The Effect of Radar Azimuth Angle on Cultural Data

he use of Synthetic Aperture Radars ( ~ A R ) for tlze stzldy of cziltural and urban scenes has been prirnarily in the area of interpretation o f tlze major aspects of land use and urban nlorphology. I n such studies the image grey tone, distribution of bright (point) returns, and linear features have been the focus of the major efforts. I t is em))hasized in this paper that tlze orientation of tlze features studied is a major factor i n the grey tone o f the I-est~lting positiue image. Thus , knowledge of this orientation, wit11 respect to the azimt~tlz angle ji.e., look-direction) of the radar antenna, is of major importance. Wi th in the Los Aizgeles urbanized region, large areas have significantly lower grey tones than adjacent areas having similar land couer. This effect i.s determined to be the result of the angle difference between the rudur azimuth und the street pattern and, more importantly, the orientation of the walls o f the structures imaged. Thus , an (1 priori knowledge o f t h i s orientation is necessary in order t o ensure acc~irate interpretation of the radar imagery. For radar systems operated from platforms which haue fixed azimuth angles (e.g., satellite systems ,such as Seasat-A), an interr~retation methodology, which considers the street patterns, is considered especially critical for proper and accurate interpretation of SAR imagery. tection) as well as for preparation of land-use systems generally have a very low resolution nlaps of more static scenes. Much of the reand imagery from the active systems (radars) mote sensing work co~ lce r~ l i~ lg such scenes has been difficult for illany investigators to has dea l t primarily wi th t h e shor ter obtain. However, this last constraint has wavelengths of the electromagnetic ( E M ) eased considerably in the past several years, spectrum, i.e., photography (0.4 1.5 p n ) , and some effb~ts have been conducted with Landsat (0.5 1.1 p l i~ ) , and infrared, both respect to the iilterpretation of urban scenes PHOTOGRAMMETRIC NGINEERING A D REMOTE SENSING, Vol. 45, No. 8, August 1979, pp. 1097-1107. PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING, 1979 from airborne radars. ,Although Seasat-A (launched on 26 June 1978) failed on 9 October 1978, it had completed about 500 synthetic aperture radar passes of two to ten minutes duration each. With the availability of these data to the scientific community, it is expected that more investigators will show an increased interest in such work. With respect to the operation, design, and utility of radars for observing, monitoring, and measuring the Earth's surface, several excellent articles are available. Jensen et al. (1977) present an easily readable paper oriented primarily toward discussing the conceptual operation of the radar and including several examples of radar imagery. Brown and Porcello (1969), on the other hand, provide the reader with a more tutorial paper concentrating on an introductory engineering approach to solving the problem of obtaining radar imagery with synthetic aperture systems. Both of these papers provide an excellent background to scientists seeking an introduction to these remote sensing systems. Several additional papers have shown some of the basic concepts of the interpretation of radar images of urban and other scenes, including Bryan (1974,1975), Coiner and Morain (1971), Henderson (1973), Lewis et al. (1969), and Rosenfeld and Kimberling (1977). Although these papers have dealt with various wavelengths and polarizations, it is noted that, generally, the types of features detectable and the methodology for interpretation remain fairly constant from one radar system to another. There are, however, several variations of the data which can occur within one given (radar) system and, as mentioned, with the advent of the Seasat-A data, some of these variables may be of major importance. Two of these aspects of the radar system, which are essentially positional or geometrical in their effect on the resulting radar imagery, are the incidence angle and the azimuth anzle. With respect to large geographic and geologic features, especially for radars operating near grazing angles, the areas in shadow on the far side of a feature will be areas of no (surface) information. This does not, however imply that the shadows themselves do not provide some information for, if the investigator knows some of the operating parameters of the radar (e.g., altitude, depression angle), the height of the feature may be determined from the length of the shadows (LaPrade and Leonardo, 1969). This has been demonstrated by Lewis and Waite (1973) with respect to large topographic elevations in Panama and, for small features (river bluff), by Bryan and Hall (1976). In urban areas, however, where the elevations of the individual features (e.g., buildings) are generally much less, and especially with respect to the resolution ofthe radar, such an effort would probably be of minimal utility. Stereo radar also increases considerably the ability to determine height and slope angle, as demonstrated by Leberl (1973), but such data are even less frequent than those from single or multiple look-direction radars. Finally, Coiner and Morain (1971) note for agricultural scenes that there is a ". . . necessity to account for changes in radar backscatter (grey level shifts) generated by angular dependencies in the scene." They refer primarily to the depression angle ofthe radar and not to the azimuth angle, but do alert the interpreter to the fact that the geometry of the scene is of great importance when dealing with radar imagery. The azimuth angle is defined as the direction, on a plane tangent to the Earth's surface, in which the radar beam is pointed. Thus, although we may refer to the azimuth angle (also sometimes termed the lookangle, look-direction, or by Koopmans (1975) as the scan direction) with respect to the points of the compass, this alone is of little utility. The more important consideration is to determine the azimuth angle with respect to the orientation of the object or target being illuminated by the radar energy. Again, radar interpreters have noted the increased ease with which linear features may be distinguished by orienting the antenna of the radar at slightly different azimuth angles. Generally, this requires an additional aircraft pass for the antenna is, for all practical purposes, securely mounted to the aircraft. Several studies have been conducted to determine the effect of the azimuth angle for geologic studies (e.g., Jefferies, 1969; Dellwig et al., 1968, MacDonald et al., 1969). It is often emphasized that for best results there is a need for multiple and orthogonal azimuth angles when searching the data for geologic faults (e.g., Harwood, 1967). For urban areas we shall demonstrate that orthogonal look-directions (or azimuth angles) do not necessarily produce the optimum radar data. Eppes and Rouse (1974) have quantitatively studied the problem of the effect of azimuth angle variations on the detectability of surface features. They dealt exclusively with linear features which have some topographic expression. The results of their laboTHE EFFECT OF RADAR AZIMUTH ANGLE ON CULTURAL DATA ratory study indicate that proper spatial filtering in the Fourier plane of the radar image having a single azimuth angle could provide a detectability of the linears equivalent to that obtained by several sets of radar data taken at multiple azimuth angles. Obviously, this approach can effect a considerable savings in data collection and, although increasing processing costs, the overall result will be a lower expenditure for essentially the same information. To date there has been little interest in the radar interpretation and application community to continue the approach initiated by Eppes and Rouse (1974). However, the need for a revitalization may be imperative due to the type of data which are to become available in the near future from Seasat-A and Shuttle Imaging Radar-A (SIR-A). A set of SAR data of the Los Angeles area was obtained on 25 May 1977 as part of a joint EuropeaniUnited States aircraft Space Shuttle simulation program. The radar system used was the NASA L-band (1.215 Ghz) radar operated by the Jet Propulsion Laboratory and mounted on the NASA CV-990 aircraft. Only HH polarization and 20 m resolution radar imagery are discussed. Figure 1 shows the flight tracks of the aircraft when obtaining the radar data over Los Angeles. The cover figure is a mosaic composed FIG. 1. Flight tracks flown by NASA CV-990 while collecting Los Angeles data. 25 May 1977. of the ten north-south data runs shown on Figure 1. Density data derived from the mosaic, which is composed of densitometrically and geometrically corrected SAR data (Bryan e t al . , 1977), were used for this present study. Within this mosaic a series of target areas of residential and commercial land use was used to test the effect of radar azimuth angles on the resulting image tone. The following discussion describes the results of this study. Large sections of the Los Angeles urban area (cover) have decidedly different grey tones. By concentrating on those areas which are of the mid-grey tones, thus discounting the very dark (gravel pits, airports, flood control areas) and the very bright (industrial areas), we note that the tonal changes are generally well-defined, have geometric patterns with linear boundaries, and are many resolution cells in size. The cause of the formation of the boundaries and the included grey areas is, then, an interesting aspect of the imagery to be considered. Changes in the land cover across the boundaries seen on the SAR imagery are not apparent. All of the study sites which will be discussed were composed of residential andior small co~nmercial (2-3 story) structures. The only contrast detected through auxiliary information (e.g., ground truth, maps, aerial photography), which correlates with the linear boundaries ofthe grey areas as seen on the SAR data, is that of the street patterns. Thus, although the street patterns within the Los Angeles urbanized area are generally rectilinear, not all are oriented in the same direction with respect to true north. This is a result of several historical factors, including (1) the development of individual towns around isolated centers and their growth into a complex of suburbs, (2) the street pattern following the original rancho boundaries, and, in several cases, (3) the fitting of streets to the topography. Consequently, by using radar data obtained at a constant azimuth angle and noting the changes in the grey tones, we can identify those areas in which the azimuth angle of the radar is critical for the change in grey tone. Large-scale vegetation has been previously noted to be an important factor in the radar backscatter of urban areas (Bryan, 1974, 1975). That effect is primarily to mask the various point targets. Consequently, the result is the smoothing of some ground detail. Although this also occurs in the Los Angeles data under discussion (because large areas and patterns are being studied), PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING, 1979 the effect of vegetation is not p-ronounced. Also, because the radar system used has a 25 m resolution, the detailed effect of vegetation has been reduced. Six sites, labelled A to F and listed in Table 1, were selected for this study. Site A: (Figure 2, Burbank). In this scene the majority of the streets are oriented in three directions, at 044O, 06B0, 000" (and orthogonally) although there are some variations such as Clybourn Avenue (340" and 350"). The dominant land use is single family residential with major strip commercial areas along Burbank Boulevard, Magnolia Boulevard, and Olive Avenue. The strongest returns are from areas where the streets are oriented true north-south or east-west. The radar orientation is 090" (Table 1). The extremely bright point return in the south central portion ("D" on Figure 2) is from the buildings of the Disney Studio which are oriented north-south within an area where all other buildings and streets are skewed to the north-south grid. Freeways and major streets appear as dark (no return) areas. Site B . This area is quite similar to Site A with respect to street pattern and radar return. In the areas of darker tones the two major street orientations are 021" and 042" (and their orthogonal directions). The light toned areas have streets oriented to the cardinal points. Site C: (Figure 3, Sun Marino, Sun Gabriel, Temple C i t y ) . For the first two examples, the radar azimuth was 090"; for Site C this angle has been rotated to 26g0. Again, those areas with streets oriented to the cardinal voints of the comDass are high return areas,and areas within which the streets are Radar Aircraft Azimuth Site Run Track Angle Location A 12 000" 090" Burbank B 8 000" 090" Altadena 7 179" 269" San Marino & San Gabriel D 11 18%" 278' Los Angeles E 11 188O 278' Inglewood F 7 179" 269' Pico Rivera & Whittier FIG. 2. Study Site A Burbank, California. Radar azimuth is 090" true. (a) L(HH) SAR image; (b) Sketch map of street patterns. B: Burbank Stud~os D: Disney Studios LGCH: Lakeqide Golf Course of Hollywood NHP: North Hollywood Park SJMC: Saint Joseph Medical Center FIG. 3. Study Site C Alhambra, San Gabriel, San Marino, and Temple City, California. Radar azimuth is 269' true, (a) L(HH) SAR image; (b) Sketch map of street patterns. AMCC: Alhambra Municipal Golf Course SGCC: San Gabriel Country Club THE EFFECT OF RADAR AZIh4UTH ANGLE ON CULTURAL DATA not parallel or orthogonal to the radar azimuth angle have significantly lower return. The dark grey area to the east is a portion of Temple City in which the streets are oriented at 075@, 07QU, and orthogonally. The boundaries of this area are Duarte Avenue, Baldwin Avenue, Encinita Avenue, and Longden Avenue. The southern boundary of this grey zone is quite diffuse because there are very few streets in this area, which is composed primarily of large storage yards and light industry (in the vicinity of the Southern Pacific Railroad and Eaton Wash). In the west portion of the study site is a similar grey area composed primarily of portions ofthe Alhambra and San Marino political divisions. Botmdaries here are Garfield and Atlantic Avenues (to the west) and the Southern Pacific Railroad (to the south). The eastern boundary follows a series of interconnecting street patterns, Two golf courses, the Alhambra Municipal and the San Gabriel Country Club, are easily identified as areas of very low returns within a zone of cardinally oriented streets. Also of major interest are the bright linears which are portions of Eaton Wash. This is a normally dry wash (25 feet wide and 10 feet deep) with vertical concrete walls and bordered by a chain link fence. This feature is easily followed where it is paralIel to the flight track (and therefore normal to the radar azimuth direction). As in the case of the railroads, it can often be traced as a dark linear due to the relatively broad right-of-way through the majority of the San Gabriel portion of the image. Site a: (Figure 4 , Los Angeles). This scene, from the central portion of Los Angeles and centered on the University of Southern California (USC) and Exposition Park (EP), provides an opportunity to note the fine degree to which the street patterns can be identified using the L-band, 20 m resolution radar. USC and Exposition Park are both large dark grey areas on the image with individual buildings providing some point returns within these dark zones. East of Exposition Park and north of Santa Barbara Street is an area of major interest. For these data the aircraft track was 188", the radar azimuth was 278", and the radar was oriented off the normal with respect to the northsouth oriented street grid and approximately 12" off normal with respect to the majority of the other streets, The small area bounded by Figueroa Street and the Harbor Freeway, an area of north-south streets, is only two city blocks wide but is of sufficient width to appear as a bright pattern on the SAR imagery. I