineated as valley regions. If valleys are found, then the presence of valley floors are inferred. Within each valley region, areas of very low curvature (i.e., flat), low elevation, having high accumulated drainage values are delineated as valley floors. The m37p23s1 and m37p19s1 data sets were processed. Figures 2 and 6 show the delineated valley floors overlaid on a depiction of the undersampled elevation data that shows dark areas as higher elevations and light regions as lower elevations.

The Basin and Range knowledge base further infers that if valley floors are found, then the presence of playas is possible. Since playa surfaces typically consist of light materials such as sand and silt, the knowledge base constrains the search by applying the fineSand _siltSignature Region function to the valley region. Any resulting delineated regions are retained as playas only if they significantly overlap the valley floor (but are not necessarily completely contained in it) and indicate a high level of accumulated drainage within. The resulting playa regions for the m37p23s1 and m37p19s1 data sets are delineated in figures 4 and 8. Figure 9 shows the m37p19s1 lower resolution playa delineation overlaid on the high resolution radar image. For clarity the radar image has been reversed, such that light areas are non-reflective and darker areas are more reflective in the radar signal domain. One can see that the playa cannot be determined by the elevation data alone. Conversely, the playa cannot be determined directly from the spectral data in that very similar sand sheet returns can be found in other areas (see the similar toned areas in figures 3 and 6). However, a region having a sand/silt signature that is further constrained to lie mostly within a valley floor is highly suggestive of a playa.


Results - Alluvial Fans

Alluvial fans in the Basin and Range characteristically have curved, lobate surfaces that are usually found in  the valleys between the mountain crests and the valley floors. Thus, if valleys exist, the knowledge base infers the presence of alluvial fans and looks in these areas for shapes having a convex plan derivative. These regions are delineated using the elevation data and maintained as potential alluvial fan objects. The mountain regions are then processed for pass shaped areas. A mountain pass region likely indicates potential sources of flow out of the mountains into the adjacent fan (i.e., the fan apex) and thus reinforces the fan landform inference. Any potential alluvial fan objects with higher elevation areas (i.e., up slope fan regions) that are not adjacent or do not overlap a pass region are eliminated. The results for the m37p23s1 data set are shown in figure 11. The curved contours in figure 10 can be seen to coincide with the lobed fan features.

 The presence of grus/sand/silt or varnish spectral characteristics within the alluvial fan region is reinforcing evidence of an alluvial fan landform. The grus_sand_silt SignatureRegion and varnishSignature Region functions were used to search the spectral image within the fan objectís spatial boundary for evidence of these signatures. If either one or both of the resulting regions covered a large percentage of the objectís total area it was retained as a fan. These spectral regions and the fan boundary are shown in figures 14 and 15 overlaid on the high resolution radar image. On the basis of these spectral regions the larger fan object was further classified into two classes of alluvial fan F2 and F3. The F2 class is an old deposit that has been in place long enough to have a varnished lag surface. The F3 class contains younger reworked material. Figure 16 shows the varnish spectral region outlined on the high resolution elevation data variance map. The variance map was generated by applying a range pixel algorithm which indicates significant surface variation to the full 2.5 meter resolution elevation data. This algorithm replaces the pixel in the center of a moving 3x3 matrix with the difference between the maximum and minimum elevation in the matrix. The results were thresholded such that values of less than 2 meters in vertical variation are light and more than 2 meters are dark. While the hard surface of the F2 may be more desirable for trafficability the variances in elevation indicate a dissected surface.





Copyright © Knowledge Sciences, Inc. PO Box 3385, Ponte Vedra Beach, FL 32004-3385
904-280-7478 bleighty@knowsci.com