Assessing pre- and post-deformation in the southern Arava ... · Assessing pre- and...

www.elsevier.com/locate/rse

Remote Sensing of Environment 86 (2003) 141–149

Assessing pre- and post-deformation in the southern Arava Valley segment

of the Dead Sea Transform, Israel by differential interferometry

Francesco Sartia,*,1, Yaacov Arkinb, Jean Chorowiczc, Arnon Karnielid, Teresa Cunhae

aESA/ESRIN, Earth Observation Program, V. Galileo Galilei, Casella Postale 64, Frascati 00044, ItalybGeological Survey of Israel, Jerusalem, Israel

cLaboratoire de Tectonique, Case 129, Universite Paris 6, FrancedThe Remote Sensing Laboratory, Jacob Blaustein Institute for Desert Research, Ben Gurion University of the Negev, Beersheba, Israel

eGeology Department, Instituto Geologico e Mineiro, Alfragide, Portugal

Received 7 October 2002; received in revised form 4 March 2003; accepted 8 March 2003

Abstract

Differential radar interferometry, using archived ERS data over the region of the Dead Sea Transform, allows to detect ground movement

(subsidence or uplift) in playas within the southern Arava Valley segment of the Dead Sea Rift. These measurements are consistent with a

mean displacement rate of about 0.4 cm/month, in the direction of the radar beam, during the 8-month period preceding the Nuweiba

earthquake of 22 November 1995. In the 3 years following the earthquake, the measured rate was smaller by a factor of 10. These movements

are not related to salt diapirs or water pumping activities in the area. The exact location, along faults, suggests a possible correlation with pre-

seismic and post-seismic fault deformation.

A simple fault model consistent with the observed phenomena associates the observed subsidence/uplift to right and left stepping en-

echelon fault patterns related to inter-seismic tensional accumulation along the faults. Further observations are necessary on this site and

similar fault areas to corroborate the correlation between seismic activity and the observed phenomena. Monitoring of these sites should

continue with differential Global positioning system (GPS) measurements and radar interferometric analysis using Envisat and Radarsat as

well as archived data (including J-ERS).

D 2003 Elsevier Science Inc. All rights reserved.

Keywords: Interferometry; En-echelon fault; Dead Sea Transform; Dead Sea Rift; Digital elevation model; Global positioning system; Hue intensity saturation

1. Introduction

Synthetic Aperture Radar differential interferometry is a

well-established technique allowing the precise measurement

of ground displacement down to a few millimeters over large

areas with a good spatial resolution. This technique is based

on the measurement of phase variation between two succes-

sive radar acquisitions. Topographic effects are corrected

using a digital elevation model that may be produced from

a dedicated interferogram. The main general limitations of

this technique are the loss of fringe clarity due to a poor

surface preservation (coherency loss) and the bias introduced

0034-4257/03/$ - see front matter D 2003 Elsevier Science Inc. All rights reserv

doi:10.1016/S0034-4257(03)00066-X

* Corresponding author. Tel.: +39-6-91480-409; fax: +39-6-94180-

552.

E-mail address: [email protected] (Y. Sarti).1 Work performed while seconded at CNES, Radar System Department,

Toulouse, France.

by atmospheric artifacts (Fruneau & Sarti, 2000; Massonnet

& Feigl, 1998). Classical examples of application of this

technique to ground displacement mapping can be found in

the literature (Feigl et al., 2002; Massonnet et al., 1993, 1996;

Peltzer et al., 2000). The Dead Sea Transform corresponds to

an active, left lateral, fault system (Fig. 1) where several pull-

apart mechanisms exist, for example the Dead Sea and Sea of

Galilee (Ben-Menahem, 1981; Garfunkel, 1981; Garfunkel,

Freund, & Zak, 1981). Chorowicz and Deffontaines (1993)

referred to transform faults on a pull-apart model in the Rhine

Graben in a similar situation. Khair, Karakaisis, and Papadi-

mitriou (2000) discussed the seismic zonation of the Dead

Sea Transform; each study contributing to the overall under-

standing of the phenomenon.

The present study, using ERS radar interferometry, ana-

lyzes the deformation of a well-limited region (Arava Valley

segment of the Dead Sea Transform) prior to and following

the Nuweiba earthquake of 22November 1995. The epicenter

ed.

Fig. 1. Dead Sea Transform fault system, schematic tectonic setting and location of study area (compiled from several authors, see references).

F. Sarti et al. / Remote Sensing of Environment 86 (2003) 141–149142

was 80 km south of Eilat–Aqaba, 3 km off the Egyptian

shoreline of the Gulf. The earthquake measured 6.2 on the

Richter scale and was followed by 2000 aftershocks; the most

severe occurred on 24 and 25 November measuring 5.5 and

4.8, respectively (Saffarini, 2000). Baer, Sandwell, Williams,

and Bock (1999) estimated a moment magnitudeMw=7.1 for

this earthquake and produced a co-seismic model with an

epicenter located in the Gulf of Eilat close to Nuweiba.

Although no Global positioning system (GPS) data were

available, the model showed that the largest co-seismic

ground displacement was in the region of Nuweiba and no

fault displacement north of Eilat–Aqaba was recognized.

Even though co-seismic and post-seismic deformation was

measured by our research, this paper focuses on deformation

that took place along the fault shortly before the earthquake

(which might be associated to precursor motions).

In the attempt to measure inter-seismic movements along

the Dead Sea Transform, the present interferometric anal-

ysis was extended to cover the region between the Dead

Sea and Red Sea during the time interval 1995–1999, thus

covering an 8-month pre-seismic period and a 3-year post-

seismic period. Subsiding and uplifting areas, within playas

F. Sarti et al. / Remote Sensing of Environment 86 (2003) 141–149 143

aligned along the Dead Sea Transform, have been recog-

nized and studied (Fig. 2). Uplifting and tilting is recog-

nized in the Yotvata playa (Arkin et al., 2000). Topographic

inversion in the Evrona Playa has also been shown to occur

(Amit, Zilberman, Porat, & Enzel, 1999). The former

phenomena are visible on several interferograms (before

and after the seismic event) within a very flat topography of

relative elevation of around 15 m. The low sensitivity to the

relief and the stability of these locations, on several

interferograms, exclude effects other than ground displace-

ment.

Interferometric coherence in the Arava remains very high

for long periods due to the particular land cover (sparse

vegetation and arid soil) and climate conditions (Arkin,

Ichoku, & Karnieli, 1999). In this situation, the precise

Fig. 2. Interferometric fringe location over the Evrona and Yotvata Playas (Arava V

ERS1 21874 (20/09/95)–ERS1 22876 (29/11/95). (c) AQ8: ERS1 22876 (29/11

western and eastern sides of each interferogram, other fringe patterns (A, B, C, D)

traced across the fault line in this figure and Fig. 4; with (d– f) being the Hue Inte

respectively.

measurement of slow inter-seismic movement requires par-

ticular attention regarding the potential presence of atmos-

pheric artifacts.

2. Acquisition and analysis

The selected ERS data set for the Eilat–Aqaba region

(Frame 585/track 343) is shown in Fig. 3a, with a portion

centered on the Dead Sea Rift. Radar intensity images of the

area are shown in the background of Fig. 2d–f (gray level).

The available Digital Elevation Model was D-TOPO-0,

sampled over 100 m with a vertical accuracy estimated at

100–200 m. Several interferometric pairs were formed,

spanning the years 1995–2000, based on the available

alley). (a) AQ7: ERS1 19369 (29/03/95)–ERS1 21874 (20/09/95). (b) AQ6:

/95)–ERS2 19736 (28/01/99). Apart from relief-correlated effects on the

are visible, aligned along the main fault. Profiles 1–2, 3–4,. . .,13–14 are

nsity Saturation images corresponding to interferograms AQ7, AQ6, AQ8,

Fig. 3. (a) Coverage of the selected data set. In bold gray: portion centered on the Dead Sea Rift and on the Arava Valley. (b) Retained interferometric pairs over

the playas: AQ7: ERS1 19369 (29/03/95)–ERS1 21874 (20/09/95). HA=2500 m, time interval: 5.8 months. AQ6: ERS1 21874 (20/09/95)–ERS1 22876 (29/

11/95). HA=350 m, time interval: 2 months. AQ8: ERS1 22876 (29/11/95)–ERS2 19736 (28/01/99). HA=450 m, time interval: 38 months.


ERS-1, -2 archives and on the orbit data. Three interfero-

metric combinations (AQ6, AQ7, AQ8) were retained (Fig.

3b) for the detection and discrimination of pre-seismic and

post-seismic displacement, as follows.

2.1. Interferogram AQ7: ERS1 19369 (29/03/95)–ERS1

21874 (20/09/95) (Fig. 2a)

This interferogram is computed for the 6-month time

interval (T0—8 months, T0—2 months) where T0 is the date

of the Nuweiba Earthquake (22 November 1995). Two types

of fringes are distinguished: relief-correlated and fringes on

flat terrain. The former, situated on both sides (eastern,

western side) of the Dead Sea Rift, are attributed to a

tropospheric effect (topographic residuals are excluded

because of the high altitude of ambiguity, 2500 m, that

would result in less than 0.1 fringes even with topographic

errors of 200 m). These fringes are common to the two

interferograms AQ7 and AQ6 (Fig. 2b), with an inverted

sign, indicating that they originate from the atmosphere of

the common date 20 September 1995 (Massonnet & Feigl,

1995).

Instead, an atmospheric origin is excluded for the fringe

patterns common to the three interferograms, AQ6, AQ7

and AQ8, aligned along the Dead Sea Rift: several ‘‘anoma-

lies’’ A, B, C, and D, not correlated with topography, can be

detected along the main fault. These fringes are located on a

very flat area (altitude variations of less than 15m) and they

cannot be attributed to topographic residuals (the local

relative vertical accuracy of the DEM in this flat area is of

the order of 20–30 m, with an altitude of ambiguity 100

times larger), nor to standard atmospheric artifacts (no

correlation with topography), nor to variable atmospheric

artifacts (because of the stable location on several different

interferograms with no common dates). We can therefore

exclude effects other than ground displacement.

Their shape, size and stable location on the playa are

superimposed on the radar intensity images in the Hue

intensity saturation representations of Fig. 2d–f.

These anomalies therefore correspond to ground dis-

placements. A phase sign analysis reveals that the patterns

in anomalies B and D correspond to a path shortening

towards the satellite pass (such as uplift), whereas those in

A and C correspond to a path increase (such as subsi-

dence).

Fig. 4b,d shows a zoom of these areas on the interfero-

gram. Their respective location on the radar intensity image is

shown in Fig. 4a,c.


22876 (29/11/95) (Fig. 2b)

Fringes similar to those in the previous interferogram are

detected, as well as atmospheric artifacts appearing on the

relief. Fringes over the anomalies A and B, measured

numerically, are two to three times smaller here than on

the previous interferogram AQ7 (same sign) and fringes

over anomaly D, although smaller, are still visible. This

interferogram is computed over a 2-month period (T0—2

months, T0+7 days). Though this interval also contains the

seismic event, a linear deformation model derived for this

time interval is consistent with the amount of deformation

measured by AQ7. Therefore, these fringes are not attrib-

uted to a co-seismic effect but are supposed to originate

from the same deformation mechanism observed in the

previous interferograms AQ7.

Fig. 4. Radar intensity and interferometric phase over the playas. Zoom of interferogram AQ7: ERS1 19369 (29/03/95)–ERS1 21874 (20/09/95) over the

playas. (b), (d) show subsidence areas called A, B, C, D (compare to Fig. 2) with their respective location on the radar intensity image (a), (c).



19736 (28/01/99) (Fig. 2c)

This interferogram is computed over a 3-year time inter-

val. Despite a partial coherency loss along the fault, due to

the long, post-seismic, time interval, a similar phase trend

can be detected along the fault as in the previous interfero-

grams (a path increase in C and a path shortening in D).

A loss of interferometric coherency along the fault and in

the playas was measured; the loss increasing with the time

between acquisitions. The Hue intensity saturation compo-

sition in Fig. 2d–f is obtained from the ERS amplitude

images and the phase interferograms. Atmospheric effects

were masked by filtering out fringes above the areas of

relief. Phase values corresponding to a very low coherency

were masked, too.

The same type of analysis, performed on the Dead Sea

region 150 km north of Eilat–Aqaba (track 78, frame 2979,

Fig. 3a), for the period 1993–1999, does not show any

ground displacement along the fault.

3. Interpretation

Displacements measured by interferometry can be quanti-

fied by computing phase profiles along segments traced

across the fault line (Figs. 2 and 5). Ground displacements

on the interferogram AQ7 are two to three times larger than

Fig. 5. Unwrapped phase profiles along segments 1–2 (region A), 3–4 and

5–6 (region B), 7–8 (region C), 9–10, 11–12 and 13–14 (region D),

traced across the fault line of Fig. 2a, for AQ7 (a), AQ6 (b), AQ8 (c).

Horizontal axis expressed in hundreds of meters. Vertical axis expressed in

mm (variation in satellite-to-ground distance). The regions A and C

correspond to a satellite-to-ground path increase, whereas B and D to a path

shortening. AQ7 shows a mean satellite-to-ground distance increase of

about 18 and 24 mm in the A (1–2) and C (7–8) areas and a mean decrease

of about 8 and 25 mm in the B (3–4) and D (11–12) areas. AQ6 shows 6,

4, 0 and 13 mm, whereas AQ8 shows 9, 4, 15 and 12 mm, respectively, on

the same areas A, B, C, D.

Fig. 6. (a) Tectonic model of playa area, e.g. Evrona


on AQ6. This is consistent with a linear deformation rate of

0.4 cm/month in the 8 months preceding the earthquake.

Though with a lower coherency (reliability), phase profiles

for AQ8 in the same areas and for the 3 years following the

seismic event show (assuming a linear model) a much lower

rate, close to 0.04 cm/month.

Thus, these interferograms are consistent with the

hypothesis that the rate of ground displacement along the

fault in the Arava Valley is about 10 times larger in the

months preceding a seismic event than after. Clearly, a

larger number of measurements are needed in order to

confirm this hypothesis and to derive a more precise model.

With regard to the direction of ground displacement, the

displacement measured by interferometry is one-dimen-

sional in the satellite-to-ground direction. This direction is

expressed in the local reference system by a vector called

sensitivity vector (Massonnet, Briole, & Arnaud, 1995) equal

to N, 0.01, E, 0.3 and S, 0.9 in the case of ERS at the latitude

of Aqaba. Thus, the measured displacement is the projection

of the real displacement vector onto the sensitivity vector. As

a result, a North–South displacement along the Dead Sea

Transform is barely observed by interferometry and one

fringe (2.8 cm in the radar direction) can correspond to

3.08 cm of vertical displacement, or to 9.24 cm of East–West

displacement, or to a combination of the two components. In

the following geological model (Fig. 6), the observed dis-

placement is interpreted as being vertical (subsidence/uplift).

Playa. (b) Tectonic sketch of the Arava Valley.

Fig. 7. Geological map of the Arava Valley: location of the Evrona and Yotvata Playas.



4. Geological model

In a segmented rift valley floor such as the Dead Sea Rift,

transcurrent movement at depth leading to shearing in

specific areas, at or near the surface, form discontinuous

zones where the main fault line is not observed. These zones

correspond, in places, to continental and lagoonal playas as

seen in the Arava Valley (Amiel & Friedman, 1971), as

shown in Fig. 7. These playas are made up of unconsoli-

dated fine clastics of clay and silt, sand, some conglomerate

and evaporitic deposits formed at the surface following

seasonal flooding (Arkin et al., 2000).

During strike/slip movements these zones act as shear

boxes developing en-echelon faults at the surface. These

faults are the surface expression of movement at depth in the

form of small-throw subsurface swarms that connect in

depth with the main fault forming positive or negative

flower structures. This configuration is shown in Fig. 6a

that represents a fault model consistent with the above

interferometric measurements. The playas correspond to

the subsiding areas that have been deformed by subsequent

movements such as the Nuweiba earthquake. According to

this geological model, the regions A and B of Figs. 2 and 4

are modeled as subsiding areas, whereas C and D corre-

spond to uplift. These movements appear to be restricted to

the playas and the deformation is on the whole more ductile

than brittle. Both tensional and compressional areas exist

where dewatering effects seem to induce vertical move-

ments sufficient to be observed by interferometry. This

configuration may provide an insight into the overall tec-

tonic pattern of the Arava Valley as seen in Fig. 6b (Bartov,

Avni, Calvo, & Freislander, 2000).

5. Conclusions and suggestions

Pre- and post-seismic ground displacement is recognized

within the playa areas, aligned along the Dead Sea Trans-

form in the Arava Valley. Unambiguous detection by differ-

ential interferometry with ERS data gives a displacement

rate that is 10 times higher before the Nuweiba earthquake

(0.4 cm/month) than that occurring after the earthquake

(0.04 cm/month). Furthermore, there is no evidence that

these areas have been affected by other phenomena such as

salt diapirs or water pumping.

Tectonic deformation in the form of subsidence and

uplift, within the Arava Valley along the Dead Sea Trans-

form, had occurred before the Nuweiba earthquake. These

movements formed the basis for the development of the

playas. The present day playas have reacted to the general

strike/slip movement along the Dead Sea Transform by

forming domal (e.g. Yotvata Playa) or en-echelon faulted

ridge structures (e.g. Evrona Playa). Some of the en-echelon

faults curve left northward, forming releasing bends where

the subsidence occurs. These structures have been modified

by local compression and uplift such as the tilting of the

Yotvata Playa (Arkin et al., 2000) and the topographic

inversion of the Evrona Playa (Amit et al., 1999).

Monitoring should be carried out regularly by GPS and

radar interferometry in order to strengthen the case for the

relationship to seismic activity. The above model should be

improved by further research using other historical sensor

measurements (J-ERS and Radarsat archives) to derive

deformation information during other time intervals along

the Dead Sea Transform.

Acknowledgements

The authors acknowledge the support of the Earth

Research Administration, Ministry of National Infrastruc-

ture, Israel; the Geological Survey of Israel; the Remote

Sensing Laboratory, J. Blaustein Institute, Ben Gurion

University of the Negev; the University of Paris, Tectonic

Laboratory; the Israel Space Agency; the French Space

Agency (CNES) Radar System Department for their

support within the framework of French–Israeli scientific

collaboration. Special thanks are due to Dr. D. Massonnet

for the useful discussions and comments and to Mr. Eli

Ram for his technical support in the field throughout the

project.

References

Amiel, J. A., & Friedman, G. M. (1971). Continental Sabkha in the Arava

Valley between the Dead Sea and Red Sea. Significance for origin of

evaporites. The American Association of Petroleum Geologists, 55,

591–592.

Amit, R., Zilberman, E., Porat, N., & Enzel, Y. (1999). Relief inversion in

the Evrona Playa as evidence of large-magnitude historical earthquakes,

southern Arava Valley Dead Sea Rift. Quaternary Research, 52, 76–91.

Arkin, Y., Ichoku, C., & Karnieli, A. (1999). Fault traces in the arid Arava

Valley floor Israel, revealed by RADARSAT surface roughness classi-

fication. Canadian Journal of Remote Sensing, 25, 302–310.

Arkin, Y., Karnieli, A., Hall, J. K., Chorowicz, J., Fleury, J., & Sarti, F.

(2000). Uplifted and tilted Yotvata Sabkha, Arava Valley, Israel. 14th

Int. Conf. Applied Geologic Remote Sensing, Nov., Las Vegas.

Baer, G., Sandwell, D., Williams, S., & Bock, Y. (1999). Co-seismic

deformation associated with the November 1995, Mw=7.1 Nuweiba

earthquake, Gulf of Eilat (Aqaba), detected by synthetic aperture

radar interferometry. Journal of Geophysical Research, 104(B11),

25221–25232.

Bartov, Y., Avni, Y., Calvo, R., & Freislander, U. (2000). The Zofar Fault—a

major rift feature in the Arava Valley. GSI Current Research, 11, 27–30.

Ben-Menahem, A. (1981). Variation of slip and creep along the Levant Rift

over the past 4500 years. Tectonophysics, 80, 183–197.

Chorowicz, J., & Deffontaines, B. (1993). Transfer faults on pull apart

model in the Rhine Graben from multi-source data. Journal of Geo-

physical Research, 98(B8), 14339–14351.

Feigl, K., Sarti, F., Vadon, H., McClusky, S., Ergintav, S., Durand, P.,

Burgmann, R., Rigo, A., Massonnet, D., & Reilinger, R. (2002,

February). Estimating slip distribution for the Izmit Mainshock from

Coseismic GPS, ERS-1, RADARSAT and SPOT measurements. Bul-

letin of the Seismological Society of America, 1, 138–160.

Fruneau, B., & Sarti, F. (2000). Detection of ground subsidence in the City

of Paris using Radar interferometry: isolation of deformation from at-


mospheric artefacts using correlation. Geophysical Research Letters,

27(24), 3981–3984.

Garfunkel, Z. (1981). Internal structure of the Dead Sea leaky transform

(rift) in relation to plate kinematics. Tectonophysics, 80, 81–108.

Garfunkel, Z., Freund, R., & Zak, I. (1981). Active faulting in the Dead Sea

Rift. Tectonophysics, 80, 1–26.

Khair, K., Karakaisis, F., & Papadimitriou, E. E. (2000). Seismic zonation of

the Dead Sea Transform fault area. Annali di Geofisica, 43(1), 61–79.

Massonnet, D., Briole, P., & Arnaud, A. (1995, June 15). Deflation of

mount Etna monitored by spaceborne radar interferometry. Nature,

375, 567–570.

Massonnet, D., & Feigl, K. (1995, June 15). Discrimination of geophysical

phenomena in satellite radar interferograms. Geophysical Research Let-

ters, 22(12), 1537–1540.

Massonnet, D., & Feigl, K. (1998). Radar interferometry and its applica-

tion to changes in the Earth’s surface. Review of Geophysics, 36(4),

962–972.

Massonnet, D., Feigl, K., Vadon, H., & Rossi, M. (1996). Co-seismic

deformation field of the M=6.7 Northridge, California earthquake of

January 17, 1994 recorded by two radar satellites using interferometry.

Geophysical Research Letters, 23(9), 969–972.

Massonnet, D., Rossi, M., Carmona, C., Adragna, F., Peltzer, G., Feigl, K.,

& Rabaute, T. (1993). The displacement field of the Landers earthquake

mapped by radar interferometry. Letters to Nature, 364, 138–142.

Peltzer, G., Crampe, F., Hensley, S., Rosen, P., & Rogez, F. (2000). Eight

years of crustal deformation in Southern California observed with ERS

SAR interferometry. ERS-ENVISAT Symposium: looking down to Earth

in the New Millennium. Gothenburg, Sweden: ESA.

Saffarini, S. H. (2000). Groundmotion characteristics of the November 1995

earthquake. Engineering Structures, vol. 22 ( pp. 343–351). Elsevier.

Assessing pre- and post-deformation in the southern Arava ... · Assessing pre- and...

Documents

Transcript of Assessing pre- and post-deformation in the southern Arava ... · Assessing pre- and...