Seven seismic units (SUs) are recognized in offshore sub-bottom profiles that were collected along shore-normal transects and along the shore platform edge (Fig. 2a). The denser units SU1–SU3 are identified in the deeper-penetrating reflection seismic profiles (Fig. 3c, d), whereas the units SU4–SU7 of lower acoustic impedance are better visible in the CHIRP profiles (Fig. 3a, b). The pinger had poor penetration in this area, likely because of the high sand content in the sediments.

Three radar units (RUs) are recognized in the onshore GPR profiles, which were recorded along survey lines that were oriented to continue the offshore seismic survey lines, and along perpendicular survey lines that followed the arc-shaped shoreline and the shore-parallel road 500–800 m inland (Fig. 2a). The penetration depth of the GPR electromagnetic wave in the studied deposits is ca. 12 m (Fig. 4a). The available drill logs from the survey area show, however, that the crystalline bedrock can be covered by more than 50 m of sand with silty and gravelly interbeds and lenses. The GPR profiles, therefore, capture only the upper part of the sandy deposits. The groundwater table is observable as a high-amplitude continuous reflector, which is at sea level in the seaward ends of the profiles, but rises to ca. 12 m a.s.l. within 1 km distance to the shoreline. The groundwater is hosted in unit RU1. The collected GPR profiles are available in the Supplement to this article.

The 900 m long refraction seismic profile was collected along the shoreline on the sandy beach (Fig. 2a). A refraction surface at a depth of ca. 50 m in the western part of the profile rises to ca. 20 m depth in the eastern part of the profile (Fig. 4c). This surface is interpreted to be the bedrock surface that is unconformably overlain by the sand-dominated subaqueous ice-contact fan deposits. This interpretation is in line with the available drill logs from the area, which show that the bedrock surface can be buried deeper than 50 m below sea level. Another refraction surface at sea level is interpreted to be the groundwater table.

Multibeam bathymetry over the survey area shows a gently undulating seafloor with water depths ranging between 5 and 25 m (Fig. 2b). The seafloor is covered with brackish-water mud (SU7), except erosional exposures of till (SU2) at small elevations in the southwestern and eastern parts of the area. An arc-shaped shore platform with sandy sediment extends 100–250 m seaward from the shoreline, sloping gently to ca. 4 m water depth (Fig. 2b). From the platform edge, the seafloor slopes to ca. 17 m depth within ca. 50 m distance. The shore platform was too shallow for R/V Geomari to navigate, and therefore no seismoacoustic surveys were carried out there.

The multibeam image shows approximately twenty pockmarks up to 25 m wide and 2 m deep (Fig. 5) on the shore platform slope and at the base of the slope down to ca. 16 m water depth (Fig. 2b). In CHIRP profiles, the pockmarks are incised in the topmost unit SU7 (Fig. 6d). Sub-vertical columns of disrupted reflectors, which are interpreted to be groundwater conduits, extend down from the pockmark base through SU7–SU5 to the upper well-stratified part of SU4 (glaciolacustrine rhythmite) at least. The seismic signal is attenuated below this stratigraphical level, likely because of the higher sand content, and the full vertical extents of the disrupted reflector columns are unclear in the profiles. The columns tend to be localized at elevations in the SU4 top, which may have acted as natural leak-off points for overpressured pore fluids (Cartwright et al., 2007). Reflectors are commonly folded up along the column margins in SU4 and SU5, in line with upward fluid flow. In the upper part of SU6 and in SU7 (brackish-water mud drifts), however, the reflectors are folded down along column margins, recording the upward migration of pockmark base with sediment deposition (Cartwright et al., 2007).

The box-core sediment sample (MGBC-2018-2) collected from pockmark D is composed of structureless fine sand with very low organic content (Fig. 7a). The sample MGBC-2018-1 from pockmark B is very similar. In contrast, the sample MGBC-2018-3 collected from pockmark E is covered with a 7 cm thick surface layer of organic-rich mud, which is underlain by structureless fine sand of very low organic content (Fig. 7b). The surface of MGBC-2018-3 is overgrown by Stuckenia spp. Caesium-137 activity contents in the sand in MGBC-2018-3 are below 20 Bq kg-1, which is a normal level for sediments deposited before the 1986 Chernobyl nuclear disaster, whereas the 137Cs contents are substantially higher in the surficial mud layer (range 27.9–44.2 Bq kg-1). However, these values are still low compared to the present values exceeding 1000 Bq kg-1 in sediments that were deposited off the Finnish south coast soon after the disaster (e.g. Jokinen et al., 2015; Vallius, 2015). It appears that the 137Cs-enriched mud layer that covers pockmark E was deposited sometime after 1986, possibly within the past several years. The measured LOI and 137Cs values are available in PANGAEA (Virtasalo et al., 2019).

Profiles measured by the CTD90M multiprobe show that the seawater is stratified with respect to temperature, with the thermocline located at 5–8 m depth (Fig. 6a–c). The temperature range is 13.2–15.9 ∘C above the thermocline, and 7.8–10.2 ∘C below. The salinity range is 4.61–5.13 PSU in the upper layer, whereas the bottom layer has a narrower range of 5.44–5.54 PSU. The water mass is well oxygenated with generally higher values in the bottom layer. Turbidity is low with slightly higher values in the upper layer. The measured hydrographic data are available in PANGAEA (Virtasalo et al., 2019).

The mean 222Rn activity concentration in the seawater surface is 16.9 Bq m-3, with locally slightly elevated concentrations of up to 49.9 Bq m-3 near the shoreline (Fig. 2c; Table 1). These elevated concentrations may result from localized groundwater seepage through the beach sand and/or from the local upwelling of 222Rn-bearing water from below the thermocline. The measured surface seawater 222Rn concentrations are available in PANGAEA (Virtasalo et al., 2019).

Significantly higher 222Rn activity concentrations were measured ca. 1 m above the bottom at pockmarks D and B (134.7 and 156.9 Bq m-3, respectively; Fig. 6b, c; Table 1). Above pockmark B, an elevated 222Rn concentration (39.5 Bq m-3) was measured in the upper part of thermocline at 5.1 m water depth, and a concentration of 21.8 Bq m-3 was measured at 1.5 m water depth, which is in the range of surface seawater concentrations. Apparently, the thermocline reduces the mixing of 222Rn-enriched bottom waters with surface waters. Notably, 222Rn concentrations measured ca. 1 and 2 m above the bottom of pockmark E (47.9 and 21.5 Bq m-3, respectively) are in the range of seawater surface and thermocline concentrations, and significantly lower than at pockmarks B and D. Therefore, 222Rn measurements indicate no recent SGD from pockmark E because the measured concentrations may result from 222Rn diffusion from sediments and/or advection. Flow directions measured below the thermocline at pockmark E (Fig. 6a) range between west and north, which makes 222Rn advection from pockmarks D and B in the west likely.

Radon activity concentrations in groundwater, measured in the observation well HP101 (12 129 Bq m-3) and in the obsolete water station that is located 7 m inland from the shoreline (37 433 Bq m-3), are substantially higher than those measured in seawater (Table 1).

Marine seismic reflection and onshore GPR profiling methods are similar in that both are based on wave propagation through water and sediments, and on wave reflection from subsurface structures and boundaries. However, they are different in the type of wave used for profiling. The seismic wave is reflected from interfaces where the acoustic impedance of sediment sharply changes (e.g. Sheriff and Geldart, 1995), whereas the GPR electromagnetic wave is sensitive to water content, and the ability of sedimentary structures and layers to hold water governs GPR reflections (Annan, 2009). The seismic refraction method is different from the reflection methods in that the seismic wave returns to the surface by refraction at subsurface interfaces with a strong acoustic impedance contrast (e.g. Sheriff and Geldart, 1995). Because of the technological differences, profiles obtained by these methods record different properties of the subsurface, and, therefore, are not inter-comparable. Significant unconformities in sediments, however, are typically associated with a change in the sediment acoustic impedance and water content, which is expressed in the different profiles by strong reflectors that can be correlated between the profiles following an allostratigraphical approach.

The use of allostratigraphy (North American Commission on Stratigraphic Nomenclature, 2005) is recommended for glacial and glacially influenced strata, which typically have frequent unconformities and high lithological heterogeneity that complicates lithostratigraphical classification (e.g. Virtasalo et al., 2007, 2014; Räsänen et al., 2009). Indeed, sediment deposition and erosion in the study area have been controlled by several independent processes, namely the retreat of the Fennoscandian continental ice sheet and the dynamics of the ice margin, the relative water level fall due to postglacial land uplift, the mid-Holocene marine incursion and the associated short-lived transgression as well as the establishment of brackish-water conditions, and, finally, coastal and aeolian processes (Virkkala, 1963; Glückert, 1986; Fyfe, 1990; Kujansuu et al., 1993; Virtasalo et al., 2007, 2014, 2016). As a result of the complex depositional history, allostratigraphy has recently gained popularity as a stratigraphical classification approach both in the Baltic Sea (e.g. Virtasalo et al., 2007, 2014; Tsyrulnikov et al., 2012; Hyttinen et al., 2017; Jensen et al., 2017) and on the surrounding land areas (Räsänen et al., 2009; Ojala et al., 2018).

In this study, allostratigraphy is used, for the first time, for the correlation of unconformities recognized in offshore seismic profiles to those in onshore GPR and refraction seismic profiles in order to establish the local stratigraphic architecture across the shoreline. Five significant unconformities are recognized in the offshore seismic profiles (Fig. 3), whereas two are recognized in the onshore GPR profiles and one in the refraction seismic profile (Fig. 4). The offshore unconformities are correlated with those onshore based on the interpreted composition and sedimentary environment of their bounded units in the deglacial to postglacial depositional succession (Fig. 8).

The lowermost unit with identifiable bottom and top reflectors in the offshore reflection seismic profiles is till (SU2; Fig. 3). The till layer is not discernible in the GPR profiles (Fig. 4a, b), which may be due to (1) the insufficient depth penetration of the electromagnetic wave, (2) the thinness or even absence of the till layer because of the higher bedrock elevation on the land area, or (3) the poor penetration of the electromagnetic wave in tills with a fine-grained matrix (e.g. Sutinen, 1992). Likewise, no surface that could be associated with a till layer covering the bedrock is observed in the refraction seismic profile (Fig. 4c). Geological mapping shows that bedrock elevations in the area often lack till cover (Kujansuu et al., 1993). Furthermore, unpublished drill logs of groundwater observation wells in the study area document no till layer on the bedrock.

The only unit that can be traced from the offshore seismic profiles to the onshore GPR profiles is the subaqueous ice-contact fan on land (RU1) and its most distal part offshore (SU3; Fig. 8). Such subaqueous ice-contact fans form a significant part of the First Salpausselkä ice-marginal formation (Virkkala, 1963; Glückert, 1986; Fyfe, 1990; Kujansuu et al., 1993). The unit is erosionally exposed at seafloor on the shore platform.

The glaciolacustrine, postglacial lacustrine, and lower brackish-water deposits (SU4 to SU6) were deposited over the whole study area (Virtasalo et al., 2007, 2014), but have since been removed from the GPR-profiled land area as a consequence of erosion by waves, currents, and winter ice during emergence from the sea. The coarse lag sheet (RU2) in the GPR profiles is the product of this erosion and associated preferential removal of fine-grained material. The overlying beach ridge deposits (RU3) were formed at the beachface during emergence and by later aeolian processes. The upper brackish-water mud unit (SU7) represents modern mud drift deposition at sea.

The main coastal aquifer on the Hanko Peninsula is hosted in the First Salpausselkä ice-marginal formation (e.g. Luoma et al., 2015), which is a ridge of contiguous subaqueous ice-contact fans and local feeding eskers (Virkkala, 1963; Glückert, 1986; Fyfe, 1990; Kujansuu et al., 1993). The studied SGD site is situated in the distal part of a subaqueous ice-contact fan (RU1-SU3; Fig. 8). Based on low reflector amplitudes in the offshore seismic profiles (SU3; Fig. 3c, d) and available unpublished drill logs from the land area, the subaqueous fan deposits are dominated by fine sand, which typically has moderate hydraulic conductivity (e.g. Bear, 1972). However, GPR profiles and the drill logs show interbeds and lenses of coarse sand and gravel in the subaqueous fan. Slug tests carried out in the area demonstrate high hydraulic conductivities for such coarser beds (K<13.8 m d-1; Luoma and Pullinen, 2011). It is thus likely that the interbedded coarser deposits in the fine-sand-dominated distal part of the subaqueous fan provide conduits for localized groundwater flow to pockmarks that are visible on the shore platform slope (Figs. 2b, 2d, 5). In addition, coarse couplet layers in the superimposed glaciolacustrine rhythmite can act as groundwater conduits to the pockmarks (Söderberg and Flodén, 1997; Virtasalo et al., 2007). The offshore stratigraphical units – till (SU2), postglacial lacustrine poorly bedded silty clay (SU5), and organic-rich brackish-water mud (SU6 and SU7) – have low hydraulic conductivities, and the silty clay and mud units likely act as confining layers (e.g. Bear, 1972). No faults are observed in the studied profiles, and the groundwater conduits do not seem to be linked with faulting.

Significantly elevated radon activity concentrations measured 1 m above pockmarks D and B strongly indicate SGD from these locations (Fig. 4b, c). In comparison, 222Rn concentrations measured above pockmark E are not elevated but similar to those measured in the thermocline above pockmark D and in surface seawater (Figs. 2c, 4a; Table 1), which precludes significant SGD from pockmark E at the time of measurements. It can thus be concluded that pockmarks are the most significant locations of SGD in the study area, and that not all of the pockmarks observed in the multibeam and side-scan sonar images are active. Although pockmarks are known to form also as a result of gas seepage from seafloor (e.g. Hovland et al., 2002), the significant groundwater (radon) flux from the studied pockmarks implies submarine groundwater discharge as the most likely mechanism of their formation. Furthermore, low organic contents make significant methane production less likely in the glacigenic deposits.

In order to roughly estimate the rate of SGD from pockmark B, the 222Rn inventory of ∼781 Bq m-2 is calculated on the basis of concentration gradient between pockmark B bottom and the thermocline upper boundary. Assuming steady state and that SGD is the only major source of radon, the calculated 222Rn inventory will decay at the rate of 142 Bq m-2 d-1, which must be balanced by SGD. Using the measured 222Rn concentrations of groundwater in the well HP 101 and the obsolete water station (Table 1) we arrive at SGD rates between 0.4 and 1.2 cm d-1. This simple model does not consider the lateral export of 222Rn due to bottom currents, for example. The 222Rn inventory and SGD rates, therefore, must be considered lower limits. These first SGD rate estimates can be improved, for example, by a more robust 222Rn mass balance as outlined in Burnett et al. (2006) by advection and diffusion modelling of porewater Cl profiles (e.g. Schlüter et al., 2004), and by three-dimensional groundwater flow modelling (e.g. Luoma and Okkonen, 2014).

Sediment samples collected from pockmarks D and B are composed of fine sand, whereas pockmark E is covered with soft organic-rich mud (Fig. 7). Considering the lack of SGD from pockmark E as demonstrated by the low 222Rn activity concentrations, it would seem plausible that SGD prevents permanent organic-rich mud deposition in pockmarks D and B. Indeed, our unpublished video observations show that percolating subsurface fluids cause the resuspension of fine sediment in the pockmarks (see also Schlüter et al., 2004). The resuspended fine sediment is then removed by near-bottom flows, resulting in lower relative sedimentation rate and coarser grain size in the pockmark (Hammer et al., 2009). The absence of significantly elevated rims around the pockmarks (Figs. 5, 6d) further supports the efficient lateral sediment transport and precludes significant subsurface sediment mobilization and transport of particles to the seafloor (Loher et al., 2016). 137Cs activity contents in the organic-rich surface mud layer in pockmark E are at the post-Chernobyl level (Fig. 7b), which suggests that SGD from the pockmark ceased several years after 1986.

SGD depends on the hydraulic head gradient between the elevated part of the aquifer on land, and the top of the permeable unit at sea (Burnett et al., 2003; Moore, 2010). The GPR profiles show that groundwater is hosted in the subaqueous fan deposits (RU1; Fig. 4). Seaward groundwater flow at the study area could have initiated only after the subaqueous fan deposits had been uplifted sufficiently high above sea level in order to produce the required hydraulic head. The change in the direction of folding of reflectors along the margins of groundwater conduits from a folded-up to a folded-down direction in the upper part of SU6 in the CHIRP profiles indicates a shift from the post-depositional folding by upward fluid expulsion to the upward migration of the pockmark bottom with sediment deposition (Fig. 6d; Cartwright et al., 2007). This timing of the formation of pockmarks and the initiation of SGD must have taken place after the beginning of the brackish-water mud (SU6) deposition at ca. 7600 years ago (Virtasalo et al., 2007). According to shore displacement studies, the subaqueous fan deposits that today rise as much as 20 m above sea level in the area began to emerge from the sea ca. 5000 years ago (Eronen et al., 2001), which is in good agreement with the reflector-folding interpretation of CHIRP profiles. Probably it took several hundreds of years before the subaqueous fan deposits were uplifted sufficiently high above sea level for the initiation SGD and the formation of first pockmarks.

Radon activity concentrations measured in the groundwater observation well (HP101, 12 129 Bq m-3) and in the obsolete water station (37 433 Bq m-3) are substantially higher than those measured in seawater (Table 1). The measured values are below the mean 222Rn activity concentration of 92 000 Bq m-3 in 961 springs and dug wells in southern Finland (Salonen, 1988), and well below the upper permissible limit for 222Rn in public drinking water in Finland that is set to 300 000 Bq m-3 (Ministry of Social Affairs and Health, 2001).