The thickness of Quaternary sediments in the study area is about 400–500 m. According to their lithology and hydrogeological characteristics, there are four distinct Quaternary aquifers (Fig. 1b). The first Holocene aquifer (Q4) is a phreatic or semiconfined aquifer with a bottom depth of 15–30 m, is primarily composed of fine sand and slit, and involves fresh, brackish, saline, and brine groundwater (Dang et al., 2020). The second Late Pleistocene aquifer (Q3), the third Middle Pleistocene aquifer (Q2), and the fourth Early Pleistocene aquifer (Q1) have bottom depths of 120–170, 250–350, and 350–550 m, respectively. They are confined aquifers primarily made of medium sand and gravel (Niu et al., 2019). The first aquifer is mainly recharged by meteoric precipitation and lateral infiltration of surface water (Li et al., 2013). The groundwater from the first aquifer is widely extracted for irrigation in the alluvial fan area. The largest salt farm in North China, the Daqinghe Salt Farm, uses shallow brine groundwater for salt production in the delta area, where there is only minor agricultural activity. Except in the alluvial fan area, the circulation between phreatic and confined aquifers is weak. The deep groundwater in the second, third, and fourth aquifers is mainly recharged by a surrounding mountain range and mainly discharged by human pumping (Ma et al., 2014).

Previous studies have shown that in the study region, the salt–fresh groundwater interface gradually deepens from land (depth of ∼ 5 m) to sea (depth of ∼ 100 m), as shown in Fig. 1b, with salt groundwater primarily occurring in the first aquifer of the delta area (Li et al., 2013; Ma et al., 2014). According to a stratigraphic transect along the present coastline (Fig. 2), the series stratigraphic architecture of the first aquifer consists of Late Pleistocene continental facies – Holocene marine facies – Holocene delta facies – modern continental facies or artificial fill, indicating that the sediments of the first aquifer have been deposited through lowstand continental accumulation, marine transgression, and highstand progradation since the Late Pleistocene.

The seawater did not reach the modern coastline from the Last Glacial Maximum to the early Holocene (about 30–9 ka BP). The Luanhe alluvial fan was active during this period (He et al., 2020). Since about 9000 a BP, the Holocene marine transgression has approached the present coastline (Xu et al., 2020), and Holocene marine sediments developed due to the sea-level rise from 9–7 ka BP. The Holocene marine transgression had reached its maximum inland area 20 km from the modern coastline by about 7 ka BP (Gao et al., 1981; Peng et al., 1981; Xue, 2014, 2016) (see the Holocene transgression line in Fig. 1). The modern coastal plain was formed through the accumulation of the highstand prograding delta on top of Holocene marine strata, together with artificial fill. In addition, lagoons are important components of the Luanhe River delta (Feng and Zhang, 1998). Based on the records of lagoon facies in the published cores for this region, the approximate distribution range of the buried lagoon is shown as a purple dashed line in Fig. 1a.

Except for P13-200 (TDS = 1.617 g L-1, indicating brackish water), all the deep groundwater samples in the study region are freshwater. Deep groundwater hydrochemical forms shift from Ca-HCO3 to Na-HCO3 upon moving from the land to the sea (Fig. 3). For the shallow aquifer, the horizontal salt–fresh groundwater interface corresponds more closely to the maximum Holocene transgression line (see Fig. 1a). The Ca-HCO3 type of shallow fresh groundwater is primarily distributed in the alluvial fan region. The brackish and low-TDS saline groundwaters, which vary from Ca-HCO3 to Na-HCO3 and Na-Cl types, occur mainly in the upper aquifer (depths of 0–15 m) of the delta area, while the lower part (depths of 20–40 m) contains Na-Cl-type saline and brine groundwaters with high TDS. Moreover, in terms of the horizontal distribution of salinity, the groundwater TDS tends to decrease from west to east, with the TDS values of saline and brine groundwater generally ranging from 16.57 to 125.97 g L-1 in the old delta (western delta) and from 3.26 to 52.48 g L-1 in the new delta (eastern delta).

Figure 4 shows the relationship between deuterium and oxygen-18. The global meteoric water line (GMWL, δ2H = 8 ⋅ δ18O + 10) is from Craig (1961), while the local meteoric water line (LMWL, δ2H = 6.6 ⋅ δ18O + 0.3) is based on δ2H and δ18O isotope data (1985–2003, mean monthly rainfall values) from the Tianjin station about 100 km southwest of the study area (IAEA/WMO, 2006). The deep groundwater samples exhibit depleted values of stable isotopes, with values of δ2H ranging from -75.52 ‰ to -57.06 ‰ and those of δ18O from -9.82 ‰ to -7.61 ‰. Shallow groundwater samples have higher hydrogen and oxygen isotope levels, ranging from -64.6 % to -22.46 % for δ2H and -8.74 % to -2.07 % for δ18O. While the relatively small overall values of fresh and brackish groundwater samples are similar to those of the river samples, saline and brine groundwater samples generally plotted below the LMWL or GMWL, meaning that the water was subjected to evaporation prior to becoming groundwater recharge (Gibson et al., 1993), or that multiple-end-member mixing processes were involved (Han et al., 2011).

The measured 14C activities of groundwater samples range from 0.774 to 105.9 pMC (Table S2). A plot of sampling depth versus 14C is shown in Fig. 5; this elucidates a negative correlation, showing that the variation in 14C activity can be attributed to radioactive decay in the aquifer. There are multiple processes that can impact the 14C properties, including groundwater mixing and dispersion, long-term variation in atmospheric 14C, and free 14C dilution (e.g., carbonate dissolution) (Cartwright et al., 2020). Because the relative impacts of these processes are not well established in the study area, the uncertainty regarding the correction of radiocarbon ages to real groundwater ages is very high. Consequently, we estimate the groundwater age as a residence time range. The uncorrected age is considered the maximum age, while the corrected age is considered the minimum age. Corrected ages are determined based on two hypothetical models of carbonate dissolution, which is the main influence on the 14C contents of water samples (Lee et al., 2016).

Figure 5 shows that the 14C activities in the shallow groundwater samples are within 30.6–105.9 pMC. These values indicate a relatively modern recharge before the atmospheric nuclear testing period of the 1950s and 1960s. The radiocarbon activities in the deep fresh groundwater samples are less than 12 pMC, which is consistent with the paleo-water recharge. This indicates that there are weak connections between the shallow and deep aquifers. Therefore, we assume that the shallow aquifer is an open system and the deep aquifer is a closed system. The δ13C mixing and chemical mass balance (CMB) models are used to estimate the corrective factor q (Clark and Fritz, 1997).

For the δ13C mixing model (Pearson and Hanshaw, 1970), q=(δ13CDIC-δ13CCARB)/(δ13CRECH-δ13CCARB), where δ13CDIC is the measured δ13C of DIC in groundwater; δ13CCARB is the δ13C of DIC from dissolved soil minerals (δ13CCARB= 1.5 ‰ from Chen et al., 2003); and δ13CRECH is the δ13C in water when it reaches the saturation zone. In this study, we use a δ13CRECH of -15 ‰, which has been suggested as appropriate for soils dominated by C4 plants in northern China (Currell et al., 2010). This model yields some relatively low q values (0.59 for G06-15 and 0.65 for G08-15), possibly because several unaccounted-for factors contribute to variable δ13CRECH values, e.g., local methanogenesis and pH or temperatures in the soil zones.

For the CMB model, q=mDICrech/mDICfinal, where mDICrech is the DIC molar concentration in the recharge water and mDICfinal is the DIC molar concentration in the final groundwater. mDICfinal was calculated using mDICfinal=mDICrech+[mCa+Mg-SO4+0.5(Na+K-Cl)] (Fontes and Garnier, 1979). DICrech was mainly HCO3 in the recharge water when pH values were between 6.4 and 10.3, and the carbonate equilibrium constant varied with temperature (Clark and Fritz, 1997). mDICrech was calculated from the estimated pH and temperature conditions in the recharge environment; e.g., at pH = 6 and T = 15 ∘C, mDICrech= 10 mmol L-1 (Currell et al., 2010).

The corrected radiocarbon ages are shown in Table S2. The residence time of deep groundwater ranges from 15 959 to 39 050 a BP, which is significantly longer than that of the groundwater in the shallow aquifer (9510 a BP to modern). Moreover, most brackish and fresh groundwater ages are modern, while brine has a longer residence period (5590–1245 a BP) and a broad range of residence times.

Deuterium and oxygen-18 are good tracers of groundwater origin and climatic conditions during recharge periods (Clark and Fritz, 1997). When combined with groundwater residence times, they can also identify modern recharge and paleo-recharge (Han et al., 2014).

The depletion of 18O and 2H values in the deep fresh groundwater (Fig. 4) can be attributed to a cold climate (Kreuzer et al., 2009) and the residence times (33 951–39 050 a BP) of the P15-150 and P14-300 samples, which may suggest that there was a recharge during the last glacial maximum. The stable isotopes in P16-100 are heavier, reflecting a recharge history linked to the warm climate in the previous deglaciation (Hendry and Wassenaar, 2000). The stable isotope values of river samples are similar to those of the shallow brackish and fresh groundwater samples, which are approximately modern, indicating lateral recharge of surface water locally. Meanwhile, in Fig. 4, G03-5 is close to the rainfall sample, indicating that modern precipitation is a new recharge source. The trend towards δ2H and δ18O enrichment in brine and saline groundwater can be attributed to seawater infiltration during the Holocene transgression period, which has been confirmed by other studies of the Bohai Sea coast (Li et al., 2017; Du et al., 2016). Additionally, due to the mixing of meteoric water and the subsequent nonequilibrium fractionation of hydrogen isotopes during evaporation (Clark and Fritz, 1997), the CSW sample is characterized by 18O enrichment but 2H depletion compared to seawater.

To distinguish the sources of groundwater salinity, the PHREEQC code (Parkhurst and Appelo, 2013) was used to measure and plot the theoretical seawater–freshwater mixing line (“mixing line”) and seawater evaporation line (“evaporation line”) using hydrogeochemical modeling. The two simulation effects serve as references for groundwater hydrochemical characteristics (Figs. 6 and 7). For the Na-Cl (Fig. 6a), Mg-Cl (Fig. 6b), and Br-Cl (Fig. 7a) diagrams, the measured brackish, saline, and brine groundwater samples fit quite well to modeling mixing lines, and evaporation lines follow a linear trend from the least to the most saline. This strongly demonstrates that the salt in these water samples is mainly of marine origin. The major ion concentrations in some samples (such as brine) are higher than those in seawater, suggesting that the enriched ions are associated with evaporation processes rather than seawater intrusion (Colombani et al, 2017).

Moreover, the samples deviate from the modeling lines (Fig. 6c and d), indicating that there may be other hydrogeochemical processes that are responsible for the modified ionic compositions (Giambastiani et al., 2013). (1) Ca2+ depletion is seen for the P18 and P12 samples in Fig. 6d. This phenomenon is likely explained by gypsum (CaSO4) precipitation. The evaporation line reveals that the Ca2+ composition of evaporating seawater follows a hooked trajectory (Fig. 6d). During evaporation to the point of gypsum saturation, the Ca2+ concentration of the residual CSW progressively decreases. (2) Ca2+ and SO42- are present in excess in most of the fresh and brackish samples (Fig. 6c and d), which can be attributed to mineral dissolution along with stream water recharging (such as gypsum dissolution), highlighting the occurrence of some degree of dilution with continental runoff since the Holocene regression. (3) Decomposition of the organic matter abundant in marine or lagoon facies sediments can result in the release of bromide ions, thus pushing the Br / Cl ratios of saline groundwater samples higher than the mixing line (Fig. 7b).

Figure 8 depicts the relationship between δ18O and Cl- in different water samples. There are higher Cl- concentrations and lower δ18O values in brine samples than in seawater, meaning that simple two-end-member mixing cannot adequately explain the groundwater salinization. Stable isotopes of high-TDS saline and brine samples fall between the seawater and CWS mixing lines, suggesting a potential for three-end-member mixing processes (Douglas et al., 2000). Therefore, we considered SW01 (seawater) and P18 (mostly saline, but with relatively depleted stable isotopes) as two saline end-members. P16-100, which was most likely recharged during the last deglaciation, was chosen to represent fresh end-members that could have been impacted by overlying seawater or CSW during the Holocene transgression. In Fig. 8, an inferred salinization zone is established that includes almost all saline and brine groundwater samples, demonstrating the salinization processes in which fresh groundwater mixed with either seawater, CSW, or a mixture of both.

The fresh and brackish groundwater samples, on the other hand, have low Cl- concentrations and less 18O, so they deviate from the assumed salinization zone and approach the river samples in Fig. 8, implying a river water–groundwater mixing trend. LH02 (with low δ18O) and SH02 (with high δ18O) were selected to represent river water end-members for different river recharge of groundwater in the study area, while G09-15 (saline but with river-like stable isotope levels) was considered as a groundwater end-member. It is presumed that a freshening zone could form between the two river water-groundwater mixing lines, indicating the occurrence of freshening processes, which would be in agreement with the continental runoff dilution discussed in Sect. 5.2.

The good fit between the measured hydrochemistry and simulated evaporation lines (Figs. 6 and 7) is an indicator that the brine samples were associated with the seawater that was exposed to evaporation during geological history. Previous research has revealed that the lagoon was active during the progradation of the old Luanhe River delta between 7 and 3.5 ka BP (He et al., 2020; Xu et al., 2020). Meanwhile, the region has had a relatively arid climate since 5500 a BP, which may have led to increased evaporation (Jin, 1984). The ancient lagoon would have been an ideal location for evaporating seawater that had been trapped due to storms or tides (Fig. 9b). As a result, concentrated saline water (CSW) with a salinity greater than that of seawater would have been created, and the CSW would have undergone two processes: (1) it would have infiltrated and descended to the lower part of the aquifer due to its higher density, before combining with the salinized groundwater from phase 1, resulting in a three-end-member mixing scenario in the relationship diagram (Fig. 8); (2) after reaching saturation during the later stages of evaporation, precipitation of minerals such as gypsum, calcite, and halite would have occurred, which would have been redissolved by meteoric waters or seawater, resulting in high-salinity water which would then have been subjected to the above process. The Br / Cl ratios in certain fresh or brine groundwater samples deviate from the evaporation line (Fig. 7b), which may be related to halite precipitation and redissolution. These two processes caused the groundwater salinity to rise even further, resulting in the formation of brine groundwater with 3 times the TDS of seawater, such as G03-20, with a residence time range of 4323–5590 a BP.

Since about 3500 a BP, a nearly 90∘ diversion of the Luanhe River channel in the study area has resulted in new delta development (Wang et al., 2007; Xue, 2016). There are some signs of a lagoon environment in the new Luanhe River delta (Cheng et al., 2020), and, as previously discussed, the brine groundwater sample G10-30 can be attributed to evaporation in a lagoon setting (Fig. 9c). However, some factors are likely to limit CSW formation in the study area: (1) the relatively low evaporation capacity due to the semi-humid climate since about 2.5 ka BP (Jin, 1984), (2) diluvial deposits or artificial reclamation would have filled the coastal lowlands such as lagoons, and (3) offshore levees prevent the seawater from flooding inland during storms or tides. Unlike the old Luanhe River delta, these factors may also explain why the current Luanhe River delta does not have high-TDS brine groundwater.

In addition, the brackish and low-TDS saline groundwater with a relatively modern age (e.g., G09-15) and the river-like stable isotope levels (Figs. 4 and 8) are compelling evidence that freshening processes have occurred in the delta plain. With the semi-humid paleoclimate, some abandoned channels have developed into small rivers such as the Suhe River and Shahe River following the diversion of the ancient Luanhe River (Gao, 1981). Firstly, the lateral recharge from the surface stream plays a role in washing out the salty groundwater. Secondly, due to the unsuitability of saline groundwater throughout human history, river irrigation has been commonly used for agricultural activities in the study region, which has freshened the upper saline aquifer (Fig. 9c). Some groundwater samples found above the seawater mixing line in the Ca-Cl and SO4-Cl relationship diagrams (Fig. 6c, d) may be related to mineral dissolution during river water or irrigation recharge. However, saline groundwater can be washed out over time in coastal zones with low-permeability marine layers and a low hydraulic gradient (van Engelen et al., 2019; Han et al., 2020).

In summary, the evolution of saline groundwater in the study area results from paleo-environment development such as sea-level changes, paleogeography, and paleoclimate, and is significantly affected by human activities. The coastal brine groundwater found on the Bohai Sea coast, such as Bohai Bay (Li et al., 2017) and Laizhou Bay (Han et al., 2014), is a special product of geological evolution. The change in sea level during the Late Pleistocene would have favored marine intrusion and a similar sedimentary environment on the Bohai coast, allowing this study to infer the following conditions for brine formation in this region: (1) stable evaporative environments (e.g., a lagoon), (2) suitable climatic conditions (e.g., aridity), (3) entry of seawater into evaporative environments (e.g., via storms or tides), and (4) a long time period for salinity accumulation.