Biofiltration field study for cold Fe(II)- and Mn(II)-rich groundwater: accelerated Mn(II) removal kinetics and cold-adapted Mn(II)-oxidizing microbial populations

Dissolved manganese in groundwater exists in its reduced form, Mn(II). Excess soluble Mn(II) is a common element of concern in water supplies. Under redox conditions favorable for Mn(II) oxidation, Mn(II) is converted to solid, dark brown Mn(III/IV) oxides, which results in black discoloration and lowers the esthetic quality of drinking water. The deposition of Mn(III/IV) oxides causes scaling and corrosion in water supply infrastructure (Pacini et al. 2005).

Biofiltration technology has been widely considered for the treatment of Mn-rich groundwater. Biofiltration promotes microbially mediated Mn(II) oxidation that converts soluble Mn(II) to insoluble Mn(III/IV)-oxide precipitates, which are easily filtered out by filter media in biofiltration systems. However, microbial activity is temperature-dependent. Low temperatures greatly hinder the activity of mesophilic Mn(II)-oxidizing bacteria (MnOB), resulting in long start-up periods and the need for energy-intensive biofiltration (Tekerlekopoulou et al. 2013). This is a concern, particularly in vast northern cold-climate regions (e.g., Canada, Alaska, and northern China and Europe) (Štembal et al. 2005; Cai et al. 2016). Previous studies report that Mn(II) removal promoted by MnOB activity and/or Mn-oxide surface coatings on filter materials significantly decreases below 15 °C (Berbenni et al. 2000; Cooley & Knocke 2016). The Mn(II) oxidation activity of Leptothrix discophora (MnOB) incubated at 8 °C, for example, was suppressed to 50–80% of its oxidation activity at room temperature (Boogerd & de Vrind 1987).

However, at the same time, MnOB are ubiquitous in natural and engineered environments such as the deep sea, rivers, lake sediments, fjords, and water pipes (Tebo et al. 2005). Indigenous MnOB tolerant to low temperatures may be abundant or even prevail in cold habitats (Krishnan et al. 2009). Enhancing MnOB adapted to low temperatures may be crucial for the onset of Mn(II) oxidation and the generation of biogenic Mn oxides. Those biogenic Mn oxides may then trigger autocatalytic Mn(II) oxidation during biofiltration (Burger et al. 2008; Bruins et al. 2015a), which may positively influence the kinetics of Mn(II) removal in cold environments. Yet, cold-adapted MnOB capable of rapidly generating Mn oxides at low temperatures have not been extensively reported in previous field studies. In addition, the onset, acclimation, and acceleration of Mn(II) removal under low temperature regimes have not been extensively described in biofiltration studies conducted at field sites.

In this study, a flow-through pilot-scale biofiltration unit containing virgin filter media (anthracite, sand, and gravel) was installed for the field experiment for the treatment of cold Fe(II)- and Mn(II)-rich groundwater. Fe(II) is often naturally abundant when groundwater is rich in Mn(II). Fe(II) oxidation is very thermodynamically favorable and occurs rapidly in the presence of oxygen. However, the Mn(II) oxidation kinetics are very slow or negligible in the absence of biotic and/or abiotic catalysts (Mouchet 1992). This study focuses on Mn(II) removal through a two-stage biofiltration unit at a water treatment plant in the town of Langham (Saskatchewan, Canada), which has an annual mean temperature of ∼2 °C. The temperature of the influent groundwater that continuously flowed into the biofiltration unit ranged from 4 to 8 °C, and the effluent temperature of the Mn filter ranged from 8 to 14.8 °C throughout the study. The objective of this study was to demonstrate how Mn(II) oxidation activity commenced, acclimated and accelerated in the biofiltration unit under the low on-site temperature regime. This study also addresses how Mn(II) removal kinetics is related to filter ripening. A cold-adapted MnOB consortium derived from the sufficiently aged Mn filter was investigated.

The groundwater source in this field experiment was the Floral formation of the Dalmeny aquifer in Saskatchewan (SK) (Fortin et al. 1991), accessed via a groundwater well at the Langham Water Treatment Plant. Water chemistry data for the influent groundwater used in this study are presented in Table 1.

Figure S1 in the Supplementary material (available with the online version of this paper) illustrates the schematic diagram for the biofiltration unit. The biofiltration unit was designed to remove Fe(II) in the first filter (Fe filter) and Mn(II) in the second filter (Mn filter). From top to bottom, the two filter columns contained four layers of filter media: granular anthracite, fine sand, coarse sand, and gravel. An aeration tank with an air compressor was installed between the Fe and Mn filters to aerate the groundwater. The biofiltration experiment was conducted over six months from July 2014 to January 2015, to include both a summer and a winter season. The operational parameters for the biofiltration unit are summarized in Table 1. Detailed information about the biofiltration operation is available in the Supplementary material.

During biofiltration, water chemistry data were collected on-site for the effluents of the two biofilters: Fe(II) and Mn(II) concentrations, temperature, pH, oxidation-reduction potential (ORP), and dissolved oxygen (DO). Colorimetric analyses were used to measure Fe(II) and Mn(II) concentrations (Hach Methods 8146 and 8149, respectively). Two sets of three probes were installed at the outlet of each biofilter (Hach-PD2P1, Hach-RD1P5, and Hach-5540D0C) to monitor pH, ORP, DO, and temperature. The formation of Mn oxides on the filter media in the biofiltration unit was assessed on-site using the Leucoberbelin blue I (LBB) colorimetric assay (Sigma-Aldrich Canada). For water samples collected from the Mn filter, off-site analyses were performed using inductively coupled plasma optical emission spectroscopy (ICP-OES) to precisely confirm the changes in Fe(II) and Mn(II) concentrations in the groundwater before and after the biofiltration treatment. Detailed information about effluent sampling (triplicate) and data analyses, including statistical analyses, is presented in the Supplementary material.

The number of viable MnOB, the most probable number (MPN), and adenosine triphosphate (ATP) concentrations were determined as similarly described in relevant previous studies (Tebo et al. 2007; Granger et al. 2014). Cold-adapted MnOB consortia were cultured at 8 °C using backwash water collected on day 110 from the Mn filter. Further information for the culture-dependent analyses above is presented in the Supplementary material.

Genomic DNA was extracted from the cultured consortium using PowerWater^® DNA Isolation Kits (MoBio Laboratories). Extracted DNA was sent to RTLGenomics (Lubbock, Texas) for high-throughput amplicon sequencing using the Illumina-MiSeq platform. Raw sequence reads were processed using the MOTHUR software package, version v.1.32.0, following the standard operating procedure for MiSeq analyses (Schloss et al. 2009). 16S rRNA gene amplicons (pair ends) from the enriched MnOB consortium were deposited in the European Nucleotide Archive (ENA) under project number PRJEB15479, and accession numbers ERS1360945 and ERS1360944.

To characterize the surface morphology and elements of the solid precipitates and filter materials, analyses were conducted using scanning electron microscopy with energy dispersive X-ray spectrometry (SEM/EDS; Hitachi S3000-N SEM with a Quartz PCI XOne SSD X-ray analyzer). An electron paramagnetic resonance (EPR) spectroscopy analysis was performed for the precipitates obtained from the culturable MnOB consortium (Supplementary material). Synchrotron-based powder X-ray diffraction (PXRD) analyses were conducted using the Canadian Macromolecular Crystallography Facility's bending magnet beamline (CMCF-BM, 08B1-1) at the Canadian Light Source (CLS). Detailed information about the data processing is available in the Supplementary material.

The Mn(II) concentrations in samples collected from the four sampling ports of the Mn filter were measured for the kinetic analyses. Temporal changes in the vertical distribution of Mn(II) concentrations were assessed using the four sampling ports installed vertically along the biofiltration unit at varying depths. Contour maps for Mn(II) concentrations in the Mn filter were generated using Surfer 13 (Golden Software).

Dissolved Fe(II) concentrations in the groundwater rapidly decreased through the Fe filter (Figure 1(a)). Fe(II) concentrations in the effluent stream of the Fe filter were consistently below the drinking water standard (DWS) for Fe(II) of 0.3 mg/L. Dissolved Fe(II) was removed from the groundwater prior to Mn biofiltration consistently over the course of the field experiment (183 days).

Figure 1(b) presents changes in the Mn(II) concentrations in the Mn filter effluent over time. Figure 1(c) shows the effluent temperatures monitored at the outlet of the Mn filter. Mn(II) concentrations in the Mn filter effluent were relatively stable over the first month (days 0–29) at ∼0.8 mg/L, which was only ∼0.1 mg/L less than the Mn(II) concentration in the influent groundwater (0.88 ± 0.01 mg/L). However, between days 29 and 34, Mn(II) concentrations began to decrease in the Mn filter effluent (Figure 1(b)). Meanwhile, the effluent temperatures were stable at 8 °C for the first 81 days of the biofiltration experiment, indicating that Mn(II) removal commenced in 29–34 days at the on-site temperature of 8 °C. Two additional months were required to reach the steady-state functioning of the Mn filter at 97 days, at which point the Mn(II) removal efficiency was high and stable (97 ± 0.9%). The mean effluent temperature during the steady-state functioning period was 12 ± 0.6 °C.

Mn(II) concentrations in the Mn filter effluent collected at the site were 0.03 ± 0.01 mg/L during the steady-state functioning period. The ICP-OES analyses for the effluent collected on day 110 confirmed that total Mn(II) concentrations in the effluent were below the DWS of 0.05 mg/L (Table 1). The difference in the datasets for mean total Mn(II) concentrations in the groundwater before and after biofiltration was statistically significant, with p < 0.0001 (t-test, Table 1). The LBB I assay (Figure 1(e)) provided an on-site indication of Mn-oxide formation on the anthracite and sand collected from the Mn filter on day 183 (at the end of the experiment). Mn(II) oxidation in the Fe filter was negligible, as indicated by the LBB I assay for the filter media collected from the Fe filter (Figure 1(d)). Total Mn concentrations in the Fe filter effluent (day 110) were as high as 0.84 mg/L. This is most likely due to the lower redox potential in the Fe filter (e.g., <200 mV), which allows Fe(II) oxidation but not Mn(II) oxidation (Mouchet 1992). NH₃ concentrations were as low as 0.53 mg/L in the groundwater influent, and nitrifying activity would not prevail in the Fe filter.

After Mn(II) removal began at 8 °C and shortly before the start of steady-state functioning in the Mn biofilter, the effluent temperature increased from 8 to 13 °C due to indoor heating in the water treatment facility in October (Figure 1(c)). There was no direct heat supply to the biofiltration units. After October, effluent temperatures fluctuated between 9.5 and 14.8 °C until early December (after the steady-state condition was reached), and then stabilized at 11 ± 0.6 °C for the rest of the winter period (until January), during which the Mn(II) removal efficiency of the Mn filter was still stable at 97% and was not influenced by the temperature fluctuations (Figure 1(b) and 1(c)).

The ORP in the Mn filter gradually increased as the Mn(II) concentrations decreased in the Mn filter effluent stream (Figure 2(a)). This correlation was statistically significant based on a Spearman correlation analysis (correlation coefficient: −0.84, p < 0.0001). Once the ORP in the Mn filter reached 300 and 313 mV around days 29 and 34, Mn(II) removal commenced as described above (Figure 2(a)). A steep increase in the ORP from ∼300 to ∼430 mV was observed shortly after a filter backwash with a subsequent inoculation of the Mn biofilter with the backwashed sludge. At that point, the ORP remained over 400 mV and became relatively stable between 400 and 500 mV. Once the steady-state functioning of the Mn filter was established, the ORP was less sensitive to both temperature fluctuations and backwashing frequency (Figures 1(c) and 2(a)).

Using the ORP and corresponding pH data measured from the influent groundwater and Mn filter effluent, Pourbaix diagrams (Mn-H₂O system) were constructed (Figure 2(b)). The Eh (or ORP) and pH conditions of the Mn filter notably shifted from reducing to oxidizing conditions over the course of biofiltration. As the Mn filter acclimated, Eh-pH conditions in the Mn filter gradually shifted to conditions suited for biological Mn(II) oxidation, as similarly reported by Mouchet (1992) and Pacini et al. (2005). Once the steady-state functioning of the Mn filter was reached (after 97 days), all Eh-pH data points fell within the field of biological Mn(II) removal. The shift in the redox conditions and pH in the Mn filter suggested the enhancement of biological Mn(II) oxidation.

The co-presence of Fe(II), NH₃, DO concentrations, and pH were not within their limiting ranges for biological Mn(II) oxidation (Table 1). The DO content in the Mn filter was maintained at 9 ± 1.7 mg/L. An increase in nitrate (NO₃⁻) concentrations from 0.04 to 1.8 mg/L was observed during biofiltration. However, nitrification did not appear to inhibit Mn(II) oxidation in the Mn filter, presumably due to the low initial concentrations of NH₃.

The viable populations of aerobic MnOB in the effluent and backwash water of the Mn filter (collected on day 110, early in the period of steady-state functioning) were significantly greater than in the untreated groundwater (also sampled on day 110), based on the plate-count enumerations of MnOB at the start-up temperature of 8 °C (Figure 3(a); one-way ANOVA, p < 0.0001). Similarly, the MPN data for day 106 samples showed that a greater number of culturable heterotrophs were present in the Mn filter effluent (93 MPN/mL) than in the raw influent groundwater (2 MPN/mL) and the Fe filter effluent (59 MPN/mL). The ATP concentrations measured after the experiment in the groundwater, storage reservoir, Fe filter effluent, and Mn filter effluent indicated that ATP was notably elevated in the Mn filter effluent, indicating that microbial metabolic activity continued throughout the period of steady-state functioning (Figure 3(b)).

The additional culturing experiment using backwash water collected on day 110 was conducted to identify a culturable MnOB consortium adapted to the start-up temperature (8 °C), and to visually observe biogenic Mn oxide formation. The backwash water used for the experiment was collected when the ORP of the Mn filter exceeded 400 mV (Figure 2(b)). The MnOB consortium from the backwash water was able to rapidly produce biogenic Mn oxides within five days at 8 °C (Figure 4). The EDS analysis detected Mn in the precipitates (Figure 4(d)). The SEM images showed poorly crystalized precipitates with amorphous morphology, which is characteristic of biogenic Mn oxides (Figure S2, available with the online version of this paper). The Mn-bearing solid precipitates exhibited a ΔH of 394 gauss, centered at a g-factor of 2.0, confirming the biogenic origin of the Mn oxides (Figure 4(e)). Conversely, the ΔH and g-factor of chemically synthesized birnessite mineral (standard) were 2,816 gauss and 2.0, respectively, indicating an abiotic origin for the Mn oxides (Figure S3, available with the online version of this paper). The rapid formation of biogenic Mn oxides at 8 °C reflected that the Mn filter had biologically acclimated at the varying low on-site temperatures. Even for filter media that were pre-coated with Mn oxides in other Mn(II)-removal studies, temperatures below 15 °C (especially below 10 °C) still significantly decreased the Mn(II) removal efficiencies (Jia et al. 2015; Cooley & Knocke 2016).

The Illumina-MiSeq 16S rRNA gene sequencing analysis for the microbial consortium identified known MnOB genera, Pseudomonas, Leptothrix, Flavobacterium, and Zoogloea (highlighted in bold font in Figure 4(c)). Approximately 60% of the microbial consortium is composed of known MnOB from the literature, based on the relative abundances of the genera as percentages of the amplicon sequence reads. Of these, Pseudomonas, Leptothrix, and Flavobacterium have been frequently reported in Mn biofiltration studies (Tekerlekopoulou et al. 2013; Cai et al. 2015). However, the microbial consortium includes other diverse populations (over 46 genera), including the genera Rhodocista (11.3%), Phaselicystis (7.3%), Bdellovibrio (5.1%), and Hydrogenophaga (<2.5%).

After the multiple backwashes and inoculation of the Mn filter with biologically aged backwash sludge collected at high ORP values of +400 mV – representing the strong activity of acclimated MnOB and rapid biogenic Mn-oxide formation – a drastic increase in the first-order rate constant (k) for Mn(II) removal was observed, an increase of up to five orders of magnitude (Figure 5(d)). The k values were as low as 10⁻⁶ to 10⁻⁵ min⁻¹ between day 97 (the beginning of steady-state functioning) and day 127. The k value rose steeply to 0.21 min⁻¹ immediately after the multiple backwashes and backwash sludge inoculation on days 127 to 132 (Figure 5(d)). The corresponding half-life (t_1/2) for Mn(II) removal was substantially reduced from ∼40 days on day 92 (before the acceleration) to 3.31 min on day 159 (after the acceleration).

In particular, the acceleration of Mn(II) removal kinetics in the Mn filter was observed in the middle of steady-state functioning, at a time when high Mn(II) removal efficiencies had already been observed (Figure 5(d)). This finding indicates that tracking Mn(II) removal efficiency does not adequately take into account the ripening of the filter media. Figure 6 shows that k values were as low as 0.007 to 0.068 min⁻¹ between days 132 and 148, even though the Mn filter exhibited the 97% removal efficiency due to the ripened bottom layers. Low k values reflected the incomplete ripening of the filter media. The drastically accelerating Mn(II) removal kinetics correlated to the progressive vertical ripening of the Mn filter media. The filter ripening progressed from the bottom (gravel), to the middle (sand), and to the top (anthracite), likely due to accumulating precipitates, biomass and/or sludge solids at the bottom (Figure 6). The different types of filter media did not significantly affect biofilter ripening. As the k value for Mn(II) removal reached 0.21 min⁻¹ with a short half-life of 3.31 min, Mn(II) from the influent groundwater was depleted rapidly as it passed through the topmost layers of the Mn filter (as specifically shown at 159 days in Figure 6). At this stage, the Mn filter was considered nearly fully ripened based on the maturation of Mn(II) removal kinetics.

The findings presented herein are important from a field perspective. Tracking changes in the EBCT-based rate constant for Mn(II) removal is a simple approach to indicating filter ripening that is sufficiently advanced to trigger meaningful Mn(II) removal under low temperature regimes, and to determining the effective ripening period required for a biofilter. During the period of accelerated Mn(II) removal, the Mn filter effluent temperatures were very stable at 11 ± 0.6 °C (Figure 5(c) and 5(d)). The on-site low temperature of 11 °C did not limit the acceleration of Mn(II) removal. The first-order rate constant of 0.21 min⁻¹ obtained after filter ripening in this study is comparable to rate constants for Mn(II) removal reported in previous Mn-biofiltration studies conducted at a wide range of temperatures. Some of those studies were conducted at warmer temperatures and with pre-aged Mn filter materials (Table 2). The four-month aging period in this study was shorter than those in previous studies (Table 2).

Birnessite (K-rich MnO₂; Mn(IV) oxide) and pyrolusite (β-MnO₂; Mn(IV) oxide) were detected in the anthracite collected on day 183 (end of the experiment), confirming that the Mn filter was fully ripened with Mn oxides with the highest Mn oxidation states based on the Mn oxidation pathway (Figure 7(c)). The Mn oxidation states exhibited by birnessite and pyrolusite were +3.5 to +3.9, and +4, respectively. Hausmannite (Mn₃O₄; Mn(III) oxide) with Mn oxidation state +2.7 (Bruins et al. 2015b) was detected in centrifuged backwash water collected on day 110 (early steady-state functioning period).

The ripened anthracite collected from the Mn filter on day 183 was coated mainly by Mn-rich substances with a repetitive coral-like morphology (Figure 7(b)). This surface morphology is similar to that of birnessite formed by autocatalytic Mn(II) oxidation based on visual comparisons of SEM images (Jiang et al. 2010; Bruins et al. 2015a). However, variations in the surface morphologies of Mn-rich coatings on the ripened filter media were easily observable on day 183 (Figure 7(a) and 7(b)). The SEM images of some surfaces of the ripened anthracite indicated a birnessite-like surface morphology that contains fewer coral-like patterns and appears to be fluffier (Figure 7(a)). Such a morphology has been similarly observed for biogenic Mn-oxide coatings also reported by Bruins et al. (2015a) and Jiang et al. (2010). There was no detection of Mn oxides in the virgin anthracite samples (control) (Figure 7(c) and Figure S4, available with the online version of this paper). Considering the multiple lines of evidence of enhanced biological Mn(II) oxidation, varying birnessite-like surface-coating morphologies reflected microbially mediated autocatalytic signatures of Mn(II) oxidation. Birnessite acts as an autocatalyst for physico-chemical Mn(II) oxidation (Bruins et al. 2015b). The detection of birnessite (crystalized) in the ripened filter media, which were largely coated by birnessite-like substances, may be connected to the accelerated Mn(II) removal kinetics observed in this biofiltration study.

In the Mn filter, biological Mn(II) oxidation was significantly enhanced, which was observed using multiple indicators. In the early stage of Mn(II) biofiltration, however, the Mn filter exhibited lower k values ranging from 10⁻⁶ to 10⁻⁵ min⁻¹ at the variable on-site temperatures (8–14.8 °C). Immediately after multiple backwashes and the inoculation of the biofilter with biologically acclimated backwash sludge, k values markedly increased to 0.21 min⁻¹ at 11 ± 0.6 °C, which is comparable to k values obtained at higher temperatures in previous studies. The drastically accelerating k for Mn(II) removal was correlated to the vertical progression of filter media ripening. The PXRD analyses detected birnessite and pyrolusite with high Mn oxidation states ranging from +3.5 to +4 in the ripened filter media. The varying surface morphologies of the ripened filter materials exhibited combined biological and autocatalytic signatures. The MnOB microbial consortium capable of rapidly producing biogenic Mn oxides at the start-up temperature of 8 °C was obtained from the biologically aged backwash water. These findings have important implications for improving biofiltration in cold regions, and represent a potential breakthrough for accelerating Mn(II) oxidation kinetics in cold groundwater.

We acknowledge technical and financial support provided by the Delco Water Division of Delco Automation Inc. This research was also funded by the Natural Sciences and Engineering Research Council of Canada (NSERC; EGP 468673-14 and RGPIN-2014-05902), Mitacs (Mitacs-Accelerate Graduate Research Internship, IT06847), and the Canada Foundation for Innovation (CFI; JELF#33982). The synchrotron-based PXRD analyses described in this study were performed at the Canadian Light Source (CLS), which is supported by the NSERC, the National Research Council Canada, the Canadian Institutes of Health Research, the Province of Saskatchewan, Western Economic Diversification Canada, and the University of Saskatchewan. We thank Dr Michel Fodje, Dr Viorica Bondici, and Samira Sumaila for technical support during the PXRD analyses, and the CLS for granting access to their analytical equipment in the Life Sciences Laboratory. We also thank Dr Ramaswami Sammynaiken for the EPR analyses at the Saskatchewan Structural Sciences Centre at the University of Saskatchewan.