Sustained upward groundwater discharge through salt marsh tidal creeks

Salt marshes can export considerable nutrients and carbon to the ocean through submarine groundwater discharge (SGD). However, the complicated SGD processes in salt marshes remain poorly understood. Here, we first report the phenomenon of numerous highly saline artesian springs found in a salt marsh system of East China. Multiple methods including time‐series thermal monitoring, isotope signatures, and high‐resolution electrical resistivity tomography were combined to determine their origin and trajectory. Strong evidence suggests that these springs keep discharging even during high tide and represent a long‐term re‐distribution process of the ancient marine water trapped in the unconfined aquifer. This new pattern of spring‐derived groundwater flow indicates a hidden SGD pathway and has significant implications for studies concerning SGD‐derived fluxes in similar multi‐aquifer‐aquitard coastal systems.

Submarine groundwater discharge (SGD) is the total water flux from coastal aquifers to the ocean and has been well known as an important component of the hydrologic cycle and a major seaward carrier of terrestrial materials (Moore 1996;Burnett and Dulaiova 2003).SGD is driven by different forces including terrestrial hydraulic gradients, density, tides, wave, and so on (Robinson et al. 2018).Due to the variability of coastal hydrogeological settings and hydrodynamic conditions, SGD processes can occur over multiple temporal and spatial scales in intertidal systems (Taniguchi et al. 2019).Knowledge of SGD components and their physical drivers has been developed mostly in permeable sandy aquifer sites (Santos et al. 2012).For low-permeability systems such as salt marshes, however, SGD is far less understood due to the highly complex features and interactions of hydrogeological, geomorphological, and ecological processes (Xin et al. 2022).
Salt marshes are fine-grained ecosystems widely distributed in the intertidal zone along global coastlines.Different from sand-dominated beach systems, surface water and groundwater interactions in salt marshes are often overlooked due to the low permeability of marsh sediments (Guimond and Tamborski 2021).However, salt marshes have recently been proved to generate considerable lateral SGD fluxes through the sediment-water interface of tidal creeks and export abundant nutrients and carbon that supports marine productivity and carbon sequestration in the ocean (Santos et al. 2021a,b;Chen et al. 2022).Numerical results suggest that groundwater flow happens over multiple scales due to abundant macropore structures as well as heterogeneous sediments (Xin et al. 2009(Xin et al. , 2012;;Xiao et al. 2019).Although field-based mass balance models are useful to estimate SGD-derived fluxes, large uncertainties often exist because of the simplified choice of SGD end members (Burnett et al. 2007).Therefore, further knowledge of the SGD processes in salt marshes is critical to accurately quantify SGD fluxes.
In this study, we first report the numerous highly saline rising springs found in the creeks of salt marshes located on the muddy coast of East China (Supporting Information Movie S1).Field investigation including time-series thermal monitoring, environmental isotope tracing and highresolution underground resistivity imaging were carried out.
We intend to answer the following two questions: (1) Are these springs discharging as a common tide-induced SGD process?(2) What is the origin and trajectory of these springs?We then discuss the possible drivers of this phenomenon and its implications for SGD studies in similar systems.

Study area
This study was conducted in the Chuandong salt marsh near the Chuandong Harbor in Yancheng City, Jiangsu Province, China (Fig. 1).It belongs to the Dafeng Milu Deer National Nature Reserve, which has been added to UNESCO's World Heritage List.The marsh platform is dominated by Spartina alterniflora, a plant widely distributed in global salt marshes.It is periodically inundated and dissected by numerous tidal creeks that act as conduits to the Yellow Sea.Tides are predominantly semidiurnal, with a mean tidal range of around 3.6 m.Characterized by subtropical monsoon climate, the mean annual temperature is 14.4 C and the mean annual precipitation is 1067 mm (Chen et al. 2022).The salt marsh appeared in the 1980s and expanded during recent decades, with a deposition rate exceeding 4 cm yr À1 due to the abundant sediment input (Du et al. 2018).The marsh platform is mostly characterized by silty clay overlying silty sands, with silts on the creek bank.The salt marsh area is well preserved and not affected by human reclamation.
The Jiangsu coast has undergone at least three marine transgressions and regressions from the Pleistocene to recent, leading to the formation of multi-aquifer-aquitard systems (Li et al. 2001).Groundwater is primarily hosted in the silty sand-dominated aquifers separated by silty clay/claydominated aquitards.According to the nearest geological borehole data (https://zk.cgsi.cn/)located 18 km west of the salt marsh, the unconfined aquifer is mostly contained within Holocene silty sand deposits, with a thickness of $ 37 m.There is a thin layer of silty clay within 2 m below the surface.The first aquitard layer is characterized by interbedding of silty clay and clay existing from a depth of 37-65 m.Geophysical and geochemical investigations suggest that the middle part of the unconfined aquifer in the Jiangsu coastal plain is dominated by relict marine water evolved from the paleo seawater since the latest marine regression, with groundwater age smaller than 1000 yr (Li et al. 2017;Zhan et al. 2022).Four major confined aquifers (I-IV) exist from the bottom of the first aquitard to a depth of about 400 m.Groundwater salinity decreases with depth and becomes fresh in deep confined aquifers (Zhan et al. 2021).The permeability coefficient of the aquifers is smaller than 6 Â 10 À6 m s À1 .

Field investigation and analysis
Field campaigns were conducted in two tidal creek systems located in the north and south of the Chuandong salt marsh (hereafter, NCD and SCD, respectively; Fig. 1).In December 2020, two automatic loggers (Solinst Model 3001, Canada) were set up in both creeks to record temperature and water head every 5 min at two positions $ 10 cm above and $ 10 cm below a spring outlet, respectively (Fig. 1c).The top logger recorded creek water value when it was inundated and air value when exposed.The measurements lasted 332 h (27 tide cycles) in NCD creek and 198 h (16 tide cycles) in SCD creek.
Water samples in the salt marsh system were collected and analyzed for salinity and stable isotopes ( 2 H and 18 O), including spring water (n = 11), marsh platform porewater (n = 17) from shallow bores (1.5-3.5 m depth), as well as hourly creek water samples (n = 24) covering an entire tidal cycle.To capture spatial and seasonal variabilities, sampling campaigns were carried out in the dry season and wet season for NCD and SCD creek systems, respectively.Individual analysis and interpretation were conducted for both creeks to avoid seasonal influences.Groundwater samples were collected from domestic wells (8-13 m depth, n = 3), saline wells (13-33 m, n = 3), and agricultural wells (> 200 m, n = 5) in the villages nearby (Fig. 1a).Five seawater samples of the nearby south Yellow Sea were taken from a 20-km-long transection starting from the Chuandong harbor, with a distance interval of 5 km.Water salinity was measured in situ using a calibrated handheld multiparameter meter (YSI Professional Plus).Samples were sealed in 500-mL clean plastic bottles and kept refrigerated before 2 H and 18 O analysis by a mass spectrometer (MAT253).The isotopic data were reported in standard δ-notation as parts per thousand relative to Vienna Standard Mean Ocean Water with analytical precision of 2‰ and 0.2‰ for δ 2 H and δ 18 O, respectively.Isotopic data of precipitation were from the nearest Global Network of Isotopes in Precipitation (GNIP) site in Nanjing, Jiangsu (https://nucleus.iaea.org/wiser).Some water samples were further analyzed for tritium concentration by a low-background liquid scintillation counter (Tri-Carb 3170 TR/SL) with an accuracy of AE 0.20 TU.
Electrical resistivity tomography (hereafter, ERT) was conducted in December 2020 to investigate the salinity distribution under the marsh platform.Two ERT profiles were performed (Fig. 1a, A-A': across the creek and a spring outlet, B-B 0 : along the creek upper bank) with an electrode spacing of 5 m and a measuring depth of $ 58 m.An Earth Resistivity meter (AGI SuperSting R8/IP) with an inbuilt processor for 64 multi-electrodes was used.The Schlumberger array configuration was used to achieve better horizontal and vertical resolutions.Topography of the profiles was measured by a real-time kinematic device and was included in the data inversion performed using the EarthImager2D (version 2.4.4) software package.

Temperature observations
Water depth (WD) during the time-series monitoring changed periodically (NCD: max 248 cm, mean 82 cm, SCD: max 327 cm, mean 106 cm; Fig. 2).Spring water temperature T S ranged from 9.3 C to 17.6 C (mean 15.1 C) at NCD and from 10.8 C to 16.0 C (mean 13.9 C) at SCD, much higher than the creek temperature T C above (NCD: 0.1-21.1 C, mean 9.4 C, SCD: À3.4 C to 14.6 C, mean 7.5 C).Temperature difference ΔT (= T S À T C ) was positive for more than 98% measuring records (6276 out of 6364) and exceeded 3 C for more than 77% measuring records (4945 out of 6364).For springs in the NCD creek, ΔT approximated zero only in several high tide events.In the SCD creek, however, spring water was still 2 C warmer than creek water at the highest tide when water depth exceeded 320 cm.
The central Jiangsu coast has strong hydrodynamic conditions during tide events, generating complex and quickly changing intertidal geomorphology (Zhao et al. 2019).Such strong hydrodynamic processes were expected to easily disturb and synchronize the temperatures between the loggers with a vertical distance of only $ 20 cm, which is however, not the case according to their significant temperature differences.Therefore, it is reasonable to believe that these springs kept discharging under most water-level conditions.It can be observed that T S and ΔT also changed periodically following tidal fluctuations, which should be caused by the variation of heat conduction process under different water-level conditions.ΔT decreased with the increase of water depth (Fig. 2b,d), suggesting that the spring discharge was somewhat suppressed by the increase of surface water head during flood tide.In intertidal systems, groundwater discharge could happen when surface water table becomes lower than that of groundwater (Robinson et al. 2018).However, groundwater level in the marsh platform could not provide the hydraulic condition for spring discharge during high tide.Such a discharge feature of springs suggests that they are not likely to be a phenomenon of normal tidally driven shallow groundwater flow.The spring temperature became close to the local mean annual temperature during low tide, which may imply a much deeper source of springs.

Salinity and isotopic composition
Creek water during a tidal cycle was characterized by wide ranges of salinity and isotopic composition (Fig. 3).For example, salinity of SCD creek water varied from 15.6 to 24.1 ppt (mean: 19.6 ppt).Its δ 2 H varied from À34.2‰ to À14.9‰ (mean: À26.0‰) and δ 18 O varied from À4.21‰ to À1.63‰ (mean: À3.09‰).Salinity of porewater in the marsh platform varied from 12.8 to 25.8 ppt (mean: 19.2 ppt).The δ 2 H and δ 18 O of porewater varied from À39.3‰ to À19.6‰ (mean: À27.5‰) and À5.41‰ to À1.73‰ (mean: À3.32‰), respectively.In both creek systems, creek water and porewater δ 2 H-δ 18 O plots (Fig. 3a,c) were generally scattered following the direction of the mixing line between precipitation and seawater, indicating that they were a mixture of freshwater and salt water.This salt marsh area is an interaction zone of the Chuandong River and the south Yellow Sea, resulting in a lower salinity level than that of seawater (mean: 26.1 ppt).The offset of creek water from the mixing line indicated the evaporation of surface water, which was more intense during the dry season.
Groundwater with different depths presented significantly different salinity and isotope signatures.Shallow domestic wells had a mean salinity of 0.8 ppt and mean δ 2 H and δ 18 O values of À43.5‰ and À6.40‰, respectively.They were scattered closer to the precipitation end member.The upper unconfined aquifer of the Jiangsu coast is mostly fresh or brackish due to the desalination effect of rainwater infiltration (Li et al. 2017;Zhan et al. 2022).Groundwater samples of those salty wells (13-33 m deep) had a mean salinity of up to 25.5 ppt and were much more enriched in heavier isotopes (mean δ 2 H: À17.9‰, mean δ 18 O: À2.86‰), which corresponds with the relict marine water in the deeper zone of the unconfined layer (Li et al. 2022).Deep (> 200 m) wells had a mean salinity of only 0.3 ppt and were much more depleted in heavier isotopes (mean δ 2 H: À53.3‰, mean δ 18 O: À7.26‰).The deep confined aquifers were thought to be recharged by freshwater under colder climatic conditions (Li et al. 2017).
Spring water salinity varied from 23.5 to 29.8 ppt (mean: 26.1 ppt), with δ 2 H and δ 18 O ranging from À21.9‰ to À12.8‰ (mean: À18.2‰) and À3.93‰ to À0.68‰ (mean: À2.00‰), respectively.Compared with creek water and porewater, springs had much higher salinity and isotopic values with smaller ranges, verifying that they were not recharged from the intertidal marsh platform.By comparing all water bodies, springs' salinity and stable isotopes were similar to those of seawater and saline groundwater.However, spring water had a very low tritium level (mean: 0.26 TU) suggesting its premodern (before the 1950s) origin, which is more similar to saline groundwater (mean: 0.60 TU) rather than modern seawater (mean: 1.95 TU).Therefore, salinity and isotope signatures suggested that springs were more likely to have a close relationship with the relict marine water underneath.

Origin and possible driving forces of springs
Two ERT profiles presented a similar underground resistivity distribution pattern (Fig. 4a).The near-surface layer (< 7 m deep) had an averaged resistivity of 1.17 AE 0.08 and 1.31 AE 0.24 ΩÁm for A-A' and B-B 0 , respectively.Relatively lower and more spatially variable salinity in this layer is a combined result of multiple processes including tidal flooding, rainwater infiltration and evaporation at the atmosphere-soil interface (Liu et al. 2022).Resistivity gradually increased for the lower layer (35-58 m deep), indicating a salinity transition in the aquitard between the unconfined aquifer and the confined aquifer.Between the depth of 7 and 35 m, however, a clear low resistivity layer can be observed, with mean resistivity of 1.06 AE 0.10 and 1.03 AE 0.11 ΩÁm, respectively.This layer further reveals the wide distribution of relict marine water between the near-surface desalinating layer and the first aquitard (Zhan et al. 2022).This highly saline layer was well connected with the creek bottom at $ 160 m in profile A-A 0 (Fig. 4b), verifying that the numerous springs along the creek bottom originate from the highly saline layer deeply beneath the marsh platform.All the results point to a conclusion that the sustained springs found in the salt marsh creek systems are not a common tide-induced SGD process, but represent a decades-scale discharge of ancient relict marine water trapped in the unconfined aquifer.
Here, we preliminarily discuss the possible driving forces of this phenomenon (Fig. 4c).One possible driving force may come from the deep confined aquifers.The highest confined groundwater levels have been observed near the study area, which may provide a hydraulic gradient for deep groundwater moving upward (Ma et al. 2022).Geological survey suggests that the thickness of the aquitards is not spatially homogeneous (https://www.ngac.cn),and thus the confined groundwater may discharge upward into the shallow aquifer somewhere, driving the highly saline layer discharging through tidal creeks.A second explanation is the compactiondriven flow, which occurs as upward water discharging from aquifer to the surface due to compression caused by the weight of the overlying sediment (Holzbecher 2002).The study area is characterized by rapid deposition of highporosity sediments, which benefits the formation of compaction-driven flow.Compaction-driven flow is usually ignored because of its lower magnitude than the flow driven by other forces (Santos et al. 2012).However, it may be significant for the salt marsh system where tidal creeks provide a preferential window for upward flow through the lowpermeability salt marsh platform.In addition, the silty clay layer within 2 m of the marsh platform may make the saline groundwater semiconfined, resulting in the compactiondriven flow converging as artesian springs at creeks that cut the silty clay layer.

Implications for SGD studies
In salt marshes, porewater exchange, which is driven by tidal pumping and occurs on the meter scale, is considered as the major SGD process and has been well studied by numerical modeling (Wilson and Gardner 2006;Xin et al. 2011).Using shallow porewater as an endmember, field-based studies have proved that these low-permeability systems, however, can generate considerable porewater exchange and associated nutrient or carbon fluxes (Tait et al. 2016;Santos et al. 2021a;Yau et al. 2022).Such considerable SGD fluxes are often explained by the preferential flow caused by vegetation roots or animal burrows (Xiao et al. 2019;Guimond et al. 2020).Different from traditional porewater exchange, the highly saline springs are a distinct upward decades-scale discharge of relict marine water from deeper aquifer through the creek bottom (Fig. 4).This phenomenon highlights a new SGD pattern that has not been reported or considered in previous studies.Since the highly saline layer widely exists under salt marsh platform and connects well with tidal creeks, it may not only discharge quickly through these visible springs but also slowly through the sediment-water interface.With a longer residence time than shallow porewater, this process should have a greater impact on the exchange of geochemical compositions (Tamborski et al. 2017).
It is a global phenomenon that the large river deltaic estuaries and adjacent continental shelves host multiaquifer-aquitard systems formed under multiple marine transgression and regression processes (Sheng et al. 2023).The ancient seawater trapped in quaternary delta systems or coastal aquifers has been widely observed in the world and is experiencing long-term re-distribution due to underground freshwater-saltwater interactions (Larsen et al. 2017).The sustained saline spring phenomenon reported here first proves that the trapped ancient seawater can discharge through the sediment-water interface even under low-permeability conditions.This large-scale groundwater discharge process mixes with the near-surface small-scale SGD processes and should be considered and distinguished when addressing SGD quantification in similar coastal systems.
In summary, our findings reveal a new SGD pattern in intertidal salt marshes that represents a hidden efflux pathway of subsurface materials in multi-aquifer-aquitard systems.This study shed new light on the complex SGD processes in lowpermeability coastal systems, while future studies are still required to further determine its driving forces and make quantitative assessments.

Fig. 1 .
Fig. 1.(a) Locations of the study area and field investigation sites.(b) A photo of artesian springs taken downstream of the SCD creek during low tide.(c) Setup of automatic loggers recording temperature and water depth at the spring outlet.(d) Setup of the ERT measurement line (A-A 0 ) crossing the tidal creek at the lowest tide.

Fig. 2 .
Fig. 2. Time-series temperature records at the spring outlets of the NCD creek (a, b) and SCD creek (c, d).(a, c) The variations of water depth (WD), creek temperature (T C , including air temperature), spring water temperature (T S ) and temperature difference (ΔT).(b, d) The comparisons between T C and T S under different water depth conditions (indicated by color gradient).

Fig. 3 .
Fig. 3. Plots of δ 2 H vs. δ 18 O and δ 2 H vs. salinity for different hydrological components in the (a, b) NCD creek system and (c, d) SCD creek system.The solid line is the local meteoric water line (LMWL) of the GNIP Nanjing site.The dashed line represents the mixing line between precipitation and seawater.The semitransparent symbols are the original data points.Statistical means and standard deviations (symbols with error bars) are also shown for a comparison between saline spring and its potential recharge sources.Seawater represents the samples collected from the south Yellow Sea near the study site.

Fig. 4 .
Fig. 4. (a) Underground resistivity distributions revealed by two orthogonal ERT profiles in the NCD creek system.(b) SGD processes in the tidal creeks.(c) Possible driving forces for the rising springs.Locations of A-A 0 and B-B 0 profiles can be found in Fig. 1.