1 INTRODUCTION

Seabed domes are isolated circular or elliptical broad topographic high features occurring on finegrained Quaternary sediment covered seabed, generally several hundred meters in diameter and 0.5–5 m in height (Judd and Hovland, 2007). This type of feature has been observed in many sea regions, such as the western Irish Sea (Yuan et al., 1992), the Aegean and Ionian Seas and the Patras Gulf (Papatheodorou et al., 1993; Hasiotis et al., 1996), the Rias region of northwest Spain (García-Gil, 2003), the Norwegian Sea (Plassen and Vorren, 2003), the Fraser Delta (West Canada) and the North Sea (Judd and Hovland, 2007), the Mediterranean Sea (Savini et al., 2009), the Ortegal Spur continental margin of the Northwestern Iberian Peninsula (Jané et al., 2010), the South Yellow Sea (Zhao et al., 2009), the Gulf of Mexico (Thoma, 2014), and the Opouawe Bank (Koch et al., 2015). The main mechanisms to explain their formation are as follows:

(1) Unconsolidated shallow sediment layers respond mechanically to increasing pore pressure caused by shallow gas and/or fine-grained liquefied sediment accumulation (Papatheodorou et al., 1993; Hasiotis et al., 1996; Lee and Chough, 2002; GarcíaGil, 2003; Plassen and Vorren, 2003; Barry et al., 2012; Boudreau, 2012; Koch et al., 2015).

(2) The sediment volume expands due to the formation of gas-hydrates or ice in deep water sediments and in high latitude seas (Paull et al., 2007, 2008; Chun et al., 2011; Portnov et al., 2013).

(3) Materials extrude from submarine layers driven by the overpressure caused by shallow gas accumulation or differential compaction (Duarte et al., 2007; Paull et al., 2007; Savini et al., 2009).

The first of these four mechanisms is the most widely accepted, and therefore, seabed domes are most often described as the surficial expression of shallow gas pockets and indicate the sites where relatively high pore pressure has accumulated (Judd and Hovland, 2007). The presence of shallow gas can be hazardous and risky for borehole drilling and offshore infrastructure because it decreases the mechanical stability of the seabed. Furthermore, the shallow gas accumulation may be an indicator of deeper hydrocarbon reservoirs. Recently, the concept of thin plate mechanics was introduced to describe the formation of these structures, therefore facilitating quantitative studies of the formation of seabed domes (Barry et al., 2012). Koch et al. (2015) developed the approach by introducing the buoyancy effect provided by the underlying shallow gas and compared the results to those presented by Barry et al. (2012). However, uncertainties still exist regarding the formation and evolution of seabed domes, and additional work is required to better understand the doming features.

Depressions with and without central domes have been observed in the North Yellow Sea (NYS), and previous studies have mainly focused on the depression structures (Liu et al., 2013; Chen et al., 2017). During a marine geo-environmental survey to establish the Yellow Sea research station in 2010, a number of these structures were observed within the 11 km by 17 km survey area. Our team revisited the same site in 2012 and 2013. Based on our data, the seabed domes were most obvious in this particular area, and therefore, we focused on the dome structures. Here, we present a detailed study of these dome features based on both high-resolution mapping data obtained in 2012 and geochemical data acquired in 2013. We then discuss the most likely formation process of the structures. We contribute to research on these anomalous structures by providing 1) acoustical evidence for the presence of shallow gas pockets directly underneath the seabed domes, 2) geochemical evidence of the fresh/brackish groundwater discharge, and 3) an alternative scenario for the formation and evolution of these structures.

2 OCEANOLOGICAL AND GEOLOGICAL BACKGROUND OF THE STUDY AREA

The NYS rests on a flat and tectonically stable seafloor with a water depth of less than 70 m and is semi-enclosed by the landmasses of China and Korea (Fig. 1). It represents one of the best examples of an epicontinental sea (Yang et al., 2003; Kim et al., 2005; Bao et al., 2010). Since the Late Quaternary, the NYS has experienced dramatic sea level fluctuations (Kim et al., 1998; Yang et al., 2003; Li et al., 2010). For more than half of this period, areas of the seafloor that currently have water depths of less than 60 m were subaerially exposed (Wang and Wang, 1980; Zheng, 1991) or covered with vegetation or fresh water, with extensive systems of groundwater aquifers distributed underneath that were filled with fresh/brackish water and directly recharged by meteoric water. During the Holocene transgression sea level rose rapidly; however, even today the interface between seawater and freshwater underneath the seabed is not in steady state (Person et al., 2003; Post et al., 2013).

Geologically, the North Yellow Sea Basin (NYSB) is an intercontinental fault subsidence basin of Mesozoic and Cenozoic age. Geophysical data and cores indicate that thick Mesozoic and Cenozoic sequences deposited in the basin include two sets of Mesozoic and Paleocene source rocks (Gong et al., 2000). Nearly northeast- and northwest-trending faults and folds developed within the NYSB. The spatial extension and development of the NYSB was controlled by two northeast-trending deep edge faults (Tian et al., 2007). The sedimentary and tectonic characteristics of the NYSB provided favorable conditions for the development of conduits for the migration of seabed fluids, including shallow gas and groundwater, which may be coupled.

The study area (which is ca. 17 km long and 11 km wide) is situated in the northwestern sector of the NYS and in the mid-west section of the study area of Chen et al. (2017) and Liu et al. (2013) at a water depth of 50–55 m (Fig. 1). The seafloor is mainly covered by fine-grained and unconsolidated Holocene sediment, which mainly consist of silt, clay, and fine sand.

3 METHOD

A detailed multi-beam bathymetric survey was conducted in the study area in 2012 using a multibeam echosounder (GeoSwath Plus system) operated at 125 kHz. The distance between the survey lines was 100 m, which ensured the acquisition of fullcoverage bathymetric data of the study area. Highresolution side-scan sonar images and sub-bottom profiles were obtained on the same cruise using a towed dual frequency (100 KHz and 500 KHz) Klein 3210 side-scan sonar system and a hull-mounted Bathy 2010^TM Chirp sub-bottom profiler with working frequency centered at 3.5 kHz, respectively. Location information during the surveys was obtained using a differential global positioning system (DGPS). The multibeam data were processed using CARIS HIPS/ SIPS software, and a 5-m data grid of the water depth, which was used to create a seabed relief map of the study area with Surfer 9.0, was produced. Side-scan sonar data were processed and analyzed with SonarWIZ 5 software. The sub-bottom data were also initially processed with the SonarWIZ 5 software; however, the results were not acceptable because of the poor quality of the sub-bottom profile data. For a better understanding of the shallow subsurface structure beneath the study area, the Seisee (v2.20.1) program was used to display and interpret the subbottom profile data, which were stored in SEGY format.

In total, seven gravity cores (designated a through g) with lengths ranging from 65 to 272 cm were collected to analyze the lithological features of the shallow sediments and the geochemical features of the pore waters in the study area during the 2012 and 2013 cruises, respectively (see Fig. 1 for locations). Cores b and f were collected from the top of the selected domes, whereas cores a and g were taken from the bottom of the related depressions. Core c was taken from normal seabed serving as a background control sample. Cores d and e were collected from the flank of the second dome (i.e., D2). Each of the processed cores was split in half immediately after recovery. One half was subdivided for pore water extraction with the Reeburgh-style sediment squeezers (Ussler Ⅲ et al., 2003), while the other half was described and stored for further studies. The chloride and sulfate concentrations in the pore waters were analyzed using a Dionex ICS-90 Ion Chromatograph with an error less than 5% from the Research Center of Analysis and Measurement, Institute of Oceanology, Chinese Academy of Sciences (IOCAS), Qingdao, China. Methane dissolved within the pore waters was sampled using the headspace equilibration method (Ussler Ⅲ et al., 2003), and then analyzed with a Shimadzu gc-14b gas chromatograph at the Ocean University of China, Qingdao, China.

4 RESULT

The geo-acoustic data and sediment cores were analyzed to describe the morphology and internal structures of the seabed domes. The geochemical data, including the chloride, sulfate and dissolved methane content in the pore water samples, were described and compared.

4.1 Geo-acoustic data

The seabed in the study area is flat with a gentle slope of less than 0.001° and is featureless except for the presence of the seabed dome structures, which are surrounded by depressions. Twelve of these typical domes were identified in the 3D seabed relief map within the 11 km (S-N) by 17 km (W-E) study area (Fig. 1). The domes are characteristically elliptical features that are randomly distributed on the seafloor, and they are surrounded by circular or semicircular depressions or moats. In the cross-section, the domes and their depressions display a broad "W" or "N" shape (Fig. 2b, d). The diameters of the domes range from approximately 250 to 1 700 m for the long axes, and 430–1 700 m and 250–900 m, respectively, for the short axes. The ellipticities (long axis/short axis) of the domes are in the range of 1.2 to 5.2, with a mean of 2.4. Most of the major axes are predominantly oriented NE-SW. The heights of the domes range from 0.1–1.0 m, with an average of 0.5 m. The widths of the depressions (defined as the distance from the outer edge of the depression to its dome) range from approximately 300 m to 1 250 m. Most of the depressions are asymmetric with deeper and wider southeastern or eastern sides. The large diameters and small heights of these structures indicate a slope angle of less than 1/500; therefore, these structures can easily be overlooked on a bathymetric map.

On the side-scan sonar images, the domes coincide with high to moderate backscatter intensity regions and are displayed as lighter elliptical features, which can easily be distinguished from the dark seabed areas (Fig. 2a, c). The images also show that the backscatter intensities are positively correlated with the heights of the domes (Fig. 2), i.e., the more elevated domes are correlated with stronger backscatter intensities. However, the surrounding depressions share similar backscatter intensities with the normal seabed, without any distinctive signal anomalies. The absence of acoustic shadows around the domes suggests that the topography of the domes is characterized by a low relief.

Two stratigraphic units can be recognized on the sub-bottom profiles: the upper layer (unit 1), which is bound by the seabed and the first reflector, R1, and the underlying layer (unit 2), which is bound by reflectors R1 and R2 (Fig. 3a, b). The upper layer exhibits a pattern of continuous and parallel reflections with a thickness of around 8 m. The underlying unit is characterized by chaotic reflections with partially stratified acoustic features with a thickness of approximately 10 m. Discernible acoustic anomalies, such as strong negative reflection polarities (yellow rectangles in Fig. 3c, d) and acoustic smear (highlighted yellow features in Fig. 3a, b), are observed throughout the profiles. These features indicate the presence of shallow gas within the sediments under both the normal seabed and the domed areas. Gas migration conduits are visible (Fig. 3c, d) and a main conduit beneath the gas pocket underneath D2 that cuts through the sediment under unit 2 (Fig. 3c).

4.2 Sediment cores

In total, seven cores were collected from the study area (Fig. 1). The cores consisted mainly of silt in addition to a small amount of sand and clay and ranged in color from olive gray to dark gray. Substantial shell debris was dispersed throughout the upper part of the cores and those collected from the top of the domes appeared to have a thicker shell layer (Fig. 4). A paleosol layer was discovered at the bottom of cores b and g, and an approximately 3 cm thick peat layer was found in core c. The paleosols found within the NYS formed during the Last Glacial when the sea level decreased by more than 100 m and seawater retreated from the Yellow Sea, which caused the seabed within the study area to be subaerially exposed (Liu, 1986; Zheng, 1991). A peat layer developed when sea level remained at a depth of approximately 60 m for hundreds of years around 11 ka after experiencing a rapid and sustained rise (Gao, 1986; Liu and Gao, 2005; Li et al., 2010). Therefore, the paleosol and peat layers are markers that indicate the base of the Holocene marine sediments. Isolated circular or irregular open voids that are similar to the textures originating from sediment degassing were observed in most cores, ranging from 1 cm by 1 cm to 3 cm by 3 cm. However, no H₂S odor was detected during core processing.

4.3 Geochemical data

All of the pore water samples were analyzed for their chloride (Cl^-) and sulfate (SO₄^2-) ions (Fig. 5). Changes in Cl^- concentrations within pore waters provide direct evidence for the discharge of fresh/brackish groundwater, and the SO₄^2- content is sensitive to the presence of shallow gas that is primarily composed of CH₄ (Borowski et al., 1999; Thamdrup et al., 2000; Ramírez-Pérez et al., 2015). The chloride content in cores a–c and e–g ranges from 505.96 mmol/L to 459.69 mmol/L and does not display distinct depth trends. In addition, the SO₄^2- concentrations in these two groups of cores range from 29.34 mmol/L to 20.97 mmol/L, demonstrating a slight decrease with depth. The SO₄^2-:Cl^- ratios in the cores range from 0.116 9 to 0.160 5 with an average of 0.1499, which is slightly higher than that of the Yellow Sea water. The ratio-depth curves show some similarity with those of the SO₄^2- content with depth. In contrast, the Cl^- and SO₄^2- concentrations in core d, which was recovered from the flank of the second dome, display a distinctive trend. In the upper 30 cm of core d, both Cl^- and SO₄^2- increase rapidly from 390.95 to 459.66 mmol/L (+17.6%) and from 21.61 to 23.26 mmol/L (0.76%), respectively. However, below 30 cm, the Cl^-and SO₄^2-concentrations decrease abruptly with depth and reach minimum values of 288.57 mmol/L (-26.2%) and 8.68 mmol/L (-59.8%), respectively, at the bottom of the core.

Thirty-six headspace gas samples were analyzed to determine the dissolved methane in cores c, d and e, and no significant differences were found (Fig. 5). The CH₄ content in the pore waters at all stations ranged from 12.15 nmol/L to 29.47 nmol/L, with a mean of 21.76 nmol/L. In comparison, the methane content in core d was slightly higher than that in the other two cores, and the maximum dissolved methane signal (46 nmol/L, which is nearly twice that of the other cores) was observed at the bottom of the core. It is interesting to note that there was minor variability in the Cl^-, SO₄^2-, and CH₄ content in core d at a depth range of approximately 105–150 cm below the seabed.

5 DISCUSSION

The accumulation of shallow gases in soft and cohesive seabed sediments can generate seabed dome structures (Judd and Hovlan, 2007; Barry et al., 2012; Koch et al., 2015), and this mechanism has likely occurred in the NYS study area.

5.1 Seabed domes with acoustic anomalies

During the acquisition of multi-beam and subbottom profile data, plume features, which are indicators of gas bubbles present in the water body (Bayrakci et al., 2014; Weber et al., 2014), were absent. Furthermore, the relatively homogeneous backscatter intensities of the dome areas on the sidescan sonar images indicate that no typical surficial seep features, such as authigenic carbonate crusts or biological communities, developed on the domes or in the depressions (Fig. 2a, c). These findings suggest a lack of obvious shallow gas activities within the study area during the time of the survey or in the recent past. The side-scan sonar images also indicate that the backscatter intensities are positively correlated with the heights of the domes (Fig. 2), i.e., the higher domes correlate with stronger backscatter intensity areas. This demonstrates that the seabed relief, rather than the surficial sediment characteristics, determines the back-scatter intensity of the signals in the study area.

Free shallow gas within sediments can attenuate acoustic energy and significantly change the acoustic features of the sediments, which results in different types of acoustic anomalies on the high-resolution sub-bottom profile images (Mathys et al., 2005; Mazumdar et al., 2009; Tóth et al., 2014; Koch et al., 2015). The strong negative reflection polarity and acoustic attenuation observed on the sub-bottom profile images indicate widely distributed shallow gas in the study area (Fig. 3) located beneath the domed areas and the normal seabed. In comparison, the dome structures were always underlain by pockets, as illustrated beneath domes 2, 6 and 8, which were relatively high and had typical dome morphologies. The shallow gas pockets corresponding with these domes were almost isolated with limited horizontal distribution, i.e., the horizontal extent of the shallow gas pockets was slightly greater than, or approximately equivalent to, those of the related domes. However, the horizontal extent of the shallow gas beneath D12 was much greater than that of the dome itself (Fig. 3b). An almost continuous sediment layer was observed beneath these shallow gas pockets. Therefore, the horizontal migration of shallow gas between the upper and underlying units likely occurred beneath D12, which may have decreased the overpressure beneath the dome and generated the eventual subsidence of the surrounding areas, thereby, producing the moat structures.

5.2 Accumulation atop domes and erosion within depressions

The relatively short lengths of cores a and g, which were collected from the bottom of the related depressions, and the presence of a paleosol layer at the bottom of each core indicate a thinner Holocene marine sediment cover in the depressions than in the domed areas and the normal seabed. The shell debris sections in the upper part of both cores were only a few centimeters thick. In comparison, cores b, d–f, which were recovered from domes, developed relatively thick shell debris sections and neither characteristic layers (e.g., a peat layer or paleosol), nor abrupt sediment phase changes were present, demonstrating a continuous deposition process and a relatively thick Holocene cover. These findings indicate sediment accumulation at the dome sites and erosion within the depressions. The thicker shell debris sections may have been caused by sediments that were reactivated by bottom currents or resuspended by the discharge of seabed fluid flows, and redeposited on the domes; or benefited from the activities of benthic animals favored by the stronger bottom currents and seabed fluid flows at the domed sites.

5.3 Presence of submarine fresh/brackish groundwater

The geochemical data do not exhibit any signs of shallow gas activities, likely because the cores did not penetrate deep enough to reach the gas-bearing layer. Another possible reason is that the shallow gas migrated upward along pathways of least resistance, which suggests that a relatively focused conduit system was established. Therefore, recovering a core just within or immediately adjacent to it is difficult. The lack of evidence of shallow gas activity in the geochemical data is therefore reasonable within the study area.

However, the synchronous decrease in SO₄^2- and Clin core d strongly suggests that discharge of submarine fresh/brackish groundwater occurred at the core recovery site. Offshore fresh groundwater is a global phenomenon that has been observed in many shelf regions (Person et al, 2003; Cohen et al., 2010; Polemio et al., 2011; Post et al., 2013). Since the Late Quaternary, the NYS, which is characterized by a flat and tectonically stable seafloor and shallow water depth, has undergone three major cycles of climatic change, including three transgressions and two regressions. During these regressions, seawater retreated from the region and the seabed of the NYS was subaerially exposed or covered by fresh/brackish water bodies and/or vegetation. Furthermore, aquifers located beneath the seabed were occupied and recharged by fresh/brackish water. Based on the limited geochemical and sub-bottom data used in our study, the only feature that can be confirmed is the presence of fresh/brackish or brackish water. Therefore, more information is needed on the fresh or brackish water aquifers, including their locations and the ages of the waters.

As the shallow gases migrate upward and enter the aquifers filled with fresh/brackish water, they cause the water to be expelled (Judd and Hovland, 2007; Cathles et al., 2010). Inferred from the gas pockets present beneath the domes, unit 1, which overlies unit 2, could provide sufficient capillary resistance to prevent the upward movement of the free shallow gas bubbles. However, this resistance would not impede the upward migration of fresh/brackish water driven by the overpressure established by the accumulating shallow gases. The in situ seawater that remained in the interstitial sediments would subsequently be diluted by the intruding fresh/brackish water, as shown in Fig. 5.

The upward migration of liquid interstitial water also occurs along the conduits with minimum resistance. Because of the heterogeneity of the sediment layers, the conduits in unit 1 cannot be straight and may have different radii. At some specific depth intervals, the migration and accumulation of fresh/brackish water is favored along both the vertical and horizontal directions, indicated by a more dilute Cl^- and SO₄^2- contents. The dilution gradients in the centers and peripheries of the upward migration conduits will also display different patterns, and one or both of these may have caused the levels of Cl- and SO₄^2- in core d (Fig. 5).

Submarine groundwater is relatively enriched in CH₄ compared with ambient pore waters and seawater (Tsunogai et al., 1996; Moore, 1999; Lecher et al., 2015). Therefore, the slightly higher CH₄ concentrations observed in the pore waters in core d may be reinforced by the submarine groundwater that intruded into unit 1. Submarine groundwater may also contain other components that can be delivered during discharge and that can considerably influence the physical and chemical features of the seabed environment (Slomp and Van Cappellen, 2004; AtaieAshtiani and Seyedabbasi, 2006; Paytan et al., 2006; Taniguchi et al., 2008; Moore, 2010), and subsequently affect biological activities on the seabed. Many marine farms are distributed around the northwestern part of the NYS and are located close to the study area. Therefore, it is important to understand the submarine fresh/brackish groundwater discharge in the NYS to assess the feasibility of marine farms.

5.4 Formation of the seabed domes and surrounding moats

Based on the above information, the following simple scenario is suggested:

(a) Shallow gas from underlying sediments migrated upwards along paths of least resistance, such as the existing fractures or fissures in the layer/layers underlying unit 2. The shallow gas in unit 2 may have included some in-situ biogenic methane that was produced by the degradation of organic matter contained in the peat layer during the glacial periods.

(b) The upward migration of shallow gas produced conduits with minimal resistance, ultimately connecting unit 2 and the underlying layer/layers. The formation of the conduits favored the upward migration of shallow gas.

(c) Shallow gas continuously enters unit 2 and migrates upward until it reaches the bottom of unit 1, which was most likely the hard layer referred to by Chen et al. (2017) that can act as a cap layer to prevent the upward migration of shallow gas because of its lateral continuity and sufficient capillary resistance. Eventually, overpressure builds up through the accumulation of shallow gas beneath unit 1, and a seabed dome is formed by the mechanical deformation of the local sea floor (Barry et al., 2012; Koch et al., 2015).

This formation process suggests that the first mechanism described in the Introduction could also apply here. Furthermore, the total overpressure required to deform the seabed and form the domes in the study area and the heights of the gas columns that caused the establishment of the total overpressure can be calculated using the equations described by Barry et al. (2012) and Koch et al. (2015) based on the elastic thin-plate theory. The calculated overpressure values for D2 and D8, the most typical seabed domes in the study area, were small (much less than 1 Pa) because of the large radii of both domes. This indicates that subtle overpressure can deform the seabed and form dome structures. In other words, the formation of seabed domes may be easier than expected in the conventional way. The calculated gas-column heights for D2 and D8 were 14.5 and 8.8 m, respectively. These values were comparable with those recorded on the sub-bottom profiles, which were 14.5 m and 9.0 m for D2 and D6, respectively. These results demonstrate that the thin-plate theory could also be applied to the quantitative study of the dome structures in the NYS.

In addition, the intrusion of shallow gas into unit 2 replaced the interstitial pore water and drove the water upward into the overlying layer, eventually reaching the seabed.

(d) Shallow gas is prone to migrate upward driven by buoyancy. However, the continuity of the upper layer at the study site persisted and prevented the upward movement of the shallow gas after the dome was formed. The resistance that shallow gas must overcome to migrate horizontally is likely much less than the resistance encountered when migrating upward. Therefore, the shallow gas began to migrate horizontally and form horizontal conduits, which favor the escape of shallow gas from the gas pockets. The horizontal migration of the shallow gas decreased the previously developed overpressure, and the support for the upper layer disappeared, which eventually caused the subsidence of the surrounding sediments and formed the depression or moat structures.

Even subtle changes in seabed morphology can influence the seafloor stress field that is produced by strong tidally-driven bottom currents and can activate the feed-back processes that are associated with local dome and moat evolution. Therefore, the interactions between the morphology and the bottom currents may have facilitated accumulation at the dome sites and erosion at the moat sites and resulted in the relatively deeper southeast sides of the depressions.

6 CONCLUSION

Twelve seabed domes with surrounding depressions were discovered in our study area. We described these features based on acoustic and geochemical data and sediment core samples and proposed a scenario for the formation and evolution of the features. Based on our data, these features were created by the accumulation and migration of shallow gases. We contribute to our understanding of these anomalous structures in the NYS by providing acoustic data that indicate the presence of gas pockets under the domes and establish an alternative scenario for the formation and evolution of these structures. Furthermore, the horizontal migration of shallow gases most likely played a major role in the formation of the depression structure. Notably, the geochemical data indicate the presence of fresh/brackish water in the seabed subsurface of the study area. This finding can be useful in guiding further shallow gas and SGD research in the NYS. However, the role that seabed groundwater discharge played in the formation and evolution of these features is still unknown. Furthermore, many other uncertainties remain, including the sources of the shallow gases and fresh/ brackish groundwater, and more research is still needed.

7 DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.

8 ACKNOWLEDGEMENT

We thank Peter G. Brewer for his careful editing of the manuscript and the helpful comments. We thank the officers and crew of the RV Kexue Ⅲ for their skilled assistance during our operations. We are grateful to the members of the Research Center of Analysis and Measurement, Institute of Oceanology, IOCAS, and colleagues from the OUC for their precise chemical test results. We would also like to express our sincere thanks to the reviewer for the valuable comments and suggestions, which were helpful in improving our manuscript.