Start-up of non-bioaugmented pumice biofilters in flow-through and recirculating flow regime for Mn removal Open Access

Biofiltration is considered a cost-effective and suitable technology for manganese (Mn) removal from groundwater (Marsidi et al. 2018). Compared to conventional physicochemical treatments, a significant advantage of biofiltration is that the use of chemicals is not necessary, representing lower operation and maintenance costs (Pacini et al. 2014; Cai et al. 2015). Mn removal in biofilters involves both biological and physicochemical processes (Bruins et al. 2017; Breda et al. 2019b). Particularly, Mn removal in a non-coated virgin medium is initiated biologically (Bruins et al. 2015a; Breda et al. 2019b). Mn²⁺ is oxidized by Mn-oxidizing bacteria (MOB) and deposited gradually as Mn oxide (MnO_x(s)) (Mouchet 1992). Subsequently, more Mn²⁺ is adsorbed in the MnO_x(s)-coated filter medium and is autocatalytically oxidized (Sahabi et al. 2009; Bruins et al. 2015a). Previous research confirmed that biological Mn²⁺ oxidation occurred mainly at the top of the non-inoculated biofilters during the start-up (Breda et al. 2019b). Finally, the start-up period is considered complete when the biofilter becomes functional in compliance with a drinking water criterion (Breda et al. 2016) or with a high Mn uptake (>90%) (Bruins et al. 2017). However, filter ripening typically continues even when high Mn²⁺ removal efficiencies have already been observed (Dangeti et al. 2017; Breda et al. 2019a).

Bioaugmented methods and inherent inoculation are used to start up the biofilters. The first one involves the following steps: biofilters are inoculated using a concentrated source of microorganisms, isolated, and grown in the laboratory (Ramsay et al. 2018; Zeng et al. 2019), backwash sludge (Štembal et al. 2004; Cai et al. 2015; Cheng 2016), or matured filter media (Zeng et al. 2010; Bruins et al. 2015b; Breda et al. 2019b), which are taken from other active Mn biofilters. These methods have been demonstrated to be effective, reporting rapid start-up periods for Mn removal from 2 up to 4 weeks (Štembal et al. 2004; Zeng et al. 2010; Cai et al. 2015; Breda et al. 2019b). The success of these methods relies on the proper selection and acclimatization of the biomass used as inoculum (Tekerlekopoulou et al. 2013). However, these resources may not be available, especially in zones where biofiltration technology is just emerging. In such cases, inherent inoculation of non-bioaugmented biofilters by autochthonous MOB present in raw water should be used. However, an important drawback is the long start-up period, required for virgin filter media, to achieve effective Mn removal (Ramsay et al. 2018). Typically, it takes between 1 and 4 months (Bruins et al. 2015b). Bruins et al. (2017) affirmed that several factors such as groundwater quality, iron (Fe) loading, type of filter media, and operational parameters (e.g. filtration velocity and backwash strategy) seem to influence the start-up of non-bioaugmented biofilters for Mn removal (Bruins et al. 2017). According to the authors, some strategies could be applied to reduce the ripening time of virgin filter media for Mn removal using inherent inoculation, such as to avoid or previously reduce the presence of other contaminants (e.g. Fe) and/or temporarily operating filters at a low filtration velocity.

The oxidation rates of Fe²⁺ and Mn²⁺ by oxygen, at neutral pH and above, are different, being Fe²⁺ relatively rapid oxidized, whereas the Mn²⁺ oxidation is orders of magnitude slower (Singer & Reckhow 2011). Therefore, simple aeration is commonly used for the oxidation of Fe²⁺, forming Fe (hydr)oxide precipitates. Even though the produced Fe oxides can adsorb Mn²⁺ and catalyze its oxidation, they can block the active sites in the filter media avoiding Mn²⁺ adsorption (Buamah 2009; Bruins et al. 2014). Moreover, Fe²⁺ can compete with Mn²⁺ for adsorption sites (Hu et al. 2004), and Mn²⁺ adsorption on Fe (hydr)oxide is lower than manganese (hydr)oxides (Buamah et al. 2008). Consequently, Mn removal is commonly better when Fe is previously removed or the Fe-loading prefilter run is lower (Bruins et al. 2014). According to Bruins et al. (2015b), the effect of a lower Fe²⁺ concentration in the feed water, combined with appropriate operational conditions (e.g. lower filtration velocities and backwash frequency intensity), plays an important role to accelerate the ripening of virgin filter media for Mn removal. In addition, Bruins et al. (2017) found that backwashing prolongs the ripening time of a virgin filter and therefore recommended to recirculate part of the filtrate or lower the filtration velocity to reduce the Fe-loading and subsequent backwashing. Regarding matured filters, Bruins et al. (2014) found from a statistical analysis of 100 drinking water plants that efficient Mn removal in matured aerated-rapid filters was guaranteed when Fe²⁺ was previously removed or when the Fe-loading per filter run was lower than 2.7 kg Fe·m⁻². Araya-Obando et al. (2022) also found that matured rapid sand filters exhibited effective Mn removal during 10 years of operation using an up-flow roughing filter (URG) as Fe pretreatment (removing 68% of the total Fe).

A lower filtration velocity enhances the settling and association of bacterial cells with the media surface (Donlan 2002). Slow velocities from around 1.5 m·h⁻¹ (Zeng et al. 2010) up to 5 m·h⁻¹ (Breda et al. 2016; Bruins et al. 2017) are commonly adopted during the start-up, with a constant velocity or by gradually increasing the velocity when the filter effluent reaches the treatment goal (speed-up stage) (Zeng et al. 2019). Other studies have recommended to operate the flow-through biofilters at half or one-third of the designed hydraulic load (Štembal et al. 2004; Li et al. 2005; Ramsay et al. 2018). Backwashings with a low-flow intensity (<30–35 m·h⁻¹) (Bruins et al. 2015b; Breda et al. 2016) have been recommended during the start-up to prevent bacterial detachment.

A novel approach, for inherent inoculation, could be to recirculate (between the effluent and influent of biofilter) the flow to improve the effect of inoculation of biofilters. Cheng (2016), in a bioaugmented method, recirculated the supernatant water of the sedimented backwashing sludge through a pilot-scale biofilter for 3 days and repeated it for 2 more days to improve the effect of inoculation. Then, the filters were operated in a flow-through regime at 2 m·h⁻¹ and reached the water quality criteria after 44 days. Subsequently, the velocity was increased to 3, 4, and 6 m·h⁻¹. This approach seems promising, as effective start-up of the degradation of organic compounds in bench-scale (pumice) biofilters by autochthonous microorganisms has been reported (Börnick et al. 2001; Worch et al. 2002). However, to the best of our knowledge, the influence of the recirculating flow regime, using only inherent inoculation, on the ripening time of virgin media for Mn removal, has not been reported.

In the previous work with the same groundwater source used in this study, the presence of autochthonous culturable MOB in raw water was confirmed (Calderón-Tovar et al. 2020). In addition, the feasibility of biofiltration for Mn removal using non-bioaugmented virgin pumice, anthracite, and sand media was demonstrated at bench-scale (Araya-Obando et al. 2021). Although, a long start-up period of about 80 days for Mn removal was required for the three materials, it was concluded that pumice has great potential for Mn biofiltration because it is a low-cost, porous, low-density medium and exhibited similar performance compared to sand and anthracite (commonly used in biofiltration). Therefore, to obtain further insight into the efficacy of the start-up of Mn removing in non-bioaugmented biofilters, using groundwater with relatively high temperatures (∼23 °C), this study focuses on the influence of the regime of operation on the ripening time of virgin pumice media. For this purpose, pilot-scale experiments were performed in duplicate to analyze the start-up of flow-through and recirculating gravity filter columns, with low Fe-loading and increasing filtration velocities.

To study the influence of the flow regime during the start-up for Mn removal of non-bioaugmented biofilters, two experiments were conducted in duplicate. In experiment 1 (flow-through regime), feed water passed continuously through virgin pumice columns until ripening (Mn removal >90%). In experiment 2 (recirculating flow regime), feed water was recirculated through biofilter columns until ripening, then the system was switched to flow-through regime. Both experiments were compared in terms of the start-up period, the Mn²⁺ concentration profiles registered over the depth of filter columns, the Mn-loading, and the water consumption.

The biofilter columns consisted of four identical polyvinyl chloride (PVC) columns with a diameter of 10 cm and a filter bed height of 70 cm. Each column was composed of virgin pumice with a grain size between 0.88 and 1.16 mm. To obtain further in-depth information on the start-up of the pilot-scale biofilters, four sampling points were distributed from the top of the media to the bottom at 15, 30, 45, and 70 mm (Figure 1). Filter columns were operated as downflow filters. The filtration velocity was controlled in the inlet of the columns (Figure 1), using a flowmeter rotameter with valve fit (LZM-6T, Sorekarain Company), with a flow range of 100–1,000 mL·min⁻¹. Specifically, two columns were used for the flow-through regime (experiment 1) and the other two for the recirculating flow regime (experiment 2). In both cases, the start-up period was defined as the ripening time required for the biofilters to achieve Mn removal efficiencies greater than 90% (Bruins et al. 2017).

In experiment 1, feed water from T2 flowed continuously in parallel through column 1 (C1) and column 2 (C2), and the effluents from both columns were discarded. The filtration velocities used during the experiment were determined according to Zeng et al. (2019), with slight modifications. Specifically, a low filtration velocity of 2 m·h was initially used to promote the attachment of autochthonous bacteria present in the raw water (inherent inoculation). The authors also proposed a speed-up stage. Hence, increments of about 1 m·h were performed when the Mn removal remains higher than 90%. Once the Mn removal reached 90%, the velocity was increased in steps of about 1 m·h until reaching a final filtration velocity of 5 m·h, keeping the desired removal percentage. In total, the filtration velocities of 2, 3, 4, and 5 m·h were used, resulting in empty bed contact times (EBCT) of 21, 14, 10.5, and 8.4 min, respectively. Besides, an initial supernatant water level of about 2 cm was provided above the filter media, as also used by Gude et al. (2018). Backwashing was performed using water from T1 when supernatant water level rose to 30 cm, using a filter expansion of about 10–20% for 10 min, similar to the pilot-scale study for Mn removal of Bruins et al. (2017).

Following a similar protocol used in biodegradation studies for organic compounds (Börnick et al. 2001; Worch et al. 2002), in experiment 2 (recirculating flow regime), column 3 (C3) and column 4 (C4) were also fed in parallel, but with water from the recirculation tank (T3) (177 L) that was previously filled with feed water from T2. Subsequently, the water from T3 was recirculated at 2 m/h (EBCT of 21 min) through the filter columns until Mn²⁺ concentration in water was not detectable at <0.03 mg L⁻¹. Afterward, feed water in T3 was discarded, repeating this protocol. In practice, feed water was renewed every 2–3 days. Consequently, feed water in T3 was renewed nine times in total. Then the columns were switched to flow-through regime and the speed-up stage procedure was applied until both columns operated at 5 m·h with a Mn removal higher than 90%. Initial supernatant water level and backwashing procedures were the same as described in experiment 1.

Raw water and filter column inlet and outlet samples were collected approximately two times a week. Water sampling along the filter bed was conducted weekly. Non-filtered samples were immediately acidified to pH < 2 with nitric acid and stored at 4 °C. A preliminary test showed no difference between filtered and non-filtered samples; hence, it was assumed that any Mn present was soluble Mn²⁺ (Cooley & Knocke 2016). Fe was referred here as total Fe. Mn and Fe concentrations were measured using an Analyst 800 atomic absorption equipment (Perkin Elmer, Waltham, USA) according to the Extraction/Air-Acetylene Flame Method 3111 C (APHA et al. 2005). The detection limits for Fe and Mn were 0.10 and 0.03 mg·L⁻¹, respectively. Dissolved oxygen (DO), pH, and oxidation–reduction potential (ORP) measurements were conducted using the Hach HQD30 equipment following the methods recommended by the manufacturer (Hach, USA).

Similar to Araya-Obando et al. (2021), 10 g of matured filter media were taken (in quadruplicate) from the top 10 cm of each filter column at the end of the experiments. The samples were diluted in 90 mL sterile 0.1% (w/v) peptone water isotonic solution and stomached for 1 min. Serial dilutions were made, up to 10⁻⁴, and 1 mL of each dilution was plated in duplicate on modified R2A agar with 17 mg·L⁻¹ MnSO₄ and incubated for 5 days at room temperature (<28 °C). The ability of isolated bacterial strains to oxidize Mn²⁺ was confirmed with the leucoberbelin blue (LBB) dye assay. Specifically, two drops of the LBB reagent (0.04% (w/v) dissolved in 45 mM acetic acid) were added to the colonies after an incubation period of 20 days. LBB dye is oxidized by Mn oxides (with Mn oxidation states of +3 and higher) producing a blue color (Piazza et al. 2019). MOB is referred to here as the isolated strains that resulted positive in the LBB test.

Data analysis was performed in R (R Core Team 2021). Summary descriptive statistics for all variables were done using stat.desc (library pastecs). Time series and concentration profiles were done using packages R Graphics.

Figure 3 also shows that Mn²⁺ concentration profiles after day 11 remained constant, locating the active zone for effective Mn removal at the top (<15 cm) of the pumice media bed. Inspection of the pumice medium after completion of the experiments showed dark Mn oxides, mainly deposited in the top 15 cm of the filter bed.

As shown in Figure 6(b), after filter columns were switched to the flow-through regime (day 33) and operated for 1 week, the active Mn removal zone was in the first 15 cm of then filter bed depth. As shown in Figure 5, column C4 also reduced removal efficiencies up to about 80% but recovered the higher Mn capacity uptake in only 2 days. This finding suggests that the biologically active zone in the filter columns required a short time of acclimatization to the new flow regime conditions (see Section 3.3). Full ripening, and thus effective Mn removal at 15 cm depth, was finally registered on day 40. Besides, the speed-up stage following Zeng et al. (2019), which was carried out after day 40 (Figure 5), did not affect the Mn²⁺ concentration profiles patterns, indicating that the virgin pumice media was also fully inoculated with the autochthonous groundwater bacteria.

As shown in Figure 4, both experiments were conducted using feed water with pH and Eh values within the field of action of MOB (Mouchet 1992). In addition, considering that non-catalyzed, homogeneous Mn²⁺ oxidation (by oxygen only) is very slow at pH < 9 (Granger et al. 2014) (feed water pH ∼7.9, Table 1), and biological Mn²⁺ oxidation is supposed to occur at the top of the non-inoculated sand filter during the start-up (Breda et al. 2019b), it can be concluded that Mn removal by the virgin pumice media was probably dominated by biological processes in both cases. This finding is in line with the previous studies that have also reported that Mn removal in a non-coated virgin medium was initiated biologically (Bruins et al. 2015a; Breda et al. 2019b) Furthermore, in the active zone on top of the columns, microbiological analysis showed that during flow-through experiments, MOB count was 1.9 × 10⁶ CFU·g⁻¹, whereas in the recirculating regime, the MOB concentration was 4.0 × 10³ CFU·g⁻¹ and was reduced to 2.9 × 10³ CFU·g⁻¹ after switching to flow-through regime. As explain later, the intermittent Mn-loading impacted the MOB population in the recirculated biofilters and probably slowed down the acclimation time in this system. More information about MOB diversity detected in the same well water and in a bench-scale experiment can be found in the previous studies (Calderón-Tovar et al. 2020; Araya-Obando et al. 2021).

During the start-up period in the flow-through regime (Figure 2), the Mn²⁺ concentration was always around 0.50 mg·L⁻¹. In contrast, in the recirculating experiment, the average initial Mn²⁺ concentration in T3 was around 0.52 mg·L⁻¹ but was typically reduced to less than 0.1 mg·L⁻¹ in 7–8 h (see Supplementary Material S2). The water replacement was conducted two times a week; therefore, during 48–64 h of operation, the recirculated water did not contain Mn²⁺. Therefore, the flow-through columns, through the experiment, registered a Mn-loading of about 0.16 kg Mn·m⁻² in total (until day 8 of operation, > 90% removal). During the recirculating flow experiment, the columns had a total Mn-loading of 0.11 kg·Mn·m⁻² to reach the 23 days of ripening time, comparable to the loading of the flow-through columns. The difference between both experiments was mainly associated with the intermittent provision of the Mn-loading in the recirculating regime in contrast with the continuous Mn-loading in the flow-through regime. It suggests that the total Mn-loading, in this case, 0.11 kg Mn·m⁻², was a limiting factor during the start-up of the non-bioaugmented Mn biofilters.

Comparing the Mn²⁺ concentration profiles (Figures 3 and 6), it can be observed that flow-through columns reached Mn²⁺ < 0.1 mg·L⁻¹ (compliance with the local drinking water criterion) at 15 cm depth in only 11 days, while recirculating filter columns required 32 days to reach Mn²⁺ < 0.1 mg·L⁻¹ at 30 cm depth. The latter case was probably affected by the intermittent provision of the Mn-loading that impacted the MOB population at the top of the filter columns. The ripening of filter media gradually rises from the bottom to the top of the filter depth likely due to the accumulation of inorganic precipitates and biomass on top of filter media (Dangeti et al. 2017). Usually, filter ripening continues even when high Mn removal efficiencies have already been observed (Dangeti et al. 2017; Breda et al. 2019a).

Under the present conditions, inherent inoculation using a recirculating flow regime required more time for ripening than a flow-through regime. However, an advantage of the recirculating system is that it consumed less feed water. During the first 8 days of filter operation, the flow-through column registered a water consumption of 3.19 m³, whereas the recirculating column consumed only 1.59 m³ in the 23 days of filter operation (see estimations in Supplementary Material S3). It represents a water consumption reduction of approximately 50%, thus making it a suitable strategy for groundwater applications, where spillage of water should be minimized. However, there are still chances to improve the performance of the recirculating system. For example, Worch et al. (2002) applied the recirculating approach for the biodegradation of synthetic organic chemicals (SOCs) and found that the adaptation of microorganisms depended on the contact time with the SOCs. Thus, the authors recommended stimulating the microorganisms by maintaining the SOC concentration in the recirculated water. In that sense, a more frequent refreshment of water in T3 (e.g. 1.5 or 2 days) and/or even supplementing dissolved Mn²⁺ into the recirculated water could help to reduce the ripening time in the recirculating system.

This study aimed to investigate the influence of the flow regime (flow-through and recirculating) on the ripening time of virgin pumice media in non-bioaugmented biofilters. For this purpose, pilot-scale experiments using the same tropical groundwater (∼23 °C) were performed to compare the start-up of flow-through and recirculating filter columns. The flow-through filter columns exhibited a Mn removal above 90% after 8 days of filter operation and reached Mn <0.1 mg·L⁻¹ (compliance with the local drinking water criterion), whereas the recirculating filter columns registered a start-up of 23 days. These exceptional rapid start-ups were probably obtained due to the favorable feed water characteristics (temperatures ∼22–24 °C, pH > 7.5, DO > 6 mg·L⁻¹, redox potential (Eh) ∼300–400 mV, and low Fe), and the initial filtration velocity of 2 m·h.

In addition, results suggest that inherent inoculation using the recirculating mode was less effective than using a flow-through regime in terms of the ripening time. However, an advantage of this operational strategy was that it consumed less feed water during the start-up (approximately 50%) than using a flow-through regime, making it still a suitable strategy for groundwater applications, where spillage of water should be minimized.

Finally, the start-up of the non-bioaugmented biofilters in both flow regimes was reached under comparable conditions, using the same feed water and adopting an initial low filtration velocity of 2 m·h. Both the flow-through and the recirculating flow regime required a similar total Mn-loading (0.16 and 0.11 kg·Mn·m⁻², respectively). In that sense, differences between the start-up periods through both regime flows could be attributed to the intermittent Mn-loading of the recirculating system. It seems that the intermittent Mn-loading impacted the MOB populations formed in the pumice stone medium during the start-up of biofilters operated in the recirculating flow regime. Therefore, it was concluded that the total Mn-loading, in this case, 0.11 kg·Mn·m⁻², was a limiting factor during the start-up of the non-bioaugmented Mn biofilters.

In summary, an optimal start-up protocol for recirculating flow regime should be: appropriate feed water characteristics, initial filtration velocity of 2 m·h (EBCT of 21 min); once the Mn removal reach 90%, velocity increase in steps of about 1 m·h until reaching a final filtration velocity of 5 m·h (∼ EBCT of 8.4 min). Fe-loading should be prevented using a prefilter (Fe concentrations in the influent <0.10 mg·L⁻¹).

The authors thank the Administrative Board of Condominio La Hacienda to give access to the well water used during this project and Agregados de Pómez from Guanacaste, Costa Rica, for providing the pumice material.

This research was supported by Vice rectory of Research and Extension (VIE) of the Instituto Tecnológico de Costa Rica (project number 1460–066). The authors would also like to acknowledge the Scholarship Committee of the Instituto Tecnológico de Costa Rica for the doctoral scholarship awarded to José A. Araya-Obando.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.