Introduction
Submarine groundwater discharge (SGD) is the flow of groundwater from the seafloor to the coastal sea (e.g. Burnett et al.
2003; Moore
2010). It has been observed as a source of nutrients, trace metals and other potentially harmful substances from land to coastal ecosystems (e.g. Moore
2010; Szymczycha et al.
2012,
2016; Luijendijk et al.
2020). Growing recognition of the importance of SGD on the environmental state of coastal sea areas has resulted in an increasing number of studies, aimed at estimation of the rates of SGD and associated fluxes by different kinds of methods (e.g. Burnett et al.
2003,
2008; Schlüter et al.
2004; Peterson et al.
2008; Gleeson et al.
2013; Tait et al.
2013; Schubert et al.
2014; Sadat-Noori et al.
2015; Szymczycha et al.
2012,
2016; Krall et al.
2017; Idczak et al.
2020). Among those methods, numerical groundwater flow models have been developed as a means to estimate and predict the SGD and associated fluxes under current and future conditions (e.g. Andersen et al.
2007; Langevin
2003; Kaleris et al.
2002). Three-dimensional (3D) geological models represent the geometry and stratigraphy of depositional sequences, and provide the geological framework for groundwater flow models. Successful estimates of the SGD and associated fluxes to the coastal sea by numerical flow modeling require good understanding of the aquifer 3D geometry, the distribution of porosity in the host sediments, and the groundwater flow pathways.
Shallow groundwater in northern Europe typically resides in the thin overburden of late Pleistocene and Holocene glacial and post-glacial sediments that immediately overlie the Precambrian crystalline bedrock. These groundwater areas are hydrologically complex with heterogeneous sedimentary cover including till, glacioaquatic gravel and sand, glaciolacustrine and post-glacial silt and clay, and in some areas, reworked littoral gravel, sand and clay (Saarnisto and Saarinen
2001; Donner
2010; Räsänen et al.
2009). Major glacigenic landforms with coarse-grained sediments are important groundwater reservoirs in the region, including the First Salpausselkä ice-marginal formation (SSI) that continues tens of kilometers along the Hanko Peninsula in the south coast of Finland, and to the Baltic Sea (Häkkinen
1982; Fyfe
1990; Kielosto et al.
1996). For many locations along the Hanko Peninsula, groundwater flow models imply fresh groundwater discharge from the SSI to the coastal Baltic Sea (Luoma and Okkonen
2014). Indeed, Virtasalo et al. (
2019) recently discovered the first SGD site in Finland, which is connected to the SSI. The groundwater discharge is associated with pockmarks on the seafloor at seawater depths between 4 and 17 m, with estimated SGD rates from
222Rn measurements of approximately 0.4–1.2 cm day
−1. Comparable and larger SDG rates have been documented for other sites in the Baltic Sea (e.g. Schlüter et al.
2004; Szymczycha et al.
2012,
2016). However, the significance of SGD and associated fluxes on the overall eutrophic environmental state of the Baltic Sea is poorly understood.
The objectives of this study were to construct a 3D geological model of the Hanko SGD site by integrating previously collected and new geophysical and geological data, and, based on this knowledge, to develop a numerical groundwater flow model in order to provide estimates of the SGD rate to the coastal sea. The 3D geological model is built under a geographic information system (GIS) platform and the 3D stratigraphic module in the Groundwater Modeling Software (GMS). The groundwater flow model is constructed by using MODFLOW 2005 code (Harbaugh et al.
2017). Due to the low density and low salinity of the northern Baltic Sea, the density-dependent model is not applied.
Discussion
Coarse-grained outwash channel fills are important seaward groundwater conduits in the glacigenic SSI ridge. A NW–SE oriented bedrock depression, which is the interpreted direction of the glacial meltwater paleocurrent, could represent the main route of sediment transport to the ice contact lake, as high proportions of coarse-grained sediments are found in this area. In many places along this depression, the coarse-grained sediments were deposited as high-porosity, channel-fill outwash sediments perpendicular to the SSI ridge morphology (Fyfe
1990). On the other hand, depressions of the bedrock topography in the NE–SW direction, parallel the SSI ridge, contain thick glacial and post-glacial fine-grained sediments that confine the aquifer. Fyfe (
1991) described a channel-fill outwash of highly porous gravels and sands with thickness of approximately 7 m that cut across the SSI in a gravel pit in Tammisaari, approximately 20 km north-east of the study area. These channel-fills could be the main pathways for the groundwater flow, especially in the shallow shore platform where the fine-grained layers are thin or absent. Virtasalo et al. (
2019) documented a number of pockmarks on the shore platform slope, where sandy sediments are exposed on the seafloor.
The simulation of the groundwater flow model was performed with sparse data and many constraints. The recharge was associated with the
K values during the calibration. Based on the assigned recharge, the calibrated
K values of the aquifer materials are quite low, generally less than 5 m day
−1. However, the low
K values could be caused by the origins of the sediments, especially in the deeper part of the aquifer (layer 2), where the sediments are a mixture of coarse and fine-grained sediments of the primary deposits with increasing proportions of fine-grained sediments in the distal part. Virtasalo et al. (
2019) concluded that the aquifer sediments, at least in the offshore area, are mainly composed of fine sands, and represent the distal part of a subaqueous ice-contact fan. Interbeds of coarser sand and gravel provide conduits of higher hydraulic conductivity for the groundwater flow, and lead to the focusing of the flow to the pockmark sites.
The simple groundwater discharge rate was estimated based on the Darcy equation between two known groundwater level locations as follows:
$$ V=K\times i\ {n}^{-1} $$
(4)
where
V is the seepage velocity of the groundwater flow [L T
−1],
K is hydraulic conductivity [L T
−1],
i is the hydraulic gradient [L L
−1], and
n is the porosity [L L
−1]. Simple calculations of the SGD rates at different parameters and locations are presented in Table
4. Head 1 is groundwater level at the groundwater divide during autumn 2017, and head 2 is groundwater level at the shoreline. The results show that the SGD rates vary depending on the hydraulic gradient,
K values and porosity values.
Table 3
Sensitivity of SGD rates (%) to parameter changes (10–50% increase) from the autumn 2017 condition. Positive numbers indicate an increase in SGD rate (%), and vice versa
The whole of the model domain: |
Recharge, cm day−1 | 0.0–0.082, 0.028 | 8.8 | 17.9 | 27.6 | 37.3 | 47.1 |
K value, m day−1 | 0.01–35.0, 0.86 | 1.0 | 1.9 | 2.7 | 3.5 | 4.2 |
GHB conductance, m2 day−1 | 3.02 × 10−5–1.6, 0.10 | 0.1 | 0.16 | 0.22 | 0.28 | 0.33 |
RIV conductance, m2 day−1 | 3.02 × 10−5, − | 0.01 | 0.01 | 0.01 | 0.01 | 0.01 |
The GHB (shore platform) area: |
GHB conductance, m2 day−1 | 3.02 × 10−5–1.6, 0.10 | 0.16 | 0.30 | 0.42 | 0.52 | 0.61 |
Pockmark B conductance, m2 day−1 | 0.08–0.25, 0.19 | −6.7 | −14.2 | −20.5 | −26.1 | −30.9 |
Pockmark D conductance, m2 day−1 | 0.08–0.12, 0.095 | −6.5 | −13.7 | −19.8 | −25.0 | −29.6 |
Pockmark E conductance, m2 day−1 | 0.001–0.045, 0.037 | −6.8 | −14.2 | −20.6 | −26.1 | −30.9 |
The RIV (offshore) area: |
RIV conductance, m2 day−1 | 3.02 × 10−5, − | 1.02 | 1.50 | 1.99 | 2.45 | 2.91 |
The groundwater flow simulations estimated the SGD to the Baltic Sea at the average rates of 0.22 cm day
−1 (range 0.0–1.21 cm day
−1) and 0.28 cm day
−1 (range 0.0–1.60 cm day
−1) per square meter of the seafloor for autumn 2017 and spring 2020, respectively. Virtasalo et al. (
2019) estimated the SGD rates based on the
222Rn measurements of groundwater samples from borehole HP101, and pockmarks E, D and B during autumn 2017. The SGD rates varied between 0.4 and 1.2 cm day
−1 between pockmarks E and B, respectively. These rates are consistent with the results of this study at the
K values of 0.1 m day
−1 (Table
4), which is expected for fine sands in the pockmarks. However, higher SGD rates may take place locally on the shore platform and slope areas with higher
K values. Virtasalo et al. (
2019) observed more active pockmarks in the eastern part of the shore platform. Sediment samples collected from pockmarks D and B in the east are composed of fine sand, whereas in pockmark E the fine sand is covered with soft organic-rich mud. The organic-rich mud layer could continue to the west, where the thick layer of fine-grained sediments is found in the area around SW3-Ova8 (Fig.
2). This thick fine-grained layer could prevent SGD in the western part of the shore platform. A similar setting was described by Andersen et al. (
2007) where the SGD was largely controlled by the geological structure of the aquifer and the fresh groundwater discharge predominantly occurred within a narrow zone of the upper 10–15 m of the intertidal zone along the shoreline.
Table 4
Estimated SGD rates at different parameters and locations along the shoreline (see Fig.
2 for locations)
B′ and B | 306 | 7.8 | 0.0 | 0.0255 | 0.20 | 63.7 | 12.7 | 1.27 |
D′ and D | 406 | 8.8 | 0.0 | 0.0216 | 0.20 | 54.0 | 10.8 | 1.08 |
E′ and E | 550 | 8.7 | 0.0 | 0.0157 | 0.20 | 39.3 | 7.9 | 0.79 |
HP100 and shoreline | 1075 | 13.1 | 0.0 | 0.0122 | 0.20 | 30.5 | 6.1 | 0.61 |
B′ and B | 306 | 7.8 | 0.0 | 0.0255 | 0.10 | 127.5 | 25.5 | 2.55 |
D′ and D | 406 | 8.8 | 0.0 | 0.0216 | 0.10 | 108.0 | 21.6 | 2.16 |
E′ and E | 550 | 8.7 | 0.0 | 0.0157 | 0.10 | 78.6 | 15.7 | 1.57 |
HP100 and shoreline | 1075 | 13.1 | 0.0 | 0.0122 | 0.10 | 60.9 | 12.2 | 1.22 |
The total submarine groundwater discharge was estimated at approximately 773.2 m
3 day
−1 in autumn 2017 and at 986.7 m
3 day
−1 in spring 2020 (Table
2). The values are modest compared to the estimated annual average direct groundwater discharge from the whole Finnish coastline to the Baltic Sea, which is approximately 1,000,000 m
3 day
−1 (Peltonen
2002). Direct groundwater discharge from comparable glacial meltwater-deposited landforms in Finland to the sea has been estimated at 173,000 m
3 day
−1 (2 m
3 s
−1; Mälkki
2003). For comparison, the mean annual discharge of the nearby Karjaanjoki River is 1,650,000 m
3 day
−1 (Saura et al.
2010).
The simulated flux rates of SGD for autumn 2017 and spring 2020 showed high SGD flux rates in the shallow shore platform along the shoreline, where the coarse-grained sediments are exposed directly to the sea. These agree with the GPR profiles and drill log data where the reflections of the coarse-grained materials were observed in the shallow part of the aquifer. Also, the wave reworking of the shore platform surface may have removed fine-grained sediments and increased the hydraulic conductivity of the aquifer surface materials, thereby enhancing groundwater discharge to the sea. However, groundwater influence was not detected during the field investigation of the shallow water area. The actual SGD rate could possibly be very low and the walking survey measuring the EC and temperature in the shallow sea area may not have been suitable for the measurement under high wave conditions. On the other hand, this may imply seasonal variation of the SGD. The simulation indicated strong relationship between the SGD and recharge. Based on the monitoring data, the highest recharge occurs during spring (April to early May). Virtasalo et al. (
2019) observed the melted spots on the sea ice along the shoreline during spring, which could be caused by the discharge of warmer groundwater (average temperature of 7 °C). A similar situation was described by Szymkiewicz et al. (
2020), where the rate of SGD varies seasonally and in relation to recharge, with peaks in late winter to early spring. In this study, the measurement of EC and temperature at the sea bottom along the shoreline was carried out in summer when the groundwater level and recharge are low. As a consequence, the SGD rate to the seafloor could have been low during that time. Based on the EC data, the submarine fresh groundwater portion of the total SGD flux can be estimated by using the following equation:
$$ \mathrm{Qf}=\mathrm{Qt}\times \left(\mathrm{ECm}\hbox{--} \mathrm{ECf}\right)/\left(\mathrm{ECs}-\mathrm{ECf}\right) $$
(5)
where Qf is simulated fresh groundwater discharge per unit area of seafloor per unit time [L
3 L
−2 T
−1], and Qt is simulated total groundwater discharge [L
3 L
−2 T
−1]. ECm, ECf and ECs are the EC values of the measured shore-platform water, fresh groundwater and the Baltic Sea water, respectively.
The measured EC values on the shore platform varied between 1,030 and 1,080 ms m−1. Given the EC of the Baltic Sea water of 1,080 ms m−1, the EC of fresh groundwater from observation well HP101 (inland) of 10 ms m−1, and the lowest measured value of 1,030 ms m−1 on the shore platform, the Qf was estimated at 0.0467 Qt, which is a very small fraction of freshwater discharge to the seafloor during summer.
Groundwater level monitoring data are available from only one observation well (HP101) and may not represent changes in the groundwater level throughout the study area. This caused high uncertainty in the transient simulation. By applying the same configuration of the aquifer materials, e.g. the same K values and seafloor conductance, the transient simulation results could provide the estimated SGD under different seasonal recharge rates. The long-term groundwater level monitoring data contributed to the flow modeling approach, making it possible to estimate the SGD under the sparse data condition. However, it should be noted that in the absence of any data on discharges, simultaneous calibration of permeability and recharge leads to inherently nonunique results.
The sensitivity analyses using the steady-state simulation of the autumn 2017 condition reveal that at the model domain scale, changes in groundwater recharge cause larger changes in the SGD rate compared with the aquifer
K value and the seafloor conductance. This indicates that groundwater recharge has a larger contribution to the variation in the SGD rate in this area. An increase in groundwater recharge will cause more SGD into the coastal sea. In addition, the weather and groundwater monitoring data indicate a positive correlation between groundwater recharge and precipitation. Based on the future climate change scenarios, precipitation is predicted to increase in winter, whereas evapotranspiration is predicted to increase during summer in southern Finland (Olsson et al.
2015). This would increase the winter time SGD to the Baltic Sea. The sensitivity analysis of the offshore area reveals the low impact of parameter changes on the SGD rates in the low-permeability seafloor areas. However, in the specific high-permeability seafloor regions, such as the areas around the pockmarks B, D and E, the SGD rates decrease as the seafloor conductance at the GHB increases. The increase in conductance of the seafloor could create more uniform flow in a large area, which could reduce the SDG rate in the pockmark areas. This indicates that any changes that enhance the seafloor conductance, e.g. wave erosion, could alter the SGD rates and patterns. In addition, analysis shows the high sensitivity of the SGD rate in the permeable seafloor regions, indicating the need for site investigations and sufficient groundwater discharge data for model calibration and validation.
Conclusions
A three-dimensional geological model of the shallow coastal aquifer belonging to the First Salpausselkä ice marginal formation in the northern Baltic Sea was carried out in Isolähde-Lappohja area, southern Finland. The late Pleistocene and Holocene depositional succession in the study area consists of till, glacial coarse-grained and fine-grained sediments, post-glacial fine-grained deposits, and reworked coarse-grained littoral and aeolian deposits. The aquifer is composed of the glacial and postglacial coarse-grained sediments, which were deposited in a NW–SE oriented bedrock depression, in the direction of glacial meltwater discharge. The aquifer is exposed on the shallow shore platform that extends approximately 100–250 m offshore from the shoreline, where the unit slopes steeply seawards and becomes covered by a thick layer of glacial and post-glacial muds. Groundwater flow preferentially takes place in the channel-fill outwash coarse-grained sediments and in sand and gravel interbeds that provide conduits of higher hydraulic conductivity, and have led to the formation of pockmarks on the shore platform edge and slope at water depths between 4 and 17 m.
A two-layer groundwater flow MODFLOW model was constructed based on the results of geological and hydrogeological data for both steady- and transient states. The steady-state flow model was run with the calibration of 97 groundwater level data points from the observation wells and GPR profiles measured in autumn 2017 (correlation coefficient R2 of 0.99 and RMSE of 0.37 m). The transient flow model was calibrated with the daily monitoring groundwater level data taken during February 2019 to April 2020 (R2 of 0.96 and RMSE of 0.10 m). The simulation results estimated the average SGD rate to the Baltic Sea at 0.22 cm day−1 in autumn 2017. The average SGD rate increased to 0.28 cm day−1 as a response to an approximately 30% increase of recharge in spring 2020. These values are consistent with the previous SGD rate estimates based on 222Rn measurements in this area. Results of the simple calculation using hydraulic gradient show that the SGD rates vary with the changes of the hydraulic gradients, hydraulic conductivity and porosity of the aquifer media. Furthermore, the sensitivity analyses reveal that at the model domain scale, recharge has a larger contribution to the variation in the SGD rate compared with aquifer K value and the seafloor conductance. While at the local scale in the offshore area, the seafloor conductance in the high-permeability area in the shallow shore platform has a larger impact on the SGD rates than in the low-permeability regions in the deeper offshore area.
The groundwater flow models were successfully developed with sparse data but still contain uncertainties. More site investigations with emphasis on recharge estimation, and groundwater discharge data, are required for the calibration and validation of models in the future.
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.