Ra and 228 Ra in the stratified estuary of the Krka River ( Adriatic Sea , Croatia ) : implications for submarine groundwater discharge and its derived nutrients

We conducted a field survey in the Krka River and its estuary (Croatia) in September 2014 to study the significance of submarine groundwater discharge (SGD) and derived nutrients. During the sampling period, the exchange time of the brackish water above the halocline in the Krka River Estuary (KRE) was 15 calculated to be 8.3 ± 3.5 days. Three approaches of three end-member model, mass balance model and time series observation in tidal period based on Ra isotopes were used to evaluate the SGD fluxes in the KRE surface layer. We estimated the SGD flux to be (1.3-7.8) × 10 m d, which was approximately 2.7-16.1 % of the Krka River discharge into the estuary. By establishing the nutrient budgets in the KRE surface layer, SGD dominated the nutrient sources, followed by Krka River. SGD-derived dissolved inorganic nitrogen 20 (DIN) and dissolved orthosilicate (DSi) contributed 26.8-68.6 % and 9.5-38.3 % to the total DIN and DSi fluxes into the surface waters of KRE, respectively. This indicates that SGD was likely a major external source of those nutrients in the KRE. We have identified that SGD-derived nutrients and their high N:P ratios may affect the ecosystem productivity in the KRE and nearby Adriatic.


25
A river-dominated estuary is the critical zone connecting the continent and adjacent sea.It is the primary pathway of freshwater and terrigenous materials transport to ocean through rivers.There are already numerous reports on the biogenic element processes in the estuaries and seas, for instance, trace metals, nutrients, carbon (e.g., Kelly and Moran, 2002;Hatje et al., 2003;Cai et al., 2004).More and more evidence has indicated a significant transport of solutes via submarine groundwater discharge (SGD), which is defined 30 as the flow of water from the seabed to coastal regions (e.g., Burnett et al., 2003).Radioactive isotopes have been well applied as tracers in evaluating SGD in coastal waters, especially radium isotopes ( 223 Ra, 224 Ra, Biogeosciences Discuss., https://doi.org/10.5194/bg-2017-254Manuscript under review for journal Biogeosciences Discussion started: 17 July 2017 c Author(s) 2017.CC BY 4.0 License.
upper stream of the Krka River via the estuary to the coastal water in the Adriatic Sea.Sampling locations for radium isotopes are given in Figure 1.Samples of surface waters were collected directly from a depth of approximately 0.5 m using the Niskin bottles to fill the container of 60 L per sample, whilst subsurface waters were taken also by Niskin bottles at a depth of approximately 2 m to fill the container of 20 L per sample.In the KRE, we also conducted a 24 hours-time series observation by sampling the surface water at time-series 5 (TS) station every three hours (Figure 1).Groundwater samples were collected in springs and wells along the KRE.In addition, corresponding water samples for nutrient analysis were collected in polyethylene bottles.
Vertical profiles of physico-chemical parameters of interest (salinity, temperature and dissolved oxygen (DO)) were measured in-situ at each site by multi parametric probe (Hach Lange HQ40D).

10
The water samples for Ra were passed through a column filled with about 20 g MnO2-impregnated acrylic fiber at a flow rate of around 0.5 L min -1 after the suspended sediments were removed by filtration cartridges (pore size of 0.5 mm).Radium isotopes 228 Ra and 226 Ra were determined by gamma spectrometry (Ortec, GWL-120-15-XLB-AWT) (Wang et al., 2014).Briefly, after leached from the MnO2-impregnated acrylic fiber, Ra isotopes were co-precipitated with barium sulfate, and the precipitation was sealed for more than 15 20 days before measurement.The counting time for each sample was 24 to 48 h.The 226 Ra activities were measured using the 214 Pb (295 keV and 352 keV) and 214 Bi (609 keV) peaks; the 228 Ra activities were used the 338 keV and 911 keV peaks of 228 Ac.The uncertainties of 226 Ra and 228 Ra were 1. 56 -20.20 % and 2.77 -20.80 %, respectively.Samples (50 mL) for the analysis of ammonium (NH4 + ) were stabilized by the addition (2 mL) of the phenol

Hydrological features
During the investigated period, salinity in the surface water ranged from 0.2 to 33.3 for stations KR2 to KR10.
Freshwater affected only upper ~2.5 m, thickness of which decreased approaching to the seaside (Figure 3a).
The maxima of temperature and DO concentration were detected along the edge of halocline throughout the 30 investigated estuary.That is explained by the solar radiation passage through the transparent brackish water and a slow entrainment in the marine water (Legović et al., 1994).Temperature and DO decreased gradually from the halocline towards the bottom water (Figures 3b and 3c).The halocline divided the water column into two parts.The freshwater-seawater interface (halocline) was described as a filter that makes two layers Biogeosciences Discuss., https://doi.org/10.5194/bg-2017-254Manuscript under review for journal Biogeosciences Discussion started: 17 July 2017 c Author(s) 2017.CC BY 4.0 License.

Dissolved nutrients in the Krka River and its estuary
Nutrient concentrations (μmol L -1 ) ranged from 1.00 to 5.81 for DIN, 0.21 to 0.69 for DIP and 3.36 to 32.92 for DSi in the surface water of the Krka River and the KRE (Table 1).The DIN and DSi concentrations 5 decreased from the upper stream towards the mouth of the estuary and had a significant negative correlation with salinity (Figures 4b and 4d).No obvious relationship between DIP concentration and salinity was observed (Figure 4c).The groundwater around the KRE had higher DIN and DSi than the Krka River water, but similar concentration of DIP.DIN/DIP i.e.N:P ratios in the water columns at most of the stations were below the Redfield ratio of 16 (Redfield et al., 1963) indicating a potential lack of nitrogen for the balanced 10 growth of phytoplankton.In addition, high Ra activities corresponded to high nutrient concentrations (Figure 4), indicating that SGD input was associated with high Ra and nutrient values.DIN and DSi concentrations increased with the depth up to the halocline in the upper water, and decreased in the underlying water below the halocline, while DIP showed opposite trend with minimum near the halocline (Figure 5).

15
The dissolved 226 Ra and 228 Ra activities during the sampling period ranged from 86 to 130 dpm m -3 and from 33 to 127 dpm m -3 , respectively (Figure 4a and Table 1).The measured values were lower than those observed in some other Croatian rivers (Bituh et al., 2008).The 226 Ra activity in Krka River was similar to the activity measured in some rivers worldwide, while 228 Ra activity was lower in the Krka River (e.g.Krest et al., 1999;Su et al., 2015).The reason may be that the carbonate sedimentation prevails in this karst river, resulting in 20 U (a parent of 226 Ra) becomes more enriched than Th (a parent of 228 Ra) (Cukrov et al., 2009;Cukrov and Barišić, 2006).Ra activities were low in the fresher waters in the upper KRE, the highest at the KRE mouth and again low out of the estuary (Figure 4a).There are several underground springs that flow out into the KRE, especially in the bay near the Zaton village in the lower part of the estuary (N.Cukrov, personal comm.), that may contribute to higher Ra activities of the stations KR4-KR8.25

Time series observation
The results from 24 hours-time series observation conducted in the estuary showed that all the parameters varied over a tidal cycle even though the tidal range was low, less than 0.3 m.Salinity varied consistently with the tidal height.Maximum and minimum salinities were found at high and low tide, respectively, and were in a range from 10.7 to 14.0 (Figure 6 and Table 1).The 226 Ra and 228 Ra activities over the time series  and 6b).Similar patterns were also observed in other places around the world (Garcia-Orellana et al., 2010;Wang et al., 2016).During the lower salinity period, the 226 Ra and 228 Ra activities were higher although some deviation due to hysteresis effect was observed.This was the result of more fresh water and submarine groundwater that come out bringing higher Ra activities.
Nutrient concentrations also varied with the salinity changes during the time series observation (Figure 6c).

5
High DIN and DSi concentrations occurred in lower salinity waters.Similar to that of Ra activity, DIN and DSi variations had an opposite trend to salinity despite a small hysteresis effect observed (Figure 6c).There was no obvious variation trend between DIP and salinity.Overall, both Ra activity and nutrient concentrations varied over the time series observation.The variations were on account of different material sources, including those from open seawater, river water and from SGD.

Three end-member mixing model
Plots of 226 Ra and 228 Ra activities versus salinity showed that the 226 Ra and 228 Ra activities in the surface estuarine water were higher than those expected from a conservative mixing line between Krka River water and open seawater (Figure 7), indicating that there was an excess of Ra entering the estuary through other sources, such as SGD (e.g., Peterson et al., 2008;Moore, 2010).It was particularly pronounced for 228 Ra that We used equations for water, salinity and 228 Ra balance as follows (Moore, 2003):

25
Then equations above can be solved for the fraction of each end-member: 1.00

Flushing time in the surface layer of KRE
As KRE is highly stratified we were interested in computing flushing time mainly for the surface fresh and brackish layer which was approximately 2.5 m deep.We used a method based on physical model described by Moore et al. (2006) as follows: (1 ) Therefore, the flushing time of KRE surface layer water was estimated to be 8.3 ± 3.5 days, which is comparable to the value of 10 days in September reported by Legović (1991).Based on the three end-member mixing model, we also calculated the faction of groundwater in the estuary to be 0.21 ± 0.07.Assuming that this value represents the fraction of groundwater in the total KRE surface layer, we can obtain the flux of SGD using the following equation: Therefore, the flux of SGD into the KRE surface layer was calculated to be (2.6-9.1)× 10 5 m 3 d -1 .

Ra mass balance model for estimating SGD
Generally, in a defined system with an assumed steady state, the Ra mass balance is equal to the sum of seawater mixing and Ra decay (e.g., Moore et al., 2006).Based on this, Ra mass balance model is another approach to quantify the magnitude of SGD.This model has been widely used in estuaries around the world (e.g., Moore et al., 2006;Rengarajan and Sarma, 2015).We carried out a mass balance model of 228 Ra to estimate the SGD flux in the KRE surface layer (Figure 9).The existence of the halocline in the KRE will prevent the 228 Ra diffusion from the sediments through the halocline into the surface layer water, so we 5 ignored the term of sediments diffusion.The atmospheric We can write the following Eq.( 9) for the 228 Ra mass balance model for the KRE surface layer: where the subscript river and SGD represent the Ra fluxes input by Krka River and SGD, respectively; the subscript mix represents the Ra loss by mixing with open seawater.Then SGD-derived 228 Ra flux 10 and SGD flux in the KRE surface layer can be given by Eq. ( 10):  10) and ( 11), we determined the SGD flux in the KRE surface layer to be (1.6-6.0)× 10 5 m 3 d -1 with all the parameters summarized in Table 2.

SGD derived from the tidal cycles
We also used another method based on the Ra time series observation of tidal cycles to estimate the SGD 20 flux in the KRE surface layer.Following the approach of Peterson et al. (2008), Eq. ( 12) was used to estimate the SGD flux as follows: ( ) Here we used the following steps to evaluate the SGD flux: i) As the each measured Ra activity (Ratotal) in the KRE surface layer was the result of total Ra source, we 25 calibrated the each measured Ra activity by subtracting out the estuarine background Ra activity (Rabkgd).
We chose the minimum activity from the measured values of the time series observation as the background of the estuarine water with a conservative SGD estimation.In this way, we could conclude that the excess Ra activity exclusively came from SGD. iii) The excess Ra inventory can be converted to Ra flux by dividing with the estimated flushing time of the surface estuarine water (Tf, 8.3 days).Therefore, the SGD flux in the upper Krka estuary was estimated to be (2.6-9.1)× 10 5 , (1.6-6.0)× 10 5 and (1.2-6.8)× 10 5 m 3 d -1 by three end-member mixing model, 228 Ra mass balance model and time series observation, respectively.The calculated results of these three approaches were similar and in a reasonable agreement, which gives us a confidence that the estimation of SGD flux in the KRE surface layer was 15 grounded, being in the range of (1.3-7.8)× 10 5 m 3 d -1 .In this way, the amount of SGD is accounted to be 2.7-16.1 % of the Krka River discharge into the KRE surface waters during the sampling period.By comparing with the other studies in the Mediterranean region (Table 3), we can see that the estimated SGD flux from this study is comparable to other reported values.
In addition, we followed Wang et al. (2015) to evaluate a water mass balance in the KRE surface layer under 20 the assumption that the study area of interest was a single box at steady state.The conceptual water mass balance for the KRE surface layer is presented by Figure 10.The total water inflow should be precipitation In this study, the precipitation and evaporation fluxes were 45.3 × 10 5 and 30.5 × 10 5 m 3 d -1 (data from http://www.esrl.noaa.gov/).The wastewater that spreads out from the city of Šibenik has an average outflow of approximately 0.046 × 10 5 m 3 d -1 (data from http://www.wte.de/WTE-Group.aspx).Then based on Eq.

30
(13), the residual flow out of the KRE surface layer was estimated to be 67.8 × 10 5 m 3 d -1 .From the water mass balance (Figure 10), we can see that SGD contribution to the total water inflow was very small being only 1.3-8.0%.In contrast, contribution from the Krka River was up to at least 48 %, which was the most important component.Besides, the water exchange flow or mixing flow between the KRE surface layer and the open sea ( M Q ) can be derived based on the salt balance by following equation: 35 where S1 (18.2) and S2 (36.9) represent the mean salinities of the KRE surface layer and the Adriatic Sea, respectively.We estimated water exchange flow or water mixing flow to be 99.7 × 10 5 m 3 d -1 , which included water mixing flow out of the estuary and exchange with the underlying water below the halocline.

Evaluation of SGD-derived nutrient fluxes to the KRE surface layer
Nutrients in the KRE were thought to mainly originate from the freshwater inflow of Krka River and from 5 anthropogenic sources near the city of Šibenik (e.g., Legović et al., 1994;Svensen et al., 2007).Here we determined that Krka River contributed to the estuary with 22.4 × 10 3 mol d -1 for DIN, 1.3 × 10 3 mol d -1 for DIP and 150 × 10 3 mol d -1 for DSi during the sampling time.However, recently the SGD-derived nutrients have been shown as a major component and indisputable sources in some estuarine systems (e.g., Su et al., 2011;Rengarajan and Sarma, 2015).In the groundwater of the KRE, the nutrient concentrations were on 10 average 103, 0.44 and 120 μmol L -1 for DIN, DIP and DSi, respectively during the sampling period, in which the DIN and DSi concentrations were much higher than those in the Krka River and the estuarine water.
Therefore, considering the groundwater sampled in a shallow depth, the SGD-derived nutrient fluxes to the KRE surface layer were estimated to be ( 13 Coastal Zone (Gordon et al., 1996), which has been widely used to evaluate the relative importance of external nutrient inputs versus the physical transports and internal biogeochemical processes within a body of water (e.g., Liu et al., 2009Liu et al., , 2011;;Wang et al., 2016).Except from river and SGD inputs, atmospheric deposition and wastewater were the other two sources for nutrient input to the KRE surface layer.These sources can be estimated by multiplying atmospheric deposition rate with the surface area (Markaki et al., 25 2010;Rodellas, 2015) and multiplying wastewater nutrient concentrations with the wastewater flux (Gunes et al., 2012;Powley, et al., 2016), respectively.In terms of nutrient outputs in this model, the net residual flux had a significant role.It can be estimated by ( ) Therefore, based on the above estimated results of each nutrient flux, the nutrient budgets in the KRE surface layer were shown in Figure 11.We found that the amount of nutrient inputs were greater than the outputs, showing that the KRE surface layer system was a sink of nutrients.Nutrients could be deposited down to the underlying water, taken up by the plankton community and lost by outflow to the Adriatic Sea.SGD was the dominant source of DIN and DSi, 30 observation ranged from 91 to 119 dpm m -3 with an average of 106 ± 15 dpm m -3 and from 63 to 139 dpm m -3 with an average of 92 ± 22 dpm m -3 , respectively.Activities of 226 Ra and 228 Ra showed an opposite trend with respect to salinity because of the dilution with the open seawater that has low Ra activity (Figures 6a Biogeosciences Discuss., https://doi.org/10.5194/bg-2017-254Manuscript under review for journal Biogeosciences Discussion started: 17 July 2017 c Author(s) 2017.CC BY 4.0 License.

15 had
lower effect than in the open Adriatic Sea.Therefore, a three end-member mixing model was based on salinity and 228 Ra to estimate the fractions of (1) open seawater, (2) river water and (3) groundwater in the KRE surface waters.
3) where f refers to the fraction of the open seawater (S), river (R) and groundwater (GW) end-member; SS, SR, SGW and 228 RaS, 228 RaR, 228 RaGW are the salinity and 228 Ra activity in the open seawater, river and groundwater, respectively.The subscript M represents the measured value for the salinity and 228 Ra of individual sample.
6) Biogeosciences Discuss., https://doi.org/10.5194/bg-2017-254Manuscript under review for journal Biogeosciences Discussion started: 17 July 2017 c Author(s) 2017.CC BY 4.0 License.With the three end-member values shown in Figure 7, we evaluated the fractions of open seawater, river water and groundwater in the KRE surface layer waters.The model results are shown in Figure 8.As expected in the surface layer of the KRE, the fraction of the river water was higher than those of the open seawater and groundwater.During the time series observation, lower changes (28-37 %) were observed for the open seawater fraction relative to the values of the river water and groundwater.

10 here
Tf is the flushing time, V refers to the volume of the surface estuarine water which is defined as the product of the average area and depth, T is the tidal period, P is the tidal prism and b represents the return flow into the open sea from the study region.In the investigated estuary, regular semidiurnal tidal period equals approximately 0.47 days from the time series observation.The tidal prism can be determined by multiplying the average surface area by the tidal range during the sampling period, which we estimated to be 15 2.6 × 10 6 m 3 .In this model, b is equivalent of the open seawater fraction, whilst the fraction of open seawater represents only the surface water.So, based on the salinity profiles of the KRE surface layer waters, we calculated the fraction of open seawater in the total surface layer water of the KRE to be 0.49 ± 0.21.
derived from a three end-member mixing model inputs, which are usually from river supply, sediments diffusion, SGD, and the loss, which includes open Biogeosciences Discuss., https://doi.org/10.5194/bg-2017-254Manuscript under review for journal Biogeosciences Discussion started: 17 July 2017 c Author(s) 2017.CC BY 4.0 License.
of 228 Ra in the KRE surface layer, open seawater and the Krka River, respectively.and depth of the KRE surface layer 15 (2.5 m).Tf is the measured flushing time and Friver is the Krka River flow during the sampling period.Based on Eqs. ( Biogeosciences Discuss., https://doi.org/10.5194/bg-2017-254Manuscript under review for journal Biogeosciences Discussion started: 17 July 2017 c Author(s) 2017.CC BY 4.0 License.ii) Assuming that the Ra activity of time series observation can represent the Ra activity in the total KRE surface layer, we estimated the excess Ra inventory by multiplying by excess Ra activity with the KRE surface layer depth (Hupper, 2.5m) and the estuary area (AES, 9.3 × 10 6 m 2 ).

5 iv)
Finally, by dividing the Ra flux with the Ra activity in the groundwater end-member (Ragw), which was 316 ± 24 dpm m -3 for 228 Ra, then we could obtain the SGD flux in the KRE surface layer.Similar to the above calculation, here we only chose228 Ra to estimate the SGD flux because of its lower activity in the open seawater.Based on the Ra activities of time series observation and by applying Eq. (12), we were able to determine the SGD flux for each time series sample.The range of SGD fluxes in the KRE 10 surface layer during the tidal cycles were estimated to be (1.2-6.8)× 10 5 m 3 d -1 .
of the KRE surface layer to the Adriatic Sea and the evaporation ( E Q ).Thus the total water inflow equals to the water outflow, and the water mass balance can be written by 25 the following equation: 14) Biogeosciences Discuss., https://doi.org/10.5194/bg-2017-254Manuscript under review for journal Biogeosciences Discussion started: 17 July 2017 c Author(s) 2017.CC BY 4.0 License.
.6-80.6) × 10 3 , (0.058-0.35) × 10 3 and (15.8-93.8)× 10 3 mol d -1 for DIN, DIP and DSi, respectively.The fluxes were equivalent to 60-360 %, 4.5-27 % and 10-63 % of the 15 riverine inputs to the KRE surface layer for DIN, DIP and DSi, respectively.It seems that the SGD in the study region provides a substantial contribution to the DIN, DIP and DSi loadings to the Krka estuarine system.Similar to the water balance, assuming that the study was conducted at a steady state, we established nutrient budgets in the KRE surface layer.It is based on a box model devised by Land Ocean Interactions in the 20 and C2 are the nutrient concentrations in the KRE surface layer and the open seawater, respectively.Another term of nutrient fluxes out of the upper Krka estuary is the exchange with the open seawater, here obtained as 1 2 % to the total DIN and DSi 35 fluxes into the KRE surface layer, respectively, followed by Krka River and wastewater.Biogeosciences Discuss., https://doi.org/10.5194/bg-2017-254Manuscript under review for journal Biogeosciences Discussion started: 17 July 2017 c Author(s) 2017.CC BY 4.0 License.Fund of State Key Laboratory of Estuarine and Costal Research (Grant number SKLEC-KF201505) and the Biogeosciences Discuss., https://doi.org/10.5194/bg-2017-254Manuscript under review for journal Biogeosciences Discussion started: 17 July 2017 c Author(s) 2017.CC BY 4.0 License.Natural Science Foundation of China (Grant number 41376089) are gratefully acknowledged.We thank all the colleagues in the field survey.Biogeosciences Discuss., https://doi.org/10.5194/bg-2017-254Manuscript under review for journal Biogeosciences Discussion started: 17 July 2017 c Author(s) 2017.CC BY 4.0 License.

Figure 1 .
Figure 1.Map of the sampling area of the Krka River.Diamonds represents regular sampling stations and star represents the time series observation station.Dashed box represents the study area of interest for estimating SGD.

Figure 2 .
Figure 2. The Krka River flow and monthly precipitation in the city of Šibenik in 2014.Red bars represent the daily precipitation of the September 2014.Data from Šibenik meteo organization (http://www.sibenikmeteo.com).

Figure 3 .
Figure 3. Vertical distributions of (a) salinity, (b) temperature and (c) dissolved oxygen (DO) from Krka River, along the estuary up to the Adriatic Sea.

Figure 4 .
Figure 4. Distributions of (a) Ra, (b-d) nutrients and (e) salinity in the surface water of KRE.

Figure 5 .
Figure 5. Vertical profiles of (a) hydrological parameters and (b) nutrient concentrations for station KR3.

Figure 6 .
Figure 6.Salinity, Ra activities and nutrient concentrations variation in the surface water of the KRE during the time series observation.

Figure 7 .
Figure 7. Plots of 226 Ra and 228 Ra activities versus salinity in the surface waters of the KRE.

Figure 8 .
Figure 8. Fractions of groundwater, river water and open seawater in the surface water of KRE.

Figure 9 .
Figure 9.A schematic depiction of 228 Ra mass balance (units in dpm d -1 ) in the KRE surface layer.

Table 1
Activities of 226 Ra and 228 Ra and concentrations of nutrients in the Krka River and its estuary and groundwater samples during the sampling period.Biogeosciences Discuss., https://doi.org/10.5194/bg-2017-254Manuscript under review for journal Biogeosciences Discussion started: 17 July 2017 c Author(s) 2017.CC BY 4.0 License.

Table 2
Definition and values used in the simultaneous equations for 228 Ra mass balances for calculating SGD flux in the KRE surface layer.Biogeosciences Discuss., https://doi.org/10.5194/bg-2017-254Manuscript under review for journal Biogeosciences Discussion started: 17 July 2017 c Author(s) 2017.CC BY 4.0 License.