Shear wave reflection seismics yields subsurface dissolution and subrosion patterns: application to the 5 Ghor Al-Haditha sinkhole site, Dead Sea, Jordan

Abstract

and 6 show clear first breaks as indicated by black arrows. Why were the first breaks not picked and used for refraction static correction, given their strong presence in the data? Why were the first breaks not surgically removed from the shot gathers? Skipping this step may result in them being stacked in the top parts of the seismic section. Please elaborate on this.
Author: Written above is "main data processing". Residual statics do not improve the result, because the resulting shifts of +-3 ms max are too small to get a significant effect. This was tested at a lot of shear wave data sets and also proofed by other autors operation with shear wave reflection data (e.g. André Pugin, GSC). First break picking for refraction statics calculation does not work above an inverse velocity function situation due to operation on a paved road. This first breaks in the records are not common refractions, picking and inverting them to derive a refractor model would fail. Clearly the first breaks are removed by Top Muting (not Sugical Muting) prior to CMP stacking, I added this detail in the processing sequence now and it is shown in the new Fig.7.
333 Resulting depth section of profile 1b after post stack FD time migration (top) and interpretation of the main structure elements (bottom). The reflectivity seen in the upper 100 m of the stacked section seems realistic. However, below 100 m, it needs to be justified by showing the events on the processed shot gathers.
Author: The nature of the few reflectors below 100 m in Fig. 8 remain speculative due to the missing borehole proof. They result from the statistic of 24-fold CMP stacking in mean, not from single shot gathers. Again the desire to proof in detail would result in a special processing paper. Fig 7 and Fig 9 clearly shows that events below 100 m are structural imaging instead of processing artifacts.

349
Finally, the reflection amplitude responses throughout all seismic sections represent reflection coefficients of nearly 0.1 or less, indicating materials with relatively low contrasts in seismically-sensitive material properties (elastic parameters and density) and hence in shear wave seismic velocity. In the processing part you mention "Amplitude scaling" and in here you interpret the relative reflection amplitudes. Please clarify if AGC, trace equalization or spherical divergence correction was applied. If the AGC was applied, interpreting the relative amplitude strength is dubious.
Author: True Amplitude processing was not applied until yet, this would be again stuff for a specialized processing paper. Even only AGC 220 ms is applied for amplitude scaling, the sections show detailed amplitude dynamics resulting from constructive 24-fold CMP stacking. This is common also in hydrocarbon exploration for structure imaging. True Amplitude processing would be required if the target would be AVO analysis or inversion etc..

381
Resulting depth sections of combined profiles 2 and 2b-2 after post-stack FD time migration (top) and interpretation of the main structure elements (bottom). Again, looking at the depths of ca. 100 m and below, the data seems to show migration artifacts. Have different migration algorithms been used and why was FD migration chosen?
Author: From my opinion again speculative, and again the kind of processing artifacts in charge is not specified. Except profile 2b-2 at the left (East) border, where some migration smilies disturb the imaging, FD migration did a good job. Different migration methods applied to shear wave reflection data were extensively tested since more than 10 years (see e.g. Polom et al. 2010 and, and FD migration (67 degrees algorithm) was found to be a sufficient tool, superior to Kirchhoff post-and prestack time migration methods. The results presented here show the advantageous imaging capabilities of the method regarding high-resolution structural analysis and depth penetration compared to common refraction seismic methods or common P-wave reflection analysis. Unless referring to previous reports using the P-wave reflection method, this sentence and the entire comparison is not valid. Please reformulate the sentence by referring to P-wave reflection studies at the sites vicinity (if any), or remove it. It is not valid to compare S-wave reflection surveys with P-wave refraction tomography or MASW.
Author: The sentence has been modified by referring a P-wave reflection result shown in Ezersky and Frumkin (2013). In particular, the message of the sentence does not compare S-wave reflection with P-wave refraction tomography or MASW, no wording about this. Why this is valid or not is never stated in a paper until yet from my point of knowledge, the proof of this reviewer statement remains missing. From my opinion, each subsurface imaging method can compared by another.

567
The favourable velocity-frequency relationship of the resulting wavelets and the absence of pore fluid effects enabled a meter-scale resolution and a nearly 200 m penetration depth. The results seen on the stacked sections seem reliable down to ca. 100 m. Without any deeper borehole, the strong reflectivity below 100 m needs to justified by either referring to reported studies in the site's vicinity indicating the nature of these reflections and confirming their presence in the processed shot gathers of this study to show that they are likely a real feature and not a processing artifact. This paper describes a good example of using S-wave reflection seismics in a setting where classical P-wave reflection would not have resulted in high resolution images. The review criteria are assessed below. After that, suggestions for improvement of the paper are listed. The review criteria are assessed as follows: 1. Does the paper address relevant scientific questions within the scope of SE? Answer: yes 2. Does the paper present novel concepts, ideas, tools, or data? Answer: yes, application of S-wave reflection seismics in difficult setting to image 3. Are substantial conclusions reached? Answer: yes 4. Are the scientific methods and assumptions valid and clearly outlined? Answer: partly, see comments below. 5. Are the results sufficient to support the interpretations and conclusions? Answer: yes 6. Is the description of experiments and calculations sufficiently complete and precise to allow their reproduction by fellow scientists (traceability of results)? Partly, see comments below. 7. Do the authors give proper credit to related work and clearly indicate their own new/original contribution? Answer: yes 8. Does the title clearly reflect the contents of the paper? Answer: yes 9. Does the abstract provide a concise and complete summary? Answer: yes 10. Is the overall presentation well structured and clear? Answer: yes 11. Is the language fluent and precise? Answer: yes 12. Are mathematical formulae, symbols, abbreviations, and units correctly defined and used? If present: yes 13. Should any parts of the paper (text, formulae, figures, tables) be clarified, reduced, combined, or eliminated? Partly, see comments below 14. Are the number and quality of references appropriate? Answer: yes 15. Is the amount and quality of supplementary material appropriate? Not applicable, no supplementary material.
Author: Many thanks for the suggestions! Suggested improvements for the paper: Section 2.2: 1. Line 183: typo 'experimentss'

Author: done
Section 3: 2. The target depth for imaging is not stated. Please explain why the chosen setup is suitable for the target depth.
Author: Proposed target depth 36-60 m is now included.
Section 4: 3. The readability of the paper would be improved if a table with processing steps and results of those steps were provided.
Author: I agree, processing sequence Is now provided in Table 1 4. There is Love wave energy present in the seismic data. Have you considered inverting these data (MASW) in order to obtain Vs information about the first tens of meters? In the overview of section 2.2 several MASW studies are reported. In the discussion it is stated that Bodet et al. (2010)  when this section is moved up and a couple of more lines are spent on the explanation. It is too short now.
Author: I agree, it's done.
Section 5: 6. Figures 7 -10: In the text there is reference to certain positions along the lines, but the horizontal distance is not clear in the figures. There seem to be numbers like 200 250 300 in the figure, but rather hidden in the portion above the depth sections and fonts too small. Please add a clear horizontal distance axis in each figure.
Author: I agree, it's done.
7. Figure 9: boreholes BH1 and BH2 are too far away (420 m) from the line. I would not show them in figure 9, no added value. To show them in figure 8 and 10 (200 m away) is already on the limit of preferred. 420 m is really too far off.
Author: I agree, it's done. Author: I agree, it's done.
Section 6: 9. Line 505 states that the internal structures such as topsets, foresets and indications of bottom sets are present in the seismic depth sections. It helps the non-geologists reading this geophysical paper if these are indicated in the bottom parts of figures 7-10.
And it helps the geologist to recognize these in the geophysical data.
Author: I agree, it's done.
10. Line 548: use of only one 1D Vs profile. Pleas elaborate on why you think this would be a valid approximation even if Bodet et al. (2010) reported strong lateral Vs heterogeneity. Or support this by MASW results for the observed Love waves in your data.
Author: as explained above in 4., Love waves are not helpful in this case. Using a mean 1D velocity profile derived from the data itself prevents to project velocity artifacts (e.g. from irregular ray pathes) onto the depth sections. This is the most reliable time-to-depth conversion if no other velocity information is available. In fact, the strong lateral velocity heterogeneity reported by Bodet et al. (2010)  Author: I agree, I will include words about this. We also aready did some time-lapse experiments there, which are required to monitor the subsurface processes. This will be published in a upcoming paper.
12. You postulate a new combined process model (lines 689-702). What data would you need in order to further support this model? The formation of subsurface channels and loss of cations might be monitored by a combination of time-lapse ERT, IP and SP. The arid environment might pose challenges for these techniques.

Abstract.
Near-surface geophysical imaging of alluvial fan settings is a challenging task, but crucial for understating geological processes in such settings. and carbonate minerals as well as clay silicates, become increasingly exposed to unsaturated water as the sea level declines, and are consequently destabilized and mobilized by both dissolution and physical erosion in the subsurface. This new interpretation of the underlying cause of sinkhole development is supported by surface observations in nearby channel systems. Overall this study shows that shear wave seismic reflection technique is a promising method for enhanced near-surface imaging in such

Introduction
Since around 1980 until today, thousands of sinkholes have affected specific areas along the Dead Sea shoreline , Shalev et al., 2006, Abelson et al., 2017, apparently contemporaneous with the rapid decrease of the Dead Sea level (Sawarieh and Alrshdan, 2011). The sinkhole processes continuously disrupt farming areas, houses, industrial sites, and infrastructure, and, therefore, hamper the future economic development of the whole region. Geological and geophysical sinkhole studies started 70 already in the 1990's at both the western (e.g. Wachs et al., 2000) and eastern (El-Isa et al., 1995) shorelines of the Dead Sea. An early map of the main sinkhole sites was published by Yechieli et al. (2002). The sinkholes typically appear in clusters on either alluvial fans or mud flats. Arkin and Gilat (2000) defined two different classes of sinkholes: a) gravel holes on alluvial fans that consist of highly permeable gravel and sand layers including some silt, clay and evaporites; and b) mud holes on Dead Sea 75 mud flats that consist of very fine marl, silt, clay and evaporitic minerals like aragonite, gypsum and halite.
An early hypothesis postulated that clay softening, liquefaction and mobilization in the subsurface, due to the dilution of former highly salty porewater by freshwater inflows, generates the sinkholes (Arkin and Gilat, 2000). As discussed by Ezersky and Frumkin (2013), two other factors may control the location of 80 the sinkholes at the Dead Sea: (1) the presence of a thick, massive salt layer that is exposed to a dissolution front at its edge, and (2) the presence of sub-surface faults that control fresh water inflow into, and thus enable dissolution of, a salt layer. Such controls were suggested on the basis of a variety of methodical approaches (e.g. Yechieli et al., 2002, Diabat, 2005, Closson, 2005, Abelson et al., 2006, Ezersky, 2006, Frydman et al., 2008, Closson and Abou Karaki, 2009, Ezersky et al., 2010 85 Ezersky, 2013c, Ezersky and Livne, 2013, Ezersky et al., 2017. Although many geophysical studies have been carried out at the Ghor Al-Haditha sinkhole site ( Fig. 1a) in the past 24 years, the subsurface structure and the subsurface erosion (subrosion) processes are still rather uncertain. Since the year 2000, several authors have proposed that a several metre thick, massive salt layer lies at top depth of 35-40 m below the ground surface at the site, underneath the alluvial fan 90 deposits (Taqieddin et al., 2000, Legchenko et al., 2008, Frumkin et al., 2011, Ezersky et al., 2013a,b,c, Ezersky and Frumkin, 2013, Ezersky and Livne, 2013, Ezersky et al., 2017. In this shallow salt layer model, chemical erosion (dissolution) of this salt layer by fresh water flow from the eastern mountain range is supposed to generate initial cavities in the subsurface, which subsequently move upwards due to continuous solution or collapse of material at the cavity top, up to the final collapse of the ground surface.

95
In contrast, Al-Halbouni et al. (2017) presented an alternative conceptual model for Ghor Al-Haditha, based on photogrammetric surveying, historic satellite image analysis and field observations. They propose both chemical and physical subrosion of weak material, which consists both of mud flat (including evaporite lenses) and alluvial fan sediments.

100
Shear wave reflection seismic at Ghor Al Haditha -p. 4 previous refraction seismic (El-Isa et al., 1995, Sawarieh et al., 2000 and MASW (Bodet et al., 2010) profiling. Boreholes BH1 and BH2 are reported in detail by El-Isa et al. (1995), two other boreholes are 105 reported by Bodet et al. (2010) without any further information. Background is a Pleiades satellite image from 2015 combined with an aerial orthophoto mosaic from 2016.
Our reflection seismic study resolves for the first time the fine structure of subsurface layers with highresolution at Ghor Al-Haditha, and thereby contributes to reappraising the different models suggested 110 above. We further show that shear wave reflection seismics has advantages to study highly porous, partly saturated alluvial fans with complex compositions, e.g. compared to P-wave reflection and refraction, which is influenced by the pore fluids, and MASW, which is only valid in a 1D layer case. A particular methodological question of interest is to what extent shear wave seismics can be used to identify dissolution processes at depth and early stages of collapse sinkholes.

Site of investigation
Ghor Al-Haditha is a small village at the south-eastern end of the Dead Sea in the province Al-Karak of the Hashemite Kingdom of Jordan ( Fig. 1 a&b) surrounding springs is stored in some man-made pools in the area. Khalil (1992) published the geological information of the area (geologic map sheet Ar Rabba at 1:50000 125 scale). The bedrock underlying the alluvial plain is not exposed, but on structural grounds it probably comprises limestone (some dolomitic or silicified), marl, chalk, and phosporite of the Ajun and Belqa

Borehole information
Two boreholes (BH1 and BH2) were drilled at the Ghor Al-Haditha investigation site in January-February of 1995 ( Fig. 1b), and the drillings and sample analyses were firstly described in a report of El-Isa et al. (1995). Bodet et al. (2010) reported two additional boreholes (Fig. 1b) from 2006, but without 145 further descriptions and not mentioned in other reports and publications. Figure 2 shows the borehole lithology of BH1 and BH2 based on a detailed microscopic analysis of cuttings (El-Isa et al., 1995).
There, the lithology is described as an alternating sequence of sand and gravel, with a "silt and clay" bed at the bottom down to 51 m and 45 m depth, respectively. Nothing is mentioned about a massive salt layer.

150
Several papers have subsequently presented lithologic cross sections for the Ghor Al-Haditha sinkhole area, in which a thick (>2-10 m) pure salt layer is postulated to lie at between 30 m and 45 m depth below the surface (e.g., Taqieddin et al., 2000, Frumkin et al., 2011, Ezersky et al., 2013a,b,c, Ezersky et al., 2017. The abovementioned boreholes, the reported depths of which vary from one paper to another, are used to support this finding, although no detailed descriptions of the boreholes, no details about the drilling method used, and no lithologic bars are provided. Because the indication of possible massive salt layers is important for comparison with our results, we point to these inconsistencies here.  et al. (1999) and Rix and Leipski (1991), where L is the total source-receiver spread length of 120 m.

215
Results of profile 3 (magenta line 3 in Fig. 1b) show shear wave velocities of mainly less than 400 ms -1 from surface to Zmax, while profile 4 (magenta line 4 in Fig. 1b) images a high velocity layer of more than 800 ms -1 from nearly 30 m depth to Zmax, which was interpreted as the shallow salt layer. A reflection seismic analysis result of profile 4 in Ezersky and Frumkin (2013) shows no reflection response of the proposed shallow salt layer, however. Refraction tomography analysis results are reported to be of 220 insufficient resolution (Camerlynck,et al., 2012) and are later on shortly mentioned in Ezersky et al.

230
Shear wave reflection seismic at Ghor Al Haditha -p. 8 To shed more light on these different models and to provide an independent database for the detailed mapping of the postulated salt layer, its morphology, and synclinal structures expected at its top, we started high-resolution shear wave reflection seismic surveying in 2013 at Ghor Al-Haditha. The target depth of the survey was initially focused to the proposed salt layer depth of 35-60 m and in maximum to 100 m depth.

Shear wave reflection seismic equipment and survey
A shear wave landstreamer (Inazaki, 2004, Pugin et al. 2004 consisting of 95 transverse horizontal (SH) geophones (10 Hz resonance frequency) in 1 m intervals was used as receiver unit, connected to a Geode (Geometrics Inc., 4 units of 24 channels each) recording system. Transverse horizontal (SH) waves were generated by the seismic micro-vibrator source ELVIS (Polom et al. 2011, 240 Krawczyk et al. 2012, Polom et al. 2013. Figure 3 shows the equipment in operation on site. The small size (nearly 1.5 m 3 ) and weight (nearly 600 kg) of the whole equipment enabled air cargo transportation to Jordan.  The source signal (sweep) was set to 20-120 Hz (20-80 Hz during the survey extension in 2014) linear frequency modulated of 10 s duration (Crawford et al., 1960). Data recording was set to 12 s duration and stored uncorrelated to enable processing of uncorrelated data later on, if required. In the field, vibroseis Shear wave reflection seismic at Ghor Al Haditha -p. 9 correlation processing was applied for immediate quality control. After initial tests on site, the source interval was set to 2 m to increase the statistical redundancy due to challenging, disturbed subsurface 255 conditions, noticeable by strong wave field scattering in the recordings. Typically, two records were gained at each source location by alternating the source polarity and stored separately. Only during times of stronger wind disturbances the number of records per location was increased to four. In the 2013 campaign, a total of 2011 records (9.34 Gb data volume) at 898 source locations were gained along 1.92 km of profile length. In 2014, 2000 records (9.14 Gb) at 1144 source locations were recorded along 2.1 260 km of profile length (see profile locations in Fig. 4). A variable split-spread source-receiver configuration (Polom et al. 2016) was applied to enable geometry optimization for the detection of dipping structures, and to facilitate workarounds due to obstacles in the profile track. To reduce the proportion of Love surface waves during recording, profiles were carried out 270 mostly either on asphalt paved roads or on dirt roads covered by compacted gravel, so that high shear wave velocities at the surface disable the excitation of Love surface waves. In 2013, geodetic surveying of the profile tracks was performed by using a handheld GPS system. Without differential GPS corrections, the final positioning of the profiles from the 2013 campaign required laborious optimizations by manual corrections based on the known distances along the receiver units. The horizontal accuracy 275 was improved to less than 5 cm for the profiling, and 0.5-1 m for absolute positioning. Reliable elevation Shear wave reflection seismic at Ghor Al Haditha -p. 10 data could not be restored. During the 2014 campaign, the positioning method was improved by using a Differential GPS system, leading to precise elevation data of 10 cm horizontal and 15 cm vertical error.
All seismic data recorded were checked and pre-processed in the evening of each recording day by using the VISTA 10.028 (GEDCO Inc., Calgary, CA) seismic data processing software on a notebook computer 280 (DELL Precision M65) for quality-control purposes and a first interpretation. Detailed inspections of correlated raw records showed strong Love wave scattering on profiles 1 and 3 that indicated a low velocity layer close to the surface, even though the dirt road construction at profile 3 was modified some years ago by gravel infill after a heavy damage caused by sinkholes and related subsidence.
Selected record examples along profile 1b illustrate the typical range of signal propagation responses at 285 the Ghor Al-Haditha site (Fig. 5).

295
In the Northeast sector of the profile, close to the main sinkhole area, a typical scattering of the wave propagation is visible. The wave propagation behaviour is mostly asymmetric regarding the receiver distance to the source position, which indicates heterogeneous subsurface structures. In the middle sector, even though an asphalt surface pavement of apparently continuous integrity, Love wave propagation was Shear wave reflection seismic at Ghor Al Haditha -p. 11 partly indicated (note area of Love wave reverberations marked on Fig. 5), probably caused by poor road construction. In the Southwest sector, clear reflection events were detected already in the single recordings. Along profile 2b-2 ( Fig. 6) selected record examples show a better data quality compared to profile 1b: flat first breaks and a range of clear reflection events occur. This is surprising, because the road quality along profile 2b-2 did not differ to that of the southwest sector of profile 1, and both roads 305 are of obviously similar age.

Data processing
Reflection imaging was carried out by following a general processing sequence described by, e.g., Krawczyk et al. (2012) and Polom et al. (2013). The main data processing flow of the first iteration (FD) Migration. Subsequently, depth conversion was applied by using a mean 1D velocity function.
Background of data processing applied is reported in Aki and Richards (1980) and Yilmaz (2001).

320
Since Love wave reverberation patterns are a widely observed disturbance effect in the whole data set (c.f., Fig. 5), elimination of such patterns was one of the main processing steps undertaken to enhance the desired reflection response. This required several iterative loops during the main processing sequence to improve the final result. The main data processing was carried out in two iteration sequences mainly by using VISTA 10.028 (GEDCO Inc., Calgary, CA) seismic data processing software. The first sequence 325 was carried out to establish the main database and to extract the main structures in the data. In the second iteration sequence, the time section results were used to improve the processing flow in detail towards the specific reduction of imaging artefacts and to stabilize the processing velocities for later use during depth conversion. It included reduction of harmonic distortions in the near source area and Vibroseis Spectral Balancing (Pugin et al. 2009) in a Vibrogram transformation domain (Polom 1997). Furthermore, the 330 results of the velocity analysis were improved. To derive final depth sections a mean 1D RMS velocitytime function (shown in Fig. 8) for all profiles was derived from the reflection seismic data set and applied to all profiles. Due to the lack of any reference velocity-depth function e.g. from VSP logging, the 1D solution was chosen as first depth approximation to prevent the projection of lateral velocity irregularities (e.g. due to irregular wave ray paths close to sinkhole affected areas) onto the depth 335 sections, which would result in structure imaging distortions. The detailed processing sequence is listed in Tab. 1. wave propagation along irregular, non-straight ray paths (due to the disturbed shallow subsurface structure) with more regular straight ray path responses from deeper levels (later travel times).

Resulting depth sections and structure correlations
The main NE-SW-trending profile 1b has transparent and strongly layered segments between ground surface and 200 m depth (Fig. 8).
Shear wave reflection seismic at Ghor Al Haditha -p. 14 seismically-sensitive material properties (elastic parameters and density) and hence in shear wave seismic velocity.
The main area of sinkhole activity is located immediately northwest of profile 1 (see Fig.4). This activity 385 has strongly affected the road construction along this part of profile 1b, northeast of its intersection with profile 3 (Fig. 8). A depression with several fractures is visible here along the road surface (Fig. 4), but without any sinkhole activity until yet. In detail, this part of the road crosses the south-eastern limit of a major sinuous depression that runs down to the former Dead Sea shore and that hosts the smaller scale sinkholes (for details see Al-Halbouni et al. (2017)

405
The combined depth sections of profiles 2 and 2b-2 are shown in Fig. 9. The profiles were both acquired on asphalt road. Also shown in Fig. 9 are the simplified lithological bars of boreholes BH1 and BH2 as projected perpendicularly into the profile plane from their true locations nearly 200 m to the SW (c.f., El-Isa et al., 1995). Due to this relatively large projection distance, the lithology bars may not precisely reflect the lithology in the profile plane. The sections show a nearly horizontal layering down to 50 m 410 depth below the reference level of -367 m a.s.l. in the SE part of the combined profile, which changes laterally to less well organized but mainly NW-dipping structures in the NW part of the combined profile.
Shear wave reflection seismic at Ghor Al Haditha -p. 16  Shear wave reflection seismic at Ghor Al Haditha -p. 19 The depth section of profile 4 (Fig. 11)   Nevertheless, the results of the profiles 2 and 4 both show main structural dips towards NW, which fit the dip tendency of the lithology detected in the boreholes (Fig. 2). Blue markers included along the borehole shafts in Fig. 12 determine the top of the "silt and clay" layer below the alluvium sequence consisting of gravel and sand. Since this "silt and clay" layer is the only one detected in the lithology bars of the 490 boreholes that is prone to act as an aquiclude sensitive to subrosion, and it also fits the general structural dip, it was obvious that this layer probably extends across the whole area. To interpolate the top of this Shear wave reflection seismic at Ghor Al Haditha -p. 20 layer in space (light blue colour in Fig. 8-11), it was traced from borehole BH1 along profile 1b and profile 1, from where it was further continued to profiles 2 and 2b-2, profiles 3 and 3b-2, and profile 4, with respect to the perpendicular projection of borehole BH2 into the profiles 2 and 4. Subsequently, 495 these layer continuations along the profiles were triangularly interpolated in space to get an impression of the probable layer extent and topography. Regarding the lithology of boreholes BH1 and BH2, the  2017)). It therefore represents a different subrosion level to the present 505 mud-flat surface, which indicates that alluvium and "silt and clay" sequences may lie below the presently exposed mud-flat.

Discussion
The shear wave reflection seismic survey acquired at the Ghor Al-Haditha sinkhole site in 2013 and 2014 was the first comprehensive shallow seismic investigation of the sinkhole phenomena area since the 510 refraction seismic investigation campaigns carried out by El-Isa et al. (1995) and Sawarieh et al. (2000), respectively. The results presented here show the advantageous imaging capabilities of the method regarding high-resolution structural analysis and depth penetration compared to common refraction seismic methods or common P-wave reflection analysis. In this section, we discuss firstly the general outcomes of our application of the S-wave reflection method to an alluvial fan setting and secondly the 515 implications for the nature of processes leading to sinkhole development at the Ghor Al-Haditha site.

1 The application of the shear wave reflection seismic technique to alluvial fan setting
The S-wave reflection method is especially advantageous in the area close to the Dead Sea border, because wave propagation is restricted to the matrix only, and so it is not affected by the pore space content, whether that is air in the shallow unsaturated zone or fresh-or salt water in the deeper parts 520 below the ground water level. Therefore the groundwater level itself does not act as a physical interface during wave propagation, as it is in general the case for common P-wave methods. In the S-wave profiles acquired in the area, no influence and no response of the ground water level to the wave propagation was detected. Furthermore, wave propagation velocities of S-waves (Vs) are in general significantly smaller than those of P-waves (Vp). The Vp/Vs-ratio ranges from nearly 1.7 for an ideal elastic medium (e.g. 525 perfect consolidated rock) to more than 10 (Yilmaz, 2015) for unconsolidated sediments (e.g. Holocene alluvium and soft clays). This leads to a significantly improved resolution when using S-waves, Shear wave reflection seismic at Ghor Al Haditha -p. 21 depending on the grade of matrix consolidation. In the case of the Ghor Al-Haditha site, this results into a resolution improvement factor of 8-10 below the ground water level, when one postulates a similar seismic signal frequency bandwidth for P-and S-waves.

530
The success of the application to the alluvial fan setting at Ghor al-Haditha is seen in the well-resolved shallow subsurface structure imaged in the seismic profiles, which show in general mainly NW dipping (i.e. lake-ward-dipping) reflectors typical of a prograding alluvial fan sequence into a lacustrine environment. The internal structure is complex and includes typical fan-delta elements such as topsets, foresets, and indications of bottomsets. Especially profile 2 in Fig. 9 shows typical dip structures of a 535 Gilbert-type delta (Gilbert, 1885). Intercalated, more-horizontal structures indicate lacustrine deposit layers. The topset structures, the intercalated lacustrine deposit layer and the dip directions fit the lithology of both boreholes BH1 and BH2, which show NW-dipping alluvial sand and gravel above an obviously lacustrine-type deposit denoted as "silt and clay" (Fig. 2). Below the blue line ( Fig. 8-11 Since wave propagation is controlled by the matrix only, effects of grain size coupling play a key role in the wave propagation of S-waves in contrast to P-waves. In contrast to P-wave velocity, which usually 550 increases with depth in the case of a fluid-saturated pore space, S-wave velocities often decrease with depth, e.g., if the grain cementation reduces or the pore pressure increases with depth, respectively, as both factors result in reduced grain contacts. A further reason for velocity function decrease with depth is the influence of a high-velocity layer at the surface, which is the case operating on an asphalt road surface. This effect is visible in the 1D velocity-depth function in Fig. 8 in the range 0-10 m below 555 ground surface. Diminished grain coupling can also arise from mechanical damage, e.g. by fracturing, subsidence and subrosion, which leads to breaks in the direct (geometrical) wave paths. In contrast to Pwaves, where such breaks may be short cut due to the wave propagation through the pore fluids, decoupling between grains causes additional paths for S-waves, leading to enhanced wave energy scattering and an apparent velocity reduction.

560
In the Ghor Al-Haditha data, enhanced wave energy scattering was observed in areas close to sinkhole activity, leading to transparent zones of weak reflectivity in the seismic sections (e.g., Fig. 8, 10). In the northeast part of profile 1b in Fig. 8, the zone of scattered reflections is imaged up to 60 m depth below the reference level of -367 m a.s.l. (i.e. nearly 50 m below the ground surface), indicating the depth range of the destabilized alluvium. Profile 3 in Fig. 10 shows the continuation of this low reflectivity zone to 70 565 m depth below the reference level (i.e. nearly 50 m below the ground surface) in the centre of that profile.
Beside the transparent zones of strong wave field scattering, buried syncline structures imaged in profiles 2b-2 ( Fig. 9), 3b-2 (Fig. 10) and profile 4 (Fig. 11) are in the depth range of 30-60 m below ground surface. These synclines probably indicate ancient or recent subsidence by sinkhole activity caused by subrosion below the syncline centres at different horizons. The structures are targets for time-lapse 570 monitoring by shear wave reflection seismics in the upcoming years to evaluate changes in the reflectivity response in detail.
The effect of apparent velocity reduction was also observed, leading partly to irregular interval velocities values less than zero. Such velocity irregularities can be caused by curved ray paths instead of straight rays which are the base of the CMP concept in reflection seismic. Subrosion affected zones (e.g. cavities) 575 in the subsurface may cause the shear waves to propagate around in stiffer material instead passing them along a straight path. Whereas this velocity problem is of minor importance for the CMP stacking, processing the velocity functions of the individual profiles required careful handling during the time-todepth conversion, where such zones of irregular velocities would locally compress the depth sections, leading to a distorted structure imaging. Since the discrepancies could not be eliminated due to missing 580 additional velocity information, e.g. by well logging, the final depth conversion was carried out by using a stabilized 1D velocity function for all profiles (Fig. 8), derived as mean function from profile parts with sufficient reflection responses. Therefore, the resulting depths have to be handled with an estimated error range of up to 20%.
As well known in hydrocarbon exploration and shallow seismic operations using the seismic refraction, size derived from the boreholes are in principle poor, and no information is available below the borehole bottoms due to the missing well logging. The alluvial fill rate can be estimated relatively by the borehole lithology profiles and by the seismic structure dips, but must be handled with care since geochronological data are limited and the area is close to an active, main sinistral strike-slip transform fault. It is to be expected beside the lake level variations that tectonic overprints may have changed the whole 595 sedimentation structure over time, even though there was no main fault structure detected in the seismic data.
In summary, the S-wave reflection seismic method achieved advantageous high-resolution imaging results of the alluvial-lacustrine deposit structure at the Ghor Al-Haditha sinkhole site, which are superior to common near-surface seismic investigation methods. The favourable velocity-frequency relationship of 600 the resulting wavelets and the absence of pore fluid effects enabled a meter-scale resolution and a nearly 200 m penetration depth. Furthermore, and unlike the refraction method, the shear wave reflection method is independent of an obligatory increasing velocity-depth function. In the case of the sinkhole affected subsurface structure at Ghor Al-Haditha, where strong vertical and lateral subsurface Shear wave reflection seismic at Ghor Al Haditha -p. 23 inhomogeneity occurs, it is not free from shortcomings, and, similar to other geophysical methods, it 605 requires borehole calibration to verify precise depth imaging, especially.

2 Implications of the shear wave reflection seismic results for sinkhole formation
Combining the reflection amplitude responses and the detected borehole lithologies of BH1 and BH2 ( Fig. 8-11 The previously-derived MASW results at the Ghor Al-Haditha site (Bodet et al., 2010, Keydar et al., 2011, Ezersky et al., 2013b, Ezersky et al., 2017 also show Vs of 200-400 ms -1 from surface to nearly 50 m depth (magenta profile 3 in Fig. 1) and to nearly 30 m depth (magenta profile 4 in Fig. 1 be handled with precaution. The strong lateral Vs inhomogeneity reported (Bodet et al., 2010) is additionally in contrast to the required 1D layer structure along the spread; such inhomogeneity is well 640 known to cause impaired inversion processing results for the MASW method (Forbriger, 2003).
Shear wave reflection seismic at Ghor Al Haditha -p. 24 Due to the significant difference in propagation velocities of P-and S-waves, the S-wave velocity results in this study are not comparable to the P-wave refraction velocity results of the previous studies of El-Isa et al. (1995) and Sawarieh et al. (2000). El-Isa et al. (1995) detected a deepest refractor layer of Vp max.
2500-3300 ms -1 at 40-50 m depth in nearly all of their profiles. In the profiling tracks which were repeated by Sawarieh et al. (2000) using a similar acquisition configuration, max. P-wave refraction velocities Vp of 2500 ms -1 and 3130 ms -1 were detected for similar depths, but predominantly the detected velocities were around 2150 ms -1 in nearly 40-50 m depth. In only one of 24 profiles (No. 5, Fig. 1) of Sawarieh et al. (2000), in the northwest outside the study area of El-Isa et al. (1995) and close to profile 3 of our study, Sawarieh et al. (2000) detected a P-wave velocity of 3948 ms -1 in 70 m depth, which was 650 interpreted as a salt diapir. Directly beside their profile 5, in their profile 4 (Fig 1), Sawarieh et al. (2000) detected a maximum velocity of 2245 ms -1 in 40 m depth. The area of profile 4 was subsequently affected since 2000 by massive sinkhole activity and subsidence, which prevented further investigation, whereas the area of profile 5 remained unaffected until today (see Fig. 1 and Fig. 4). Shear wave reflection seismic at Ghor Al Haditha -p. 25 The most significant horizon in both boreholes for which documentation exists is the "silt and clay" layer 680 with its top below ground surface at 43 m in BH1 and 49 m in BH2, respectively, because this is the one detected in the boreholes that is most prone to subrosion. The drilling of boreholes BH1 and BH2 both had been stopped after nearly 2 m within this layer because it was interpreted as base of the sand and gravel sequence, and because it was supposed to act as aquiclude regarding hydrogeological aspects (El-Isa et al., 1995). Profile 1b (Fig. 8) shows a change of the main reflection pattern at the location of 685 borehole BH1 from nearly the top of this layer (ca. -420 m a.s.l., 50 m below surface) to 110 m below reference depth (ca. -470 m a.s.l., 100 m below surface). So it is obvious that the thickness of this layer is either more than 2 m, or that this layer is part of a stacked sequence including several "silt and clay" layers below. Since the depth range of the "silt and clay" layer was always below the Dead Sea level, the sedimentation of this layer occurred in the lacustrine evaporite-rich environment of the Dead Sea during 690 the last 10 ka BP (Bookman et al., 2004). In analogy to exposures of lacustrine deposits visible in the sides of 1-8 m deep freshwater channels that were carved recently into today's mud plain (e.g. Al-Halbouni et al., 2017) there may be also thin evaporite layers (<1 m thick) embedded in this "silt and clay" sequence below the alluvial plain.
The mineralogy of the clay material in the "silt and clay" layer in BH1 and BH2 was not determined by 695 El-Isa et al. (1995). The clay material was only described to be of green colour, this would be an indication for illite (also called French clay), which is typically described to be of dark olive-grey colour and contains portions of potassium and water. Sawarieh et al. (2000)

710
This solution process may be in places amplified by a physical effect well known for the so-called "quick clays" in Nordic countries (Geertsema, 2013). "Quick clays" are originally deposited in a salt-rich marine environment in the Northern hemisphere during glaciation where they formed an electrostatically bonded double-layer structure including a cation (e.g. sodium) between two clay particles. When these clays become no longer subjected to salt water conditions (due to isostatic uplifts in Nordic countries, due to 715 decrease of the Dead Sea level in our study area) and fresh water infiltrates these clays washing away the cations, the clay particles do not remain in a stable bonded structure and change to a liquid behaviour.
Such a process was previously proposed by Arkin and Gilat (2000) for the Dead Sea sinkholes, although Shear wave reflection seismic at Ghor Al Haditha -p. 26 the relatively low clay content of up to 20% detected in the fine material sediments around the Dead Sea (e.g. Khlaifat et al. 2010) indicates that such a process accounts for only a minor portion of the subrosion 720 process.
In line with Krawczyk et al. (2015), we therefore propose a new, combined process model based on both chemical and mechanical erosion: the fine material stacks of marl, silt and clay within the alluvial fan change progressively from solid to liquid behaviour in the contact area of the fresh water inflow due to Thirdly, even alluvial material at the subrosion interface gets washed out. The cavities extend horizontally 730 and grow upward until the gravitation force of overlying alluvial stratum exceeds its bonding forces. This initializes a sudden sinkhole. The process is controlled by the long-term, seasonal and ephemeral movement of the salt-fresh water interface (Salameh andEl-Naser, 2000, Alrshdan, 2012), the volume of the fresh water flow and its velocity, the volume of buried soluble and/or mechanically erodible material and its mechanical properties within the alluvial fan.

Conclusions
The shear wave reflection seismic study at the Ghor Al-Haditha sinkhole investigation site was the first experiment where this geophysical investigation method was applied in the environment of the Dead Sea.

750
Based on the shear wave reflection seismic results, and supported by the drilled stratigraphy, our new model for the sinkhole process at the Ghor Al-Haditha sinkhole site suggests that the dissolution of small, distributed inclusions of marl, silt, clay, and evaporites within the alluvial fan is enhanced by freshwater intrusions from the eastern wadis. Furthermore, we propose that such chemical erosion in the subsurface exists in a feedback with mechanical erosion of weak material, especially silt and clay, and with higher flow velocities of poorly consolidated alluvial material also. The process is not restricted to the depth of the "silt and clay" layer detected in the boreholes BH1 and BH2 and outlined in the area, it can also affect shallower or deeper layer-like inclusions of weak, soluble, non-massive lacustrine material.
Consequently, geophysics-based mapping of areas prone to sinkhole hazard at the Dead Sea should consider different lithological controls on the location of sinkhole development.

760
Areas of future sinkhole development may be indicated by time-lapse monitoring using shear wave reflection seismic. The observation of time dependent structure changes in the subsurface and changes of irregular velocity zones caused by disaggregation or cavity formation recommends as a tool to indicate ongoing sinkhole development. This aspect will be the focus for investigations in upcoming experiments.
Competing Interests: The authors declare that they have no conflict of interests.   previous refraction seismic (El-Isa et al., 1995, Sawarieh et al., 2000 and MASW (Bodet et al., 2010) profiling. Boreholes BH1 and BH2 are reported in detail by El-Isa et al. (1995), two other boreholes are reported by Bodet et al. (2010)         subsidence. This may cause the weaker reflection signatures in the upper 50 m if compared to profile 2 ( Fig. 8). The main reflection pattern signature is similar to profile 2 and 2b-2, showing NW dipping events. In the SE of the profile a V-shaped structure is visible close to surface, filled with nearly Shear wave reflection seismic at Ghor Al Haditha -p. 36 horizontal reflection events, which is interpreted as a refilled channel side-cut. Due to the wide subsidence area NE of profile 3 the top of the "silt and clay" layer (blue line) was continued to NW 1095 starting from profile 1b instead using the projection of BH2.