Prof. Pasquale Murino
Instituto U. Nobile
Piazzale V. Tecchio, 80
80125 Napoli, Italy
Co-Investigators:
Lucio Castellano
Luciano Russo
Mario Ferri
SIR-C/X-SAR data analyses in Campania Test Site


OBJECTIVES

The overall objective of this investigation is the evaluation of SIR-C/X-SAR data collected over the Campania test site in geological analyses.

Specific objectives include the following:

To determine effective procedures to process SAR data to extract information useful in geological analyses, in particular lithology mapping;
To evaluate the relative utility of multifrequency quad-pol SAR data in morpho-structural analyses.

In particular a processing algorithm based on a covariance matrix approach, which permits us to take advantage of the polarization agility of dual-channel multifrequency SAR, has been developed. Unfortunately only one quad-pol single-look image was available so a great amount of work was carried out on dual-pol data which do not allow a complete electromagnetic characterization of targets.


PROGRESS

The Picentini Mountains

The Picentini mountains have been selected because of their very articulated morphology and structure. The area is characterized by many geological formations and structural typologies (faults, overthrusts, klippens, and tectonic windows). The geological analysis of SAR images has been carried out locally (the selected area does not exceed 400 km2) so, for example, structural trends have not been considered. The study has been focused on both geological details (e.g., the analysis of the limit between limestone and dolomite of the Stella mountain) and structural lineament identification.

The Phlegrean Fields

The Phlaegrean Fields area offers the opportunity to study a morphology characterized by several crateric rims having different ages and histories. For this area SIR-C/X-SAR data have been integrated by ERS-1 and JERS-1 data imagery (lower incidence angles and different looking directions).

The Somma-Vesuvius complex

Vesuvius a very interesting test site, which was observed during a number of spaceborne and airborne remote sensing missions. For that reason we are able to compare the results obtained using SIR-C/X-SAR data with those obtained using other SAR imagery acquired by ERS-1, JERS-1, and AIRSAR missions.

Available SIR-C/X-SAR data

The following data takes have been used:

Methodology

Color and B/W images have been generated from both dual-pol (11x) and quad-pol (16x) L- and C-band data takes and from fixed polarization X-band acquisitions respectively.

Dual-pol data color images have been obtained by combining the HH, HV and SPAN information on three (Red, Green and Blue) 8-bit color planes. That combination is especially helpful in the L-band where the cross-polarized return is mainly contributed by the high depolarization effect of dense vegetation. In such a way the green appearance of HV return facilitates the geologist to detect some variations in the lithology.

A more sophisticated processing has been developed and applied to quad-pol data takes. First and second order statistics of co-polarized and cross-polarized measurements can be easily extracted from scattering matrix or Stokes scattering operator and used to build a covariance matrix. If single-look data are available these statistics can be computed assuming both a temporal average of values (single look) or assuming a spatial average of values (multi look), the latter being less noisy. If HV=VH, the covariance matrix (now 3x3) can be decomposed, according to Cloude, Van Zyl, and Castellano, in the sum of three scattering mechanisms (the product of each eigenvector for its transpose with opposite phase) weighted by the respective eigenvalues. Even though the basis of such a decomposition is rotating in the target space for different targets (when the eigenvalues change the eigenvectors rotate in a plane), it can be shown that, if at a first approximation we assume a fixed basis of the decomposition, the eigenvalues will differentiate among targets with a different scattering behavior. In addition three eigenvalues (normalized or not) can be represented on three color planes for visual photo-interpretation.

Geology

The SIR-C/X-SAR images, generated as above mentioned, have been analyzed by using photo-interpretation techniques and integrated with images of comparable spatial resolution acquired by other sensors (ERS-1 ascending, descending and roll-tilt mode, JERS-1, AIRSAR) . The results have been compared and/or integrated with geological sheets and when possible with aerial photos, geological sheets and field checks.

Brightness, color, texture, shading and pattern information is used for discriminating among geological features and lithologies. Foreshortening/layover effects, in the case of the Picentini mounts and of the Phlaegrean Fields, have been verified by comparing SAR images and geological cartographic profiles.


SIGNIFICANT RESULTS

Data Processing

Tables 1 (L-band) and 2 (C-band) show some experimental results obtained by decomposing the normalized covariance matrices computed over 9 sample windows (11x11 pixels). Each target (Vesuvius cone, 1944 lava flow, pine forest and sea) has been selected at most twice in order to check about the effects of slight different looking directions and/or incidence angles. Sea samples have been extracted from AIRSAR images since on these data the incidence angle ranges from 23° to about 58°.

For each sample have been computed the ratio HH/VV (second column), the ratio VV/HH (third column), the third eigenvalue (fourth column), the module (fifth column) and phase (sixth column) of correlation coefficient between co-polarized channels, the first (seventh column) and second (eighth column) normalized eigenvalues. The target "entropy", here defined as the sum of the normalized eigenvalues, is reported in the last column.

As can be noted, if the target is constituted by sand/ash as in the case of the Vesuvius cone, the first eigenvalue represents more than the 70% and 80% of target entropy for L-band and C-band respectively. These values suggest that, also for L-band scattering, the surface effect dominates the volumetric effect. A comparison made with P-band AIRSAR data shows a probably more consistent volumetric scattering at these frequencies (a value of 60% is typical on the Vesuvius cone). The analysis of [Lambda]3 (h in the Tables) confirms these remarks; the L-band (5-6 %) and C-band (12-13%) values show clearly, for the latter, a higher depolarization effect suggesting that the pyroclastic products of the Vesuvius cone are rougher at a centimeter scale than at a decimeter scale.

For concerns of the 1944 lava flows, the depolarization effect increases sensibly with respect to the cone in the L-band as well as the C-band. The first eigenvalue is now only 50-55% of target entropy while the second and third eigenvalues reach about the 20-25%. It may be concluded that the roughness at a centimeter scale is comparable to the roughness at a decimeter scale.

Pine Forest data confirm that trend since the entropy increases and [Lambda]1 is still decreasing (40-45% in L-band and 45-50% in C-band).

This decomposition of covariance matrix can be used also in the analysis of sea scattering. Some investigations which have been carried out show that at low incidence angles (23°) the specular reflections largely dominate and target entropy is very low especially for L-band where the sea acts like a deterministic target and [Lambda]1 constitutes about the 98% of total entropy. As the incidence angle increases the entropy increases as well and the VV return becomes also up to three times the HH return (L-band at 58°).

Geology

SIR-C/X-SAR images, because of the high incidence angle (about 50°) clearly show the differences between the western side of the Picentini mounts and the Avella mountains. The first ones are characterized by a complex overlapping of structures while the second ones are simply crossed by a normal fault.

The following classes have been easily discriminated:

1) Alluvium, pyroclastic products (ignimbrite campana), colluvium, flisch, and "melange" of the sicilide unity;

2) Carbonatic-dolomitic rocks of the "Campano-Lucana" platform;

3) Bedded sands and conglomerate of the Altavilla unit.

At low incidence angles (i.e., about 20-25°) discrimination between conglomerates and sands of the Altavilla unit, and the carbonatic formations is not possible. On the investigated area, the differences between calcareous-dolomitic and alluvium/flisch terrains are simply detected while the discrimination between flisch and alluvium is a difficult task. On the Mai mountains (a part of Picentini mountains), some Trias-Giura limits seems to be actually tectonic and not stratigraphic as pointed out in the literature.

In X-SAR imagery, the morphology of the Mounts Lattari stands out clearly as well as the boundaries among carbonatic rocks, colluvium and alluvial fan and tuff or eruptive volcanic products less or more pedogenized. The alluvial fan presents a very thin texture while the Vesuvius 1944 lava flows are less evident.

In the Somma-Vesuvius area, the limit between the products of pre-Mount Somma and the undiversified pyroclastic products has been recognized by using differences of intensity, color and texture which are due to changes of morphology, vegetation cover and drainage pattern. Good discrimination is achieved among the characteristics of the recent lavas, principal quarries, Mount Somma caldera rim, the drainage pattern and the pyroclastic products of the Vesuvius cone.

A detailed analysis of the Phlegrean Fields area requires the integration of data acquired with different incidence angles and from different looking directions in order to achieve a good detection of the crateric morphology and, in some cases, for the detection of zones affected by volcano-tectonic collapses. This is the case of Mount Gauro which is characterized by eastern and western crateric slopes slightly concave toward the exterior, with the sides higher than the center. The east and west slopes of the Gauro are related, according to some authors, to volcano-tectonic collapses following explosive events that occurred about 5000 years ago. SIR-C/X-SAR images show a more realistic morphology and the shadow distribution allows a good identification of the heights of this crateric edge.

HH polarization allows a good characterization, in terms of tone and texture typologies, of terrains and anthropic constructions while in the HV images, the texture is more homogeneous.


FUTURE PLANS
Research activities are still in progress. New sub-areas included in the Campania test-site will be considered, especially for what concerns INSAR and DINSAR data processing. For example, 16 corner reflectors have been already installed in the area of Sannio-Matese and new installations are planned in the next months.

Polarimetry

Further investigations both theoretical and experimental will be carried out on quad-pol data.

In particular the following items are under study:

Interferometry

Since interferometric SLC L-band and C-band data were not yet available for Vesuvius or Etna, the work has been conducted on X-SAR SSC products which include the Vesuvius and a part of Matese area. Differential INSAR processing is under progress and the results will be compared with those extracted from ERS-1 SLC data.


PUBLICATIONS
[1] Castellano, L., A. Siciliano, and P. Murino, "Characterization of Natural Targets in Volcanic Areas Using SAR Polarimetric Covariance Matrix," 45th Congress of the International Astronautical Federation , Jerusalem, Israel, 9/14 October 1994.

[2] Murino, P., M. Ferri, L. Castellano, L. Russo, and A. Siciliano, "Using SIR-C/X-SAR Data in the Analysis of Volcanic Areas. The Campania test site (Southern Italy)," IGARSS '95 , Firenze, Italy.

[3] Castellano, L., "Nuovi Metodi di Analisi di Dati SAR Polarimetrici Ripresi da Piattaforme Aerospaziali," Ph.D. Dissertation, National Libraries of Rome and Florence (Italy), 1994.

[4] Castellano, L., P. Murino, and A. Siciliano, "Characterization of Natural Targets Using Multiparametric Spaceborne-SAR imagery," To be presented at AGARD/NATO Meeting on Remote Sensing: A Valuable Source of Information , Toulouse, France, 22-25 April 1996.

[5] Castellano, L., and P. Murino, "On the Usefulness of Interferometric SIR-C/X-SAR Data for Geological Investigations in Campania Test Site," To be presented at 47th Congress of the International Astronautical Federation , Beijing, China, October 7-11, 1996.

L-band 1/z z h (deg)


Cone 1 1.0003 0.9997 0.1210 0.8654 -2.5208 1.8655 0.1345 2.1210
Cone 2 1.0405 0.9611 0.1764 0.8119 -4.8326 1.8136 0.1879 2.1780
44' Lava 1 0.9608 1.0408 0.5562 0.3886 -7.2830 1.3914 0.6102 2.5578
44' Lava 2 0.9775 1.0230 0.5736 0.4111 -7.4226 1.4120 0.5885 2.5741
Forest 1 1.0956 0.9127 0.7312 0.2185 3.6549 1.2410 0.7673 2.7395
Forest 2 1.1336 0.8821 0.6097 0.2737 -1.3037 1.3092 0.7066 2.6254
Sea (23) 0.9580 1.0439 0.0195 0.9613 -0.8334 1.9632 0.0386 2.0214
Sea (45) 0.4371 2.2879 0.0746 0.8739 0.8260 2.6363 0.0897 2.7996
Sea (58) 0.3176 3.1483 0.1835 0.7040 7.0143 3.3137 0.1522 3.6494

Tab. 1 - Decomposition of the covariance matrix (9 sample windows 11x11 pixels), L-band results.


C-Band 1/z z h
(deg)


Cone 1 1.1558 0.8652 0.2873 0.6742 3.0451 1.7001 0.3208 2.3083
Cone 2 1.1588 0.8630 0.3033 0.6022 3.9899 1.6310 0.3908 2.3251
44' Lava 1 1.0277 0.9731 0.4983 0.4785 -5.0762 1.4796 0.5211 2.4991
44' Lava 2 0.9644 1.0369 0.4388 0.5211 -7.7925 1.5231 0.4782 2.4401
Forest 1 1.1808 0.8469 0.6669 0.2561 7.0226 1.3195 0.7081 2.6946
Forest 2 1.1566 0.8646 0.7099 0.2500 -8.8327 1.3001 0.7211 2.6002
Sea (23) 0.9222 1.0843 0.0070 0.8966 -13.3407 1.9035 0.1030 2.0135
Sea (45) 0.4770 2.0996 0.2048 0.7164 2.1040 2.3695 0.4141 2.7814
Sea (58) 0.5555 1.8003 0.5000 0.2681 0.7945 1.8556 0.5002 2.8558

Tab. 2 - Decomposition of the covariance matrix (9 sample windows 11x11 pixels), C-band results.

Table of Contents

Converted to HTML by Alvin Wong, al.wong@jpl.nasa.gov

The Jet Propulsion Laboratory
4800 Oak Grove Drive
Pasadena, Cailfornia 91109