Sinkholes and uvalas in evaporite karst: spatio-temporal development with links to base-level fall on the eastern shore of the Dead Sea

. Enclosed topographic depressions are characteristic of karst landscapes on Earth. The developmental relationship between depression types, such as sinkholes (dolines) and uvalas

The main problem for unravelling the spatio-temporal relationships between uvala and sinkhole (dolines) development in limestone karst areas is that the landform evolution is controlled by the relatively slow dissolution kinetics of carbonate minerals. Consequently, the development of these landform types is not directly observable in such areas. Furthermore, the slow rate of karstic processes means that limestone 60 areas are susceptible also to long-term geomorphic influences from changing climate and active tectonics.
Indeed, the areas in which such uvalas and dolines have been best documented occur in tectonically complex settings in which climate has varied considerably over time, such that many landforms have been modified not only by karst processes but also by fluvial and/or glacial processes.

65
An opportunity to shed new light on the geometric and genetic relationships between sinkholes and uvalas has arisen with their rapid development in an evaporite karst setting over the last 35 years at the margins of the hypersaline Dead Sea (Arkin and Gilat, 2000;Taqieddin et al., 2000;Yechieli et al., 2006;Avni et al., 2016;Abelson et al., 2017). The sinkholes, which have been the focus of previous work, now number in the several thousands. They represent a substantial geohazard in the Dead Sea region, and they 70 have already destroyed or damaged several tourism facilities, factories, evaporation pond dykes, highways, link roads, houses and farmland. Although the occurrence of uvala-like depressions has also been documented in areas of evaporite karst such as the Dead Sea (Al-Halbouni et al., 2017;Avni et al., 2016;Baer et al., 2002;Closson, 2005;Frumkin, 2013), the central uplands of Turkey (Doǧan, 2005;Doğan and Özel, 2005;Waltham, 2015) and areas of Saudi Arabia (Gutiérrez and Cooper, 2013;Youssef 75 et al., 2015), they have not been studied as extensively as their carbonate equivalents.
The Dead Sea level represents the regional hydrological base-level, and its largely anthropogenicallyforced decline at a gradually increasing rate since the late 1960s (Lensky et al., 2005) has been linked with the sinkhole formation. In theory, the base-level fall should cause a seaward shift of the 'fresh-saline 80 interface' developed between the hypersaline Dead Sea brine and less saline (i.e. relatively 'fresh') groundwater (Salameh and El-Naser, 2000;Yechieli, 2000;Yechieli et al., 2009). Such a shift should enable groundwater under-saturated with respect to halite and other evaporitic minerals to infiltrate the evaporite deposits in the subsurface, thus triggering karstification and surface subsidence. The location of karstification and sinkhole formation is controlled by the intersection of the 'fresh-saline interface' 85 with the shoreward edge of evaporite deposits in the sub-surface. A prediction of this theory is that new sinkhole development should shift seaward with time also. Evidence for such a shift on the well-studied western Dead Sea shore is somewhat patchy (Abelson et al., 2017;Avni et al., 2016), however, and it has been regarded by some authors as unconvincing (Charrach, 2018). Preferential flow of relatively fresh groundwater into evaporite-rich deposits along conductive regional tectonic faults has been proposed as 90 an alternative control on the location of sinkhole (and uvala) development (Abelson et al., 2003;Charrach, 2018;Closson, 2005;Shalev et al., 2006).
In this paper, we provide a first detailed documentation of the spatio-temporal evolution of sinkholes and uvalas on the eastern shore of the Dead Sea, at Ghor Al-Haditha in Jordan. Our aims are to discern spatiotemporal inter-relationship of these two types of karstic depression, and to examine how their 95 development relates to base-level fall. Our approach combines remote sensing data spanning the 50 year duration of base-level fall from 1967-2017 with close-range photogrammetric surveys and field observations made in 2014-2017. Our results show the most detailed insights to date into the spatiotemporal development of sinkholes and uvalas in evaporite karst settings, and the clearest yet illustration the effect of base-level fall on that development. Our study's overall aim is to contribute these fresh 100 insights to the ongoing debate about the nature and origin of uvalas and about their relationship to the sinkholes within them.

Tectonic and hydro-geological framework
The Dead Sea is the hyper-saline terminal lake of the Jordan River (Figure 1A), and it lies within the ~150 km long and ~8-15 km wide Dead Sea basin (Garfunkel and Ben-Avraham, 1996). The basin lies at 105 a left step (or bend) along the left-lateral Dead Sea Transform fault system. Maximum tectonic subsidence is ~8.5 km around the Lisan peninsula, adjacent to our study area (Ten Brink and Flores, 2012). The basin has subsided rapidly from the late Pliocene to present (Ten Brink and Flores, 2012), and in that time has hosted several palaeo-lakes of varying size and duration (Bartov et al., 2002;Torfstein et al., 2009). With respect to modern global mean sea level (msl), a high-stand of -162 m msl was reached at around 25 ka 110 ago, during the 'Lisan Lake' episode. The modern Dead Sea initiated after a major low-stand at around 10 ka (Bartov et al., 2002). As its level has declined from -395 m msl to -431 m msl , the modern lake has divided into northern and southern parts; the latter is now occupied entirely by industrial salt evaporation ponds. The base level fell at a rate of 0.5 m yr -1 in the 1970's, and at a rate of 1.1 m yr -1 in the last decade. In absolute terms, the lake level has declined by 37 m as of 2017 and is forecast to drop 115 a further 25-70 m by 2100 (Asmar and Ergenzinger, 2002;Yechieli and Gavrieli, 1998).
The Ghor Al-Haditha study area, which is about 25 km 2 in size, is situated on the southeast shore of the northern part of the Dead Sea ( Figure 1A). The area lies in a zone of tectonic complexity at the eastern basin margin, where subsidence is relayed between several major tectonic structures. From south to north, 120 these are: the N24-trending Wadi Araba fault; the N0-trending Ghor Safi fault; the Ed-Dhira monoclinal flexure; and the N80-trending Siwaqa fault. Further north again, and marking the north-east boundary of the study area, is a prominent N0-trending escarpment ( Figure 1B). This probably reflects the orientation of another major basin-bounding fault (Khalil, 1992), here termed the Eastern Boundary Fault ( Figure   1C; cf. Meqbel et al., 2013), although the exact location of the fault trace is unclear. Three major wadi 125 (dry river valley) systems, Wadi Ibn Hammad, Wadi Mutayl and Wadi al Mazra'a (the latter lies just outside the study area to the southwest) drain the uplands to the east and southeast. These wadi systems have formed an alluvial fan plain at elevations between -360m and -380m in the south of the study area ( Figure 1B). Smaller alluvial fans occur sporadically along the coastline in the central and northern parts of the study area. West and north of the alluvial plain, exposure of the former lakebed by the ongoing 130 recession of the Dead Sea has formed a 'mud-flat' or 'salt-flat'. At the transition between the two lithologies, numerous artesian springs are found, which feed surface streams that drain into the Dead Sea via numerous channels. Other surface stream channels in the study area lack these groundwater-fed springs and are instead fed by water drained from the various wadi systems during flood events.

135
The geology of the Ghor Al-Haditha study area ( Figure 1C) comprises folded and faulted sequences of siliciclastic or carbonate rocks, which are locally overlain by semi-consolidated to unconsolidated lacustrine or alluvial deposits (Khalil, 1992). Hydro-geologically, there are three principal aquifer units: (1) a lower sandstone aquifer comprising the Ram group and Kurnub formation of Cambrian to early Cretaceous ages, respectively; (2) an upper carbonate aquifer spanning the Ajlun and Belqa groups of late 140 Cretaceous to early Tertiary age; and (3) a superficial aquifer in the Lisan formation of Plio-Pleistocene age (Khalil, 1992). The Lisan formation comprises both alluvial and lacustrine deposits. The alluvium consists of poorly-sorted, semi-consolidated to unconsolidated sands and gravels interbedded with minor silts and clays (El-Isa et al., 1995;Sawarieh et al., 2000). Similar, stratigraphically younger, but unconsolidated alluvial deposits are probably equivalent to the Ze'elim formation of Holocene age (Abou-145 Karaki et al., 2016). The Lisan and Ze'elim formations also comprise lacustrine deposits, some of which are exposed on the former Dead Sea bed. These comprise laminated to thinly bedded layers of marl, clay, evaporites and silt interbedded with a spatially variable proportion of distributed lenses or layers of evaporites such as halite, aragonite and gypsum (Arkin and Gilat, 2000;Polom et al., 2018). Similar lacustrine deposits likely extend in under the alluvial fan plain (Polom et al., 2018). 150

Data and Methods 165
Our data set includes high resolution optical satellite imagery and aerial survey photographs covering the 50-year period from 1967-2017 (Table 1). We orthorectified and pansharpened the satellite imagery by using standard algorithms and workflows in the PCI Geomatica software package. All data were integrated and analysed within a Geographic Information System (GIS) software package (Q-GIS). The number and extent of remotely-sensed sinkholes represent minima, as local farmers have filled in sinkholes to mitigate disruption to their work. Therefore, we also include information from sources that undertook earlier field surveys in communication with local farmers (El-Isa et al., 1995;195 Sawarieh et al., 2000;Closson and Abou-Karaki, 2009

Base level fall, shoreline retreat and bathymetry
The Dead Sea level drop has resulted in a dramatic retreat of the shoreline in the Ghor Al-Haditha area 215 ( Figure 2A). As of 2017, the shoreline had retreated from its 1967 position by a minimum of 0.3 km in the north of the study area and by a maximum of 2.5 km in the south. The rate of retreat in the southern part of the study area accelerated from < 10 m yr -1 between 1967-1980 to an average rate of ~45 m yr -1 between 2000-2017 ( Figures 2B, C). In the north of the area, the rate of retreat has been a steadier of about 7-8 m yr -1 . The pre-recession bathymetry was steepest in northern part of the study area (Figure 3

225
The rate of shoreline retreat is correlated non-linearly with the former bathymetric slope ( Figure 2D).

Widespread subsidence of the former lakebed
Comparison of the reconstructed lake bathymetry with the 2016 DSM reveal several substantial elevation changes since lake recession (Figure 3). The negative differences in elevation reveals a pattern of 245 subsidence of 0 -6m across the former lakebed in the southern part of the study area over distances on the kilometre scale (profiles A and B). This wide-scale subsidence consistently diminishes seaward to the position of the 2017 shoreline, where elevation difference tends to zero across the study area. The southeast end of Profile A also captures more localised subsidence due to development of an uvala-like depression, of which a number of similar zones of subsidence can be seen across the central part of the 250 study area (for details, see section 4.4). Subsidence of 0 -3m occurs also in the areas of exposed lakebed in the north of the study area, as outlined by the elevation difference along profile C, although these are on a smaller spatial scale in keeping with the smaller shoreline retreat there.
Positive elevation changes occur near the 1967 shoreline in the southern and northern parts of the study 255 area. Those in the south coincide with the active alluvial fan at the mouth of the Wadi Mutayl, as well as with areas of vegetation growth (trees and bushes) and/or anthropogenic activates (e.g. earthworks at the former Numeria Mud Factory site). Large positive elevation differences in the north coincide with the main north-south highway, which was constructed in the early 1990s.

Sinkhole development and morphology
Sinkhole formation began at Ghor Al-Haditha in the mid-1980s in the southern part of the study area ( Figure 4). Initiation of new sinkhole development subsequently shifted north-northeast-ward, roughly parallel to the coastline. In detail, the sinkholes have initiated in clusters, with gaps between earlier 275 clusters filled or reduced as new sinkholes and new sinkhole clusters form. The most active area is now adjacent the Dead Sea highway in the northern part of the study area.
After initiation in a given sub-area, further new sinkhole development within that sub-area has generally   Figure  5C). The azimuths of the long axes of all sinkholes show a general E -W alignment (Figure 5D), which is broadly parallel to the average aspect of the slope for the study area. 300

Uvala development and morphology
The uvalas in the study area are gentle depressions of several hundred metres in lateral extent of nonuniform shape that enclose numerous sinkholes (Figures 4, 6 & 7). These uvalas have De/Di ratios of 320 0.016 -0.042, calculated using the method suggested by Ćalić (2011)  material-dependent. In alluvium, subsidence-related displacements are accommodated on fewer but larger fractures, whereas in mud-rich lacustrine deposits such displacements are accommodated on more numerous but smaller fractures. As shown below, these fracture systems are spatially and temporally associated with subsidence of each uvala. They are not to be confused with regional tectonic structures.

330
Development of each uvala appears to follow precursory sinkhole formation at that site (prior to the first remotely sensed ground cracking associated with the uvala). About 2 -8 years after the first sinkhole sighting (in which time many new sinkholes have generally clustered about the initial holes), ground cracks develop that no longer trend concentrically to any single sinkhole, but instead delineate a wider zone of subsidence that envelopes several sinkholes or even several clusters of sinkholes. The first uvala, 335 U1, developed between 1992 and 1999 in the south of the area, near the Wadi Ibn Hamad (Figure 4) After initiation, uvala growth is closely linked with further sinkhole formation within it. For instance, groundcracks related to U2 initiated around two spatially-discrete sinkhole clusters; these fractures sets 345 propagated and joined as sinkhole development migrated seaward (Figure 6). For both U3 and U4, two 'prongs' of coevally-migrating ground crack and sinkhole development are visible (Figures 6 and 7).
Again, the direction of uvala growth has generally been seaward. The end of uvala growth is also linked with the end of sinkhole development; this is exemplified by uvala U1, which ceased development by 2006, in tandem with cessation of sinkhole activity nearby. 350

when first visible in imagery) are grouped and coloured in four-year intervals for U2 and in two-year intervals for U4. Each uvala is linked morphologically to a highly
355 active stream that emerges on the seaward side at several meters below the surrounding ground surface.

Links between subsurface stream flow and the formation of sinkholes and uvalas
Several features of the uvalas and the formation of sinkholes within them strongly suggest a link between their development and the channelized flow of relatively fresh groundwater in the subsurface. The best example of such links is seen at uvala U2, close to the former Numeira Mud Factory site. The history of this uvala occurs in close association with the development of the system of stream channels which drain 370 the superficial aquifer within the deposits of the Wadi Ibn Hamad. This stream channel system has localised from several small channels to one 'main channel' over time. This main channel formed late in the system evolution and, unusually, it developed by rapid retrogressive (upslope, headward) erosion (Al-Halbouni et al., 2017). This new channel developed from a spring, the exact location of which has changed several times since its inception, which emerged in the middle of the saline mudflats and in association 375 with drainage of a lake hosted in U2 (Figure 6; c.f. Fig. 16, Al-Halbouni et al., 2017). The migration patterns of sinkholes within U2 over time converge at the spring location, as the two initially separate zones of subsidence coalesce as shown (Figure 6). Upstream incision at the head of the channel is spatially and temporally linked with sinkhole collapses, which occurred on a time-scale of a few days.
These collapses suggest that the water flowing in the channel reaches its head via subsurface conduits, 380 which are the cause of subsurface instability related to the surface collapse and sinkhole formation.
Further evidence of such links is seen at Uvala U4, which is linked spatially and temporally with another artesian spring which feeds a similar meandering stream channel (Figure 6), that we have termed the 'black stream' (due to the very dark-coloured water that flows within it). Initial ground cracking at U4 385 occurred proximal to a subtle linear depression (or 'blind valley') between it and the spring feeding the 'black stream', which we first observe in satellite images from 2009. The migration of sinkholes within U4 seems to follow a flow path from the intial pre-uvala sinkhole cluster to the 'black stream' head, suggesting the presence of a flow conduit beneath the depression. The additional 'prongs' of sinkhole migration and groudcracking at uvalas U4 and U3 (Figures 6 and 7) may also represent a surface 390 expression of subsurface conduit development, although no associated springs have yet been observed.

Discussion
Our reconstruction of the former Dead Sea bathymetry in the Ghor Al-Haditha study area combined with the DSMs from our photogrammetric surveys has revealed subsidence on three area scales: (1) a ~ 3 x 395 10 6 m 2 sized swath of distributed subsidence affecting the lacustrine deposits of the former Dead Sea bed; (2) several 3 x 10 4 -5 x 10 5 m 2 sized zones of more localised subsidence (uvalas); and (3) numerous 1x10 0 -7 x 10 1 m 2 sized features of highly localised subsidence (sinkholes). In this section, we focus our discussion first on the former of these phenomena; the latter two are discussed in detail subsequently.

Distributed subsidence of the former Dead Sea bed following base level fall 400
The distributed subsidence of the former lakebed increases systematically in magnitude from zero at the 2017 shore line to a maximum of about 4 or 5 m near the shoreward edge of the lacustrine deposits. The magnitude of subsidence in the lakebed area lying between the 2000 -2017 shorelines correlates with the age of emergence of the lakebed as the Dead Sea has receded (Figure 8). Data from the predominantly alluvial area lying between the 1967-2000 shorelines and from the northern-most part of the study area 405 are excluded from this figure, because these are demonstrably subject to confounding influences from alluvial erosion/deposition, karstification and, most significantly, anthropogenic landscape disturbance (infrastructural development). The data in Figure 8 are fit well by a function in which subsidence magnitude varies linearly with time, and they are fit slightly better by a function in which subsidence magnitude varies a function of the square root of time. In agreement with Baer et al. (2002), we regard 410 the most likely driving mechanism for the observed distributed subsidence of the former lakebed to be consolidation and compaction of the formerly water-logged evaporite and marl deposits upon the lowering of the Dead Sea level. A non-linear subsidence rate model, as shown in Figure 8, is more compatible than a linear model with 420 the distribution and magnitude of subsidence rates of the lakebed as reported previously from InSAR analysis (Baer et al., 2002;Fiaschi et al., 2017;Nof et al., 2019;Yechieli et al., 2015) and from differencing of LiDAR-derived DSMs (Avni et al., 2016). In general, the rate of subsidence of the former lakebed in these reports decreases landward from the contemporary shore line. The magnitude of subsidence rates previously reported are up to 0.18 -0.30 m/yr immediately adjacent the contemporary 425 shore line, with magnitudes of 0.01 -0.15 m/yr further landward. The non-linear rate of subsidence modelled in Figure 8 would give a rather high initial subsidence rate of 0.53-0.65 m/yr in the first two years after emergence, but this rate would decline to 0.10-0.13 m/yr in the 10-17 years after emergence.
The latter rate agrees well with InSAR-derived subsidence rates of 0.05 -0.15 m/yr adjacent to the 2000 shoreline around Ghor Al-Haditha and the Lisan peninsula (Fiaschi et al., 2017). Our results thus represent 430 the first ground validation of InSAR-based detection of distributed subsidence of the former lakebed at the Ghor Al-Haditha study site.

Morphological attributes of sinkhole and uvalas in evaporite karst
The sinkholes and uvalas in the evaporite karst setting of the Dead Sea are distinct in terms of their scale and morphology. The uvalas are much more irregular in plan-view than the sinkholes, and they enclose 435 numerous sinkholes, including coalesced sinkholes and clusters of sinkholes. The uvalas also have depth/diameter ratios (De/Di = 0.016 -0.042) up to two orders of magnitude lower than the sinkholes (De/Di = 0.02 -1.80).
These morphological characteristics of the evaporite karst sinkholes and uvalas are similar to those in some limestone karst settings. In data from the limestone karst of Trieste as reported by Bondesan et al. 440 (1992) have a more irregular form and have far lower depth/diameter ratios than the sinkholes (dolines). The main difference between the respective depression types of each karst setting is the absolute size.
Sinkholes and uvalas in the evaporite karst of the Dead Sea are considerably smaller than their equivalents in the limestone karst, possibly as a result of the greater material strength and potential depth of karstification in the limestone regions (Al-Halbouni et al., 2018Ćalić, 2011). 450 The relationship between surface materials and sinkhole morphology in the Dead Sea evaporite karst setting is well known (Filin et al., 2011;Al-Halbouni et al., 2017). The low vs high De/Di ratios of the sinkholes in 'mud' and 'alluvium', respectively, have been attributed previously to a contrast in the strength (Al-Halbouni et al., 2018 and/or rheology (Shalev and Lyakhovsky, 2012) of these 455 materials. Expanding upon data presented in these previous works, we show that De/Di ratios of sinkholes formed in 'salt' generally fall between those of sinkholes formed in 'alluvium' or 'mud' (Figure 5B). This probably reflects the observation that near-surface deposits dominated by 'salt' layers usually contain a considerable proportion of 'mud' layers, and so the mechanical behaviour of 'salt' is weakened.
Compared to those in mud-rich sediments or alluvium, sinkholes formed in the salt-rich sediments also 460 have generally smaller diameters ( Figure 5A). This could reflect a scaling limit imposed by the level of karstification, which field evidence locally shows is at, or within a few metres of, the surface in the saltrich material in the northern part of the area. The effect of material property on the uvalas is seen mainly in the expression of the marginal fracturing, which as for sinkholes, is more sharply defined in the higher strength alluvium (see also Al-Halbouni et al., 2018). However, the data are too few to determine any 465 material-linked differences in depth/diameter ratio of the uvalas.

The formation of uvalas and their inter-relationship with sinkholes
The processes governing the genesis of uvalas are debated in karst geomorphology (Ćalić, 2011;Kranjc, 2013;Lowe and Waltham, 1995). As summarised by Ćalić (2011) coalescence of sinkholes. Here we discuss the relevance of these three end-member mechanisms for the genesis of the uvalas related to evaporite karstification in our study area.

475
Surface dissolution and sinkhole coalescence can be precluded as major mechanisms for uvala formation in the evaporite karst setting of Ghor al-Haditha. Much of the surface material affected by uvala formation is non-karstic (non-soluble alluvium), and the hyper-arid climate conditions of the study area probably limit any solution in the vadose zone. Although sinkhole coalescence occurs at Ghor al-Haditha, resulting in compound sinkholes both within and outside of an uvala, this process is not responsible for the main 480 morphometric attributes of the uvalas. The uvalas are observed to develop as larger-scale depressions that have distinct morphometric attributes and limits, both in space and time (Figures 4, 6 and 7).
Our observations are consistent with a mechanism of uvala formation by areally-distributed sub-surface corrosion and erosion, which results in a broad lowering of the surface by subsidence. Our observations 485 further show that an uvala initiates, evolves and ceases in tandem with sinkhole development within it.
These observations are broadly consistent with modelling of subsidence produced by the development of multiple void spaces at progressively deepening levels with varying individual growth rates (Al-Halbouni et al., 2019). We propose that this indicates the same overall formation process for both types of enclosed depression, but that the morphological expression of the process differs depending on the scale at which 490 the process operates. Uvalas can be considered to be the intregrated subsidence reponse of dissolution and mechanical erosion distributed over a mechanically unstable subsurface volume (e.g. a groundwater conduit network). Sinkholes, on the other hand, represent discrete subsidence reponses within that volume to smaller-scale zones of highly localised material removal and related instability (e.g. in an individual groundwater conduit). 495 It is beyond the scope of the current study to resolve general relevance of the subsidence mechanism of uvala formation in evaporite karst for areas of limestone or gypsum karst. One must of course be conscious that similarity of form does not necessarily mean similarity of genesis. Future work might involve resurveying uvalas in limestone karst areas for structural evidence of uvala-scale subsidence, such as 500 synclinal bed rotations and marginal fractures (fissures and/or faults) that are geometrically and kinematically linked to the uvala. Although sparse, structural data presented for the Grda Draga uvala in the limestone Dinaric karst, Slovenia (Fig. 4 of (Ćalić, 2011)) seem broadly consistent with a genesis through subsidence. In limestone karst areas, regional tectonics may facilitate (and complicate) uvala formation by, for instance, providing zones of enhanced permeability for groundwater flow (Ćalić, 2011). 505 Additionally, in limestone karst, most uvalas do not have water at their bottoms (Calic 2011), whereas water has occupied (ephemerally or otherwise) the bottom of some of our evaporite karst uvalas. Hence the relationship of uvalas to the water table may vary from one karst environment to the other, in line with the dissolution dynamics of each.Future studies with more detailed datasets are required to test whether the mechanisms of uvala development in evaporite karst in our study area are applicable to limestone and 510 gypsum karst areas.

Effects of base-level fall on the spatio-temporal pattern of sinkhole and uvala development
At the kilometre scale, the spatial distribution of sinkholes and uvalas at Ghor Al-Haditha follows two linear trends: a N24° trend in the south and a N0° in the north (Figure 4). These trends match those of main regional faults in the Dead Sea transform (Figure 1A, C) and so indicate some tectonic control (cf. 515 Abelson et al., 2003;Closson, 2005;Yechieli et al., 2015), the nature of which has been debated. Some authors have envisaged that tectonic faults conduct ground water directly into evaporite deposits (Charrach, 2018;Closson and Abou Karaki, 2013), while others suggest that tectonic faults control the initial depositional geometry of evaporite deposits (Ezersky et al., 2013;Frumkin et al., 2011;Frumkin and Raz, 2001). In closer detail (hundred-metre scale), the sinkhole distribution at Ghor Al-Haditha is 520 rather non-linear (Figure 4). This non-linearity may reflect control from the distribution of salt-rich evaporite deposits at depth (Ezersky et al., 2013), and thus reflect the palaeo-shoreline, as determined by the regional fault systems on the larger scale.
The spatio-temporal development of sinkholes and uvalas at Ghor Al-Haditha shows two striking features. 525 Firstly, new sinkholes and uvalas has successively formed along a SSW NNE trend with time, i.e. roughly shoreline parallel (Figure 4). Secondly, after they have been established in a given part of the study area, the formation of new sinkholes and the growth of the uvalas occurs consistently seaward direction, i.e. roughly shoreline perpendicular (Figures 4, 6 & 7). Both of these observations are qualitatively consistent with the predicted migration of this dissolution front in the response of sinkhole 530 population evolution to base level fall, especially if the migration of the fresh/saline interface is considered to intersect obliquely with the distribution of salt-rich evaporite deposits in the subsurface (Figure 9). Although constraints on the fresh-saline interface from boreholes or geophysical techniques are lacking in the Ghor Al-Haditha study area, the systematic spatio-temporal migration of new depression development provide strong evidence that a seaward shift of the fresh-saline interface induced by base-535 level fall is a key control on sinkhole development here. A future definitive analysis of the sinkhole and uvala migration should include a borehole drilling campaign in the study area, similar to that conducted on the western shore (Abelson et al., 2017;Yechieli, 2000).

Summary & Conclusions
Our results provide, for the first time, a detailed picture of the interlinked geomorphological development of sinkholes and uvalas in an evaporite karst setting. They also provide new insights into to impact of base level fall on that development at the Dead Sea. Based on the combination of remote sensing data, photogrammetric surveying and field observations, our main findings with respect to the Ghor Al-Haditha 550 study area on the eastern shore of the Dead Sea are as follows: (1) At least 1,100 collapse sinkholes and five uvalas have formed by subsidence in the evaporite karst setting of Ghor Al-Haditha since the mid-1980s. The developments of individual uvalas and of sinkhole populations within them are intertwined in terms of onset, evolution and cessation. While many sinkholes develop initially in clusters, the uvalas develop as a larger-scale, gentler and 555 structurally-distinct depressions around such clusters.
(2) The studied evaporite-karst uvalas likely form through subsidence in reponse to distributed subsurface dissolution and erosion within a mechanically unstable subsurface volume (e.g. a groundwater conduit network). Sinkholes, on the other hand, represent discrete subsidence reponses 560 within that volume to smaller-scale zones of highly localised material removal (e.g. in individual groundwater conduits) and related instability. In agreement with inferences for examples in limestone karst settings, the uvalas in the studied evaporate karst setting do not form by coalescence of sinkholes. Surface dissolution as a mechanism for uvala formation is not significant in this hyperarid setting. 565 (3) The exposed former Dead Sea lakebed at Ghor Haditha has also undergone a wider-scale subsidence of up to several metres that deceases toward the present-day shore line. Subsidence rates estimated here by DSM differencing are in line with those estimated previously by remote sensing. This widerscale subsidence is possibly related to post-recession compaction/consolidation of the near surface 570 lakebed sediments upon withdrawal of pore-fluids as the lake level has fallen.
(4) The location of new sinkholes and uvalas migrates markedly with time, roughly parallel to the shoreline. After initiated, sinkhole clusters and uvalas also show a marked seaward growth with time.
These migration patterns of new depression development are broadly consistent with theoretical 575 predictions of a spatio-temporal control on karstification from a laterally-migrating interface between saturated/under-saturated groundwater, as induced by the base-level fall.

Data Availability
A full set of metadata is available upon request. Satellite images: some open access (Corona), but mostly commercial. Aerial images: available at discretion of RJGC. Photogrammetric surveys: raw images, 580 DSMs and orthophotos available upon consultation with the authors. Geological Map 1:50,000 Ar Rabba: available at discretion of MEMR.

Author Contribution
RAW and EPH led the production of figures and writing of the manuscript. RAW undertook the majority of the data analysis associated with the satellite imagery time series and with the 2015 and 2016 DSMs. 585 Additional satellite imagery processing and data analysis was performed by LS, DAH and EPH. DAH and LS generated the orthophotos and DSMs of the study area using SfM photogrammetry. EPH, DAH, LS, HAR, and AS undertook the field studies and close-range photogrammetric surveys in 2014 -2016.
All authors reviewed and commented on the manuscript, and they contributed to discussions of the data.

Competing interests 590
The authors declare that they have no conflict of interest.
10 Special issue statement (will be included by Copernicus)

Acknowledgements
We acknowledge MEMR colleagues for support in fieldwork and other logistical support. The comments of two anonymous reviewers have greatly improved the quality of the manuscript. Part of the work of 595 NAK was done during a sabbatical year supported by the Deanship of scientific research -The University of Jordan. The authors acknowledge financial support from GFZ and the Helmholtz Association's recent Dead Sea Research Venue (DESERVE) initiative (Kottmeier et al., 2016), especially for the associated data and fieldwork costs. Funding for RAW's masters research project, supervised by EPH, was provided by the Geological Survey Ireland under a GSI Short Call grant to EPH (Contract Number: 2017-sc-002). 600