Introduction

For four decades, running and ground waters of the Dead Sea (DS) drainage area have been extensively overexploited resulting in a strong imbalance in the water quantity of this terminal lake. The drawdown has risen from about 392 m below sea level (b.s.l.) in 1958 to 418 m b.s.l. in 2003 (e.g. Salameh and Abu Naseir 1999; Salameh and El-Naser 1999, 2000a and APC written communication). Consequently, new land has progressively emerged along the former shoreline. Less directly visible are changes in the regional groundwater level and in the hydrogeological setting of the near shore subsurface. The fresh-saline water interface migrated toward the DS, accompanying this lowering (e.g. Salameh and El-Naser 2000b). In consequence, new springs appeared along the retreating shoreline; salt layers previously immersed in DS water became exposed to freshwater and its dissolving capacity; numerous sinkholes and subsidence then appeared in many places along the DS shore.

To know which precursory elements to detect and monitor, in order to assess areas prone to collapse, we have to understand subsidence and sinkhole occurrence mechanisms. The DS shore sags occur where masses were withdrawn from underground layers (Frumkin and Raz 2001). Groundwater preferentially uses faults, joints and fractures such as conduits or flow lines. Chemical and physical erosion takes place along such weaker zones. Underground voids then appear, creating growing cavities, which come up to the surface after successive roof collapses. Hazard phenomena related to the DS lowering exist all along the Jordanian shore, usually at the tip of alluvial fans. For example, in the northern part, for years numerous landslides occurred along the muddy-gravely coast and, at the beginning of 2004, a sinkholes site affected a 40-year-old alluvial fan that had emerged. However, in this particular case, collapse features could be related to wastewater injected in the subsurface by hotels. The southern DS is by far the most affected by sinkholes and decametric to kilometric subsidence phenomena. GAH, the Lisan Peninsula (LP) and the Lynch Strait, which emerged around 25 years ago, constitute the areas of interest in the present paper (Fig. 1, right, number 1, 4 and 5).

Fig. 1
figure 1

Tectonic setting of the Dead Sea (DS) ( left) (adapted from Ben-Avraham 1997); Space Shuttle photograph focuses on the area of interest ( right). In the text, we will call the Lisan “old” part the triangular area defined by (6), (7) and (9). A recent emerged platform (REP) surrounds most of the “old” Lisan Peninsula ( LP). It appears clearly, e.g. between the former Cape Costigan (7) (−390 m b.s.l.) toward the shoreline (around −418 m b.s.l. in 2003). During the 1990s, the Arab Potash Company (APC) built two saltpans, no. 18 (2) and 19 (3) over the broadest part of this emerged zone. Nowadays, they are affected by sinkholes and subsidence (see text and Figs. 2, 3). Ghor al Haditha (5) (GAH) is an agricultural zone strongly affected by collapse features since the 1980s (Fig. 5). Until the end of the 1970s, the Dead Sea was a single water surface. Due to the lowering of the water level, the Lynch Strait (4) emerged and the lake split into two sub basins ( left). The size of the northern one is continuously reducing while salt evaporation ponds of the APC (Jordan) and the Dead Sea Works (Israel) cover the southern one. Space Shuttle photograph acquired on 2003–02–23 (Astronaut Photography of Earth—ISS006-E-30992 - http://eol.jsc.nasa.gov/)

Full size image

All along the DS shore, hazards seem, at least partially, to be structurally controlled (Abelson and others 2003; Frumklin and Raz 2001; Taqieddin and others 2000). Here, we clarify this aspect by placing hazardous features, recorded along the Jordanian southern DS shore during the last fifteen years, in their geological context. We discuss their relation to the global and local tectonic setting. To achieve this goal, we performed analyses of sinkholes and subsidence inventories as well as using the results of field observations, space borne radar interferometric outputs i.e. coherence images, Digital Elevation Model (DEM), deformation fields and also aerial photographs and interpretation of satellite images, inside a Geographical Information System (GIS).

Chronology of the most costly hazardous events that occurred during the nineties along the Jordanian southern Dead Sea shore

The global cost of the DS lowering, in term of damage caused to human infrastructures and activities, has not yet been established. However, by comparing various sources discussing the DS lowering related hazards (e.g. Abelson and others 2003; Abou Karaki 1995; Arkin and Gilat 2000; Baer and others 2002; Batayneh and others 2002; Closson and others 2003a, 2003b, 2005; Frumkin and Raz 2001; Itanar and Reizman 2000; JER 2003; Knight 1993; Taqieddin and others 2000; Tapponnier 1993; Yiechieli and others 2003), it is possible to state that the Arab Potash Company (APC) is by far the most affected industrial actor. We estimated its total losses could rise from $70–90 million. Below, we present some elements justifying this huge estimate. During the 1990s, APC extended its industrial complex with the building of two new salt evaporation ponds (SEP 18 and then 19, see e.g. Fig. 1, right, number 2 and 3) over a wide part of the recently emerged platform (REP) bounding the “old” LP in the west (Fig. 1, see number 6, 7 and 9 for the “old” LP limit).

Sep 18 problems

In 1992, the Lynch Strait (Fig. 1) was flooded during an exceptionally rainy winter. During the summer, APC’s workers pointed out an alignment of sinkholes. In October, near Cape Molyneux, new features destroyed one of the roads in the APC industrial complex over about 100 m (Knight 1993; Tapponnier 1993; Taqieddin and others 2000). The real problem for the company was that the damaged road should have turned into a part of the 12-km-long dyke planned for the future $32 million SEP 18. Consequently, designers reshaped this 95 million m3 pond and built the dyke eastward of this hazardous zone. While the pond was being completed at the end of 1997, a leak in the new dyke was discovered. Despite this threat, the pond was filled at the beginning of 1998. Since then, APC has tested several solutions to repair the leak while working on taking precautionary measures to prevent the collapse of the dyke. According to APC security engineers the leak could be due to the formation of an artesian water basin or to a sinkhole that formed in 1996 (JER 2003). In 1999, a decametric hole perforated the bottom of the basin. Security engineers decided to fill it quickly (APC, personal communication, 2004). This costly operation consisted of building an earthwork bridge of more than 100 m long, 40 m wide, with a height of more than 10 m, to allow trucks to approach the collapsed area and unload hundreds tons of landfill into it. From October 2002, after only 4 years of activity, SEP 18 had been partially dried up (Fig. 1) because of a dyke failure. During the same year, APC invited international tenderers to carry out remedial work (APC 2002). APC estimated the work would be completed in 17 months.

Sep 19 destruction

On March 22, 2000, 2 months after its completion, a part of SEP 19 ($38 million) collapsed over 1,650 m during the period of filling (Closson and others 2003a, 2003b, 2005). The destruction took place 2 km NW of Cape Costigan (Figs. 1 and 2). For over 4 years, other segments of this pond were damaged, increasing the number of hazardous places.

APC’S annual loss

Between 2000 and 2002, subsidence and sinkholes hazards related to the DS lowering caused the loss of $38 million and stopped the exploitation of a $32 million production unit. Moreover, 3 years later, at the end of 2003, APC lost a lawsuit against SEP 19 contractors (N. Abou Karaki, personal communication). Compensation could be $14 million (JER 2003). This amount (32+38+14=$84 million) does not reflect the annual losses of the APC.

To calculate the annual loss from the period of filling, we should compare the (expected) annual income generated without taking into account the problems of sinkholes with the real annual income. The case of SEP 19 is the easiest. It never generated income. After the destruction of SEP 19 and during the lawsuit (2000–2003), APC restricted visits to the whole peninsula and data in relation to saltpans became confidential. Hence, the annual loss for SEP 18 since 1998 is difficult to calculate. Nevertheless, the above-mentioned events shed a light on the problems encountered by security engineers in keeping the saltpan in operation.

Tectonic setting

Global setting

The DS fault system is a sinistral transform plate boundary between the African Plate and the Arabian Plate. It links the Red Sea sea-floor spreading centres with a segment of the Alpina Orogenic Belt in the Taurus Mountains zone. The transform dates back to the Cenozoic when the African-Arabian continent broke up (e.g. Al-Zoubi and ten Brink 2001). Faults are the dominant structural elements along the transform. The main strike-slip faults, North and South of the DS, are arranged en-echelon so that the DS is a pull-apart basin formed by left-lateral slip (Quennell 1958; Garfunkel and others 1981). From topographic and bathymetric data (Hall 1997; Bartov and others 2002), the DS pull-apart basin can be divided in two compartments: a “southern” and a “northern” one. The normal transverse En Gedi fault separates them (Fig. 1, left). On the basis of subsurface seismic data (e.g. Ben-Avraham 1997; Gardosh and others 1997), two other transverse faults, at intervals of 20–30 km, have separated the southern unit into three fault-bounded blocks. From South to North, they are respectively:

The greatest number of sinkholes and subsidence sites encountered along the whole Jordanian DS coast belong to the Lisan block. It is worth noting that most of the sinkholes affecting the western shore also appeared over this block, but not necessarily all. The reason can be found in the geomorphic difference between the areas of the Lisan block and the northern compartment of the DS basin i.e. narrow, steep seashores to the north, and wide, low angle shores, alluvial fans, and a dry sea floor to the south (Itanar and Reizman 2000).

The Lisan block

The Lisan block is a structurally complex rhomboid-shaped area. From the Moab rift escarpment, in the East, to the WMFN in the West, we distinguish: (Fig. 1, left and Table 1 for faults’ abbreviations)

Table 1 Main faults characteristics from Ben-Avraham 1997; Gardosh et al. 1997; Niemi and Ben-Avraham 1997
Full size table

  • In the intersection zone between the AF and the BF (Fig. 1, left, numbers 3 and 5), the AF separates into two parallel sections: the GSF and the WDF plus a third, the EDF, encompassing the almond-shaped Adh-Dhira basin eastward and westward (Fig. 1, left, number 8, 9, 10). This basin is a consequence of the active fault running on from the AF (Tapponnier 1993):

  • The NNE-N striking WDF extends along the Jordanian DS coast until it intersects the GF and then continues northward (Fig. 1, left, numbers 8 and 7). Inside the Lisan block, it also corresponds to the eastern limit of the El-Mazra’a—GAH graben that separates the LP from the Moab rift escarpment (Fig. 1, left). The GSF crosses through the graben toward the North (Ben-Avraham 1997; Garfunkel 1997). Bathymetric contours allow identification of a deep and wide submarine canyon that could result both in the passage of the GSF (see references map-1 and map-2) and in the erosion resulting from a (near) complete drying up of the DS (Neev and Emery 1967). Alluvial fans occupy both sides of the graben. They are more apparent along the eastern side of the graben while a gentle slope characterizes the western border up until the LP eastern fault cliff;

  • The LP occupies the centre of the Lisan Block. Fault scarps marked by highly eroded cliffs bounded the area that emerged before the appearance of the REP along much of its length (hereafter we will call this zone the “old” part, see also Fig. 1’s legend). This triangular tongue-shaped zone caps the biggest salt diapir of the DS basin (Fig. 1, right, number 6, 7 and 9. The hydrographic network is globally radial, emphasizing the elliptic uplifted area (Fig. 1, right). In the SW, a dendritic network of gullies dissect a wide collapsed part, while two linear networks of ravines affect the NE and NW sides;

  • The Lynch Strait corresponds to a graben separating the LP from the WMFN (Fig. 1, right);

  • Similarly to what happens on the eastern side, in the area of intersection between the BF and the SF, the SF separates into two sections (Fig. 1, left, numbers 4, 6, 11). The WIFS is generally viewed as the principal southern extension of the Jordan Valley strike-slip fault. It crosses through the Lynch Strait toward the En Gedi diapir before turning off northward (Ben-Avraham 1997; Niemi and Ben-Avraham 1997; Garfunkel 1997). The second section is the WMFN, clearly visible all around the citadel of Masada (outside the area of study).

The study of faults, fractures and flexures allows a description of the present day dynamics of the area. The Lisan block lies within the Jordan DS Transform zone which mainly functions in left lateral shear with a normal component in that area as suggested by the focal mechanism of the mb=5.1 earthquake of 23 April 1979 (mechanism established by Arieh and others 1982). The seismicity of the basin is asymmetric. Most of the activity concentrates along the eastern side of the northern compartment. A cluster of events, just north of the LP, can be associated with the En Gedi fault. Dorbath (1998) and Klinger (1999) worked out the focal mechanisms for the area’s earthquakes from 1982 onwards. The fault plan solutions were found to mainly reflect strike-slip faulting where four out of nine events were left lateral, three were right lateral, one reverse and one normal faulting. The right lateral faulting probably occurs on the faults that cross the basin in the E–W direction.

Sinkholes and subsidence in relation to the tectonic setting of the Lisan block

The northern part of the Lisan Peninsula and salt evaporation pond 19

Mapping of the main hazard features

Figure 2 gathers some of the most pertinent elements with regard to hazards affecting SEP 19. The data set was produced by using various approaches and techniques such as field observations, photo-interpretation, space borne radar differential interferometry and interferometric DEM analysis.

Fig. 2
figure 2

Hazards affecting salt evaporation pond 19 (SEP 19) (see also Fig. 1). Major faults location derived from Ben-Avraham (1997). Bathymetric contours come from Hall 1997, and see reference Map-2. The relations between elements of this figure are discussed in point 4

Full size image

By applying a differential interferometric technique to the European Remote Sensing (ERS) radar images, Baer and others (2002) Closson and others (2003a), Derauw (1999) and Shimoni and others (2002) mapped two kilometric subsidence zones (Fig. 2) separated by a narrow passage, several hundreds meters wide, characterised by an absence of deformation.

From photo-interpretation, we analysed the northern REP and inferred the existence of a highly disrupted zone north of Cape Costigan (Closson and others 2005). We interpreted this one as a consequence of the intersection of two converging local fault networks bounding respectively the east and the west margins of the “old” Lisan area (e.g. Bartov and others 2002; Sunna 1986).

From grey contrast variations over panchromatic aerial photographs and SPOT images, we delineated a particular geological feature with a lozenge-shape (LSF) (Closson and others 2003a). Inside a GIS, we observed a connection between the locations of the LSF and the narrow passage connecting the two subsidence areas (Fig. 2). We interpreted it as a fault-bounded hectometric block slightly tilted toward SW. Two independent data sources deriving from radar interferometry processing agree with this interpretation:

  1. 1.

    Interferometric radar coherence images highlighted a sub horizontal outcrop of distinctive salt rich deposits. From variation in shape of this layer, two kilometers North of Cape Costigan, we deduced an offset of rock material with similar properties in relation to the resistance of topsoil erosion (Closson and others 2005);

  2. 2.

    We studied subtle topographic variations of the bottom of SEP 19 from a very accurate Digital Elevation Model (DEM) computed with an ERS “tandem” pair of acquisitions (Closson and others 2005; see Derauw 1999 about the vertical accuracy). In the northern part of the LP, we detected lineaments defining several hectometric-long collapse features (Closson and others 2005). Field surveys confirmed the existence of such basins. One Aster image, acquired a few months after the destruction of the saltpan, showed remaining circular and elongated ponds of brine in different places, thus validating the existence of topographic depressions. These observations, as well as the ones exposed below, support the idea of dissolution phenomena under the bottom of SEP-19.

Dyke collapse and new damage

Because of continuing subsidence movements detected from the beginning of the 1990s (Shimoni and others 2002), after the March 22, 2000 collapse, we could expect that the remaining dyke segments “record” new hazardous events. We postulated that some days or weeks after the rapid emptying of the saltpan (55 million m3.hr−1), isostatic movements would have occurred, creating breaks in some places along the dyke. Discussions with APC’s engineers did not support this supposition. However, during field surveys in 2002 and 2004, they mentioned several places where decimetric to metric cracks (Figs. 2 and 3), perpendicular to the dyke, appeared later, months to years after the incident. We also observed a decametric landslide.

Fig. 3
figure 3

Salt evaporation pond 19 (SEP 19) was destroyed on March 22, 2000. A shows the area where 55 M m3 of brine previously pumped went back into the Dead Sea in less than 1 h. Two black arrowheads indicate sinkholes in the extension of the dyke. B shows a decimetric crack affecting the dyke. Others exist mainly concentrated in the areas of subsidence (Fig. 2). They appeared several months to years after the breaching of the dyke. It implies that ground displacements are always active and the destruction of SEP 19 was most probably unavoidable. The inset is a Space Shuttle photograph acquired on 2003-02-23 (Astronaut photography of Earth—ISS006-E-30992 - http://eol.jsc.nasa.gov/)

Full size image

Because most of the cracks, but not all, are concentrated in the two areas where subsidence is the strongest (by comparison with the whole peninsula, see Baer and others (2002) or Shimoni and others (2002)), we think that it emphasizes the contribution of subsidence to the destruction of the pond. From such fractures (Fig. 3, B) we deduce three important facts in the frame of SEP 18 and 19’s rehabilitation: (1) ground displacements have been and are very active since 2000; (2) the whole dyke segment along the shore is prone to collapse and the most hazardous places correspond to the two subsidence zones (Fig. 2); (3) the destruction of SEP-19 was most probably unavoidable because of the number of new cracks affecting the earthwork dyke.

Furthermore, we detected two metric sinkholes in the extension of the remaining dyke segment (Fig. 3, A). This finding confirms that unsaturated water, with respect to salt, flows under SEP 19. As previously mentioned, underground flow take place preferentially along faults and cracks. Consequently, the mapping of lineaments (Closson and others 2005) and sub surface discontinuities is of importance in order to map SEP 19’s threats in greater detail.

The Lynch strait and Sep 18

Formerly, the Wadi Araba poured out into the DS southern sub basin, now occupied by salt evaporation ponds (Fig. 1, right, number 11). Today, the brackish-water discharge of the Wadi Araba catchment is guided into the northern DS sub basin by a truce-line flood channel (Fig. 1, right, number 8). The river thus crosses the Lynch Strait and created a new delta in the northern part (Fig. 1, right, number 10). When entering the drying up Lynch Strait, parts of the brackish water percolate. The underground water is thus sub-saturated with respect to salt and dissolution phenomena can take place. Parts of the mud holes affecting the Lynch Strait can be explained in such a way (Fig. 4) (Knight 1993; Tapponnier 1993; Taqieddin and others 2000). Collapses appeared without threat to human lives or infrastructures until 1992. Investigators have shown that the 2-km sinkholes lineament that destroyed an APC access in 1992 occurred due to fluctuations of the DS level combined with sudden Wadi Araba fresh water incursions after the major flood in March 1992. The collapsed zone corresponded to former subsurface water flow paths. Alignment and angular change direction suggested possible tectonic control of the phenomena. Twelve years after this incident, new evidence supports this assumption. A few kilometers north-westward, sinkhole clusters generally display a linear shape along the shore (Arkin and Gilat 2000; Itanar and Reizman 2000; Abelson and others 2003; Yiechieli and others 2003). Strikes are similar in two predominant directions (24°±15° and 340°±16°) of the WMFN (Table 1; Sagy and others 2003).

Fig. 4
figure 4

Hazards affecting the salt evaporation pond 18 (SEP 18) and the Lynch Strait. The relations between elements of this figure are discussed in point 4

Full size image

From aerial photographs integrated inside a GIS, we deduced that 1992’s sinkholes are globally aligned in a NNE direction compatible with one of the main fault directions (24°±15°). On closer inspection, two segments, hundreds of meters long, with two different strikes, compose this alignment of sinkholes (Fig. 4). We interpret this angular change of direction as a possible consequence of the intersection of the WIFS (Fig. 1; Niemi and Ben-Avraham 1997) and the local fault zone (Bartov and others 2002), bounding the western margin of the “old” Lisan (Figs. 1 and 4).

The mapping of the deformation fields

We investigated the deformation field affecting SEP 18 some years before its completion in 1997–98. We applied SAR differential technique to a pair of ERS acquisitions spanning 11th June 1992 to 05th August 1993. The differential interferogram showed inside the pond’s perimeter lineaments, parallel and perpendicular to the dominant strikes but, above all, two subsidence areas with several collapse features inside (we reproduced outlines over Fig. 4). The smallest zone was not far from the eastern limit of the area flooded in 1992, while the second zone affected a wide area in the northern extremity of the pond. Consequently, already by 1992–1993, two hazardous places threatened the future SEP 18. In 1999, one decametric sinkhole perforated the bottom of the saltpan in the immediate vicinity of the biggest subsidence area. This fact corresponds to the differential interferogram interpretation.

The Ghor al Haditha area

GAH attracted the attention of researchers (Abou Karaki 1995, 1999; Batayneh and others 2002; Closson and others 2003b, 2005; Salameh and El-Naser 2000b; Taqieddin and others 2000). The monitoring of subsidence and collapse features dates back to 1991 (Pr. N. Abou Karaki, personal communication). Gravel holes occur in clusters, scattered inside a kilometric ribbon-shape hazardous area, around 200 m wide, NNE striking (Fig. 5), roughly parallel to the DS shore, at the tip of an alluvial slope resulting from the growth and merging of several fans.

Fig. 5
figure 5

The graben of Mazra’a – Ghor al Haditha (GAH). Sinkholes and subsidence occur in a 200-m-wide ribbon shape NNE-striking between the Ghor Safi (GSF) and West Dhira faults (WDF) (see Fig. 1). Young and old collapse features also affect the eastern Lisan faulted margin. The relations between elements of this figure are discussed later

Full size image

From 2000, we focussed on a specific case regarding an elliptic, near-circular network of cracks in two different places (Closson and others 2005). They had already been detected in 1991, and then monitored and photographed each year (Pr. N. Abou Karaki, personal communication). The circular systems evolved from a few meters long, 2 cm wide to hundreds of meters long and one meter wide in several places by the year 2004. We noted subsidence acceleration from the year 2000. In 2002, clusters of decametric sinkholes occupied the bottom surface (Closson and others 2005). They are thus compounded structures. Similar features were observed along the western shore too (e.g. En Gedi, Frumklin and Raz 2001). In 2003, a new subsidence area in the alignment of the two first places, several hundred meters northeastward, was mentioned (M. Barjous of the Natural Resources Authorities, personal communication). In 2004, we located, with a global positioning system, numerous sinkholes sites affecting the REP in the continuation of the alluvial slope, toward the WDF (Fig. 5; Fig. 1, left). Thus, we detected the widest sinkhole ever seen in the area with a diameter of around 60 m. It clearly appears over high-resolution satellite images and astronaut photographs. We also recorded new collapse features along the sinkholes lineament toward the GSF (Fig. 5; Fig. 1, left). The geographical distribution of collapses complements our view of the shear zone extending between GSF and WDF (Fig. 5). Dissolution phenomena related to the DS lowering revealed the existence of at least one young, 5-km-long buried fault. Beyond GSF (southwestward), we have not detected any sinkholes or subsidence (so far). At first glance, this suggests that the hazardous lineament connects only GSF-WDF. However, we think that it could go on until it reaches the eastern Lisan margin. Fig. 5 (question marks) presents two potential extensions toward the LP. One proposition connects with a relatively old major circular depression (Fig. 5, Birkat el Haj) while the other connects with a very young sinkholes site.

Indeed, for years, even before the Dead Sea lowering, collapses affected the highly faulted boundary zone between the LP and the Mazra’a – GAH graben (Figs. 5 and 6). Dissolution phenomena occur because of underground brackish water coming from the LP, flowing along the faults toward the northern DS sub basin. When focussing our attention on the “Birkat el Haj” area (literally “the swamp of Haj”) (Fig. 1, right, number 9 and Fig. 5), from the analysis of an InSAR DEM computed with the ERS tandem pair of 15th to16th December 1995 (Closson and others 2005; Derauw 1999), we delineated two hectometric circular collapses and lineaments bounding two small grabens (Fig. 5). One of them is precisely in the extension of the sinkholes lineament (thus with the same strikes). The lineament could cross over the tip of the alluvial fan north of Mazra’a (Fig. 5). The wide circular collapse structures of Birkat el Haj are relatively old because they are present over 1953’s aerial photographs. They are thus not related to the present day DS lowering. During a field survey, we found one young and massive sinkhole two kilometers northward of Birkat el Haj. According to APC engineers, many others exist in this area. From a discussion with Pr. E. Salameh, we suggest a second possible connection with the main hazardous alignment.

Fig. 6
figure 6

Location of relict fossil sinkholes and recent collapses detected by radar interferometry over the “old” part of the Lisan Peninsula (LP). Lineaments deduced from satellite images, aerial photographs interpretation and the geological sketch map of Sunna (1986)

Full size image

Young collapses and relict fossil sinkholes of the Lisan Peninsula

From stereoscopic analysis of aerial photographs acquired in 1953 (September, 1:25,000) and 1992 (September; 1:30,000), we detected lineaments and clusters of relict fossil sinkholes as well as hectometric circular depressions on the east side (Figs. 1 and 5, Birkat el Haj). We also mapped most of the linear features as indicators of potential faults, joints, fractures, bedding and folds. Previous works, in particular the findings of Sunna (1986) mainly based on field observations, complement our interpretation (Fig. 6). Sunna (1986) observed that several detected lineaments were faults, while others were either controlled by lithological change or resulted from erosion along a set of fractures and joints. We improved on this research by analysing a high precision interferometric DEM (Derauw 1999) generated from two ERS acquisitions on 15th and 16th December 1995.

Differential inteferograms computed for the period 1992 to 1999 (Cornet and others 1998; Derauw 1999; Closson and others 2003a) have shown linear subsidence and collapses over the southern uplifted part. The uplift of the Lisan salt dome has created an oval pattern of anticlines and synclines as well as faults and cracks in the southern part of the peninsula. Sinkholes areas underline the weakest places i.e. highly fractured rocks. In this southern zone, they are thus a consequence of the diapirism. Their global distribution has equally adopted an oval pattern.

Baer and others (2002) observed similar features and proposed that three phenomena could explain the movements: dissolution of specific salt layers, collapse within a layer that underwent enhanced extension at the margin of the dome and dissolution of salt layers in the underground. Cavities affecting the “old” part of the LP can be divided in two categories in relation to their geographical position (Fig. 6). On the one hand, a cluster of young collapses in continuing evolution, related to the uplift of the Lisan salt diapir. On the other hand, all the other collapse features can be considered as relict fossil sinkholes.

Discussion

Origins of hazards affecting SEP 19

The impact of the Lisan sub dome diapir uplift

Several authors (e.g. Baer and others 2002) detected an uplifted area interpreted as a possible sub dome of the Lisan diapir (Figs. 2 and 7). We mapped this displacement field from two interferometric pairs of ERS images acquired in descending mode, track 078/frame 2979, from 11th June 1992 to 05th August 1993 and 29th July 1995 to14th July 1996 (see Closson and others 2003b). Inside a GIS, we noted a correspondence between the subsidence/uplifted areas and three absolute minima of the Bouguer anomalies (Hussein 1997): −130 mGal for the potential diapir sub dome; −126 mGal and −128 mGal for the subsidence areas along the shore. This supports the sub dome diapir interpretation. Because of the proximity between the subsidence/uplifted places, we suggested that the sub dome could control the mechanism of subsidence at its immediate periphery.

Fig. 7
figure 7

Relation between various damaged places (sinkholes, collapses, dyke breaching) and possible lineament intersections affecting the Lisan Peninsula (LP) and its vicinity. The radial hydrographic network of the Lisan Peninsula underlines the diapir uplift and show directions of related faults and fractures. The main faults of the Lisan tectonic block derived from Bartov and others 2002, Ben-Avraham 1997, Closson and others 2004, Sunna 1986

Full size image

From field observations (2004), we possess a new argument supporting the sub dome diapir interpretation for the uplifted area. Figs. 2 and 3B show cracks affecting the earthwork dyke in several places, but two of them are outside the subsidence areas. To explain such an occurrence, we suggest relating them to the Lisan diapir sub dome uplift (Fig. 7). Classically, the manifestation of diapir uplift appears through faults and fractures systems emanating from a common central point. Such systems exist a few kilometres southward from the main dome of the Lisan diapir. They are either visible through lineaments or partially shown by the hydrographic network (Figs. 5 and 7).

The effects of compaction phenomenon

Baer and others (2002) proposed another explanation for subsidence. They related the subsidence to the decrease of the water level by analysing the relationship between changes in DS water levels and deformation of the aquifer network. They suggest that subsidence along the DS shore occurred where the total thickness of the fine-grained marl layers is between 5 and 20 m in the upper 30 m below the surface.

The effects of subterranean unsaturated water

During a field survey (2004), we observed sinkholes affecting the collapsed area of SEP 19 (Fig. 3, A). From this fact, we deduced another possible cause, never previously mentioned, which would explain the collapse of the dyke i.e. the presence of unsaturated water flowing in the underground of SEP 19. Fig. 3 (A) indicates that unsaturated water exists underground and causes the dissolution of residuals DS water in the aquifer matrix, creating cavities, even as far as north of Cape Costigan. This in turn results in ground collapses liable to provoke leakage in an earthwork dyke.

To fill the underground aquifers of this desert area with unsaturated water, we think that the fault networks bounding the “old” Lisan area by the east and the west could be used as conduits. Intersection fault zones north of Cape Costigan (Fig. 1) should concentrate fresh or brackish water. The topography of this zone (Closson and others 2004) characterised by several subsidence features supports the idea of a higher rate of dissolution. Due to the fresh-/DS water interface migration, intersection zones located in the vicinity of the foreshore should present the greatest risks of collapse occurrence.

The origins of sinkholes and subsidence affecting SEP 18

Sinkholes affected saltpan 18 more frequently than 19. Hence, we deduce that more unsaturated water circulates under the bottom layers and in its vicinity. Water comes from two directions: (1) from the wadi Araba, owing to the truce-line flood channel, as attested by the 1992 sinkholes alignment studies (Knight 1993; Tapponnier 1993); (2) from the LP, probably owing to the Lisan diapir’s radial faults system. APC’s engineers emphasize the role of the underground brackish water coming from the “old” Lisan part. Several springs exist along the foreshore bounding the pond and APC divers mentioned an underwater sulphur spring close to the new brine intake station, between SEP 18 and 19 (personal communication during a field survey in 2004).

From topographic maps, aerial photographs or various satellite images, it is easy to delineate linear kilometric narrow canyons dissecting the “old” part of the LP (Fig. 4). The analysis of an accurate InSAR DEM computed with the ERS tandem pair of 15–16th December 1995 (Closson and others 2004; Derauw 1999) and the geological sketch map of Sunna (1986), show that canyons are in the extension of kilometric lineaments diverging from the Lisan diapir main dome. Such faults and fractures networks affect both the “old” LP and the REP. WNW striking linear weakness zones cross through SEP 18 and could be used as groundwater conduits. It is worth observing that the most affected areas are in the extension of the main canyons (Fig. 4).

Ben-Avraham (1997) and Bartov and others (2002) mapped two fault zones, NNE striking, crossing through the Lynch Strait and bounding the Lisan’s western margin respectively. Intersections between such faults networks and the Lisan diapir radial fault system created intersections in various places all along the western REP.

We previously mentioned that such intersections probably correspond to the most hazardous places along the former shorelines. The distribution of the hazardous events recorded during the last 12 years support this hypothesis (Fig. 7).

Mechanism of subsidence in Ghor al Haditha

Several processes can be invoked to explain subsidence and collapse features. On the one hand, comparison of the fresh water’s conductivity at the apex and the tip of the most affected place showed that, in a few kilometres, the concentration of salt increased to 40 g/l. This value confirms the existence of soluble deposits in the underground. Given an average debit at the apex of 120 l/s (E. Salameh, personal communication; Salameh and El-Naser 2000b), wide bowl shaped subsidence features could be explained in such a way. On the other hand, the size of the bowl shape subsidence also suggests a lowering of porosity by the repacking of unconsolidated grains. Celik and Asfin (1998) investigated similar subsidence features in Turkey. They underlined the importance of compaction in their case study. The triggering factors of such phenomena are the decrease of pore pressure and the vertical equilibrium of sediments in the vadose zone. Both are related to the drop in levels of the DS and surrounding water table. Consolidation of the underlying materials (i.e. laminated calcareous silts interbedded with layers of halite and gypsum) is affected by the sudden increase of effective stress. Hence, subsidence at the surface occurs several years after the decrease of the ground water level and could be delayed with increased depth of the aquifer. The monitoring of cracks of which dates back to 1991 supports this possible explanation (Fig. 8). As previously mentioned, circular fractures and subsidence have slowly developed from 1991. Subsidence acceleration occurred from 2000 and then massive sinkholes appeared in 2002 (Closson and others 2005).

Fig. 8
figure 8

Location map is a subset of Fig. 5, focusing over Ghor al Haditha (GAH). Pictures 1–6 show the evolution of a crack segment belonging to one of the three-bowl shape subsidences, from 1991 to 2002. Strong acceleration occurred from 2000 and several massive sinkholes perforated the bottom from around 2002 (Closson and others 2004). Photographs courtesy of Najib Abou Karaki

Full size image

From observations along the western coast, in front of the Lisan, Abelson and others (2003) suggested that a mechanism exists combining gradual land subsidence with the formation of massive sinkholes. Differential compaction could generate and localize shear strain within clayed layers along or parallel to pre-existing faults. These faults may then serve as conduits for sub-saturated water to percolate to the salt layers and promote the development of sinkholes. The gradual land subsidence features and sinkholes could be controlled by tectonics. They reflect the young and possibly active faults concealed within the fill of the DS basin.

Conclusion

Forty years ago, sinkholes along the DS coast were most probably related to seasonal short-term fluctuations accompanied by parallel variations of groundwater levels rather than to the DS lowering. Since that time however, the number of hazardous events has increased dramatically and, during the nineties, more and more damage has been recorded on both sides of the DS.

In this paper, we emphasize:

  • The role of conduit played by structural elements in underground water circulation;

  • The contribution of the Lisan diapir radial faults network which intersects the fault systems affecting the boundaries between the LP and the eastern (Mazra’a – GAH) and western (Lynch Strait) grabens;

  • The role of intersection fault zones in explaining the higher dissolution rate in specific places along the former shorelines. Fault intersection zones located close to the shore concentrate unsaturated water liable to dissolve salt layers previously immersed in DS water;

  • The role of fresh-/DS water interface migration in explaining the location of sinkholes sites along the shore.

We have also showed that the boundaries of the subsidence areas complement our view of the tectonic structure buried in the REP. We located relict fossil sinkholes over most of the oldest part of the LP and recent collapses related to the diapir uplift in the southern part. Such a study should be a warning to all contractors and decision-makers concerned with the industrial development or the expansion of tourism along the Jordanian DS shore, and more specifically in the northern part where serious problems seem to be appearing.