The configuration of the constant-head permeameter is shown in Fig. 1. Experiments were conducted in three cylindrical acrylic columns with 10 cm lengths and 7 cm diameters. The columns were packed with 0.07–0.87 mm diameter graded Ottawa sand (silica classified #C109 ASTM #C778), supported by stainless steel mesh placed inside the column end caps. Tap water was used as the source of influent; it was dechlorinated by adding granular activated carbon (200 mg/L) and sparging with air for 2 h prior to use. After dechlorination, tap water was again sparging with nitrogen gas to remove dissolved oxygen (DO) for simulating conditions of low dissolved oxygen in groundwater. Characteristics of tap water were as follows: 6.7 ± 0.02 of pH, 217 ± 6.1 μS/cm of electrical conductivity (EC), 0.1 ± 0.05 of DO, 0.17 ± 0.06 of total nitrogen, and 0.6 ± 0.08 NTU of turbidity. Local soil provided the inoculum in the columns. Approximately 100 g of soil was collected from the premises of Kangwon National University, Chuncheon, Korea and mixed with 300 mL of tap water to simulate microbial growth in a natural soil environment. The mixture was shaken for 10–15 min in an incubator at 150 rpm (30ºC) and 0.5 mL of supernatant was added as inoculum to each column.

Three columns (C1, C2 and C3) were operated under water-saturated conditions at constant room temperature. Using an adjustable peristaltic pump, the influent was fed at a flow rate of 0.94 mL/min, which approximates the flow of velocity of typical groundwater in sandy aquifers [28]. Empty bed contact time for all columns was calculated to be 57.7 h. The contents of glucose and nitrate in the influents for C1, C2, and C3 during the experiments are summarized in Table 1. Column 1 (C1) was fed with tap water alone as a control. Columns 2 (C2) and 3 (C3) were fed with 5 mg/L and 100 mg/L of glucose, respectively, which simulate recharge water that contains relatively low and high contents of nutrients. Until day 84, all columns were operated with glucose only and then 20 mg N/L of nitrate was introduced until the end of operations (day 112). All influents were refreshed every 2 d to prevent degradation of glucose before they were applied to the sand grain columns. All experiments were performed in triplicate and the average values are reported in the results.

Hydraulic conductivity (K;cm/s) was analyzed based on Darcy’s Law (1) on days 1, 21, 42, 63, 86, 105, and 118:

where Q represents the rate of flow or volume of water per unit time (cm³/s), L is the length of the sand column (cm), h is the constant head difference (cm), A represents the cross-sectional area (cm²) and t denotes time (s).

For chemical analyses, 50 mL of sample was collected from the influent and effluent ports of each column on day 1, 21, 42, 63, 86, 105, and 118, and chemical oxygen demand (COD), glucose, and nitrate (only on day 86, 105, and 118) were measured. COD was determined based on standard methods for water and wastewater [32]. Glucose concentrations were analyzed using the phenol-sulfuric method [33]. For nitrate analysis, approximately 20 mL of sample was passed through a 0.45 μm syringe filter (Smart Pore; Woongki Science Co. Ltd., South Korea) and stored at 4ºC prior to ion chromatography analysis (DX-100, AS14 column; Dionex Corp, USA). The numbers of bacteria were measured by cell counting using a Neubauer chamber on days 21, 84, and 118.

All data were analyzed using Microsoft Excel 2016 and graphed using Microsoft Excel 2016 and Sigma Plot 10. To evaluate the statistical significance of the results, p-values were analyzed based on the t-test using the method in a study of Welch [34].

The effect of glucose on bioclogging was evaluated based on the K values during operations of C1, C2, and C3 with no introduction of nitrate. The K values in C1 were statistically constant (p-value < 0.05) until day 63 with an average value of 0.0227 ± 2.16 × 10⁻⁴ cm/s. While overall K values in C2 (on average, 0.0221 ± 3.45 × 10⁻⁴ cm/s) were slightly smaller than those in C1, they also remained statistically unchanged (p-value < 0.05) until day 63. However, the K values in C3 significantly decreased from 0.0235 ± 2.52 × 10⁻⁴ cm/S to 0.0168 ± 4.16 × 10⁻⁴ cm/s during the first 21 days and then they were not notably changed until day 63. This change in the K values in C3 corresponds to the earlier reports about bioclogging [35–39]. For example, Jeong et al [39] likewise reported that hydraulic conductivity sharply decreased early in the operation of a sandy column under continuous flow and then no further significant drop in hydraulic conductivity was discovered. They considered that in the early stages of bioclogging, aerobic and anerobic microorganisms multiplied rapidly to form colonies and occupied the pore channel, causing blocking water flow and a sharp decrease in hydrodynamic dispersion. However, when microbial colonies developed to a certain level, water flow transformed from uniform to non-uniform flow, resulting in no more significant decrease in hydraulic conductivity.

The data from bacteria counting (Fig. 3) confirms that the variation in hydraulic conductivity between C2 and C3 was mainly attributed to microorganism stimulation in sand grains. Bacterial counts in C1 and C2 at day 21 accounted for 1.43 × 10⁴ ± 1404 and 1.57 × 10⁴ ± 1501 cells/mL, respectively; whereas the value for C3 was considerably higher, 5.97 × 10⁴ ± 3113 cells/mL. It is speculated that this greater microbial stimulation in C3 is associated with fermentation. As Fig. 4 demonstrates, C3 showed a significant glucose removal (approximately 40%); however, no COD removal was observed in C3. Moreover, while C1 and C2 commonly showed fairly similar pH values between influents (on average, 6.5 ± 0.2 in C1 and 6.7 ± 0.1 in C2) and effluents (on average, 6.6 ± 0.2 in C1 and 6.5 ± 0.1 in C2), only C3 presented a substantial drop in pH between influents (on average, 6.6 ± 0.1) and effluents (on average, 3.1 ± 0.3). Given that electron acceptors such as dissolved oxygen and nitrate rarely exist in the sand grains, these results suggest that fermentation was the main cause for glucose removal and fermentation by-products such as organic acids and alcohols attributed to the no COD removal and decrease in pH in C3.

The above results of the K values in C1, C2, and C3 until day 63 indicate that the content of glucose highly impacts hydraulic properties by stimulating microorganisms. When recharge water contains low carbon concentration (C2), hydraulic conductivity was not substantially worsened. On the other hand, recharge water that possesses high carbon content (C3) leads to a significant decrease in hydraulic conductivity in the system. This observance suggests that control for organic matter in recharge water is necessary for sustainable operation of MAR.

Hydraulic conductivity in C1, C2, and C3 before and after introduction of nitrate is compared to investigate the effect of nitrogen on bioclogging. In C1 and C2, slight decreases in K values were observed after introduction of nitrate, on average, from 0.0228 ± 2.16 × 10⁻⁴ cm/s to 0.0218 ± 3.09 × 10⁻⁴ cm/s and from 0.0221 ± 3.44 × 10⁻⁴ cm/s to 0.0209 ± 3.16 × 10⁻⁴ cm/s, respectively (Fig. 2). Results of bacterial counts before and after introduction of nitrate in C1 and C2 were also minimally changed (Fig. 3). For example, before nitrate was introduced, the average bacterial counts in C1 and C2 accounted for 1.43 × 10⁴ ± 1404 and 1.57 × 10⁴ ± 1501 cells/L at day 21, respectively, which slightly increased to 1.57×10⁴ ± 1285 and 1.67 × 10⁴ ± 1137 cells/L at day 84, respectively, with introduction of nitrate.

Unlike in C1 and C2, the K values in C3 before and after introduction of nitrate were fairly varied. Before introduction of nitrate, the average K value in C3 was 0.0179 ± 3.73 × 10⁻³ cm/s, which decreased to 0.0133 ± 1.11 × 10⁻³ cm/s after introduction of nitrate (Fig. 2). Data of bacterial counts in C3 supports that this decline is attributed to microbial reproduction. Before nitrate was introduced, the result of bacterial counts in C3 was 5.97 × 10⁴ ± 3113 cells/mL at day 21; with introduction of nitrate, this value increased to 8.15 × 10⁵ ± 3139 cells/mL at day 84 (Fig. 3). It appears that this increase in bacterial counts is relevant to stimulation of denitrifying bacteria that utilizes nitrogen as an electron acceptor. Fig. 4 presents that introduction of nitrate led to COD removal 57% in C3, which did not occur before nitrate was introduced. Furthermore, a decrease in pH between influents (on average, 6.6 ± 0.1) and effluents (on average, 6.4 ± 0.2) was no longer observed in C3 with introduction of nitrate. These results suggest that fermentation did not occur after nitrate was introduced. This is most likely because the introduced nitrate served as an electron acceptor. In addition, the finding that 44% of the introduced nitrate was removed only in C3 (Fig. 5) also supports that denitrifying bacteria were stimulated.

The findings from the above analyses demonstrate that nitrogen can be a factor affecting bioclogging in MAR. While introduction of nitrate was not able to stimulate significant microbial growth under conditions of insufficient organic content, nitrate boosts microbial reproduction, particularly that of denitrifying bacteria, when sufficient organic matter is provided. This suggests that control of both electron donors (i.e., organic matter) and acceptors (i.e., nitrogen) in recharge water can provide a more effective measure for avoiding excessive bioclogging in the MAR.

The current study investigated effects of nutrients in recharge water on bioclogging in a sand grain MAR. The results demonstrate that high contents of carbon and additional introduction of nitrogen resulted in significant stimulation of fermentation and denitrifying bacteria, leading to serious bioclogging in a sand grain MAR. As discussed in the introduction, sources for recharge water include not only highly treated water such as effluents from water treatment facilities but also less or none treated water including stormwater runoff and irrigation flow. The present study demonstrates that employment of untreated water to MAR can adversely affect sustainability of MAR. However, reliable guidelines for quality of recharge water to prevent bioclogging in MAR have not been developed yet. Most earlier attempts have been based on limited research data [40], and others have recommended treating recharge water to a level of drinking water quality [5]. Given that most MAR operations utilize reclaimed water, achieving such costly high-quality recharge water may be inefficient. Hence, further evaluation for relations between various water quality parameters and bioclogging in MAR systems is necessary, which will aid in developing efficient guidelines for sustainable operation of MAR.