Geophysical analysis of an area affected by subsurface dissolution – case study of an inland salt marsh in northern Thuringia, Germany

. The subsurface dissolution of soluble rocks , also called subrosion, can affect areas over a long period of time and pose a severe hazard. We show the beneﬁts of a combined approach using P-wave-and S H -wave reﬂection seismics, electrical resistivity tomography, transient electromagnetics, and gravimetry for a better understanding of the subrosion (cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58) dissolution process. The study area, ’Esperstedter Ried’ in northern Thuringia, Germany, located south of the Kyffhäuser hills, is a large inland salt marsh that developed due to dissolution of soluble rocks at approximately 300 m depth. We were able to locate 5 buried subrosion structures , subrosion zones,

understand the components and controlling factors associated with it. The more boundary conditions that influence :::::::: constrain the processes and the structures that can be determined, the better, e.g., dynamic models (Augarde et al., 2003;Shalev et al., 2006;Al-Halbouni et al., 2019) can be adapted in order to make better predictions of risk :::::::: hazardous areas.
The presence of salt springs and the occurrence of sinkholes and depressions in the near-surface indicate soluble rocks in the underground such as the Zechstein formations, and they show that Bad Frankenhausen and Esperstedt are affected by subrosion 90 ::::::::: subsurface ::::::::: dissolution (Reuter, 1962).

Stratigraphy
Five boreholes are used for the later correlation of seismic reflectors and stratigraphy (Figs. 2 & 3). The Zechstein formation (z) is the oldest one drilled. In the research area, the Zechstein consists of the Werra-, Staßfurt-and Leine Formations (z1-z3).
North and south of the Esperstedter Ried, the Zechstein formations are much closer to the surface than in the central part of

Data acquisition methods
East of Bad Frankenhausen in the Esperstedter Ried, P-and S H -wave reflection seismic surveys as well as ERT, fixed loop TEM and gravimetric measurements were carried out along several profiles (Fig. 3).

ERT & TEM
The goal of the electrical resistivity tomography (ERT) and the transient electromagnetic (TEM) surveys was to investigate the subsurface resistivity distribution to determine zones of subrosion ::::::::: dissolution : and areas of potential saltwater ascension ::: rise.
Whereas the large-scale direct current measurements provide robust information about the general resistivity structure, but 135 with comparatively poor resolution, TEM is especially suited to resolve good conductors down to a few hundred metres depth (Milsom & Eriksen, 2011). The ERT survey was conducted with a dipole-dipole configuration using 26 pairs of electrodes with approximately 200 m spacing along the 4.5 km N-S profile. The transmitter provided up to 30 A of source current, voltages were recorded with nine remotely-controlled data loggers ( Fig. 4) developed at LIAG (Oppermann & Günther, 2018).
The ERT was restricted by power lines, roads and a gas pipe. By choosing a fixed-loop setup for the TEM survey, it was 140 possible to cover parts of the profile that are nearly unaffected by strong artificial electromagnetic noise. Receiver positions had 50 m spacing in N-S direction and are placed across four large transmitter loops with 250 m × 250 m dimensions. For more information about the survey and specific details of sensors and data analysis, : we refer to Rochlitz et al. (2018).

Gravimetry
The gravity survey (TLUBN, 2017) , which was carried out by the company Geophysik GGD mbH in 2013 on behalf of the  Figure 4. Seismic (a-d) and ERT (e-g) equipment of LIAG used during the field campaigns. For the seismic surveys hydraulic P-wave and S-wave vibrators (a) were used as active seismic sources, vertical/horizontal geophones (b & d) were used as receivers, and a Geometrics Geode recording system (c). For the ERT surveys, a mobile power source consisting of a generator (e), electricity power inputs plugged into the ground (f), and an adapted recording system (g; photo taken from Oppermann & Günther (2018)) were utilized.
forward modelling, where structural information from reflection seismic interpretation is available as a constraint. The focus was more on the regional basin structure than on local, surface inhomogeneities. The gravity survey was carried out along a profile with a station spacing of 100 m, using a Scintrex CG-5M quartz gravimeter with nominal ± 0.005 mgal standard 150 deviation of repeated measurements. The positions and elevations were determined by differential GNSS or by total station surveys with a standard deviation of ± 0.02 m. To enable a map-based qualitative interpretation, supplementary gravity data from the regional survey of Thuringia, acquired in the second part of the 20th century, with a mean point distance of 1 km to 2 km and standard deviations between ± 0.01 mgal and ± 0.06 mgal (Conrad, 1996) completed the dataset.

S H -wave reflection seismic
The processing of the S-wave ::::::: S H -wave : data was carried out using the processing software VISTA Version 10.028 by Gedco
2018) with a correlation length of 800 m for the horizontal direction and 70 m for the vertical direction, in order to account for the predominant layering of the geological strata.
Processing of TEM data mainly includes noise removal by selective stacking and logarithmic gating. 1D inversion of the processed data was challenging due to affection by strong atmospheric and anthropogenic noise. The two obtained resistivity 200 distributions based on the coil and SQUID receivers are overall similar, but the inversion results based on the SQUID receiver exhibited a higher consistency between neighboring stations, greater penetration depths and less artifacts caused by anthro-pogenic noise (Rochlitz et al., 2018). Nevertheless, using resistivity constraints from ERT and structural information from seismics, the reliability of inversion results could be evaluated.
In P1 and S1 (Fig. 6a, d) from the surface down to ca. 100 m depth and between ca. 1.20 km and 1.50 km profile length, mostly 255 horizontal and continuous reflectors with partly high amplitudes are imaged, which represent Quaternary and Tertiary deposits ( Fig. 6). These impedance contrasts represent layer boundaries. The high-amplitude reflector at ca. 10 m to 25 m depth, which is visible in both sections and traceable throughout the entire profile of S1, represents the boundary between the Quaternary gravel and silt, and the Tertiary clay. In section P1 (Fig. 6a), in about 100 m depth, within the Tertiary deposits, another highamplitude reflector is imaged. It is mostly continuous and traceable throughout the entire profile, which we interpret as Tertiary 260 brown coal and use as a marker horizon (Fig. 6b). In S1 (Fig. 6d), the brown coal shows no distinct reflector, but instead the internal structures of the Quaternary and Tertiary deposits can be observed in more detail compared to the P-wave section (for details see section 5.3 and Fig. 8). Below the top Triassic, at ca. 250 m depth in P1, a horizontal reflector showing a partly strong impedance contrast is imaged in the east of both sections between 1 km and 2.40 km profile length, especially in P1 (Fig. 6a, b). This is interpreted as the boundary between the Lower Triassic sandstones , and the Permian evaporites. The sandstones of the 265 Lower Triassic show almost no internal structures and the area below the top of the Permian contains only poor reflections, probably due to :: the : limited penetration depth of the seismic waves and the resolution limits.
To the west of P1 and S1, between 0 km and ca. 0.75 km profile length, shallowly-dipping reflectors form a bowl-shaped structure ca. 0.70 km in length at 100 m to 300 m depth within the Tertiary, Triassic and Permian deposits (Fig. 6a, c, d). The brown coal marker horizon was used to support the interpretation, since no boreholes are available in this area. Profile S1 270 shows only the eastern margin of this structure, but it gives more detailed information on the internal features of the formations with respect to the P-wave profile (Fig. 6d, e). Onlapping silt and clay layers of the Tertiary are observed above the coal, and further above are horizontal Tertiary and Quaternary deposits. Between ca. 0.80 km and 0.90 km profile length at ca. 100 m to 280 m depth V-shaped reflectors are observed in P1 (Fig. 6c). The same zone is characterized by synclinal reflectors and low reflectivity in S1 (Fig. 6d).

275
The Triassic sediments show local thinning and a general decrease in thickness from east to west. Numerous faults and fractures in the Tertiary, Triassic and Permian formations are imaged (Fig. 6f). In P1 steeply-dipping normal and reverse faults, which transverse the seismic profile, are identified, and in S1 not only the faults of P1 are imaged, but also other steep faults and many nearly vertical fractures within the bowl-shaped structures, which are not all :::: some :: of :::::: which :::: have :::: been : drawn in the interpretation (Fig. 6). These are not be observed in P1 since their scale is below the resolution limit of the P-wave profiles 280 (Fig. 6e).
We interpret the large bowl-shaped structure (Fig. 6b, e, f) to be a former ::::::: collapse sinkhole that opened during the Tertiarydue to subrosion. The nearly vertical fractures within and below the depression ::::::::: subsidence ::::::: sinkhole that crosscut the Triassic and the Permian indicate collapse of an underground cavity. Since the brown coal layer dips and is crosscut by some of the fractures, the sinkhole must have occurred after the deposition of the organic material. This is supported by the onlapping Tertiary and 285 the horizontally-layered Quaternary sand and gravel above. The small structure more to the east seems to be a second collapse ::::::: sinkhole with steep margins (Fig. 6e, f).
The most northern part of P2 between ca. 6.35 km and 6.60 km profile length from the surface down to about 400 m depth ( Fig. 7a, b) shows irregular and discontinuous reflectors of low amplitudes. They represent the Permian Zechstein formations 295 of the Kyffhäuser hills. This area is separated from the south by a steep, northward-dipping thrust fault, the KSMF.
South of the KSMF, at 50 m to 100 m depth (Fig. 7a, b), a continuous, high-amplitude reflector is imaged, which is traceable throughout almost the entire profile. Just as in P1 and S1, this impedance contrast is interpreted as the boundary between the Quaternary and Tertiary deposits. A second, mostly continuous reflector with high impedance contrast is visible between ca. 2.90 km and 6.35 km profile length at 100 m to 200 m depth, which is interpreted as the top of the brown coal ( Fig. 7a, b,   sandstones below show no internal structures, but at ca. 250 m depth, a discontinuous, high-amplitude reflector is imaged. This is interpreted as the top of the Permian Zechstein, which is traceable throughout the entire profile P2 (Fig. 7a, b). Repeatedly, between 3.20 km and 6.35 km profile length the reflector shows low reflectivity, but between ca. 2.40 km and 3.20 km profile length it is continuous and has even higher amplitudes. These two areas are separated by a steep northward-dipping normal 305 fault (Fig. 7b, e, f).
Directly south of the KSMF, between 5.10 km and 6.35 km profile length, dipping reflectors that form a bowl-shaped structure within the Quaternary to Permian deposits are imaged (Fig. 7a, b, f). In contrast to the two structures observed in P1 and S1, the Quaternary is affected too. The deepest point of the bowl-shaped structure is at ca. 400 m depth and coincides with a low reflectivity zone in the Permian Zechstein. Another 0.30 km wide depression was identified to the south between 310 2.10 km and 2.40 km profile length at 200 m to 350 m depth within the Quaternary to Permian sediments (Fig. 7a, b). A third bowl-shaped structure was imaged at shallow depth, in the near-surface between 3.30 km and 3.45 km profile length at ca. 40 m to 100 m depth. This was pointed out as the area of interest for the S H -wave reflection seismic survey S2 (Fig. 7c, d, e). The high-amplitude reflectors of the top of the Tertiary are also observed in S2, but the impedance contrast is weaker than in P2. In S2 details of the internal structure of the different formations can be recognized (for details, see section 5.3 and Fig. 8).

315
Besides the KSMF, other steep faults are identified below the northern depression and below the near-surface depression ::::::: structure to the south (Fig. 7f). The profiles P2 and S2 reveal that the fault below the near-surface depression ::::::: structure, between 3.30 km and 3.45 km profile length at ca. 40 m to 100 m depth, is not a single fault, but instead is a fault zone and nearly vertical fractures within the Lower Tertiary sediments are also imaged in S2.

Seismic facies analysis
The procedure of the seismic facies analysis is based on Roksandic (1978). A total of six seismic facies types (A-F) and four seismic features (SF1-SF4) can be identified in the P-and the S H -wave seismic data on the basis of configuration, continuity, amplitude, and frequency content (Fig. 8). A comparison of P-wave and S-wave was not possible for all facies and feature 330 types, because for some locations only P-wave data was available.
The three facies A-C are characterized by continuous to semi-continuous reflectors and represent more or less undisturbed layers. Facies A consists of Quaternary and Upper Tertiary with continuous, horizontal, and parallel reflectors. Compared to the S-wave sections, the P-wave sections show low amplitudes in the uppermost part due to resolution limits. As a result, the silt, sand and gravel layers of the Quaternary are not imaged, and the deeper parts are of high amplitudes. The S-wave data, 335 however, shows :::: show : generally high amplitudes with a high frequency content and the differentiation of individual reflectors within the two formations is more detailed, because of ::: due :: to : the improved resolution. Facies A might be a good water conduit for horizontal water flow, due to the permeable gravel and sand layers, but noticeable vertical fluid pathways, which are important for subrosion ::::::: sbsurface :::::::::: dissolution, were not found ::::::: identified. Facies B shows the internal structures of the lower Tertiary silt, gravel, sand, and clay, and it has semi-continuous, wavy to sigmoid-parallel reflector patterns. Both P-and S-wave 340 data show a high frequency content, but the amplitudes in the S-wave are higher for this facies and the reflectors are thinner and more detailed compared to the P-wave data. Facies B, with its semi-continuous reflector patterns, could favour vertical water  flow. Facies C shows the Tertiary fill of a subrosion-induced depression :::::::::::::::: dissolution-induced ::::::: sinkhole : with oblique to parallel reflectors. P-wave and S-wave data show slightly different images. In the P-wave data : , : the Tertiary fill is visible as semicontinuous reflectors of medium amplitudes and medium frequency content, and in the S-wave data an onlap fill is observed 345 with continuous reflectors, high amplitudes, and medium frequency content. In contrast to the Tertiary deposits of Facies B, the depression fill does not seem to be strongly fractured.
The three facies D-F are characterized by mostly discontinuous reflectors and represent disturbed layers. Facies D consists of Lower Triassic sandstones of the Buntsandstein, and the pattern configuration can be described as hummocky-clinoforms. In the P-wave data this facies is discontinuous, of low amplitudes and low frequency content, and in the S-wave data semi-continuous Two differentiations for the undisturbed evaporite can be made. High amplitudes might indicate thicker salt layers that generate 355 a stronger impedance contrast against the Triassic sandstones above, compared to medium amplitudes of possibly evaporite (anhydrite, gypsum) that would generate a weaker impedance contrast. Facies E2 shows the disturbed case with discontinuous, hummocky to chaotic reflection patterns, low amplitudes, and low frequency content. Both P-and S-wave data show the same characteristics and this facies is interpreted as fractured and leached Zechstein formations, and are used for the determination of subrosion zones ::::::: potential ::::: zones ::::::: affected ::: by ::::::::: dissolution ::::::::: processes. Facies F images the interior of the Permian Zechstein The seismic feature SF1 consists of semi-continuous reflector patterns that form a bowl-shaped structure in both P-and S H -wave data, although the amplitudes in the S H -wave data are generally higher, as is the frequency content. It is interpreted as a broad ::::::: collapse : sinkhole with more or less horizontal layers above. In the S-wave data a divergent fill and a fractured 365 underground ::::::: fractured ::::: rocks beneath the sinkhole are identified. Seismic feature SF2 shows different characteristics in P-wave and S-wave. In the P-wave data, discontinuous reflectors form V-shaped troughs of medium amplitudes and medium frequency content, whereas in the S H -wave data semi-continuous reflectors form parallel, synclinal structures of low amplitudes and low frequency. It is interpreted as another subrosion-induced, steeper :::::::::::::::: dissolution-induced, ::::::: steeper, ::::::: collapse : sinkhole. Seismic feature SF3 consists of multiple troughs of continuous to semi-continuous reflectors of high and low amplitudes and medium to 370 low frequency content. It is interpreted as a subrosion-induced depression :::::::::::::::: dissolution-induced :::::::::: depression, :::: also ::::: called ::::::: sagging ::::::: sinkhole, : and the sagging is either still ongoing or was active until recent times, which is supported by the fact that all formations from the Permian to the Quaternary are affected ::::::: involved. Seismic feature SF4 is similar to SF3, but in contrast only a U-shaped trough is observed in SF4. Just like SF3 all formations from the Permian to the Quaternary are affected, which is interpreted as a subrosion-induced depression :::::: sagging :::::::: sinkhole, where sagging is ongoing.   Figure 9. TEM profiles (a) on seismic profile P2 and ERT profile (b) with interpreted seismic structures in black. The profile lengths were adapted so that they match with the gravimetric profile to enable better comparability of interpretations, thus, e.g., the ERT profile starts at 2.1 km and ends at 6.6 km. The vertical exaggeration is around 2:1 and the reference datum is 150 m a.s.l. for the ERT and TEM data.
Although P-and S H -wave reflection seismic data image the same subsurface structures, the seismic facies imaging can be different due to the variations in penetration depth and resolution of the seismic waves types, which can influence the interpretation of the reflection pattern. Therefore, we suggest a combined analysis of P-wave and S H -wave seismic data.  Figure 9 shows the inversion results of TEM and ERT in combination with the interpreted geological structures of seismic 380 profile P2. The target depth of the TEM method is between 300 m and 600 m, depending on the subsurface conductivity characteristics (Rochlitz et al., 2018), and the target depth of the ERT is of ca. 400 m. For the TEM survey at the fixed loop positions 2 and 3, a five-layer model instead of the original four-layer model by Rochlitz et al. (2018) was evaluated to be more suitable to explain data as well as subsurface resistivities. The profile length annotations were adapted to match with the gravimetric profile to enable better comparability of interpretations, thus e.g. the ERT profile starts at 2.10 km and ends at 385 6.60 km.

Interpretation of ERT & TEM
In the near-surface within the Quaternary deposits the electrical resistivity ranges from ca. 10 Ωm to 40 Ωm between 2.10 km and 5.80 km profile length down to 30 m depth. In the north, the KSMF is located and due to the thrust fault, leached Zechstein is found only a few meters below the surface showing resistivities of several hundreds of Ωm. North of the KSMF, the salt is completely dissolved/eroded.

390
Below ca. 100 m and 300 m depth a broad zone of low resistivity with less than 10 Ωm between 2.30 km and 6.10 km profile length is observed. This zone corresponds with Tertiary and Triassic deposits, as interpreted in the seismic sections P2 and S2. Within this zone is an extremely conductive thin layer of approximately 1 Ωm, which is in particular resolved by the TEM data. The boundaries of this layer coincide well with main seismic reflectors. It is interpreted to match with saline aquifers between the lignite and :: the : Triassic Buntsandstein. Unfortunately, there are no continuous TEM-soundings across the entire 395 profile available. In contrast, the ERT inversion data smears this layer over greater thickness due to limitations of resolution, but proves its continuity.
From 3.10 km to 2.10 km profile length the extremely conductive layer vanishes and resistivities between 100 m and 300 m depth increase slightly, as shown by the ERT and TEM FL 1, which is probably related to the fault zone at 3.20 km. The fault might hamper the lateral groundwater flow, and therefore, the lateral distribution of the salt water coming from the 400 leached ::::::: dissolved : Permian deposits, which migrated upward along the faults and fractures to the north due to artesian-confined groundwater conditions.
The top Permian at 250 m to 300 m depth, as indicated by seismic interpretation, correlates with the top of the TEM basement layer with high resistivities of more than 100 Ωm. This is also visible in the ERT result, but the contrast is smoother and does not reveal a unique basement depth. At 5.50 km profile length, the ERT shows lower electrical resistivities reaching 405 depths of 500 m. This coincides well with the interpreted fault. This area correlates also with the position of subrosion-induced :::::::::::::::: dissolution-induced structures, like the sinkholes above. Therefore, it is an indicator for salt-water ascension :: rise, due to dissolution of Zechstein evaporites and artesian-confined groundwater conditions. The ascending ::::: rising salt water is the reason for the development of the inland salt marsh. A similar low-resistivity area at such great depths is located at 2.10 km to 2.50 km on the profile. However, since the profile ends there, detailed interpretation is speculative.

410
Overall, the ERT and TEM results are in very good agreement with the seismic interpretation, although no joint inversion was carried out, which would improve the quality of the datasets and their interpretation.

Interpretation of gravimetry
The prominent features on the Bouguer map (Fig. 10a)  Several spectral domain filters were applied in order to locate possible sources of gravity anomalies and to highlight fine changes in the gravity field. As an example, ::: The :::::::: tilt-angle ::::: filter : (Fig. 10bshows the result of the so called 'tilt derivative' or 420 'tilt angle' filter (Miller & Singh, 1994), which is defined as where VDR is the first vertical derivative of the gravitational potential, which describes the vertical gravity component, and THDR is the total horizontal derivative, which describes the combined horizontal gravity components of the field vector in xand y-direction. This filter process generates maxima centered above the source of the anomaly. Its zero crossing is located 425 close to the edges of source bodies. All amplitudes are restricted to values between +π/2 and -π/2 (+90 • and -90 • ) , thus suppressing strong amplitudes and amplifying weak amplitudes. The tilt-angle filter ) : reveals an elongated source beneath the Kyffhäuser hills and a local low in the southern part of the Esperstedter Ried. In general, negative values correlate well with the location of Quaternary fluvial sediments of low density and areas with presumed mass loss due to dissolution of soluble rocks, while the majority of positive values surround Triassic, Permian and Carboniferous outcrops.

430
Quantitative interpretation was accomplished by iterative 2.5D forward modelling of the gravimetric profile GR (Fig. 10c, d).
This simplified approach is suitable for the S-N profile, since the Bouguer anomalies depict an elongated structure of roughly 10 km length. It strikes nearly perpendicular to the profile, and therefore exhibits, as a first-order approximation, a 2D character.

455
suggests that locally limited areas with varying densities must be located in depths of e.g. 150 m to 250 m. Further to the north of the inland salt marsh one large-scale minimum is observed, which correlates with a basin-like structures identified in seismic section P2 (Fig. 7). Based on our results, we assume that it developed because of syn-sedimentary sagging due to slow mass movement induced by subrosion :::::::::::::::: dissolution-induced :::: mass ::::::::: movement. The syn-sedimentary development enabled the accumu- Figure 10. Bouguer anomalies (a) show a positive anomaly for the Kyffhäuser hills and the tilt derivative (b; e.g. Miller & Singh (1994)) shows a negative anomaly for the Esperstedter Ried because of mass movement due to subrosion :::::::: subsurface :::::::: dissolution and the accumulation of, e.g., Quaternary sediments with lower density. In the local Bouguer anomalies of the profile GR (c) three small-scale and one large-scale local minima are observed (d), which correlate with subrosion-induced :::::::::::::: dissolution-induced : mass loss identified in the seismic sections P2 and S2. The legend contains the description of the modelled formations and their assumed densities taken from literature. The blue line in (a) and (b) is the gravimetric profile GR. The vertical exaggeration is 2:1.
lation of higher amounts of unconsolidated Quaternary sediments and also Tertiary deposits with lower densities, compared to 460 the surrounding areas. As a result, a large-scale minimum is seen in the gravity data at this location.
In the Esperstedter Ried the P-and S H -wave reflection seismics and the gravimetric investigation revealed large-and smallscale structures, like faults, fractures and thickness variations, delivering a structural model of the area. With regard to the fault 475 development, it is known that from the Upper Cretaceous to the Lower Tertiary, the Kyffhäuser hills were upthrusted and the KSMF developed (Freyberg, 1923;Seidel, 2003). This fault has a Hercynian strike, which can be observed in the geological map of Bad Frankenhausen (Schriel & Bülow, 1926a, b) and in the gravimetry data (Fig. 10). This preferred Hercynian strike direction of northern Germany can also be observed in the formations shown in the reflection seismic profiles P2 and S2 ( Fig. 7), in which other faults were identified that transverse the profiles, cut the Lower Tertiary, and show approximately a 480 NW-SE strike. It has to be noted that the Bouguer anomaly maps shown in this study have a resolution limit of 1 km and therefore cannot image this scale of faulting.
Overall, the Permian, Triassic and Tertiary formations show thickness variations across the study area. Except for local variations, a general decrease in thickness from south to north, towards the KSMF, is observed for the Permian and Triassic deposits. The thickness variation of the Triassic is probably a result of erosion after deposition and a varying accommodation space during deposition, but the thickness variations of the Zechstein are mostly the result of subrosion :::::::: dissolution. Since faults are able to enhance subrosion ::: such :::::::: processes : (Closson & Abou Karaki, 2009;Del Prete et al., 2010;Ezersky & Frumkin, 2013a;Wadas et al., 2017), the leaching ::::::::: dissolution process is more intense close to the KSMF and more salt and gypsum are dissolved 500 leading to mass movement and a decrease in thickness of the Zechstein formation. Other reasons, such as active diapirs and salt movement, as reasons for the Zechstein thickness variations can be excluded. According to the geological map, the top Carboniferous is found in ca. 550 m depth and the thickness of the Permian is expected to be between ca. 350 m to 400 m. The salt in the Permian deposits needs to be much thinner, so even if the top Carboniferous varies in depth it is highly unlikely that the salt layer would be thick enough to form active diapirs (Schultz-Ela et al., 1993;Jackson et al., 1994). Salt movement due 505 to increased differentiated load is also unlikely, because areas with a thicker Triassic Buntsandstein do not correlate with areas of thinner Permian deposits (Schultz-Ela et al., 1993), instead the opposite is observed. However, the thickness of the Triassic sandstones does have an influence on the formation of subrosion-induced :::::::::::::::: dissolution-induced : structures. Great thicknesses of a compact rock are relatively stable against subrosion-induced subsidence and collapse, because even when cavities are formed in the Zechstein formations beneath, the sediments above the cavity would form a structural arch (Waltham et al., 2005), which 510 prevents ::::: should ::::::: prevent collapse and subsidence. Whereas low thicknesses of the Triassic sandstones increase the possibility of cavity collapse and local subsidence, which is the case in the Esperstedter Ried. Further evidence for the long lasting subsidence is given by the Tertiary brown coal that was deposited during the Oligocene, and which shows a varying thickness and a dip variation of 10°to 60° (Frank, 1845). It is unlikely that the brown coal was deposited with such a high dip, so we assume that the brown coal thickness variations are a result of continued subrosion-induced :::::::::::::::: dissolution-induced : sagging during deposition, 515 with thinner brown coal at the basin margins and thicker brown coal at the basin centre.
A better understanding of especially the recent local subrosion ::::::::: dissolution processes, however, also requires detailed knowledge about the fluid pathways and the localization of subrosion ::::::::: dissolution : zones. This was accomplished using ERT/TEM, and seismic facies analysis of P-and S H -waves. The top of the soluble Permian rocks was detected at 250 m to 300 m depth, so near-surface subrosion ::::::::: dissolution, as it is the case for the town of Bad Frankenhausen, north of the KSMF (Wadas et al., 2016;520 Kobe et al., 2019), is not possible. The unsaturated groundwater needs to reach greater depths in order to leach the evaporites such as salt and anhydrite. One of this groundwater levels is detected in :: at 150 m to 200 m depth according to the TEM data, which shows a low resistivity zone in this depth. Regarding this, downward water flow through a fracture network is required for subrosion :: its ::::::::::: enlargement ::: and :::::: further ::::::: solution :: in :::::: soluble ::::: rocks : (Billi et al., 2007). Although this process takes place mostly at shallower depths, a correlation between fractures and subrosion ::::::::: dissolution features is observed for the Esperstedter Ried from 525 the seismic facies analysis. The low resistivity zone of the TEM data is connected with the soluble rocks through a fractured seismic facies of the Triassic and Lower Tertiary, as shown in the S-wave :::::::: S H -wave data (Fig. 8). Faults are also very important for vertical water flow, and this is shown by the ERT data. At faults the water can migrate downwards and dissolve the deeper soluble rocks, as observed in seismic profile P2 and the ERT (Figs. 7 & 9), where below the sagging structures steep faults crosscut the Tertiary, Triassic and probably Permian deposits, and serve as large fluid pathways. The water can also migrate upward along the fault planes (Legrand & Stringfield, 1973) and fractures Westhus et al., 1997), mostly due to artesian-confined groundwater flow towards the surface, as is the case for the Esperstedter Ried. The ERT identified an area of low resistivity at the surface between 3.9 km and 4.7 km profile length, which is the centre of the inland salt marsh ( Fig. 9). Besides the fluid pathways, possible subrosion :::::::: dissolution : zones have also been identified by the P-wave reflection seismic profiles and the ERT. Areas affected by subrosion :::::::: dissolution : show a less pronounced top Zechstein reflector, and a 535 lower resistivity. The leaching and the mass movement destroy the layer boundary and the layering of the Zechstein, therefore no continuous reflector can be observed (e.g. below the large sagging structure to the north of P2). The opposite is visible in the south of P2, where a strong impedance contrast without a low resistivity zone is imaged for the top Zechstein, indicating less subrosion ::::::::: dissolution.
Our recommended workflow (Fig. 11) will be explained in the following, including conditions and limitations of the proposed methods that have to be kept in mind.

550
Reflection seismics is the preferred method to identify sagging-and collapse structures, and also faults and fractures, even in urbanized regions, as other studies have shown for P-waves (e.g. Steeples et al. (1986) wave reflection seismics small-scale structures can be identified. But since subrosion :::::::: dissolution : results in strong vertical and lateral variations of the underground, : a densely-spaced seismic survey has to be acquired in order to image these variations.
Furthermore, the penetration depth of the seismic waves has to be considered. Shear-waves can image the underground at higher resolution than P-waves in the case of water-saturated and unconsolidated sediments, but the penetration depth is lower.
As a result, with the equipment used in this study, shear-wave reflection seismics is able to image the underground down 560 to ca. 300 m depth and the P-wave reflection seismics delivers images from ca. 30 m below surface to ca. 400 m to 500 m depth. Another seismic source with a higher energy transmission might solve this problem, and to detect smaller subrosion ::::::::: dissolution features at the near-surface, : higher sweep frequencies, and at least a smaller source and receiver spacing : , should be used. The seismic facies analysis has shown that one and the same facies unit can have different characteristics, depending on the wave-type applied, but both, P-and S H -wave reflection seismic methods combined, can give valuable information about subrosion-::::::::: dissolution-: and fracture zones, and therefore potential fluid pathways.  (2020)). The advantage of TEM over ERT is the improved resolution in the near-surface down to ca. 300 m depth, but without ERT the information about the electrical resistivity in greater depths e.g. 400 m to 500 m would be missing, where subrosion :::::::: subsurface ::::::::: dissolution :::: and :::::::: erosional :::::::: processes : at faults may occur at greater depths. In general,

575
ERT is better suited to detect lateral resistivity variations, whereas TEM is highly sensitive to conductive structures. Similar to seismic methods, the source and receiver spacing is the determining factor for the possibility to detect aquifers, fluid pathways, or subrosion ::::::::: dissolution : zones. In this study, the spacing of 200 m for the ERT and of 50 m for the TEM survey was sufficient to recover larger subrosion :::::: solution : features (e.g., the low resistivity zone at the fault in profile P2 at 5.60 km profile length), but targeting near-surface, small-scale subrosion :::::::: dissolution : zones would require survey layouts with denser source and receiver 580 spacings. It is also necessary to note , that ERT has a restricted applicability in urban areas due to disturbances by electromagnetic noise. In contrast : , Rochlitz et al. (2018) demonstrated that the inversion results of a SQUID-based TEM exhibits a higher consistency between neighboring stations and less artifacts caused by anthropogenic noise compared to classical coil-based TEM. The identified aquifers and salt-water bearing areas combined with the structural model can than :::: then be used to derive a hydrological model. From this hydrological model, together with the reconstructed sedimentary facies and depositional history 585 of the region, the fluid pathways and subrosion ::::::::: potentially :::::::::::::::: dissolution-affected zones can be derived.
Both reflection seismic-and electromagnetic methods do not deliver information about local mass movements or cavities induced by subrosion ::::::::: dissolution ::: and :::::::::: subsurface :::::: erosion. For this gravimetry is the preferred method for near-surface investigations in order to depict density contrasts and local gravity anomalies (e.g. Butler (1984); Kersten et al. (2017); Kobe et al. (2019)). In our study, we show that structures and zones affected by subrosion :::::::: dissolution : as identified by the reflection seismic-590 and electromagnetic methods correlate with local minima of the Bouguer anomaly due to reduced densities. Therefore, the gravity data together with the structural model can be utilized to get ::::: obtain : a density-distribution model of the subsurface from which the local mass movement can be deduced. The station spacing for the gravimetry profile in this study was 100 m, which was appropriate to detect mass movement on a larger scale (e.g. the near-surface collapse structures producing local minima had a lateral extent of at least ca. 100 m), but . :::: But in case of the detection of small-scale subrosion, ::::::::: dissolution ::::::: features, :::: e.g. as

Conclusions
The initial trigger of subrosion :::::::: dissolution : at the inland salt marsh of the Esperstedter Ried were the tectonic movements during the Tertiary, which led to the uplift of the Kyffhäuser hills and the formation of faults parallel and perpendicular to the low mountain range. The faults and the fractured Triassic and Lower Tertiary deposits serve as fluid pathways for groundwater to leach :::::: dissolve : the deep Zechstein deposits, since subrosion is ::::::::: dissolution :::: and ::::::: erosional ::::::::: processes ::: are more intense near faults.
Data availability. The seismic, ERT and TEM data are the property of LIAG. The data are available from the first author upon request.
Please contact Sonja H. Wadas for details. The gravimetry data is the property of the Thuringian State Institute for Environment, Mining and Conservation.

635
Author contributions. The seismic data was conducted by SW, HB and UP, and LIAG's seismic field crew. Data processing of the SH -wave seismic data was carried out by SW and data processing of the P-wave seismic data was carried out by HB. The seismic facies was analyzed by SW. The ERT and TEM data was surveyed by RR, TG and MG, and inversion of the ERT data was carried out by TG and processing of the TEM data was performed by RR. The modelling of the gravimetry data was carried out by PS. All authors contributed to the interpretations.
SW created all figures except for figure 8 which were created by RR and TG. SW prepared and discussed the results with all co-authors. SW wrote the manuscript and all co-authors inspired and improved the manuscript.