Response of redox zonation to recharge in a riverbank filtration system: a case study of the Second Songhua river, NE China

Bank filtration induced by groundwater pumping results in redox zonation along the groundwater flow path. Besides the river water, recharge from other sources can change local redox conditions; therefore, redox zonation is likely to be complex within the riverbank filtration (RBF) system. In this study, hydrodynamics, hydrogeochemistry, and environmental stable isotopes were combined together to identify the redox conditions at an RBF site. The recharge characteristics and redox processes were revealed by monitoring the variations of water level, δH and δO, and redox indexes along shallow and deep flow paths. The results show that local groundwater is recharged from river, regional groundwater, and precipitation. The responses of redox zonation are sensitive to different sources. In the river water recharge zone near shore, O2, NO 3 , Mn(IV), Fe(III), and SO 2 4 are reduced in sequence, the ranges of each reaction are wider in deep groundwater because of the high-velocity deep flow. In the precipitation vertical recharge zone, precipitation intermittently drives O2, NO 3 , and organic carbon to migrate through vadose zone, thereby decreasing the groundwater reducibility. In the regional groundwater lateral recharge zone in the depression cone, the reductive regional groundwater is continuously recharging local groundwater, leading to the cyclic reduction of Mn(IV) and Fe(III).


INTRODUCTION
Riverbank filtration (RBF) involves water extraction by pumping wells at riverside fields to enhance water supply by stimulating river recharge to groundwater. This can attenuate or degrade pollutants, such as suspended solids, inorganic or organic substances, poisonous heavy metals, pathogenic viruses, and bacteria; hence, it is considered an efficient and natural treatment technology for water quality improvement (Hiscock &  In addition to river water, groundwater recharge from other sources can also change the local redox conditions, and hence, redox zonation is likely to be more complex within the RBF system (Kedziorek & Bourg ; Buzek et al. ).
The Kaladian well field, located in northeastern China, is characterized by regional groundwater rich in Fe and Mn with Mn 2þ and Fe 2þ contents of over 6 and 10 mg/L, respectively. The local groundwater in recent 3 years has concentrations of NH þ 4 , Mn 2þ , and Fe 2þ at 0. 98-3.62 N mg/L, 4.71-6.82 mg/L, and 8.54-12.74

Study area
The Kaladian riverside well field, located on the alluvialproluvial plain of the Second Songhua River in northeastern China, is characterized by flat topography with elevations varying from 124 to 129 m above sea level. The unconfined aquifer is mainly composed of sand with the thickness of 17-20 m, and a continuous and stable clay-laminated layer of ∼24 m thickness forms the impervious base of the unconfined aquifer. The upper and lower parts of the unconfined aquifer consist of fine sand and medium sand with a thin layer of silty-clay in the middle, respectively. The long-term pumping wells (C1-C8) operating intermittently at a total pumping rate of 6,000-10,000 m 3 /day have formed a stable groundwater depression cone centered around well C5 and C6.
The well field is located in a cold temperate zone that experiences a continental semi-arid monsoon climate with a dry and windy spring, hot and rainy summer, cool autumn with early frost, and a cold and long winter. Monthly average temperatures in January and July can reach À17.5 and 23.3 C, respectively, while the annual average temperature is 4.7 C. Annual mean precipitation is 425.7 mm, of which more than 70% is concentrated from June to August, while the annual mean evaporation is 1,689.8 mm, with the maximum in May at 316.5 mm. The Second Songhua River, flowing from southeast to northwest, is controlled by climate and upstream water conservancy projects and has an annual average runoff, river surface width, and depth of 476.0 m 3 /s, 400-450 m, and 3-7 m, respectively.
Herein, 22 long-term monitoring wells with a depth of 6-9 m were constructed to monitor the groundwater level and quality since December 2015. Based on the monitoring results, a dense monitoring section from the river to the depression cone center was constructed in August 2018, as shown in Figure 1. The hydrodynamic and environmental conditions are susceptible to change within 30 m of the shore; therefore, six near-shore monitoring points NS1, NS1,NS2,NS3,NS4,NS5,and NS6 were established at 0,2,5,8,14, and 30 m from the shore, respectively. Furthermore, six far-shore monitoring points FS1,FS2,FS3,FS4,FS5,and FS6 were located at 80,200,420,700,850,and 1,110 m from the shore. Each monitoring point has a shallow (7.5 m) and deep (14 m) well with a screen interval of 6-7.5 and 12.5-14 m, respectively. Furthermore, two regional groundwater monitoring wells, RG1 and RG2, are located at 2,100 and 2,700 m, respectively, from the shore.

River and groundwater sampling
River water and groundwater samples were collected in September 2018. River water was collected 0.5 m below the surface. Groundwater was pumped at a rate <5 L/min from the wells, and the samples were collected after flushing the volume of the well pipe three times. Sampling bottles were filled to the brim and then sealed. Samples for SO 2À 4 , NO À 3 , and NH þ 4 were stored in polyethylene bottles, while concentrated sulfuric acid was added to the samples for NH þ 4 to

Aquifer medium sampling
Sample sites of the vadose zone and aquifer medium were located in the vicinity of point NS5, FS1, and FS3 and samples were collected using percussion drills in September 2018. The collection horizons were 0-0.5, 1.0-1.5, 2.0-2.5, and 6.5-7.5 m below the land surface. Samples were stored in 500 mL glass bottles and were immediately transported to the laboratory at À20 C for testing total nitrogen (TN), soil organic carbon (SOC), and ion exchange forms of NO À 3 /NH þ 4 (IEF-NO À 3 /NH þ 4 ). TN content was tested at 120-124 C in an alkaline substrate using K 2 S 2 O 8 to oxidize all forms of nitrogen in the medium to NO À 3 , then tested using a continuous flow analyzer. The loss-on-ignition method was used to obtain the organic matter content, and then by using the Van Bemmelen factor of 1.724, was converted to the SOC content. The IEF-NO À 3 /NH þ 4 in the 2 mol/L KCl extraction medium was tested using a continuous flow analyzer.

In situ permeability test for riverbed sediment
The standpipe falling head test method, also known as the tube test, was used to determine the permeability of the riverbed sediments; four test points R1, R2, R3, and R4 were arranged, respectively, at 5, 30, 150, and 300 m from point NS1, as shown in Figure 1. The length of the standpipe was 1.50 or 2.80 m for different river depths, with an inner diameter of 0.04 m. After being inserted into the depth L v of the riverbed sediment, the water head change process in the tube was recorded after adding water from the upper end, and the calculation was carried out by Equation (1) (Chen ), where K v is the vertical hydraulic conductivity of riverbed sediments, m/day; L v is the length of the riverbed sediments in the test tube, m; t 1 and t 2 are the instantaneous moments at the beginning and end of the test days; and h 1 and h 2 correspond to the water level in the test tube, m. The trend of regional groundwater level changes was consistent with that of the river stage, and the annual change was generally less than 1.5 m. The local groundwater level of the well field was lower than the river level throughout the year, indicating year-round river filtration.

River stage and groundwater level dynamics
In point NS1-NS6 near the shore, the groundwater level could quickly respond to the change in the river level, showing an annual variation of 2.8-3.0 m. In point FS1-FS4 far from shore, the annual variation in the groundwater level was low, at 0.9-1.8 m. During winter when the pumping intensity was high, variations in the water levels in point FS3 and FS4, located within the depression cone, were relatively large, indicating that the far-shore groundwater level was less affected by the fluctuations in the river level and was instead mainly affected by pumping.
Spatial distribution of δ 2 H and δ 18 O in river water and groundwater River water and regional groundwater are two stable lateral recharge sources, and two major end members of the groundwater in the well field. The Second Songhua River originates in the Changbai Mountains (elevation 2,750 m), and the elevation difference with respect to the well field is more than 2,000 m, leading to lower δ 2 H and δ 18 O values in the river water. The western part of Jilin, where the well field features low terrain and shallow groundwater depth, is easily affected by evaporation; hence, the δ 2 H and δ 18 O values of the regional groundwater are relatively high.
In the NS1-FS4 monitoring section, the distribution ranges and average of δ 2 H and δ 18 O values for shallow and deep groundwater lie between those for the river water and regional groundwater, as shown in Figure 3 Eh and DO at the GRZ and DCZ were À126.25 mV and 1.18 mg/L, respectively, which were higher than those of the shallow groundwater at the RFZ and regional groundwater, with a tendency of increasing along the direction of flow until reaching their highest values at the DCZ.    pH. The pH value of the river water was 7.72 as shown in Figure 4(c). During river filtration, pH continuously decreased to 7.28 in well NS1-1, then slowly increased along the direction of flow, while the average shallow groundwater pH in the RFZ was 7.64. Average regional groundwater pH was 7.18. Average shallow groundwater pH at both the GRZ and DCZ was 7.31, which is between the shallow groundwater at the RFZ and regional groundwater values, with a tendency of decreasing along the direction of flow; the lowest pH was recorded in the DCZ.
The contents of NO À 3 and NH þ 4 in river water were 2.274 and 0.124 N mg/L as shown in Figure 4(d) and 4(e), respectively. During river filtration, the shallow groundwater NO À 3 decreased rapidly, while NH þ 4 increased rapidly, until in well NS1-1 at the shore, where NO À 3 then dropped to 0.762 N mg/L, having been consumed up to approximately 70%. Meanwhile, NH þ 4 increased to 0.613 N mg/L, a 4-fold increase, until in well NS4-1 that is 8 m from shore, where NO À 3 dropped below the detection limit, while NH þ 4 was 0.865 N mg/L. Regional groundwater NO À 3 was very low, mostly below the detection limit, while average NH þ 4 was relatively high at 1.881 N mg/L. In the shallow groundwater of the GRZ and the DCZ, average NO À 3 was 0.759 N mg/L, which was higher than the shallow groundwater at the RFZ and regional groundwater, while average NH þ 4 was 1.188 N mg/L, which was between that of the shallow groundwater at RFZ and regional groundwater.
Mn 2þ /Fe 2þ . The contents of Mn 2þ and Fe 2þ in the river water were extremely low as shown in Figure 4(g) and 4(h). During river filtration, shallow groundwater Mn 2þ and Fe 2þ began to increase until reaching well NS1-1; with Mn 2þ peaking at 1.12 mg/L in well NS3-1, which is 5 m from the shore. Fe 2þ peaked at 3.42 mg/L in well NS6-1, which is 30 m from the shore. Both Mn 2þ and Fe 2þ began to decrease after peaking until they dropped to a minimum value in well FS1-1, which is 80 m from the shore.
Average regional groundwater Mn 2þ and Fe 2þ were 5.98 and 12.21 mg/L, respectively. Average shallow groundwater Mn 2þ and Fe 2þ at the GRZ and DCZ were 3.79 and 6.38 mg/L, respectively, ranging between the shallow groundwater at the RFZ and the regional groundwater.
Meanwhile, Mn 2þ and Fe 2þ at the DCZ were closer to that of the regional groundwater.

SO 2À
4 . The SO 2À 4 content in the river water was 29.72 mg/L as shown in Figure 4(f). Average shallow groundwater SO 2À 4 in well NS1-1-NS4-1 that is within 14 m from shore was 29.98 mg/L without obvious change. In well NS5-1, which is 30 m from shore, SO 2À 4 began to decrease, reaching its lowest value of 8.97 mg/L in well FS1-1, which is 200 m from shore at the GRZ. Average regional groundwater SO 2À 4 was 58.36 mg/L. Average shallow groundwater SO 2À 4 at the DCZ was 62.29 mg/L, which is higher than the shallow groundwater at the RFZ, GRZ, and regional groundwater. Meanwhile, the content of HS À was relatively low at all monitoring wells and was maintained below the detection limit.
DOC. The DOC content of the river water was 23.14 C mg/L as shown in Figure 4(i). During river filtration, at the RFZ and GRZ within 80 m from the shore, average shallow groundwater DOC of well NS1-1-FS1-1 gradually decreased along the direction of flow with a value of 9.33 C mg/L on average, which is approximately 60% less than that of the river water. The average regional groundwater DOC was 16.38 C mg/L.
At the GRZ and DCZ within 80-700 m from the shore, the average shallow groundwater DOC of well FS1-1-FS4-1 was 17.75 C mg/L, which is higher than that of the shallow groundwater at the RFZ and regional groundwater.

Redox indexes in deep groundwater
Eh/DO. The Eh values and DO contents of deep groundwater were relatively stable as shown in Figure 5 which was closer to that of the regional groundwater.
NO À 3 /NH þ 4 . Compared with the shallow groundwater, the deep groundwater had lower NO À 3 content and higher NH þ 4 content as shown in Figure 5(d) and 5(e), respectively. Average deep groundwater NO À 3 and NH þ 4 in well NS1-2-NS6-2, which are within 30 m from shore at the RFZ, were 0.394 and 2.081 N mg/L, respectively. Average NO À 3 and NH þ 4 in well FS1-2-FS4-2, which are 80-700 m from shore at the GRZ and DCZ, were 0.452 and 2.230 N mg/L, respectively, which are higher than those of deep groundwater at the RFZ and regional groundwater.
Mn 2þ /Fe 2þ . As shown in Figure 5(g) and 5(h), the contents of Mn 2þ and Fe 2þ in deep groundwater at the RFZ and GRZ began to increase in well NS1-2, with Mn 2þ in well NS6-2, which is 30 m away from the shore, peaking at 1.18 mg/L, and average Fe 2þ in well FS1-2 and FS2-2, which are 80-200 m from shore peaking at 3.76 mg/L. After peaking, Mn 2þ began to decrease in well FS1-2, which is 80 m from shore, while Fe 2þ did not show a significant decline, and both average Mn 2þ and Fe 2þ gradually increased in well  and Fe 2þ contents were closer to those in the regional groundwater than to the water near the depression cone center. DOC. The content of DOC in deep groundwater was higher than that in the shallow groundwater as shown in Figure 5(i).

SO
In deep groundwater, average DOC in well NS1-2-NS6-2, which is within 30 m from shore at the RFZ, was 24.75 C mg/L. Average DOC in well FS1-2-FS4-2, which are 80-700 m from shore at the GRZ and DCZ, was 25.85 C mg/L, which is higher than the value recorded for the deep groundwater at the RFZ and the regional groundwater; DOC at the GRZ increased significantly.

Spatial distribution characteristics of carbon and nitrogen in aquifer
Vertical distributions of N and C contents in the vadose zone and aquifer medium are shown in Figure 6. The points NS5,

Recharge of local groundwater
To more precisely clarify the recharge conditions along the direction of the two flows, the hydrogen and oxygen stable isotopes of the river water and groundwater were calculated and analyzed based on mass conservation. Relative contributions of the river water and regional groundwater could be estimated using an end member mixing model which can be determined using Equation (2), δ 1 n 1 þ δ 2 n 2 ¼ δ(n 1 þ n 2 ) n 1 þ n 2 ¼ 1 ( 2) where δ 1 , δ 2 , and δ are the isotope values for river water, regional groundwater, and local groundwater, respectively; and n 1 and n 2 are the proportions of river and regional groundwater, respectively.  Thus, it can be concluded that the contribution rate of the river to deep local groundwater is greater than that of the shallow groundwater, indicating a closer relationship between the river and deep groundwater. Meanwhile, along the direction of groundwater flow, with increasing distance away from the river, the contribution rate of the river to both shallow and deep groundwater continuously decreased, especially at the DCZ, where the contribution of the regional groundwater was quite high.
The amplitude of the contribution rate of the river at shallow flow was higher than that at deep flow, indicating that the recharge condition was more complex for shallow groundwater. In point FS2 at the GRZ, where the permeability of the vadose zone is greater than that in point NS5 and FS4, precipitation could more easily pass through the vadose zone to recharge the shallow groundwater. In points FS1 and FS2 at the GRZ, the δ 18 O values of shallow groundwater are lower than that of deep groundwater, as shown in Figure 3, which implies that the shallow groundwater at the GRZ could be affected by precipitation.

Principle and identification processes
River water is the most important recharge component of the RBF system. During river filtration, the main sequence of reduction processes indicates DOC as the electron donor, as shown in Table 1 Based on this, regional groundwater makes the second highest contribution toward the quantity of recharge, which is continuous and stable, and shows spatial variability. The variations in Mn 2þ and Fe 2þ contents in groundwater might reflect the influence of regional groundwater lateral recharge in cases when: (1) the regional groundwater has high concentrations of Mn 2þ and Fe 2þ ; (2) the Mn 2þ and Fe 2þ contents in groundwater at well FS1-1, which is 80 m away from shore, are lowest because of the sequential redox processes; and (3) mixing of groundwater with two different reducibilities suggests a new environment wherein the concentrations of Mn 2þ and Fe 2þ in the far-shore groundwater are influenced by groundwater mixing and reaction.
Precipitation recharge may also contribute to the quantity of groundwater. However, variations in NH þ 4 , NO À 3 , and DOC contents in groundwater might reflect the influence of the vertical recharge of precipitation, in cases Table 1 | Redox processes in a closed system (modified after Stumm & Morgan 1995;Champ et al. 1979) Reaction Equation

Mn(IV) reduction zone
When the Eh values of groundwater dropped to a desired range, manganese minerals in the riverbed sediment and aquifer medium were released into the groundwater by organic complexation or reductive dissolution, as symbolized by the variation in Mn 2þ content, which increased slowly to reach a peak value along the flow direction and then decreased because of oxidation and precipitation (Greskowiak et al. ; Kedziorek & Bourg ).

Fe(III) reduction zone
Similar to Mn(IV) reduction, under favorable Eh values of groundwater, ferrous minerals in the riverbed sediment and aquifer medium were released into the groundwater by organic complexation or reductive dissolution, as symbolized by the variation in Fe 2þ content similar to that in Mn 2þ content. However, this range of this zone could be assigned in well NS4-1-NS6-1, 8-30 m from the shore for shallow groundwater and well FS1-2-FS2-2, 80-200 m from the shore for deep groundwater.

SO 2À 4 reduction zone
This zone is symbolized by an obvious decrease in SO 2À 4 contents, as shown in Table 1; the ÀΔG (W) values of Fe(III) and SO 2À 4 reduction were relatively similar, leading to a non-definite boundary. The HS À produced by SO 2À 4 reduction would precipitate upon reaction with Fe 2þ , which is one of the reasons for the decreasing levels of Fe 2þ and low concentrations of monitored HS À as mentioned above.
The ranges could be assigned in well NS6-1-FS2-1, 30-200 m from the shore for the shallow groundwater, and well FS2-2, ∼200 m from the shore for the deep groundwater.

Redox conditions within the PVZ
The contents of Mn 2þ and Fe 2þ reduced, but stable values were observed between well NS6-1 and FS1-1 for shallow groundwater during RBF with a relatively little influence of precipitation and regional groundwater as shown in Figure 8(a); therefore, shallow groundwater in well FS1-1 and the regional groundwater were determined to be the two Mn 2þ and Fe 2þ mixing end members for far-shore groundwater. By using the end member mixing model as shown in Equation (2), the mixing ratio of water using δ 18 O was calculated; then, the Mn 2þ and Fe 2þ mixed lines for the shallow and deep flows were obtained. Results are shown in Figure 9(a) and 9(b).
The redox background at the PVZ, which is represented by point FS2, was the SO 2À 4 reduction environment. The measured Fe 2þ concentration was lower than that in the Fe 2þ mixed line for both the shallow and deep flows, indicating the occurrence of reactions that consume Fe 2þ . Possible reasons for this might be (1) Fe 2þ combined with HS À produced by SO 2À 4 reduction and (2) oxidizing by O 2 recharge occurred vertically through the vadose zone as Fe 2þ was readily oxidized, except for DOC and HS À as shown in Table 2. However, in the shallow groundwater, HS À was consumed almost completely in well FS2-1 and DO content was still relatively high as seen in Figure 4(a), thus O 2 would act on the consumption. In the deep groundwater, SO 2À 4 began to reduce in well FS2-2, while DO content was maintained at low levels as seen in Figure 5(a); thus, it is inferred that HS À and O 2 contributed to the decreasing of Fe 2þ .
Unlike Fe 2þ , the measured Mn 2þ concentration was obviously higher than the Mn 2þ mixed lines of both shallow and deep groundwater, indicating high Mn(IV) reduction.
The redox environment was still reductive despite an increase of Eh in comparison with the SO 2À 4 reduction zone. Therefore, it is inferred that the actual lowest Eh values in shallow and deep groundwater should be located between the controlling values of Mn(IV) and Fe(III) reduction. Under this relatively weak reductive condition, nitrification would hardly occur in the presence of HS À and Fe 2þ . Therefore, NO À 3 could be recharged and accompany O 2 vertically, and both would be reduced before Fe(III) and SO 2À 4 . This shows that the vertical recharge affected the sequential redox zonation mainly because of the recharge of electron accepters with higher Gibbs free energy and electron donors.
Redox conditions within the regional groundwater lateral recharge zone At this zone, under conditions where NH þ 4 , NO À 3 , and DOC along the flow direction did not increase, (NH 4 ) 2 SO 4 fertilizer used at the surface caused significant accumulation of The points FS3 and FS4 represent the groundwater of RGZ, where the measured Mn 2þ and Fe 2þ concentrations were obviously higher than the mixed lines at both shallow and deep groundwater as shown in Figure 9(a) and 9(b), indicating that Mn(IV) and Fe(III) reduction were both Mn(II) oxidation 2Mn 2þ þ O 2 þ 2H 2 O ¼ 2MnO 2 þ 4H þ À40.3