Estimating submarine groundwater discharge in Jeju volcanic island (Korea) during a typhoon (Kong-rey) using humic-fluorescent dissolved organic matter-Si mass balance

We examined the residence time, seepage rate, and submarine groundwater discharge (SGD)-driven dissolved nutrients and organic matter in Hwasun Bay, Jeju Island, Korea during the occurrence of a typhoon, Kong-rey, using a humic fluorescent dissolved organic matter (FDOMH)-Si mass balance model. The study period spanned October 4–10, 2018. One day after the typhoon, the residence time and seepage rate were calculated to be 1 day and 0.51 m day−1, respectively, and the highest SGD-driven fluxes of chemical constituents were estimated (1.7 × 106 mol day−1 for dissolved inorganic nitrogen, 0.1 × 106 mol day−1 for dissolved inorganic phosphorus (DIP), 1.1 × 106 mol day−1 for dissolved silicon, 0.5 × 106 mol day−1 for dissolved organic carbon, 1.6 × 106 mol day−1 for dissolved organic nitrogen, 0.4 × 106 mol day−1 for particulate organic carbon, and 38 × 106 g QS day−1 for FDOMH). SGD-driven fluxes of dissolved nutrient and organic matter were over 90% of the total input fluxes in Hwasun Bay. Our results highlight the potential of using the FDOMH-Si mass balance model to effectively measure SGD within a specific area (i.e., volcanic islands) under specific weather conditions (i.e., typhoon/storm). In oligotrophic oceanic regions, SGD-driven chemical fluxes from highly permeable islands considerably contribute to coastal nutrient budgets and coastal biological production.

The plots of DIN, DSi, DOC, and DON concentrations versus salinity show conservative mixing for salinity ranging from 0 to 34 in Hwasun Bay (Fig. 1). These results indicate that the sink and source of nutrients and DOM are negligible in this bay, perhaps owing to rapid seepage rates along the coast of Jeju Island (0.14-0.82 m day −1 ) 9 . The concentrations of DIN, DIP, DSi, and DON were higher in groundwater than those in seawater, whereas DOC concentrations were lower in groundwater than those in the bay seawater during all sampling periods (Fig. 1). In addition, the average concentrations of DSi and DON in groundwater increased after the typhoon (Fig. 1). The increase in DSi appears to be due to enhanced silicate weathering rates and DON appears to originate from the soil matrix after typhoons. Previous studies reported that extreme weather, such as typhoon, induced mechanical weathering and increased sediment and soil supply to channels [35][36][37] .
The FDOM H intensities in the brackish groundwater of Hwasun Bay ranged from 1.3 to 5.7 QSU on October 4 (avg.: 4.0 ± 1.1 QSU), 1.4 to 6.3 QSU on October 7 (avg.: 3.7 ± 1.3 QSU), and 2.4 to 5.3 QSU on October 10 (avg.: 3.9 ± 0.9 QSU), which were significantly higher than those in seawater (avg.: 0.2 ± 0.1 QSU) but lower than those in fresh groundwater (avg.: 5.6 ± 0.5 QSU) ( Supplementary Fig. S1g). The FDOM H intensities, indicating humic sources such as terrestrial, anthropogenic, and agricultural sources 31 , decreased with increasing salinity in all sampling campaigns (Fig. 2). However, in the high salinity zone, FDOM H showed deviations from the seawater and the fresh groundwater mixing line in the subterranean estuary, which might be a result of infiltration and transformation of marine organic matter in the beach sediments during tidal inundation 12,33 . However, the nonconservative FDOM behavior in this saline zone differs from the lower salinity zone where FDOM H generally behaves conservatively. This conservative behavior is highly dependent on the balance between freshwater supply rates and mixing relative to the biological production rate of FDOM.
FDOM H showed good positive correlations with DON ( Fig. 3a) and good negative correlations with DOC ( Fig. 3b). Coble 38 reported similar correlations between DOC and DON and peak C for the "humic-like" component, which were observed in all seasons in Hwasun Bay 22 . These high concentrations of nutrients and DON in groundwater can be mainly attributed to terrestrial sources, including agricultural activity or domestic wastewater, whereas DOC showed positive correlations with salinity, implying a marine origin.

Estimating SGD in Hwasun Bay using FDOM H and Si mass balance models. Ra isotopes and 222
Rn have served as the most powerful tools for gauging the magnitude and mechanism of SGD because Ra isotopes and 222 Rn are chemically conservative in seawater and enriched in groundwater 2,39-41 . Although it has received less attention than radioisotope tracers, DSi is also a useful tracer for determining SGD when DSi is highly enriched in fresh groundwater and it shows a conservative behavior in coastal aquifers 28,29 . In this study, the concentrations of DSi in fresh groundwater showed the highest values in fresh groundwater samples and exhibited good negative linear correlations with salinity. Thus, DSi is expected to be a good tracer for estimating SGD flux.
In this study, we applied FDOM H as a tracer to estimate SGD flux. Although DOM characteristics vary depending upon the various environments and FDOM is known to account for 20-70% of DOM (generally represented by DOC), it has the highest values in coastal regions, where freshwater inputs are dominant 31 . In this study area, FDOM H showed strong negative correlation with salinity and overwhelmed by the overall dilution of the terrestrial FDOM H relative to the internal production rate of FDOM. Thus, it has been used as a good SGD tracer in areas with high SGD rate in volcanic islands 32,34 . All previous studies conducted in Jeju Island consistently showed good negative correlation between FDOM H and salinity 22,32 , which indicates apparent recalcitrant FDOM H sources from terrestrial inputs. Thus, we assumed that terrestrial origin FDOM H behaves conservatively in the subterranean estuary, and fresh groundwater flux was calculated using FDOM H as a tracer also with DSi. In the steady state, the mass balance of FDOM H and Si could be expressed as follows: where the terms on the left side of the equation indicate input fluxes arising from diffusion from sediments (first term) and submarine groundwater flow (second term) and mixing with open ocean water (third term).
The diffusion from sediments was calculated for an area (1.90 × 10 7 m 2 ) in Hwasun Bay via the regeneration rates of FDOM H (1.4 × 10 5 μg QS m −2 day −1 ) and Si (5 mmol m −2 day −1 ) in sediments 42,43 . SGD flux was calculated based on average concentrations of FDOM H (4.03 ± 1.13 g QS m −3 on October 4, 3.68 ± 1.34 g QS m −3 on October 7, and 3.93 ± 0.89 g QS m −3 on October 10) and Si (102 ± 41 mmol m −3 on October 4, 98 ± 27 mmol m −3 Mixing with open ocean water was evaluated based on the differences in concentration between bay seawater and open ocean water for FDOM H (0.12 ± 0.06 g QS m −3 on October 4, 0.14 ± 0.12 g QS m −3 on October 7, and 0.08 ± 0.06 g QS m −3 on October 10) and Si (3 ± 2 mmol m −3 on October 4, 4 ± 1 mmol m −3 on October 7, and 4 ± 2 mmol m −3 on October 10), the water volume of the bay, and the unknown exchange rate between bay seawater and open ocean water.
We estimated the seepage rate of groundwater and water residence times simultaneously by solving Eqs.  www.nature.com/scientificreports/ and the freshwater end member mixing-line may not be determinant in this calculation. In addition, our estimation already included the uncertainty of FDOM H intensities in groundwater samples. The water residence time in this study was slightly lower than that (2.5 days) calculated using the tidal prism model, whereas the seepage rate of groundwater in this study was relatively higher than that (0.12 m day −1 ) obtained using the 222 Rn-Si mass balance model 13 although this study was conducted on the same survey area as that employed by Kim et al. 13 .
In general, the main driving forces of SGD were affected by hydraulic gradients between the land and ocean, tidal and wave pumping, convection-driven processes, and precipitation 1,5,44 . In particular, during storms and typhoons, wave pumping rates can increase by orders of magnitude exceeding the rates of fresh water inputs from runoff and SGD 7 . According to the Korea Meteorological Administration (KMA, https ://web.kma.go.kr/eng/ index .jsp), the amount of rainfall three days before each sampling campaign in Hwasun Bay was 3.4, 337, and 0 mm on October 4, 7, and 10, respectively. Thus, the difference in the seepage rate of groundwater appears to be associated with heavy rainfall and wave pumping arising from the typhoon and/or the uncertainties associated with different methods. Previous studies have reported that there were approximately 50% to > 100% uncertainties associated with SGD estimation using 222 Rn and Ra tracers [45][46][47] owing to the natural variability of isotope tracers in the groundwater endmember and loss by mixing with outer-bay water in coastal regions.  (Table 1). In this calculation, DIP showed non-conservative behavior and determined which concentrations were highly dispersed with the highest uncertainties (~ 57%). The highest SGD-driven fluxes of nutrients and DOM were obtained one day after the typhoon and were higher than those (0.3 × 10 6 mol day −1 for DIN, 0.003 × 10 6 mol day −1 for DIP, 0.2 × 10 6 mol day −1 for DSi, 0.1 × 10 6 mol day −1 for DOC, and 0.1 × 10 6 mol day −1 for DON) reported by Kim et al. 13 and Kim et al. 22 .

SGD-driven nutrient and organic matter fluxes in Hwasun
The input fluxes of nutrients can be attributed to diffusion from bottom sediments as well as SGD in Hwasun Bay. The diffusion fluxes of DIN, DIP, DSi, DOC, and DON from bottom sediments were calculated by multiplying the area of Hwasun Bay by the previously reported rates of regeneration of nutrients and diffusive DOM fluxes from bottom sediments 42,[48][49][50] . The estimated diffusion fluxes of DIN, DIP, DSi, DOC, and DON from bottom sediments were approximately 0.03 × 10 6 , 0.01 × 10 6 , 0.10 × 10 6 , 0.04 × 10 6 , and 0.004 × 10 6 mol day −1 , respectively. The fluxes of DIN, DIP, DSi, DOC, and DON through SGD, based on the overall nutrient fluxes into the bay, contribute approximately 98%, 88%, 89%, 89%, and 100% of the total fluxes, respectively. Thus, SGD appears to be an important pathway as a nutrient and DOM source in Hwasun Bay.
The fluxes of DOC through SGD in Hwasun Bay were one order of magnitude higher than diffusion fluxes of DOC and higher than the fluxes of POC through SGD (Table 1). This result indicates that SGD-derived DOC in Hwasun Bay could be the most important source of carbon. However, in situ production by biological activities rather than by SGD in this bay may play an important role in determining the carbon budget, which can be confirmed by the concentrations of DOC in groundwater being lower than those in the bay seawater and the negative correlations between DOC and FDOM H (Fig. 3b).
The fluxes of FDOM H through SGD (24 × 10 6 g QS day −1 on October 4, 38 × 10 6 g QS day −1 on October 7, and 21 × 10 6 g QS day −1 on October 10 with combined uncertainties of ~ 110%; Table 1) in Hwasun Bay were one order of magnitude higher than the diffusion fluxes of FDOM H and two orders of magnitude higher than those in Jochun Bay 32 (northern part of Jeju Island; 0.1 ~ 0.4 × 10 6 g QS day −1 ) owing to the relatively low SGD flux (4.1-6.9 × 10 4 m 3 day −1 ), calculated using the 222 Rn mass balance model 25 . The fluxes of FDOM H through SGD into the bay contribute approximately 80% of the total input fluxes. These results highlight the possibility of SGD being an important hidden source of FDOM H in the volcanic island. In oligotrophic oceanic regions, coral reefs are highly productive ecosystems that should be protected from the damaging effects of solar UV radiation 51 . Thus, SGD-derived FDOM H could be beneficial to the sustenance of coral ecosystems considering their ability to protect coral reefs from bleaching under harmful UV radiation in surface water 51,52 . In the last three decades, seaweeds have replaced corals, leading to the global decline of coral reefs in association with ocean acidification and changing nutrient dynamics 53 . However, high loads of FDOM H arising from SGD provides favorable conditions for coral ecology.

Conclusions
The seepage rate of groundwater estimated using an FDOM H -Si mass balance model was approximately 2-4 times higher than that estimated using the 222 Rn mass balance model reported by Kim et al. 13 . This difference may be attributable to the high level of rainfall and wave pumping owing to the typhoon rather than uncertainties associated with the use of each method. Owing to its several advantages, including relative simplicity, low cost, chemical conservativeness in seawater, and enrichment in groundwater relative to seawater, the FDOM H -Si mass balance model can be effective for estimating SGD in coastal areas of a highly permeable zone without any continuous river or stream discharge. The larger SGD-driven nutrient, DOM, and FDOM H fluxes in Hwasun Bay during typhoons could play an important role in biogeochemistry linked to oceanic production and carbon fluxes. Nevertheless, more extensive observations are necessary to evaluate SGD and nutrient fluxes through SGD depending on geophysical processes.  (Fig. 4). Seawater samples from the outer bay were collected from five stations (S4-S8) on October 11, 2018 because sampling surveys could not be conducted in the outer bay during the sampling period because of the typhoon. Fresh groundwater samples were collected from groundwater wells along the coastline. Brackish groundwater samples were collected from shallow pits dug into nearshore sandy sediments above porous basaltic rocks. Seawater samples from the inner bay were collected in the low-tide line using a plastic beaker. Seawater samples from the outer bay were collected in Niskin bottles attached to a CTD rosette on the R/V A-Ra of Jeju National University, Korea.

Materials and methods
Water samples for FDOM, DOC, total dissolved nitrogen (TDN), and dissolved nutrients analyses were collected and filtered immediately in the field using a Whatman 0.7 μm disposable syringe filter. FDOM samples were stored in pre-combusted amber glass vials and kept refrigerated (< 4 °C) until analysis. The subsamples for DOC and TDN were transferred into pre-combusted glass ampoules (in the furnace at 500 °C for 4 h) and acidified with 6 M HCl (pH ~ 2); the ampoules were then flame sealed. The subsamples for dissolved inorganic nutrients were stored in HDPE bottles (Nalgene) and frozen until analysis. Samples for POC were collected in 1-L of HDPE bottle and were filtered through a 0.7 μm GF/F filter. The filter papers were placed in a petri dish and frozen for storage. Analytical methods. Salinity was measured in situ using an YSI Pro Series conductivity probe. Fluorescence measurements of FDOM were conducted using a spectrofluorometer (SCINCO FluoroMate FS-2) in the scan mode. Emission (Em) spectra (250-500 nm) were collected at 2 nm intervals at excitation (Ex) wavelengths of 250-360 nm (5 nm intervals). Water Raman scattering was eliminated by subtracting the daily fresh distilled water signals from the sample data. Data intensities, obtained in counts per second (cps), were normalized with quinine sulfate standards (fluorescence spectra of quinine sulfate standard solution in 0.1 N H 2 SO 4 at Ex/Em of 350/450 nm) and expressed as parts per billion of quinine sulfate equivalents (ppb QSE). Excitation-emission matrices (EEMs) for all data with smoothing were obtained using MATLAB with Savitzky-Goray filters. The PARAFAC model was applied to our 3D EEMS data and validated using split-half analysis and core consistency test 59 . Three components were statistically identified as component 1 (Ex max /Em max = 300/370 nm), component 2 (Ex max /Em max = 315/340 nm), and component 3 (Ex max /Em max = 340/428 nm) (Supplementary Fig. S2). According to Coble 31 , component 1 and 3 are indicative of marine FDOM H (peak M) and terrestrial FDOM H (peak C), respectively, and component 2 is found to be a FDOM P (peak T).
Inorganic nutrients, including NO 3 − , NO 2 − , NH 4 + , Si(OH) 4 , and PO 4 3− were analyzed using a nutrient autoanalyzer (Alliance Instruments, FUTURA II +). In this study, we define DIN as the sum of NO 3 − , NO 2 − , and NH 4 + . Artificial seawater (salinity: 35) was used as the matrix for the blank and standard. The analytical uncertainties were within 2% for DIN, DIP, and DSi according to certified reference materials (MOOS-1 from National Research Council, Canada and DSR from University of Miami, USA).
DOC and TDN concentrations were analyzed using a TOC-V CPH analyzer (Shimadzu, Japan). Based on the calibration curves of acetanilide (C:N = 8), DOC and TDN measurements were standardized. The measured values of 44 μmol L −1 for DOC (n = 6) and were 32 μmol L −1 for TDN (n = 6) agreed well within 5% for the certified values. DON concentrations were calculated by subtracting the DIN concentrations from TDN concentrations.