GLORIA Imagery of the U.S. EEZ

Contents

Introductory Notes
Preface
EEZ-SCAN 1984
The GLORIA II System
Digital Processing Techniques

Introduction
Geometric Corrections
Radiometric Corrections

Digital Mosaicking
Sonar Image Interpretation
Navigation
References Cited

Introductory Notes

This document is the World Wide Web version of the standard text that appears on all USGS Atlases of GLORIA - EEZ Scan data. A typical paper atlas (or CD-ROM) contains imagery, bathymetric maps, and other geophysical data (seismic reflection profiles). Because of space limitations, initially only imagery will be presented. Images on CD-ROMs are stored as "raw" data. A typical file will be a rectangular array of ascii characters, each character representing a grey scale value between 0-255. The NASA (JPL) program IMDISP can be used to process these files. Other commercial or shareware image processing software May be used as well. All images on the WWW sites are stored as GIF. If raw images are served, they will be compressed using the "ZIP" format. Because the "raw" binary format is just a stream of numbers, it can only be read if you know the number of "lines" and "samples" in the file. This information is available in the .lbl file that accompanies each compressed image.

You may view photos and diagrams of GLORIA deployment. (300 Kb)

Preface

Because of the size of the image files (typically 20 megabytes) significant "de-sampling" has been needed to display the images over the World Wide Web. If you wish to work with the full resolution images, please obtain the CD-ROM of these data when it becomes available.

EEZ-SCAN 1984 and the U.S. Geological Survey's Mission

On March 10, 1983, President Reagan unilaterally declared an expansion of the United States' sovereign rights to all natural resources, within a zone extending to 200 nautical miles beyond the shoreline. This newly proclaimed Exclusive Economic Zone (EEZ) provided a mandate for further exploration. The general bathymetry was known, but the detailed physiography of the EEZ was not well known. Only with such detailed knowledge of the seafloor could we begin to appraise its resource potential and the consequences of exploitation or other activities where man might disturb physical, biological, or chemical systems of the oceanic or seafloor environments.

In 1984, the USGS launched a program using a long-range sidescan sonar system (GLORIA) to study the entire EEZ. During the summer of 1984, USGS and IOS (Institute of Oceanographic Sciences, U.K.) scientists surveyed the EEZ off California, Oregon, and Washington, an area of about 850,000 square km. The results of this survey are thirty-six 2-degree sheets, at a scale of 1:500,000 (see note 1). The survey cost approximately one penny per acre. The acoustic images produced by the program are no less remarkable than the first photographs from the far side of the Moon.

A cursory glance at the GLORIA imagery reveals a plethora of geologic features: volcanic edifices, fault scarps, channels, levees, slump scars, large sediment bedforms, crustal lineaments, and textural or tonal differences that reflect varying sediment types. These images provide the framework, a "road map", to direct more detailed investigations. As land surveys commonly rely on various types of remotely sensed data, so the clearer perception of submarine features provided by GLORIA enable marine geologists to focus on specific features of interest (see note 2).

The GLORIA II System

The GLORIA system is a long-range sidescan sonar developed at the Institute of Oceanographic Sciences. The system was specifically designed to map the morphology and texture of seafloor features in the deep ocean. Acoustic images of the seafloor are formed by transmitting sound pulses at the seafloor to either side of the towed vehicle and recording the backscattered sound waves. Two arrays of transducers mounted back to back are inside the towed vehicle. The arrays can emit pulses of energy at 20-, 30-, or 40-s intervals and, between transmissions, record the echoes from as far away as 30 km. The acoustic data are sequentially obtained from narrow strips of seafloor, and thus successive transmissions over adjacent strips of seafloor are used to construct an image line-by-line.

The maximum swath width largely depends on the prevailing acoustic-propagation conditions. For GLORIA, the swath width can be as great as 30 km on each side of the track. Under normal conditions, however, it is usually somewhat less. If acoustic conditions are unfavorable and the water depth is less than about 1,500 m, then the range may be less than 10 km. The maximum range for the surveys reported in this atlas is 22.5 km to either side of the ship's track.

Several factors influenced the acoustic frequency that was selected for the GLORIA system. A low frequency is desirable because sound-wave energy is absorbed in seawater, and the rate of absorption rapidly increases with the frequency of the sound wave. However, low-frequency acoustic arrays have to be large and heavy to achieve the prescribed beam shape. The GLORIA array was designed as a compromise between these factors so as to achieve both low noise levels and great stability. The operating frequency of the GLORIA system is about 6.5 kHz with the port array at 6.8 kHz and the starboard array at 6.2 kHz. Each array is 5.3 m long by 40 cm high, a configuration that gives a horizontal acoustic beam 2.7 degrees wide and a vertical beam of 35 degrees. The beam width is specified between half-power points, and considerable energy actually radiates outside these limits. The array is designed to confine the energy as nearly as possible to the plane perpendicular to the track and to fill the quadrant from the nadir (the point on the seafloor directly beneath the towed vehicle) up to near horizontal.

The acoustic energy backscattered from the seafloor is recorded in digital format. Each pixel of the image has a size, measured along the track, that is proportional to the range and extends to hundreds of meters at extreme range because the sound beam diverges at 2.7 degrees. The recording system was designed so that one complete scan is subdivided into 982 pixels. The cross- range pixel size represents about 50 m, which is smaller than the along-track pixel size, and thus features in the raw data are elongated parallel to the track, particularly at extreme range.

A few of the important technical features of the GLORIA system are provided here. For more detailed information and specifications, see Somers and others (1978). Each array consists of a total of 120 transducers, 30 to a row, 60 to each side. The vehicle is 8 m long, weighs 2.25 tons in air, and is neutrally buoyant. The array is towed by 400 m of heavy double-armored cable with 36 electrical cores for the signals and services. There is no active depth control and, at the normal survey speed of 8 knots, the vehicle depth is about 50 m. The arrays are electrically split into six sections each, and each section has its own circuit to the ship. The arrays can handle 10 kW of power at about 90 percent efficiency. The vehicle has a compass that allows the acoustic beam to be stabilized against yaw by introducing compensating delays to the section signals. The signals are detected, amplified, filtered, and smoothed before being digitized and recorded on magnetic tape.

Digital Processing Techniques

Introduction

Sonographs are a record of the acoustic-backscattering properties of the seafloor and near surface sediments; those from GLORIA represent the backscatter produced by a 6.5-kHz frequency. The strength of the acoustic backscatter is a convolution of several functions, the four major ones being

1. the slope angle of a feature relative to the incident sonar signal (topographic characteristics);
2. seafloor roughness factor, the minimum being 4 cm for the GLORIA system (determined by the wavelength of the sonar and the grazing angle of the sonar ray to the seafloor; Sabins, 1978);
3. the variation in physical properties of the upper few meters of the seafloor; and
4. the water "atmosphere", which attenuates the strength of the sonar signal as well as produces background noise.

The backscatter was recorded as a digital number (DN) with a range of 0 to 255 discrete values (8- bit data generated from a compression of the original 12-bit data). The GLORIA system was configured during these cruises to transmit a 2-s pulse of sound every 30-s to give a maximum range of 22.5 km on each side of the ship's track. The reflected sound waves were recorded on a time basis so that the data are in a slant-range rather than a ground-range (true geographic) geometry. Also, the variations in the ship's speed with time generated variations in the size of the footprint (the area on the seafloor) of each pixel in the along-track direction. These distortions, as well as others discussed below, were corrected so that the sonograph images represent orthorectified and true plan views (assuming a flat seafloor) of acoustic backscatter patterns on the seafloor.

Geometric Corrections

The major sources of geometric distortions in the data are:

1. water-column offset;
2. slant-range geometry;
3. aspect or anamorphic ratio distortion; and
4. changes in the ship's speed.

Each of these major distortions has been eliminated from the data presented in the atlas. Below is a brief explanation of the procedures used to correct for these geometric distortions.

The GLORIA system starts recording data as soon as the transmitted acoustical wave is terminated, and therefore the original images include pixels to both sides of the nadir that contain non-image data. The number of non-image pixels varies along the trackline as a function of the water depth directly beneath the ship. The processing software merges the navigational and bathymetric data with the image data. Included as part of the header information for each image line of sonar data are the date, time, latitude, longitude, and water depth. Knowing the size of each pixel in the across-track direction and the water depth directly beneath the ship, the water-depth offset can be calculated in pixels to predict the location on each side to which the nadir pixel was mapped.

The GLORIA system is a time-based imaging device that collects image data using a near-range depression angle of approximately 90 degrees and a far-range depression angle of 5 to 10 degrees and generates images in slant-range geometry (Somers and others, 1978). The extreme differences in depression angles cause the original image to have major distortions. The program that corrects for water-depth offset also simultaneously corrects for slant-range geometry. Slant-range to ground-range corrections, based on depression angles at both the near-range, or nadir and far-range locations, are applied to the image to properly map each pixel to a ground-range geometry. In this particular correction, the ray path of a sound wave in water is assumed to be a straight line. Another major geometric distortion present in GLORIA images is the aspect or anamorphic ratio that exists between the along- and across-track directions. The sampling interval in the across-track direction for the 30-s pulse- repetition rate generated pixels with an approximate resolution of 45 m. The program that corrects for the water depth and the slant-range distortions generates 50-m-resolution pixels in the across-track direction. However, the resolution in the along- track direction is dependent not only on the 30-s pulse- repetition rate but also on the ship's speed The average resolution in the along-track direction is approximately 125 m, which produces images with an aspect-ratio distortion of about 2.5. This generates a raw image that is distorted or stretched in the along-track direction. Another related source of geometric distortion is introduced by any change that occurs in the ship's speed while it is collecting the image data. The ship's speed is mainly influenced by the direction and strength of current and wind speeds relative to the ship's course, and whether the ship is heading in a straight line or is in a turn. During EEZ-SCAN 84 the speed varied from about 7 to 10 kts, which caused the pixel resolution in the along-track direction to vary from approximate 110 to 140 m. This introduced an "accordion" effect into the geometry of the image in the along-track direction.

The aspect-ratio distortions discussed above were removed by using the latitude and longitude values extracted from the header of each record to compute the distance travelled by the ship every 30 min (unless a turn is detected, in which case the program uses a 10-min interval). Given the distance travelled and the desired pixel size, the number of pixels required for the particular 30-min segment was calculated. To simultaneously correct for the aspect-ratio distortion, a 50-m-resolution pixel size was generated on the output image in the along-track direction. This pixel size was selected so that information in the across-track direction would not have to be omitted. This procedure corrects the image data for aspect-ratio distortions and any distortion introduced by changes in the ship's speed.

Radiometric Corrections

The second major category of processing steps deals with radiometric corrections. These steps change the DN value of a pixel rather than its spatial location, as is the case with geometric corrections. The DN value of a pixel can be changed as a function of its current DN value, as a function of its position, or both. The radiometric corrections used for the GLORIA data in this atlas include four different corrections:

1. a shading correction for the attenuation of the sonar energy in water as a function of range;
2. a power correction for very near nadir because of slow buildup;
3. speckle-noise correction; and
4. removal of striping noise.

The power drop-off problem is relatively severe in sonar images because the acoustic wave is attenuated in water as a function of approximately 1/range to the third power. The method employed to correct this radiometric problem uses a two-pass algorithm. During the first pass through the data, the average DN value is computed for each column of pixels of the digital image in the along-track direction. These values are then normalized by the average of all the column averages (the overall average of the image) to generate correction coefficients for each column. The correction coefficients are then applied to each pixel during the second pass through the data. The coefficients are nonlinear and a function of range; they effectively remove attenuation in the across-track direction. This technique has the characteristics of a spatial filter that removes large horizontal low-frequency patterns that are present because of the radiometric problems introduced by the imaging system (Chavez, 1986). By normalizing to the average of the image rather than to a set DN value, backscatter comparisons can be made between different areas or different images. The correction also allows areas within the image with lower or higher backscatter characteristics to be properly identified and mapped. This was not possible before the correction because the DN values were strongly modified as a function of their range position. Profiles of different areas in the across-track direction can now be used for backscatter comparisons.

The method used to remove speckle noise was to apply a small (2 samples by 2 lines) smoothing filter to the entire image. This is the best approach with GLORIA data because, besides removal of speckle noise, it helps smooth, by pixel duplication, the blocky appearance that the image would have because of the 2.5X digital enlargement that is introduced when correcting for the aspect- ratio distortion.

Another radiometric or noise problem often present in data collected by scanning devices is striping in the scan direction. A combination of high- and low-pass spatial filtering was used to remove the striping. Two separated images were generated from the input data; one composed of the high-frequency components minus the noise frequency, and the other composed of low-frequency components minus the noise. The two resultant images were then digitally added to produce an image similar to the original but without the noise (Chavez and Soderblom, 1974). The filter shapes used to remove the striping noise from the GLORIA data were a 1- line by 71-sample high-pass filter plus a 9-line by 71-sample low-pass filter.

Finally, there is a halo effect that occurs in the along-track direction near the nadir that was removed using a 151-line by 31- sample high-pass filter plus a 151-line by 401-sample low-pass filter.

Digital Mosaicking

After the geometric and radiometric corrections were made the data were digitally mosaicked into 2-degree by 3-degree quadrangles. This procedure entailed taking individual straight line segments of image data and removing the header information at the beginning of each scan line record, tone stretching the data to ensure homogeneity with other line segments in the quadrangle, and running the stretched data through transformation and geometric programs to locate the line segment's position and orient it within the quadrangle space. The line segment was then mosaicked together with other image files into that space. Finally, because there was intentional overlap of data between line segments, the overlapping data of lesser quality were cut and discarded using a digital process called stenciling. This process was repeated multiple times for each quadrangle depending on the number of line segments lying within each quadrangle's coordinates.

Sonar Image Interpretation

The mosaic is a gray-scale image of the acoustic backscatter of the seafloor. White represents the greatest acoustic backscatter and black represents the lowest acoustic return. The darkness or lightness of a feature or an area on the mosaic, therefore, is a function of how much sound is backscattered from the seafloor. Backscattered acoustic energy, in turn, is controlled by the relief of the seafloor (height, abruptness), microtopography and roughness of the seafloor (sediment waves, bedrock irregularities), and physical properties of the upper 10 m of the seafloor (lithified rock and medium-size sand and clay deposits all have different backscatter coefficients, for example). When viewed from the trackline, a positive-relief feature, such as a volcano, usually appears as a bright zone (the facing slope) with a dark zone (the acoustic shadow) behind it. Conversely, a negative-relief feature, such as a canyon, usually appears as a dark zone (the near wall is shadowed) followed by a bright zone (the far wall is facing the sonar beam). In general, exposures of seafloor crust such as seamounts, volcanoes, and oceanic ridges tend to be bright, or at least brighter than the surrounding seafloor, on most images. However, not all bright targets are indicative of seafloor basalt. Bright patterns are also caused by certain sedimentation patterns and by steep sediment-covered slopes facing the sonar beam. Dark patterns also have a variety of causes, of which shadows and certain sediment facies are the most common.

Navigation

Navigation for all the EEZ-SCAN 84 cruises was with Loran C in the hyperbolic mode with independent satellite fixes. The two navigation inputs were recorded on magnetic tape and merged into a new file using only good satellite fixes and correcting the trackline positions between good satellite fixes using the Loran C data. Our experience using the west coast Loran C net is expected to be accurate to within 200 m.

In addition to the at-sea navigation accuracy, the positional accuracy of features in the image mosaics was affected by the digital production of the mosaics. However, we believe that the accuracy of the technique is within the accuracy of the navigation data.

References Cited

Chavez, P.S., 1986, Processing techniques for digital sonar images from GLORIA: Photogrammetric Engineering and Remote Sensing, v. 57, no. 8, p. 1133-1145.

Chavez, P.S., and Soderblom, L.A., 1974, Simple high-speed digital image processing to remove quasi-coherent noise patterns: AMERICAN Society of Photogrammetry Symposium, 41st Annual Meeting, 1974, Wash. D.C., Proceedings. p. 595-600.

EEZ-SCAN Scientific Staff, 1991, Atlas of the U.S. Exclusive Economic Zone, Bering Sea: USGS IMAP-2053, 145 p.

EEZ-SCAN 84 Scientific Staff, 1986, Atlas of the Exclusive Economic Zone, Western Conterminous United States: USGS Miscellaneous Investigations Series I-1792, 152 p., scale 1:500,000.

Sabins, F.F., 1978, Remote sensing, principles and interpretation: San Francisco, Calif., W.H. Freeman, 426 p.

Somers, M.L., Carson, R.M., Revie, J.A., Edge, R.H., Barrow, B.J., and Andrews, A.G., 1978, GLORIA II - an improved long range sidescan sonar, in Proceedings of the Institute of Electrical Engineering on Offshore Instrumentation and Communications, Oceanology International Technical Session J: London, BPS Publications Ltd., p. 16-24.

Note 1 - The reference to sheets in the text refers to the 2- degree by 2-degree quadrangles in the 1986 printed atlas publication. The equivalent of seventeen of those original thirty-six quadrangles or 2 degree sheets have been digitally mosaicked to date and comprise the GLORIA image dataset included in this release.

Note 2 - This is an edited excerpt from a more extensive preface written by David G. Howell for the "Atlas of the Exclusive Economic Zone, Western Conterminous United States."