Discharge of groundwater flow to the Potter Cove on King George Island, Antarctic Peninsula

There is only a small number of recent publications discuss glacial runoff in Antarctica and even fewer of them deal with the groundwater flow discharge. This paper focuses on the groundwater flow aspects and is based on a detailed study performed on a small hydrological catchment, informally called "Potter Basin", located on King George Island (KGI; Isla 25 de Mayo), South Shetland Islands, at the northern tip of the Antarctic Peninsula. This region has experienced drastic climatological changes within the past five decades. The basin is representative for the rugged coastline of the Northern Antarctic 5 Peninsula, and is discussed as a case study for possible future evolution of similar basins further to the South. A conceptual hydrogeological model has been defined using vertical electrical soundings (VES), geological and hydrogeological surveying methods, geomorphological interpretation based on satellite imagery, permeability tests, piezometric level measurements, meteorological, geocryological and glaciological data sets. The transmissivities of the fluvial talik aquifer and suprapermafrost aquifer range from 162.0 to 2719.9 10−5 m s−1, and in basaltic fissurated aquifers from 3.47 to 5.79 10−5 m s−1. The 10 transmissivities found in the active layer of hummocky moraines amount to 75.23 ·10−5 m s−1, in sea deposits to 163.0 ·10−5 m s−1, and in the fluvioglacial deposits they were observed between 902.8 and 2662.0 10−5 m d−1. Finally, the groundwater flow discharge was assessed to 0.47 m s−1 (only during January and February), and the total groundwater storage was estimated to 560 ·103 m. This data can be used to adjust the local glacial mass balance and to improve the understanding of coastal sea water processes in Potter Cove and their effects on the local marine biota, as a consequence of the global climate 15

Ermolin and Silva Busso (2008) observe a mean annual soil temperature of -2.0 • C which is slightly warmer than elsewhere on the AP. Lisker (2004) proposes an average homogeneous geothermal gradient for the AP of 0.03 • C considered representative 5 for continental Antarctic areas. Based on this information, the maximum thickness of permafrost in the hummocky moraines and side blackberries is estimated to between 70 -80 m.
The permafrost found here is comparatively warm (mean annual ground temperatures are greater than -2.0 • C) and thin (less than 80 m). Assuming a homogeneous medium, a geothermal gradient of 0.03 • C/m is considered representative for continental Antarctic areas (Lisker, 2004). This yields a maximum permafrost thickness of about 70 -80 m in bottom moraines and terminal 10 Holocene moraines in the study area, and about 40 and 60 m in the present moraine near the border of Fourcade glacier. In the fluvioglacial plain, gravity slopes (that are not solifluction slopes) and wetland, the permafrost thickness is estimated to less than 40 m. Ermolin and Silva Busso (2008) found the beach zones to act as limits to the extension of permafrost. The depth where temporal thawing takes place is determined by the climatic boundary conditions and different lithology types. It is directly linked with the formation of surface and suprapermafrost water. Ermolin and Silva Busso (2008) suggest 15 that the development of permafrost and groundwater flow is determined by the periglacial geoforms and cryogenic processes.
The collected data show that the thickness of the active layer ranges from 0.5 to 0.8 m in the hummocky moraines. It reaches 1.0 to 1.5 m in the fluvioglacial plain and gravity slopes. On the periphery of the fluvial talik anbd the coastal wetland located at ca. 2 m a.s.l., the influence of surface and groundwater causes an increase of the active layer thickness to 2.5 m.
A prominent feature of the study area are the ice-free areas and above-melting-point summer air temperatures, which leads 20 to exposed ground surfaces during summer months that are prone to austere cryogenig and exogenic processes, e. g. frost jacking and sorted stripe, typical for the different sediment types of the study area. Frost jacking is a characteristic for slope desposits and wetlands, and it is prevailing in the study area's bottom moraines of presence and Holocene that contain the suprapermafrost aquifer. The characteristic feature in the fluvioglacial plains, wetlands and fine-textured bottom moraines, is the sorted stripe visible as nets with a pattern size of about 0.5 to 1.0 m. Short-term diurnal turnover in soil temperature around 25 freezing point temperature in the wetlands and fine clastic soils found mostly in lake depressions result in a smaller-size pattern.
Slope deposits and lateral moraines that contain ice-rich permafrost and buried ice stem primarily from local bedrock subjected to landslides. The intense weathering and highly dynamic processes of freezing and thawing in addition to the observed high wind speeds in the study area, lead to formation of colluvial slopes and vertical cliffs. Ermolin and Silva Busso (2008) identify thawing of ice-rich permafrost and buried ice as the agents to the prevailing processes of thermoerosion in fluvioglacial plains 30 as well as lake formations. In general, the fluvioglacial deposit areas present slightly lower gradients, from which it is possible to infer a higher permeability. These units have gravel sediments with few fine sediment content. The volcanic rock outcrop areas are fissured aquifers and show significantly lower gradients, from which it is possible to infer a lower permeability. These areas are distributed throughout the subsurface region including Potter basin.
Silva- Busso (2009) found that the groundwater discharge from flow from suprapermafrost and interpermafrost aquifers to be the main drivers of the hydrodynamic balance in the Potte catchment. Silva- Busso (2009) observed that streams and lagoons in a close-by small hydrological basin (Matías Basin) are supplied by the suprapermafrost aquifers.

Data and methods
This study analyses the direct relation of groundwater discharge and hydrogeological processes to glacier ablation. The aim of 5 this paper is to establish a conceptual hydrogeological model for this processes based mainly on the field work conducted in the austral summer of 2011 in the area of Potter Peninsula. In order to accomplish this goal, several analytic methodologies have been applied including in-situ observations and satellite remote sensing data. A Quick Bird image was used for correcting the geologic and geomorphologic interpretation of the Potter basin from field data. Figure 1 shows the basin boundaries and divides, as well as watercourses, derived from a topographic map of the area (Lusky et al., 2001) and complemented by aerial images 10 and own in-situ GPS data. The instrumentation used for this study included a resistivity meter Fluke 80 (Fluke Corporation, Netherlands) to measure the electrical current and potential difference to calculate electrical resistivity which allows for a quantification of water and/or ice content distribution in the respective layer. Conductivity and pH-values were obtained with a conductivity meter WTW315 and a pH meter WTW320 (Xylem Analytics, New York, USA). The water depth was measured with a piezometric sonde. Optical leveling was conducted with a FOIF 57 (FOIF Co, Suzhou, China), and distances were 15 measured with a Telemeter Nikon Forestry Laser PRO 550 AS (Nikon Inc., Japan). All observational and measurement sites are mapped in Fig. 1.
Geologic units were mapped and the hydrogeology aspects were analyzed using vertical electrical soundings (VES) as local geologic and hydrogeologic survey methods. The VES were conducted using a digital resistivitymeter with manual telluric compensation. The sensitivity of the device is 0.01 mV and 0.1 mA, using a continuous power supply of 160 V and 1000 mA.

20
A Schlumberger geoelectric line device with an electrode spacing of AB/2 = 44 m was used, with AB being the maximum electrode spacing (in all VES) to investigate the local quaternary deposits. The subsurface geology can be assessed by geoelectrical studies. The geo-electric techniques used are indirect methods to identify changes in ice content, water salinity or lithological change and have been previously applied and validated for subsurface studies in Antarctica Muñoz Martín et al. (2000); Silva Busso et al. (2013). Eighteen VES were performed in the Potter basin (called VES12 -VES29, see Fig. 1) and 25 interpreted. From these profiles, five resistive horizons and eight resistive layers were identified. Tables 1 and 2 summarize the characteristics of the found resistive layers, i.e. layer thickness and resistivity, and the geological interpretation. This method provides the layers thicknesses, depth, and distribution of underground ice or water content. It is the basis for the estimation of the hydrogeological cross sections of the outlet and the mid basin. The location of VES surveys, permeability tests and static level groundwater measurements points in Potter basin are shown in Fig. 1. From the interpretation of these data (see Tab. 1), 30 a hydrogeological model was derived (see section 4.1).
Thirteen wells were drilled with a hand drilling system for the deep groundwater level measurements and the slug test, which measures the permeability by the salinity method (Custodio and Llamas, 1983). Observations of specific permeability by slug tests were conducted at the locations shown in Fig. 1 (see permeability test points) identifying layers with low and high permeability (i.e. fluvioglacial deposits) according to: where K is the permeability, V the Darcy velocity, φ the specific yield and ∇ i the hydraulic gradient.
The fluvial talik is the layer showing the highest permeability. Since the water in the active layer is fresh water (ca. 400 5 µS cm −1 ), another test to quantify the permeability was conducted with NaCl-tracer NaCl (Custodio and Llamas, 1983;Hall, 1996). Results are shown in Fig. 3. In less permeable layers (like moraine deposits and basaltic rocks), the slug test was performed following the Lefranc method with variable levels (Hall, 1996): with h 1 and h 2 the water levels at the start and the end of the test, respectively; t is the time passed for the water level going 10 from h 1 to h 2 ; L is the longitude and d the diameter of the drill.
It should be noted that the application of these units, the non-uniformity and poor accessibility made it very challenging to construct the wells or conduct a test by traditional pumping. The data enabled us to derive a suprapermafrost piezometric map of the talik and hydrological active layer area. From the slug test data, the permeability was estimated in the different geologic unit outcrops. This method has been demonstrated to be effective in volcanic rocks and moraine deposits and it is recommended 15 for use on low permeability aquifers. Due to the high permeability found at this research site, this method was not applicable for the fluvioglacial deposits, but instead a salt permeability test was used (Custodio and Llamas, 1983). The groundwater hydraulic gradient was calculated on the basis of the different hydrogeologic units obtained from the piezometric map. The meteorological, permafrost and glaciological data sets were used for a complementary analysis of the hydrogeological model presented in here.

20
The calculation of the groundwater discharge requires the use of the Darcy's Law, assuming that summer groundwater discharge has completely saturated the sediments and that the runoff from the (adjacent) Fourcade Glacier is high. These assumptions are valid during 1 to 1.5 months in the austral summer (presumably January and February). For a proper assessment of underground discharge flow and local hydrodynamics, it is essential to define the hydrogeological units, which depend on the geological deposits present in the basin. To achieve this goal, a geological interpretation was performed on a Quick The Potter basin consists of two main channels, North Potter stream and South Potter stream. It is a hydrological system that is mainly driven by the discharge from polythermal glaciers that form a part of the Warszawa Ice Field. In both sub-basins, the appropriate places for discharge measurements were chosen (see Fig. 1). The contemporary glacial retreat and active permafrost processes are found to be the main drivers of runoff and groundwater processes in the study area as discussed in the following 5 sections.

Geological deposits
The Potter Basin contain areas of volcanic rocks, moraines of different events, marine and fluvioglacial deposits (see Fig. 2).
As noted by Fourcade (1960), features are mainly hummocky moraine with different ice content and the latter is determined as of different stages of genesis (Silva Busso and Yermolin, 2014). Although there are still missing elements to propose suitable 10 chronological order, the oldest units are undoubtedly older basaltic-andesitic or volcanic rock as dated by Pankhurst and Smellie (1983). These are present and intensely fissured in many places, covered by the cryoeluvium product of cryoclastic weathering.
They have also been identified in the beds of the river channels in the middle of the Southern Potter Basin ( Fig. 1 and 2), which underlines the regional continuity. Fourcade (1960) generally interpretes the till deposits as hummocky moraine and does not differentiate each imposed lithology. The moraine in the south to south-western part of the map (see Fig. 2) can be assumed 15 to be the lateral boundary of the complete Potter basin. It is the oldest deposit identified here as cycle 1, and probably the former glacier front or side moraine. The remaining moraines noted on the map are interpreted as hummocky moraines. These moraines have variable ice contents (including buried ice) that originate at least from three different periods of glacial advances and retreats. The hummocky moraines (defined as cycle 1 in, Fig. 2), are called "old till" but they include deposits generated by other more recent processes. They are part of the lower basin above approx. 15 m.a.s.l, and have a discontinuous permafrost 20 with interstitial ice content of about 12-14% (Silva Busso and Yermolin, 2014). This area is, where the active layer and the suprapermafrost aquifer is formed.
While it is difficult to establish a precise limit and also taking into account that the active layer can be developed without a free aquifer French (2017) These identified with the processes of tidal and wave rework (the latter of lower intensity) and originate from clastic supply from moraines.

Hydrogeologic units
The shoreline is formed by a gravel coastal cordon that controls an inland lagoon with deposits of fine sand and silt (Barión et al., 2019). This is a transition zone between the talik and the cryopeg (Silva- Busso, 2009), although the latter is likely to be 10 replaced by a wedge of fresh water coming from the creeks due to the magnitude of groundwater flow discharge.
Resistive variations are found to be mainly due to lithological changes or changes in ice content. These two factors determine the probability of development of an active layer or a suprapermafrost aquifer. Moreover, cross sections of subsoil can be derived from these data and groundwater discharge sections quantified. This method allows to detect a layer of non-outcropping sediments in the area of the lower Potter basin (Layer V). These sediments are thicker than the other geologic units, e.g. VES

15
12 profiles show a layer thickness of non-outcropping sediments of 13 m at a base depth of 22 m. These resistive layers can be interpreted as old till deposits of more ancient hummocky moraines or previous fluvioglacial events.
Permeable deposits are interpreted here as fluvial talik that correspond to layers I, III and V, summarized here as Group A (see Tab. 2). Layers III and V showed similar resistivity and can be assumed to be lithologically the same. Layer I has a lower resistivity because it contains marine deposits or salt water due to its close vicinity to Potter cove, however, it is a 20 thawed and permeable zone. Layer II (cryoeluvium) contains deposits and is located on the southwestern slope of the old moraine. Although water can move by snow infiltration, it is not in contact with the Fourcade Glacier, which is outside of this group despite their similar resistivity. Group B is defined only as the cryoeluvium. Another group (Group C) is defined by the hummocky moraines or discontinuous deposits with low ice content for the layer IV. From the layers of this group evolves the active layer, containing the summer suprapermafrost aquifer outside the talik and the coastal zone. The observed resistivity is 25 significantly higher than for Group A (up to 3689 Wm in VES 28) indicating areas with low ice content and distinguishing it from other areas. Group D is represented by layer VI, attributed to deposits with hummocky moraines with buried ice where an active layer develops discontinuously or not at all, thus it is assumed as a layer of low permeability. Group E contains layer VII which consists of weathered basalt, where an aquifer with low ice content forms. This originates in the polythermal bedrock glacier (Fourcade Glacier) which receives the subglacial water from the whole catchment area. The increased resistivity of 30 igneous rock is evident in this layer. Finally, unaltered basalt is found in layer VIII, with the highest resistivity signal, defining the regional aquifuge (Group F).
Another important aspect is the determination of in-situ permeability of each group with aquifer characteristics (talik, and fissurated suprapermafrost aquifer). Observations of specific permeability by slug tests and NaCl-tracer were used to identify layers with low and high permeability. The fluvial talik is the layer showing the highest permeability. This is considered to be representative for all layers in Group A. Group A contains layers of psamitic marine deposits whose permeability is less than that of the fluvioglacial deposits (gravels). Figure 3 presents results of the permeability tests. Table 2 summarizes the interpretation of these results as the local hydrostratigraphic groups and their corresponding permeabilities. In groups without aquifer development no action was performed. 10 Figure 4 shows the longitudinal cross section through the whole basin with its axis from the coast to the glacier front (cross section AA' as defined in Fig. 4). Figure 5 displays the cross section downstream through the permeable deposits on the coast (BB' section as defined in Fig. 4). These cross sections give an idea of the relationship between the hydrostratigraphic sets, but  Table 3 and the obtained values are within the typical ranges for such types of lithologies.

Suprapermafrost aquifer
In the lower and medium Potter basin 11 manual drillings of boreholes (see Fig. 1) were made with the purpose of verifying 20 the continuity of the suprapermafrost aquifer. Water depth was measured to produce a piezometric map that allows for the calculation of hydraulic gradients. In the upper Potter basin, manual drilling proved to be impossible due to the existence of continuous permafrost with a strong ice layer and poor development of the active layer. These findings suggest that there is no development of a suprapermafrost aquifer. Considering the geological field observations and analysis of satellite imagery, there are strong indications that above 45 m.a.s.l the possibility of a development of suprapermafrost aquifer is negligible. Therefore, 25 the only detailed piezometric map covers the area of elevation between 0 -45 m.a.s.l (shown in Fig. 6). The groundwater flow network shown here is well integrated in NW direction following the topographic gradient and discharging into Potter Cove. This piezometric surface is well integrated in Groups A and C (fluvioglacial, marine and hummocky moraine deposits summarized in cycle I, Fig. 2 and Tab. 1). Group E contains the layers of fissured aquifer basaltic rock where no observation wells were drilled (only two for the slug test) making it impossible to identify its extent or its piezometry. However, there is  (2005):  Khrustalev (2005) were calculated in its original units (Tab. 4) but all other results were translated to SI units. 25 The normative thawing depth H d was calculated from the 42-years' meteorological time series at the Russian Bellingshausen Base (Martianov and Rakusa-Suszczewski, 1989;AARI, 2016). From this data time series t p and τ p were obtained. In the period between March and November, which is the austral winter time, the amount of positive degree days per month does not reach 2%. The application of the method after Khrustalev (2005) requires a percentage of positive degree days per month higher than that value. The values for the physical properties of permafrost and soils ( λ c , λ d , t 0 , t c , ρ, C c , C d , γ c and γ d ) 30 for the study sites were taken from Silva Busso and Yermolin (2014). The weights of wet and dry soil (w t and w n ) were 10 https://doi.org/10.5194/hess-2020-422 Preprint. Discussion started: 15 September 2020 c Author(s) 2020. CC BY 4.0 License.
determined specifically for the sample wells for the piezometers (see Fig. 1). The results are listed in Tab. 4. In summary, the maximum summer normative thawing depths calculated according to Khrustalev (2005), amount to values between 3.21 m and 4.29 m. Analysis of the geoelectric measurements leads to an estimate of the thawing depth of about 1.74 m to 5.2 m for austral summer months. The two methods show comparable results, meaning that the estimated thickness and the aquifer cross section are plausible.
and subtracting it the total porosity (φ t ) This was done by particle size analysis of samples obtained from the wells (permeability test sites see Fig. 1). The total 20 porosity amounts to φ t = 0.17m 3 /m 3 for Group A and φ t = 0.14m 3 /m 3 for Group C, respectively. The calculated effective porosity is then calculated as φ e = 0.12m 3 /m 3 for Group A and φ e = 0.06m 3 /m 3 for Group C. As for fissured aquifer  Fig. 4 allows for the calculation of the area of each aquifer sections. With this information, the monthly discharge can simply be estimated by Darcy's law: where Q is the flow through the permeable porous section in m 3 s −1 , S is the saturated area in m 2 , and ∇ i the hydraulic gradient of the aquifer section in m/m. It is also possible to calculate the Darcy velocity (V d ), which can be estimated by an 5 approximation of groundwater flow velocity by Applying Eq. 10 and 11 for each aquifer results in the average monthly flow in m 3 s −1 and is listed in Table 5. Based on the analysis for the study period, the set of aquifers (Groups A, C and E) have a monthly average discharge of Q = 0.47 m 3 s −1 into Potter Cove during the austral summer months 2011.

10
It is noteworthy to analyze the Darcy velocity, as it represents approximate transit times. The total time it takes for a water parcel to traverse the whole groundwater reservoir along the direction of the discharge was measured for each hydrostratigraphic set. In the case of Group A aquifers, the transit time was calculated to up to 14 days, meaning a high velocity of the water flow. For group C aquifers, the calculated velocities are significantly lower with estimated transit times of the order of months to a maximum of 4 months. For a short summer, the observed transit times might not be sufficient for the water flow 15 to cross the aquifer resulting in transit times of over a year. Finally, the aquifers of group E show the lowest velocities and a determined annual flow rate that can be on the scale of decades. In this group, it is important to consider its location at the top of the bed rock of Fourcade glacier.
According to Paterson (1994), polythermal glacier water circulation at the base occurs throughout the year. This means for the fissurated basaltic rock aquifer of group E, that it is the only one with the possibility to recharge and discharge throughout 20 the year. Otherwise, if it is only active during summer, the transit of water through this layer would have significantly longer resident times.
In summary, it is possible to differentiate between three main flow regimes of distinct discharge time scales (daily, monthly and decadal). Finally, the total groundwater storage (R t ) is determined by the total water capacity that drains the aquifer gravitationally (Custodio and Llamas, 1983). The storage capacities can be calculated by for each group, where A is the extension area, b sat the saturated thickness, and φ e the effective yield.
This allows for estimation of the respective storage terms derived from the effective porosity. The results are listed in Tab.
6. The sum is the total groundwater storage and amounts to 560 · 10 3 m 3 .

Sensitivity analysis of groundwater discharge and storage terms
The characteristics of the applied methods and results suggest the application of a sensitivity analysis of the parameters used to estimate groundwater discharge (Q) and total storage terms (R t ). A ceteris paribus approach analyzes the impact of one parameter on a reference outcome, holding constant all other parameters. However, in this case it is very difficult to obtain alternative values for comparison. For this reason, it was decided to establish a reference condition, although no longer arbitrary, that allows us to study the possible dispersion of the results. The sensitivity ratio (S s ) is given by and where Q max and Q min are the maximum and minimum plausible discharges, and R t,min and R t,max are the maximum and 10 minimum plausible total groundwater storage terms, respectively.
The most sensitive parameters have been highlighted in Tab. 5 and 6. In the assessment of groundwater flow discharge three sensitive parameters have been detected. The first is the hydraulic gradient of group A, particularly in the fluvioglacial deposits where variations between 1.10-2 -5.10-2 m m −1 are sufficient to overcome the condition in Eq. 13. However, since these layers exist in a well-defined integrated piezometry with defined hydraulic gradient, the variability of hydraulic gradients are limited 15 and it is therefore considered a robust parameter. The quantified discharge of Group A is also sensitive to the permeability.
Values for the permeability vary on a daily basis between 2.31 and 11.57 ·10 −3 m s −1 in the old till deposit layers. This is a complex issue as this layer is only detectable in deep VES that have not been applied in the field. Therefore, the measured values in the actual fluvioglacial deposits cannot be considered a robust parameter. The permeability could be higher, resulting in a possible underestimation of groundwater storage terms.

20
The third parameter is the aquifer cross section of the old till deposits. Variations to double or half size of the aquifer cross section fulfill the condition in Eq. 13 for the same reasons mentioned above. It is, however, assumed a robust parameter since the VES data on the basis of which the layer geometry was determined, and the calculation of the depth of seasonal normative thawing were well defined for the considered areas and layer depths. The calculation of total groundwater storage has proven to be sensitive to the parameterization of specific porosity within group E (fissured aquifers). The problem is related to the 25 type of porosity (fissures) and the difficulty to assess the layer depth. Assuming a higher confinement or higher values of specific porosity as condition in Eq. 13 is rapidly reached. It is difficult to estimate an appropriate value for this parameter.
The permeability measured in areas of outcrops give us a similar magnitude, but several authors have mentioned a nonlinear relationship between permeability and porosity (Kozeny, 1927;Carman, 1938). A value of φ e = 0.01 m 3 m −3 has been applied according to the empirical relationship proposed by Bourbié et al. (1987), although the lithologies descrived there would be 30 inconsistent with the ones found in our study area. This parameter can thus not be considered as very robust. In all cases, the 13 https://doi.org/10.5194/hess-2020-422 Preprint. Discussion started: 15 September 2020 c Author(s) 2020. CC BY 4.0 License.
condition stated in Eq. 14 is fulfilled. Based on the geological profiles observed in the ravine channel and in hydrogeological surveys, we propose a hydrogeological schematic of the area in the Potter Basin on King George Island shown in Fig. 7.

Discussion and Conclusions
There are several conclusions from this work related to the groundwater discharge in Potter Cove (geology, permafrost conditions, glacial water supply and others). The applied methodology is simple and represents a classical groundwater study in the 5 field, as well as applied criteria and concepts that are well proven in hydrogeology. First, the Potter basin is a very young basin and the volcanic rocks contain substrate with a very low permeability. Above these are glacial or fluvioglacial clastic sequences.
The latter correspond to different progression and retreat events of the Fourcade Glacier in recent times. As a result of these processes, the discontinuous permafrost occurs only in the lower and middle basin and continues at the top of the upper basin.
The discontinuous permafrost zone, talik fluvial areas, cryopeg and active layer constitute the local aquifer suprapermafrost.

10
It has been possible to define a separate hydrostratigraphy in clastic (Groups A and C) and fissured (Group E) aquifer layers.