Abstract
The two flights of the integrated SIR-C/X-SAR radar module aboard the Space Shuttle Endeavour in April and October 1994 opened a new dimension in microwave remote sensing. This third Shuttle Imaging Radar concept (SIR-C) was the first multifrequency and multipolarization SAR system monitoring the earth from space.
In this study the SIR-C/X-SAR data were used to evaluate and verify the tectonics of the Timna Valley in the south of Israel close to the Dead Sea Rift. The geologic setting of the test area is characterized by a magmatic Precambrian massif surrounded by Paleozoic and Mesozoic sediments. The orginal enrichment of the copper and manganese mineralization is mainly associated with the Cambrian sedimentary Timna Fm. while the final and major enrichment was caused by hydrothermal activity along faults and joints related to the late Cenozoic rifting or older rejuvenated faults.
The elaboration of the tectonic inventory was the first goal of
the study presented here. In a first step the three frequencies
were analysed separately and it could be shown that they represent
different information concerning the tectonics. Further it could
be demonstrated that different processing steps like the calculation
of principal components which can only be carried out having multidimensional
SIR-C/X-SAR data, could improve the interpretation of tectonic
structures clearly. In order to compare the results with existing
geological and tectonical maps, the SIR-C/X-SAR data had to be
transformed into the reference geometry of these maps. The data
were geocoded following the rigorous range-Doppler approach. This
geocoding process additionally offers the opportunity of radiometric
correction of the data to reduce the relief induced distorsions
of the backscattering which improves the identification of the
structures in the SAR images. Before the correction the great
scale topography is dominant accentuated by the slant range geometry
of the SAR data. After correction small scale texture phenomena
become dominant and emphasize the interpretation of tectonical
structures being masked by the relief induced distortions of the
signal before.
Introduction
The Timna Valley has been studied intensively since 1947, mainly during the operation of the Timna Copper Mines (Segev et al., 1992). The first activities are reported from ancient King Salomon.
Due to the location close to the Dead Sea Rift and the Tamad fault, the Timna Valley has an intensive tectonic history. In this study the potential of the SIR-C/X-SAR instrument for the detection of structural elements in this test site was evaluated. Additionally, the tectonics of the Timna Valley were to be revised and the gaps of the present structural model suggested for this area had to be fulfilled by applying radar imaging. Several faults and structures which were not registered in the existing geologic maps could be mapped in the SIR-C/X-SAR data and verified on ground.
Due to the slant range projection of Synthetic Aperture Radar (SAR) systems this kind of data has characteristic geometric and radiometric distortions. The distortions strongly depend on the terrain and increase rapidly, especially in mountainous areas like the Timna Valley resulting in extreme changes in local mean intensity. The correction of these phenomena is a strong requirement for the investigations of surface phenomena and the tectonics. Without correction it is not possible to evaluate the radar data on the base of their backscattering due to the strong relief induced distortions, while layover and radar shadow effects produce signals lacking interpretable information (Häfner et al. 1993).
Terrain geocoded images were used throughout to enable comparison with existing geologic maps and to integrate the SIR-C/X-SAR data in geographical information systems with a well-defined reference geometry.
In this study the SIR-C/X-SAR data were geocoded via the rigorous range-Doppler approach and additionally radiometrically corrected on the base of the local incidence angle map after Holecz et al. (1993) derived from a high resolution digital elevation models (DEM). Taking into account a spatial resolution of 25 x 25 m referring to the DTM, resampled to 12.5 x 12.5 m, the relief distortions with size greater than 25 m can be removed. In addition to texture information induced by the small scale relief (<25 m), the surface roughness is the dominant part influencing the variance of the SAR images. The influence of soil moisture can be neglected since the Timna Valley is situated in the arid Negev desert.
Information about the tectonics is on the one hand controlled by the great scale morphology (>25 m). On the other hand the relief induced radiometric distortions are strongly influenced by the topography thereby masking essential information about the tectonics controlled by small scale relief (<25 m). Therefore, for a complete mapping of the structural inventory, radiometrically uncorrected- as well as corrected data have to be used for interpretation (Wever & Frei, 1996).
For an improved reconnaissance of the structures secondary derivative
images, representing ratio- or principal component images were
also used. They describe differences between the responses at
the frequencies or polarizations, and not absolute backscattering
values. Using this image processing technique multidimensional
data are required which are only available by the SIR-C/X-SAR
instrument. From the Timna Valley in the southern part of Israel
SIR-C/X-SAR data at mode 11 with the VV polarized X-band as well
as the like and cross polarized (HH, HV) C- and L-band was available.
Geology
The Timna Valley is situated on the western margin of the Dead Sea Rift, and north of the junction with the Tamad fault, an E-W regional fault running from the Gulf of Suez to the Arava Valley. The Tamad fault is probably a Precambrian structure rejuvenated in the Miocene predating the Dead Sea Rift. The Dead Sea Rift is a strike-slip fault with a left-lateral movement, where the eastern Jordanian block moved about 100 km northward. The suggested sinistral movement took place during the Miocene (60 km) and in the Plio-Pleistocene (40 km). Also extensive uplifting of the new plate margins took place, which has a vertical distance of 1-2 km along the Arava Valley, and increases towards the south up to 3-5 km. Recently the activities along the Dead Sea Rift have been still going on, the last earthquake was dated Nov. 1995.
The internal structure of the Timna Valley is caused by older tectonic movements from the Precambrian to the lower Cretaceous [Beyth & Segev, 1983; Garfunkel, 1981). The main fault directions are WSW-ENE, NW-SE, interpreted as conjugated shearzones to the Dead Sea direction. Beside these normal faults there are reverse faults with parallel joint systems and dike directions. Additionally NNE-SSW directions occur matching with the direction of the Dead Sea Rift. Eyal & Reches (1983) defined two tectonic stress fields, each relative uniform in both space and time. The Syrian Arc stress field has a dominating maximum horizonal compression trending W to WNW, in the Late Cretaceous to Eocene rocks in the folds and plateaus west of the Dead Sea rift. The second field called the Dead Sea stress has an dominating horizontal extension trending E to ENE, in all rocks inside the rift and proximal thereto.
The geologic setting of the test area is characterized by a magmatic
Precambrian massif surrounded by Paleozoic and Mesozoic sediments.
The magmatic core (Har Timna) can be classified into basic/ultrabasic
peridotite and norite, intermediate diorites and monzonites as
well as asidic granites and syenites. The sediments mainly consist
of sandstones, limestones, dolomites and subordinately of silty
shales. Mineralizations occur as beds, veins and nodules of secondary
hydroxides of copper and manganese minerals accompanied by phosphate
lenses with uranium, mainly associated with the Cambrian Timna
Fm.. The enriched source rocks are of sedimentary origin, the
final and major enrichment was caused by hydrothermal activity
along faults and joints of the late Cenozoic rifting or rejuvenated
faults (Beyth, 1987). The indications for hydrothermal epigenetic
activity are veins of manganese, copper, phosphate with uranium
cross cutting the whole stratigraphic section from the Precambrian
to lower Cretaceous mainly along the NNW-SSE and NE-SW directions.
The mineralized veins can also be found along NNW trending fault
lines penetrating sediments of a recent terrace.
Data Processing
High precision terrain geocoding of the image is necessary to
transform the data into the chosen reference geometry and for
the removal of relief induced radiometric distortions using the
local incidence angle. Cartographic as well as sensor specific
transformations were used for parametric geocoding of the SIR-C/X-SAR-data
based on a DEM. Sensor as well as processor characteristics must
be taken into account, and therefore a rigorous range-Doppler
approach (Meier 1989, Meier et al. 1993, Häfner et al. 1993,
1994) is required. For each backscattering element the following
two equations have to be fulfilled:
Rs = slant range fo = carrier frequency
= sensor-, object-position vector fD = Doppler frequency
= sensor-, object-velocity vector
c = velocity of light
If the sensor position or some transformations are not known with sufficient accuracy, the distance-Doppler relationship is derived via ground control points. The accuracy of the parametric geocoding finally depends on the exact determination of these points. In this study the procedure of parametric geocoding was carried out using a DEM digitized from a topographical map (scale: 1 : 50 000) and the exact distance-Doppler relationship was derived from 55 ground control points. The resolution cells were calculated to 12.5x12.5 m, which corresponds closely to the real spatial resolution of 2-look SIR-C/X-SAR data.
The radiometric correction of the data bases on the local incidence
angle map after Holecz et al. (1993). Especially in mountainous
areas like the Timna Valley this parameter varies in a wide range
which dramatically reduces the geophysical interpretation of backscatter
phenomena. The SAR data have to be normalized to a uniform local
incidence angle calculated as follows (Clapp, 1946):
= normalized backscattering coefficient = incidence angle
° = normalized backscattering coefficient i
= local incidence angle
Additionally, principal components of second order and ratio images
were calculated. The first principal component is associated with
the highest eigenvalue and is generally a weighed average of all
input data. It describes an image of the albedo and topography
in the range covered by the remote sensing system (Drury, 1987).
The second principal component essentially presents the differences
between the L-band in HV-polarization and the shorter wavelengths
X-band in VV-polarization. The noise in the data, being the only
completely uncorrelated component, strongly increases in the higher
order components. Due to this fact only the second principal component
is suitable. Concerning the ratio images the selection of frequencies
/ polarizations with low correlation coefficients will lead to
images with high contrast and low noise. Table 1 gives an overview
of the correlation coefficients of the available frequencies and
polarizations of the system. In general, the correlation coefficient
between different polarizations is much higher than between different
frequencies. Hence, in this case the combination of L-HV and X-VV
(Tab. 1) is advisable for producing ratio- or difference images.
Tab. 1: Correlation coefficients of the available frequencies
and polarizations of the SIR-C/
X-SAR system after filtering
Results
The interpretability of the structural inventory can clearly be enhanced by the use of multifrequency SAR-instruments like the SIR-C/X-SAR system. Dependent on the surface conditions relative to the wave-length different structures and directions can be detected in the available frequencies. Figure 2 represents the structural interpretation of the SIR-C/X-SAR data and documents the two postulated stress fields. Several faults and structures which were not registered in the existing geologic maps could be mapped in the different image products of the SIR-C/X-SAR data and verified on ground.
By geocoding processes using the rigorous range-Doppler approach, the data and secondarily derived products can be transformed into the chosen reference geometry. This process is necessary for the exact localization of the interpreted structures and for the comparison with existing maps. Additionally the availability of a high resolution DEM offers the opportunity to correct the data radiometrically in order to reduce the relief induced distortions in the images. The process of radiometric correction of the data strongly reduces the influence of the topography on the radar backscatter and enables the detection of structures which were masked before the correction by the dominant information of this topography.
Different image processing procedures are also able to improve the interpretability of structures. In the present study principal components of second order as well as ratio images were calculated. The secondary derivative images describe differences between the responses at the frequencies or polarizations used, and not absolute backscattering values.
On the base of these image processing products similar effects can be observed in the radiometrically corrected images. The information about topography is strongly reduced and the information about the surface becomes dominant. This effect is based on the fact that the short X-band can not analyse the different degrees of roughness because in the X-band nearly all surfaces appear rough. Due to this fact the surface information is represented very uniform in the images while the influence of topography is dominant on the backscattering. By the combination of different frequencies presented here, this effect can be used for the elimination of the dominant influence of the topography and subsequently in the accentuation of the structural inventory.
The SIR-C/X-SAR data show a very distinct differentiation of the
lithological units. Especially the cristallin Har Timna complex
can be clearly distinguished from the surrounding sediments. The
information content of the SAR data depending on the surface conditions
offers a new dimension in the differentiation of the lithology.
The main structural features can be seen more or less clearly
in all frequencies and polarizations of the SAR data. In the X-band
and similar in the C-band the alluvial areas and very fine grained
sedimentary surfaces can be well differentiated. The structural
information in X-band images is less than in C- and L-Band. All
together the C-Band gives a similar differentiation of the rock
units as the X-Band but the structural information are better
documented. The L-Band offers the greatest potential in mapping
the tectonics and lithology in this area, e.g. this frequency
is able to separate the magmatic Har Timna complex from the surrounding
sediments.
a) b) c)
Fig. 1: Rose diagrams the interprated structures representing the Syrian Arc- (SAS) and the Dead Sea stress field (DSS): a) number of structures
b) total length of structures
c) average length of structures
Conclusion
The interpretation of the structural inventory can clearly be enhanced by the availability of several frequencies and polarizations like the SIR-C/X-SAR instrument | |
Terrain geocoding is necessary for the integration of the data into geographic information systems with a given reference geometry | |
In mountanous areas like the Timna Valley the information about the topograpgy is dominant and masks the information about the tectonics and surface phenomena | |
To reduce the relief induced distortions the SIR-C/X-SAR data have to be corrected radiometrically | |
All structures and faults existing in the geologic map of the Timna Valley could be verified in the SIR-C/X-SAR data | |
Additional structures could be identified in the SIR-C/X-SAR data and verified via ground truth | |
A wide range of roughness can be detected by the availability of different frequencies | |
A discrimination of geologic units is possible by use of the backscattering and texture phenomena of several frequencies and polarizations | |
L-band is the most favourable frequency for geologic applications. X- and C-band are highly correlated, there is only a need for one shorter frequency in addition to the L-band | |
Multifrequency data are more useful than multipolarization data with the exeption of quad-pol data which are required for inversion algorithms | |
There is a need for high resolution DEMs for geometric and radiometric correction of the data especially in mountenous areas |
Acknowledgments
The funding support for this research was provided by the German
Space Agency DARA GmbH in the framework of the SIR-C/X-SAR missions.
The author would also like to thank all NASA/JPL colleagues for
good cooperation and A. Altendorf for digitizing various maps.
References
BEYTH M., 1987, Mineralizations related to Rift systems: Examples
from the Gulf of Suez and the Dead Sea Rift. Tectonophysics, pp.
141, 191-197.
BEYTH M. & SEGEV A., 1983, Lower Cretaceous basaltic plug
in the Timna Valley. Israel Journal of Earth Sciences, 32, pp.
165-166.
Clapp R.E., 1946, A Theoretical and Experimental Study of Radar
Ground Return, Report No. 1024, Radiation Laboratory, Massachusetts
Institute of Technology.
DRURY S.A., 1987, Image Interpretation in Geology,
Allen and Unwin (Publishers) Ltd., London, UK.
EYAL Y. & RECHES Z., 1983, Tectonic Analysis of the Dead Sea
Rift Region since the Late-Cretaceous Based on Mesostructures,
Tectonics, Vol. 2, No. 2, pp. 167-185.
GARFUNKEL Z., 1981, Internal Structure of the Dead Sea Leaky Transform
in Relation to Plate Kinematics. Tectonophysics, 80, pp. 81-108.
H&AUML;FNER H., HOLECZ F., MEIER E., N&UUML;ESCH
D. & PIESBERGER J., 1994, Geometrische und Radiometrische
Vorverarbeitung von SAR-Aufnahmen für Geographische Anwendungen,
Zeitschrift für Photogrammetrie und Fernerkundung (ZPF),
Bd. 4/94,
pp. 123-128.
H&AUML;FNER H., HOLECZ F., MEIER E., N&UUML;ESCH
D. & PIESBERGER J., 1993, Capabilities and Limitations of
ERS-1 SAR Data for Snowcover Determination in Mountainous Regions,
Proceedings of 2nd ERS-1 Symposium., Hamburg, Germany, on 11-14
Oct. 1993, ESA SP-361, pp. 971-976.
HOLECZ F., MEIER E. & N&UUML;ESCH D., 1993, Postprocessing
of Relief Induced Radiometric Distored Spaceborne SAR Imagery.
In SAR Geocoding: Data and Systems, edited by G. Schreier, (Karlsruhe:
Wichmann Verlag) pp. 299-352.
MEIER E, 1989, Geometrische Korrektur von Bildern orbitgestützter
SAR-System. Remote Sensing Series, 15, Geographisches Institut,
Universität Zürich.
MEIER E., FREI U. & N&UUML;ESCH D., 1993, Precise Terrain
Corrected Geocoded Images. In SAR Geocoding: Data and Systems,
edited by G. Schreier, Karlsruhe: Wichmann Verlag,
pp. 173-186.
A. SEGEV, M. BEYTH & M. BAR-MATTHEWS, 1992, The Geology of
The Timna Valley with Emphasis on Copper and Manganese Mineralization-Updating
and Correlation with the Eastern Margins of the Dead Sea Rift.
Geological Survey of Israel, Report G.S.I/14/92.
WEVER T. and FREI M.; 1996, Integrated Use of Optical and Microwave
Data for Geologic Applications in Israel, Proceedings. of 11th
Thematic Conference on Applied Geologic Remote Sening, Vol. I,
Las Vegas, Nevada, on 27-29 Feb 1996, pp. 355-364.