Artificial groundwater recharge (AGR) is a widely applied technique for producing drinking water by the infiltration of surface water into an appropriate geological formation, such as an esker. The continuous infiltration of surface water imposes the conduction of large quantities of natural organic matter (NOM) into an aquifer. Designing and implementing new areas for AGR requires a thorough understanding of the underlying biogeochemical and hydrological processes, including the mechanisms involved in the attenuation of NOM during infiltration. This study aimed to address the role of biodegradation in NOM removal and changes in microbial community structure during AGR for drinking water production. A bench-scale sand column and a fluidized-bed reactor (FBR) were used to quantify the proportion of mineralisation from dissolved organic carbon (DOC) removal and to reveal the influence of temperature on biodegradation. The spatial and temporal changes of microbial communities in the experimental systems were evaluated using culture independent molecular biology methods. The NOM removal and microbial community changes during infiltration were followed at three AGR sites in Finland; Hämeenlinna (site A), Jyväskylä (site B) and Tuusula (site C). In addition, extracellular enzyme activities and nutrient availability at the Tuusula site were determined.
The study showed efficient NOM removal along the flow path in both the experimental sand column and at the full-scale AGR sites. Reductions of total organic carbon (TOC) at sites A, B and C were 91%, 84% and 74%, respectively, in the winter, and 88%, 77% and 73%, respectively, in the summer. Similarly, total P decreased along the flow path at site C down to the detection limit. The proportion of organic N from total N decreased from 45% in the infiltration basin to 4 - 15% in the production well, most likely as a result of microbial metabolism. The average ratios of C:N:P indicated P limitation along the aquifer flow path and the natural groundwater. The nutritional conditions in the aquifer thus became highly oligotrophic. In the saturated sand column, on average 76% and 81% of TOC was removed within the first sampling port at 0.6 m and the entire length of 18.5 m, respectively. No accumulation of NOM was detected despite the long recharge period of 941 d. Increasing the hydraulic load from 0.3 m3(m2d)-1 to 3.1 m3(m2d)-1did not affect TOC removal. This was likely due to the washout of NOM from the sand matrix due to the sudden increase in the flow rate. Large molecular size and aromatic NOM fractions were preferentially removed in the beginning of the flow path, which was likely due to sorption. However, no accumulation of smaller fractions was observed.
Biodegradation covered a substantial proportion of NOM removal in the sand bed. The δ13CDICanalysis revealed 32% to 52% DOC mineralisation in the sand column, depending on the temperature and hydraulic load. In both the continuous-flow FBR and the FBR batch tests, the highest average dissolved oxygen (DO) consumption rate was reached in summer (June-August), when lake water temperature was at its highest, followed by the fall, spring and winter. In the FBR batch tests, the DO consumption at low temperatures followed a typical kinetic curve for NOM biodegradation; i.e., a deep initial linear curve followed by a lower gradient. This illustrates the difference between rapidly and slowly biodegradable and non-biodegradable fractions. In the batch tests, a Q10 of 2.3 was found, which reveals the strong temperature dependency of NOM biodegradation. The analysis of δ13CDICrevealed 27% and 69% mineralisation of DOC in the FBR batch tests at 23 °C and 6 °C over 65 min and 630 min, respectively.
In the infiltration basin sediments at site C, substantially higher biomass content, measured as volatile solids (VS), was present in the surface layer (from 0.7 mg VS g-1 to 94 mg VS g-1) as compared to the bottom sediments (from 0.5 mg VS g-1 to 3.0 mg VS g-1) (measured to a total depth of 10 cm). Similarly, the chlorophyll-a content was higher at the surface than in the bottom sediments, the quantity depending on the season. In the sand column, the VS showed continuous adsorption and desorption of biomass at the 1.2 m sampling port with the average quantity of 6.4±0.8 mg VS g-1 dw sand. Most of the NOM was removed before the 1.2 m sampling port. In the FBR, on the other hand, the quantity of carrier biomass increased until a certain level, after which no more increase occurred during the course of time of the research experiment. The maximum biomass amount in the FBR was 13.1 mg VS g-1 dw carrier. A substantial decrease in cell counts occurred in aquifers at sites A, B and C. The average cell counts in raw waters varied from 7.4 × 105 cells ml-1to 24.0 × 105 cells ml-1 and in extracted groundwaters from 0.5 × 105 cells ml-1 to 1.0 × 105cells ml-1. In the sand column, the average decrease in the cell counts by 0.6 m distance was from 20.2±5.7 × 105 cells ml-1 (n=23) in the raw water to 4.3±1.1 × 105 cells ml-1 (n=3) at 0.6 m distance. In the FBR, on average 35±5% fewer cells were present in the outlet water compared to the inlet water.
At sites A, B and C, the bacterial communities in raw waters and extracted groundwaters were diverse. Changes occurred during infiltration, which was shown by DNA extraction followed by the PCR of 16S rRNA genes and denaturing gradient gel electrophoresis (DGGE) fingerprinting. While the natural groundwater microbial community was diverse, it was different from that of the extracted groundwater in the AGR area. In the sand column, both the DGGE fingerprinting and the length heterogeneity analysis of amplified PCR products (LH-PCR) showed a change in the bacterial community already by 0.6 m distance as the community composition shifted from an Actinobacteria-dominated population to a diverse, mainly Proteobacterial community. Concurrently, a substantial decrease in DOC concentration and cell counts had occurred by that stage. The original lake water community changed overnight in the FBR feed tank amended with phosphate and nitrate. The feed tank community differed from the FBR outlet water community. While the water phase was dominated by Actinobacteria, Proteobacterial groups dominated in the biofilm. However, the dominance of the specific groups was not constant, which illustrates the dynamic nature of both communities.
Compared to raw water, substantial increases were detected in the specific extracellular enzyme activities (EEAs) of α-D-glucosidase, β-D-glucosidase, phosphomonoesterase, leucine aminopeptidase and acetate esterase when measured in the AGR aquifer at site C. This was paralleled with decreasing nutrient concentrations shown by strongly negative correlations between the measured EEAs and the nutrient pools. The trend of increasing EEA along the flow path thus indicates a decrease in the availability of nutrients for bacteria. The EEA in the basin sediment (down to 10 cm) and the pore water samples were of the same order of magnitude as the basin water, indicating similar nutritional conditions. The EEAs had strong positive correlations with each other, suggesting synergistic and cooperative functions.
In summary, the study demonstrated the substantial role of biodegradation in NOM removal during AGR. The biodegradation was shown to depend on seasonal changes in raw water characteristics. A change in environmental conditions was shown to be reflected by changes in the composition and the physiological functioning of the microbial community. The study contributed to the understanding of NOM removal mechanisms in AGR and provides information for different interest groups of water production.