Submarine Groundwater Discharge helps making nearshore waters heterotrophic

Submarine groundwater discharge (SGD) is the submarine seepage of all fluids from coastal sediments into the overlying coastal seas. It has been well documented that the SGD may contribute a great deal of allochthonous nutrients to the coastlines. It is, however, less known how much carbon enters the ocean via the SGD. Nutrients (NO3, NO2, NH4, PO4, SiO2), alkalinity and dissolved inorganic carbon (DIC) in the submarine groundwater were measured at 20 locations around Taiwan for the first time. The total N/P/Si yields from the SGD in Taiwan are respectively 3.28 ± 2.3 × 104, 2.6 ± 1.8 × 102 and 1.89 ± 1.33 × 104 mol/km2/a, compared with 9.5 ± 6.7 × 105 mol/km2/a for alkalinity and 8.8 ± 6.2 × 105 mol/km2/a for DIC. To compare with literature data, yields for the major estuary across the Taiwan Strait (Jiulong River) are comparable except for P which is extremely low. Primary production supported by these nutrient outflows is insufficient to compensate the DIC supplied by the SGD. As a result, the SGD helps making the coastal waters in Taiwan and Jiulong River heterotrophic.

Recently the salinity and major ions such as Ca, Mg, K, Na, Cl and SO 4 in the submarine groundwater samples around Taiwan have been measured 15 . However, nutrients and carbon in the SGD have never been reported in the SGD from this part of the world. In fact, only a handful of studies have reported nutrients and carbon in the SGD in China 7, [16][17][18][19] . In this study, SGD samples were collected from 20 locations around the subtropical island with an area of 35,873 km 2 . DO, nutrients (NO 3 , NO 2 , NH 4 , PO 4 , SiO 2 ), N 2 O, CH 4 , DOC, pH, TA, and DIC were measured and pCO 2 calculated. For comparison, we also collected data in the Jiulong River in China across the Taiwan Strait. Whether the SGD helps making the coastal waters autotrophic or heterotrophic is evaluated.

Results and Discussion
Concentrations of chemical parameters. The average concentrations of chemical parameters for a total of 278 submarine groundwater measurements and for the local surface seawaters above the SGD sampling sites are given in Table 1. Salinity data are taken from Chen et al. 15 . Out of the 20 sampling sites 15 showed evidence of some submarine freshwater outflow (those marked is black on Fig. 1). The salinity of the SGD varies widely between 0.008 and 34.8 with an average of 21.92 ± 11.43, which is considerably lower than the average of the corresponding local surface seawater (32.5 ± 2.42). Previous study 15 has indicated that the SGD export from Taiwan is about 1.07 ± 0.7 × 10 10 t/a; of which, 0.38 ± 0.48 × 10 10 t/a is the freshwater component. These values are, respectively, about 14% and 5.2% of the total river outflow from Taiwan, and fall within the ranges reported elsewhere. We realize that the SGD sampling sites were not distributed evenly and that seasonal data were not obtained for most sites. However, Moosdorf et al. 20 provided the only other estimate of annual fresh groundwater discharge from Taiwan at 5,486 m 3 per m of coastline. This compares with our freshwater SGD component of 3,200 ± 4,000 m 3 /m/a. Considering the large uncertainty, the agreement is reasonable. We thus went on to look at chemical species with the understanding that an uncertainty of a factor of two is to be expected. The percentage DO saturation (DO (%)) of the SGD and local surface seawater are plotted vs salinity in Fig. 2. Again, the average DO (%), 67.6 ± 21.9, is considerably lower than that of the local surface seawater (95 ± 6.47; Table 1). In general, the DO (%) in the SGD is lower when the salinity is lower ( Fig. 2a; p < 0.001) because of the consumption of DO due to decomposition of organic matter in the subterranean environment isolated from the atmosphere.
Decomposition of organic matter leads to the release of nutrients. As a result, the average NO 3 concentration, 27.4 ± 54.4 μM, is much higher than that of the average local surface seawater (4.84 ± 5.08 μM; Table 1). The NO 3 in the SGD increases when the salinity decreases ( Fig. 2b; p < 0.001). In fact, the average NO 3 concentration in the local surface seawater is higher than those generally found in waters surrounding Taiwan 21-23 (<2 μM). Although there are other sources of NO 3 such as riverine input and acid rain 24 Table 1). The presence of NO 3 and NO 2 , however, indicates that much, but not all, of the nitrogen released as a result of organic matter decomposition has been reduced to NH 4 . Concentrations of NO 2 , NH 4 and N 2 O are much higher in the SGD compared to those found in the local surface seawater (Table 1). Higher NO 2 and N 2 O values are generally found when the salinity is lower (Fig. 2c, p < 0.01 for NO 2 , and Fig. 2e, p < 0.025 for N 2 O, respectively) but the pattern is not as clear in the case of NH 4 (p < 0.25; Fig. 2d). For PO 4 the average concentration in the SGD (0.88 ± 2.44 μM) is higher than that in the local surface seawater (0.55 ± 0.48 μM; Table 1) with waters surrounding Taiwan having the lowest value 21 (<0.4 μM). Worth noting is that the average (NO 3 + NO 2 + NH 4 )/PO 4 ratio of 136 is much higher than the Redfield ratio of 16. This is consistent with the notion that the surface waters of rivers entering the East China Sea and the South China Sea have an average N/P ratio higher than 100 25,27 . In addition, P is removed from groundwater more easily than N 27 .
Along with nitrogen and phosphorus silicate is also a major macronutrient in the oceans, and silicate concentrations in the SGD also increase with decreasing salinity (p < 0.001; Fig. 2g). The average SiO 2 concentration in the SGD (64.2 ± 58.3 μM) is also significantly higher than the average concentration in the local surface seawater  Table 1). Several high values above 50 μM are found in the local surface seawater compared with the generally low value of <5 μM found in waters surrounding Taiwan 21 . This is an indication that phytoplankton uptake is not fast enough to consume the SiO 2 released by the SGD near its source.
In the reduced environment CH 4 is generated. We do not have sufficient CH 4 data (n = 13) to see a clear trend relative to the salinity but the average CH 4 concentration (523.1 ± 1,231 nM; Table 1) in the SGD is clearly higher than that in the local surface seawater (240 ± 554 nM; Table 1). Note the CH 4 concentrations in waters around Taiwan are around only 5 nM 28 . This indicates that the SGD inputs of CH 4 do not have sufficient time to oxidize or to be released to the atmosphere near their sources.
Decomposition of organic matter generally lowers pH in the aerobic environment 29 such as found in our case so the average pH (7.81 ± 0.29) of the SGD is slightly lower than that of the local surface seawater (8.10 ± 0.14; Table 1; Fig. 3a), which is similar to the pH of the waters surrounding Taiwan [30][31][32][33] . As for TA and DIC their values increase with the dissolution of calcareous rocks and decomposition of organic matter and it is indeed what was found. The average TA and DIC in the SGD (3,438 ± 1,417 and 3,193 ± 1,373 μmol/kg, respectively) are significantly higher than those found in the local surface seawater (2,343 ± 358 and 2,040 ± 363 μmol/kg, respectively; Table 1) and higher values are found at lower salinities (p < 0.001 in both cases; Fig. 3b,c). These TA and DIC values are more than 1,000 μmol/kg higher than those found in waters near Taiwan [33][34][35] .
The SGD has a high average pCO 2 of 4,729 ± 13,163 μatm (Table 1; Fig. 3d) compared with the average of the local surface seawater (477 ± 479 μatm). The pCO 2 shows a weak negative (p < 0.05; Fig. 3d) correlation with salinity. These values are higher than the pCO 2 of surface waters found near Taiwan 31,33,35,36 . But, whether the surface seawater receiving the SGD is heterotrophic, i.e., whether the water is a source or sink of CO 2 depends on the balance between carbon-consuming primary production and the excess DIC supplied by the SGD. The average C/N and C/P ratios of particulate matter in NW Pacific marginal seas are 8.8 and 152, respectively 36,37 . Based on this stoichiometry the average amount of nitrogen and phosphorus supplied by the SGD in Taiwan for each kg of water may consume 1,070 μmol/kg and 133 μmol/kg DIC, respectively. These values are much lower than the average excess DIC supported by the SGD. That is to say, primary production supported by the nutrient input from the SGD is insufficient to compensate for the high DIC and pCO 2 supplied by the SGD. As a result, the SGD around Taiwan leads to a CO 2 source for the atmosphere. Similar situation applies to the Jiulong and Pearl River Estuaries. Finally, decomposition of DOC also releases CO 2 . Indeed, the higher pCO 2 values in local surface seawaters (477 ± 479 μatm; Table 1) relative to the atmosphere (~400 μatm) support this conclusion. Similar conclusion has been reported for the Pearl River Estuary 16 and elsewhere 38,39 . The number of DOC data is also small (n = 31) but there seems to be a trend showing high values at low salinities (p < 0.25; Fig. 3e). The average DOC in the submarine groundwater (114 ± 112 μM) is slightly higher than that in the local seawater (84 ± 27 μM; Table 1). The waters surrounding Taiwan generally have a DOC concentration below 75 μM 40,41 . Note Fig. 3e seems to indicate that the DOC is removed, hence becoming a source of nutrients and pCO 2 .
It is critical to point out that the C/N and C/P values of the SGD (Fig. 3f,g) are much higher than the Redfield Ratio. To re-iterate, the excess nutrients supplied by the SGD are insufficient to consume the excess carbon thus the SGD helps making the coastal waters heterotrophic.
Of note is that the SGD-derived DIC flux is greater than the TA flux in the Pearl River estuary, indicating that the SGD serves to reduce the CO 2 buffering capacity of the local seawater 29 . Yet, submarine groundwaters around Taiwan the TA flux is slightly higher than the DIC flux. As a result, the SGD from Taiwan serves to increase slightly the CO 2 buffering capacity of the local seawater. Even so, the high pCO 2 and the high C/N and C/P ratios of Taiwan's SGD makes it a contributor of heterotrophic nearshore waters.
The percentage saturation for aragonite (Fig. 3h) and calcite (Fig. 3i) reaches a mean value of four and six, respectively for the local surface seawater but are slightly lower in the SGD. There is no doubt that the higher TA, DIC and pCO 2 of the SGD compared to the local seawater is due to the dissolution of calcareous rocks and decomposition of organic matter in the groundwater. The decomposition of DOC must also be at play hence increasing pCO 2 . Since the submarine groundwaters do not become anoxic sulfate reduction probably has not occurred to a great extent. Figure 4 shows ΔHCO 3 plotted vs ΔCa (local seawater is taken to be with HCO 3 = 2.3 mM and Ca = 10.3 mM at a salinity of 35; Ca data taken from Chen et al.) 16 . The samples falling around the HCO 3 /Ca = 2 line reflect the dissolution of CaCO 3 . Much of the data shows an excess of HCO 3 and Ca but the pattern is not obvious. Figure 5 shows the saturation state of aragonite and calcite plotted vs pH. Lower saturation state corresponds to lower pH, indicating that the decomposition of organic matter leads to the dissolution of calcareous rocks. The end result, however, is that the submarine groundwater is mostly highly super saturated, especially those with a pH above 7.5. The saturation state of aragonite and calcite even reach 12 and 20, respectively.

Fluxes of nutrients and carbon.
Since the properties of groundwater are not expected to show much seasonal variation as compared to the flux (e.g. Szymczycha et al. 39 ) the above conclusions represent reasonable averages. Based on the rudimentary SGD flux value reported by Chen et al. 15 the annual amount of nitrogen, phosphorus, silicate, TA and DIC export due to the SGD around Taiwan are 1.18 ± 0.83 × 10 9 , 9.3 ± 6.5 × 10 6 , 0.68 ± 0.48 × 10 9 , 3.43 ± 2.4 × 10 10 and 3.17 ± 2.22 × 10 10 mol, respectively (Table 2). Based on the river flow (http://gweb.wra.gov.tw/wrweb/) and the N, P data (http://wgshow.epa.gov.tw/) of the 25 largest rivers in Taiwan the total N and P fluxes are 1.12 × 10 10 and 0.12 × 10 10 mol/a, respectively. Simply stated, the SGD outflow is as much as 10.5% of the river outflow for N but only 0.78% for P.
Our study in Taiwan (Table 2) results in a SGD discharge of about 1.18 ± 0.83 × 10 9 mol/a N (NO 3 + NO 2 + NH 4 ), which is equivalent to a yield of 3.28 ± 2.3 × 10 4 mol/a N per square kilometer of the total catchment area. This value is similar to the annual yield of Jiulong River at 3.90-8.23 × 10 4 mol/km 2 /a calculated based on the size of the catchment area and the flux of N reported by Wang et al. 17 . The flux of P resulted from this study for Taiwan is 9.3 ± 6.5 × 10 6 mol/a, and the yield is 260 ± 180 mol/km 2 /a. The P flux and yield for Jiulong River at 2.9-6.1 × 10 5 and 20-41 mol/km 2/a (Table 2), respectively, are surprisingly low. As reported above, the N/P ratio obtained from this study for the submarine groundwater is 136. The flux data of Wang et al. 17 for Jiulong River translate to a N/P ratio of 2000 which is extremely high although we realize that P is removed from groundwater. Our work for the Jiulong River (Table 3) results in an N (n = 11) to P (n = 9) ratio of 51 in the river basin and an N (n = 7) to P (n = 7) ratio of 27 in the estuary. Of note is that the average groundwater N and P concentrations calculated from the data of Wang et al. 17 are 495 and 22.75 μM, respectively. The resulting N/P ratio is only 21.8, similar to what we found in the Jiulong River estuary but way below the reported ratio of 2,000 for the SGD by these authors.
Liu et al. 16 reported the total annual flux of P for the Pearl River at 30-680 × 10 6 mol and Liu et al. 42 obtained similar values. Liu et al. 42 also reported the N fluxes for the Pearl River at 3.65-157 × 10 9 and 0.95-40 × 10 9 mol in summer and winter, respectively. They reported the Si fluxes at 1.9-91.3 × 10 9 and 0.51-23.4 × 10 9 mol in summer and winter, respectively ( Table 2). The Pearl River (Fig. 1) has a large catchment area of 453,700 km 2 which is located at the same latitude as southern Taiwan. The P flux of Liu et al. 16 translates to a yield of 66-1,500 mol/ km 2 /a and the results of Liu et al. 42 are similar (Table 2). These results are comparable to those from Taiwan but much higher than those from the Jiulong River. The annual DIC flux of Liu et al. 16 for the Pearl River is 15.3-34.7 × 10 10 mol and the yield is 3.37-7.65 × 10 5 mol/km 2 which is comparable with our result in Taiwan. The reported SGD N and P in the literature are also given in Table 2, and the ranges are high. Our fluxes per m 2 for N and P in Taiwan are at the low end of these reported values.  The total flux of Si for Taiwan is 0.68 ± 0.48 × 10 9 mol/a compared with the larger flux of 0.96-2.0 × 10 9 mol/a (Wang et al.) 17 or 2.94-3.14 × 10 9 mol/a (Hong et al.) 19 reported for Jiulong River. As for the yield the value of 6.53-13.6 × 10 4 mol/km 2 /a calculated based on the flux data of Wang et al. 17 , and the value of 0.65-0.69 × 10 4 mol/ km 2 a based on the data of Hong et al. 19 bracket the yield of Taiwan at 1.89 ± 1.33 × 10 4 mol/km 2 /a. Liu Table 3. Concentrations of NO 3 , NO 2 , NH 4 and PO 4 for riverine (S < 2) and estuarine water (S ≧ 2) of Jiulong River.
reported the total flux and yield of Si for the Pearl River. Although their total fluxes are high their yields also bracket our results for Taiwan. The Si fluxes reported for a subtropical bay in south China 43 are slightly higher than those for Taiwan ( Table 2). The TA and DIC fluxes for Taiwan are 3.43 ± 2.4 × 10 10 and 3.17 ± 2.22 × 10 10 mol/a, respectively. These values compare with 0.75-1.6 × 10 10 and 0.88-1.82 × 10 10 mol/a, respectively, for Jiulong Rvier based on the data of Wang et al. 17 (Table 2). The yield of TA for Taiwan at 9.5 ± 6.7 × 10 5 mol/km 2 /a is slightly higher than those for Jiulong River, at 5.1-10.9 × 10 5 mol/km 2 /a. The reported TA fluxes elsewhere bracket our results ( Table 2). The yield of DIC for Taiwan is 8.81 ± 6.17 × 10 5 mol/km 2 /a which falls between the slightly lower value of 6-12.4 × 10 5 mol/ km 2 /a (Wang et al.) 17 and the slightly higher value of 14.6-18 × 10 5 (Hong et al.) 19 for the Jiulong River. The Pearl River basin also has an abundance of calcareous rocks. The DIC yield (3.37-7.65 × 10 5 mol/km 2 /a), nevertheless, is smaller compared to our result in Taiwan. This is perhaps because the weathering is weaker in the less steep Pearl River basin. This points to the difficulty of comparing the total flux or yield. It is yet not possible to compare data per unit area of the ocean floor or per unit length of the coastal line. In terms of flux per m 2 of the seepage area, however, our DIC flux falls in the range reported elsewhere as shown in Table 2. Our DOC flux (9.5 ± 9.3 mol/m 2 /a) is also comparable with those reported in the literature (Table 2).

Conclusions
The concentrations, fluxes and yields of N, P, Si, TA and DIC for the SGD in Taiwan have been reported for the first time, and these values are broadly comparable with the data in the literature. The nutrients supplied by the SGD are insufficient to compensate the DIC supported at the same time. As a result, the SGD around Taiwan leads to a source of CO 2 for the atmosphere in the coastal seas. Similar situation exists in the Jiulong and Pearl River estuaries in Southeast China, and perhaps in other coastal regions around the world as well.

Methods
Geologically Taiwan is relatively young. The collision of the Philippine Arc and the Asian continent gave rise to the Central Range of Taiwan, and the orogenesis is still going 44 . The population is 23 million. The western part of Taiwan is mainly covered by undeformed sediments, and is heavily populated. Less populated is Southern Taiwan where the coasts are largely covered by coral reefs. Eastern Taiwan has a coastal range, and the less populated coasts are mainly rocky.
Preliminary sampling of the SGD in Taiwan was performed from 2004 to 2016. Twenty sampling sites around the coastal areas Taiwan are shown in Fig. 1. Measurements of SGD fluxes were reported in Chen et al. 15 . Submarine groundwater samples for chemical analysis were drawn by a device designed by Zhang and Satake 45 mostly on the sandy coast. At one site 350 m off SW Taiwan divers collected freshwater (S = 0.008) at a water depth of 8 m. The corresponding local surface seawater sample was also collected. Samples for NO 3 , NO 2 , NH 4 and PO 4 were collected in Jiulong River and its estuary in 2008. Preserved samples, with saturated HgCl 2 added, were brought back and measured in the laboratory with details given in Chen 21 , Yang et al. 41 and Tseng et al. 46 and 2017 28 . The percentage DO saturation (DO (%)) was calculated based on the solubility equation of Chen 47 . The HCO − 3 and pCO 2 were calculated based on pH and TA using the CO2SYS program.