Start-up strategies for nitrification and manganese oxidation on a single stage RSF for drinking water production Open Access

Rapid sand filters (RSFs) are a common treatment process in drinking water production, for selective separation of unstable elements in groundwaters. They can combine particle removal by filtration with biological and physical-chemical processes to eliminate mostly nitrogen compounds, iron and manganese (Tekerlekopoulou et al. 2013). Presence of dissolved oxygen allows biological nitrification, which has to be complete, with the conversion of all ammonium and nitrite into nitrate, due to the toxicity of nitrite at low levels. Manganese removal can proceed through two pathways: (1) biological manganese oxidation in presence of oxygen, and (2) physical-chemical oxidation of soluble manganese (Mn²⁺) into particulate manganese oxides by adsorption on reactive sand composed or coated by manganese oxides. An important operational issue with these biofilters is the long period required to achieve a stable and efficient elimination of the targeted compounds, leading to a considerable loss of water. In addition, physical-chemical oxidation of manganese requires frequent chemical regeneration of the reactive sand oxidation potential, which represents industrial costs and constraints.

Nitrification is a well-known process in water treatment, commonly described in two-step biological processes involving bacteria and archaea (Niu et al. 2013). In presence of oxygen, ammonia-oxidizing bacteria (AOB) (the majority being part of the Nitrosomonadaceae family under the Betaproteobacteria class) (Klotz & Stein 2011) and ammonia-oxidizing archaea (AOA) (among them, the phyla Crenarchaeota and Thaumarchaeota) (Pester et al. 2011; Urakawa et al. 2011; Kitzinger et al. 2019), oxidize ammonium to nitrite. Then, nitrite oxidizing bacteria (NOB) (including Nitrobacter and Nitrospira genera) (Ehrlich et al. 1995; Tränckner et al. 2008) oxidize nitrite to nitrate. For several years, microorganims named commamox (of the Nitrospira genera) have also been identified to operate complete nitrification (Daims et al. 2015; Kessel et al. 2015).

Manganese adsorption and oxidation on reactive sands allows rapid and efficient manganese removal. However, regular interruptions of the process and addition of chemical agents (strong oxidant agents such as Cl₂ or KMnO₄) are needed to regenerate the catalytic oxidative potential of the reactive sand. Biological manganese oxidation is a slow process which leads to particulate manganese oxides which in turn promote the chemical oxidation pathway similarly to reactive sands (Bruins et al. 2015a; Breda et al. 2019). Evidence of such processes has been reported for Pseudomonas manganoxidans (Gage et al. 2001) and Leptothrix dischopora SS1 (Zhang et al. 2002) or for Bacillus specie in the case of sewage activated sludge (Abu Hasan et al. 2012).

Biological manganese oxidation is interesting because it limits the addition of chemical agents. Moreover, co-elimination of ammonium and manganese is feasible (Bruins et al. 2014; Cai et al. 2015; Zeng et al. 2020). However, a negative influence of nitrification on biological manganese oxidation has been reported in the literature. The rate of biological manganese oxidation could be reduced in the presence of nitrification, due to oxygen competition, nitrite accumulation (Vandenabeele et al. 1995; Cheng et al. 2016) and/or pH and oxidation reduction potential (ORP) conditions (Abu Hasan et al. 2012; Han et al. 2013). In all cases, biological activities and biofilm development for nitrification and manganese oxidation are dependent on environmental and/or operating conditions (such as temperature, inlet concentrations, pH and oxygen) (Bruins et al. 2014). In real conditions, low temperature and low concentrations of ammonium and manganese lead to slow start-up of industrial biofilters (Ciancio Casalini et al. 2020). Durations between 25 and 40 days for complete nitrification were reported by (Vandenabeele et al. 1995; Cheng 2015; Albers et al. 2018) and between 15 and 75 days for complete manganese oxidation (Bruins et al. 2015a; Cheng et al. 2018; Breda et al. 2019). Hence, high volumes of water are wasted during this settling phase, as it is unsuitable for production. To reduce start-up time, acceleration of the biological activities in terms of microbial filter colonization and growth must be achieved. Strategies to reach this goal must therefore be investigated.

Thus, for initial biofilm settlement enhancement, two different strategies can be considered: (1) start with inoculated sand from another location to enhance the initial biomass quantity (X) (Cai et al. 2015; Albers et al. 2018; Ciancio Casalini et al. 2020); (2) modify the initial environmental conditions (T, [S]) to improve biological rates (Queinnec et al. 2006). For natural mineral waters, biomass addition is impossible due to regulations that forbid any modification of natural microbial ecosystem (Directive 2009/54/EC of the European Parliament). Hence, only some modifications of environmental conditions are acceptable. Therefore, to improve initial biological rates, enhancement of temperature and/or substrate concentration can be preferentially considered. These modifications can only be done off production during the start-up period. Temperature is a well-known parameter, efficiently influencing biological rates of nitrification (Antoniou et al. 1990; Kaelin et al. 2009) and of manganese oxidation (Tipping 1984; Zhang et al. 2002). However, the temperature dependency of these biological rates is not known and depends on the microbial populations presents in the system (Tipping 1984). Similarly, specific growth rate dependency on the substrate concentration in connection with the value of substrate affinity has not been studied in the case of groundwaters. However, it is dependent on the type of microbial population.

Our study focused on the evaluation of start-up strategies to accelerate nitrification and manganese removal without added exogenous biomass. Start-up strategies based on an increase in water temperature or in substrate concentration were examined in an experimental approach. A biofilter fed with groundwater from industrial boreholes was operated with the possibility to modulate inlet temperature and substrate concentrations (ammonium, nitrite, or manganese) leading to various start-up experiments. Pulses of nitrite (injections of nitrite during a set period of time) were applied in some experiments to investigate nitrite influence on manganese oxidation. The start-up experiments were compared to reference ones where the same biofilter was operated in conventional natural conditions (no modification of the inlet water) and using the same borehole. Results were confirmed for boreholes situated in different locations.

Figure 1 presents the experimental pilot composed of two identical stainless-steel columns (height 1.5 m, inlet diameter 0.15 m) filled with sand. The height of the sand bed was 1.25 m with a porosity of 60.2%. The inlet flow rate was fixed to obtain a water velocity identical to the industrial sand filters used at the borehole to adsorb manganese. Sampling ports were placed along the height of the filters at 0.15 m, 0.53 m, 0.73 m, 0.93 m and 1.130 m from the top. To ensure aerobic biological activities, oxygen was provided continuously by a venturi system, sucking air through a sterile filter to avoid any microbial contamination of water. The column was cleaned before every experiment, using chlorine, or a mix of peracetic acid and hydrogen peroxide, followed by a thorough rinsing.

First, reference experiments and start-up experiments were operated on an industrial borehole, labelled borehole A. Then, the start-up strategy was repeated with an identical pilot on an industrial borehole located in another location, labelled borehole B. Table 1 presents the main properties of the groundwater filtered on these two sites. As a side note, nitrate concentration was around 500 μgN-NO₃⁻/L; this parameter was measured to validate the complete oxidation of the nitrogen and validate nitrogen mass balances (data not shown).

Table 2 presents the key conditions during the implantation of the biological activities for each experiment. The experiments are labelled with a code name: ‘Ref’ indicates an experiment in natural conditions, ‘SU’ indicates a start-up experiment, and A or B indicate the borehole location). Due to the low concentration of phosphate in the groundwaters, the start-up experiments were conducted with a continuous addition of phosphate (5 μgP/L) (Kors et al. 1998). The experimental plan (Table 2) shows experiments with temperature increase, with or without substrate dosing, with or without recirculation of the filtered water, for two industrial sites. In all cases the start-up was considered completed when the following parameters were achieved: ammonium below 7 μgN-NH₄⁺/L and nitrite below 2 μgN-NO₂⁻/L for nitrification experiences, and manganese below 10 μgMn²⁺/L for Mn oxidation. Substrates used were under the chemical forms of (NH₄)₂SO₄, NaNO₂, and MnSO₄(H₂O).

During experiments SUA2 and SUA3, nitrite pulses were applied to obtain nitrite concentration in all the depth of the filter, to evaluate the impact of different nitrite concentrations on manganese oxidation. Nitrite and manganese profiles were realized before, during and after the nitrite pulses. The nitrite concentration applied for the pulses ranged from 110 μgN-NO₂⁻/L to 1,280 μgN-NO₂⁻/L, to be compared with Cheng et al. (2016) who observed nitrite influence on manganese oxidation for nitrite concentration up to 930 μgN-NO₂⁻/L.

During the experiments, water volumes were sampled at different times after the beginning of the experiment at the inlet and at the outlet of the filter columns and at different column heights. Before and during sampling, the stainless steel tubes of the sampling valves were regularly sterilized with a blowtorch to avoid any contamination from outside. Samples were stored at 5 °C and analyzed in a maximum of 3 h from the sampling time.

Ammonium, nitrite and manganese concentrations were analyzed by spectrophotometry using a Spectroquant Ammonium Test kit 114752 (0.007–3.00 mgN-NH₄⁺/L), Nitrite Test kit 114776 (0.002–1.00 mgN-NO₂⁻/L), and Manganese Test kit 114770 (0.010–10.00 mgMn²⁺/L), respectively, on the Spectroquant Pharo 300 spectrophotometer from Merck.

As nitrate is the final product of the nitrification, measurements were realized to validate the completeness of nitrification by mass balance. For all experiments, nitrites were fully converted into nitrate, thus data for nitrate concentration are not presented.

Oxygen was controlled by measurement at the inlet and the outlet of the filter by electrochemical sensor (Memosens COS81D). The system was aerobic during all the experiments, except for some short periods due to technical incidents.

pH, temperature, and conductivity were measured at the inlet and the outlet of the columns for control by sensors (Tophit CPS471D, Easytemp TMR35 and Indumax CLS54D, respectively), with no significant variations, hence data are not provided.

Virgin sand (composed of a minimum of 96% silica) from CAS filtration was used in the pilot filters (granulometry: 1.18–0.60 mm; mineral hardness: 6–7 Moh; bulk density: 1,560–1,600 kg·m⁻³; specific gravity: 2.65).

Before filtration experiments, virgin quartz sand was exposed in a batch reactor to ammonium and then manganese concentrations to investigate the potential interactions between the sand and these compounds. The ammonium and the manganese concentrations did not change during these tests. No physical-chemical interaction was observed between the sand and the targeted compounds.

Two reference experiments, RefA1 in 2019 and RefA2 in 2020, were operated in reference conditions at borehole A, for 530 days and 80 days, respectively. These experiments were purposely run at different times to assess the reproducibility of reference start-up data; in addition, these experiments were run at the same time as the optimized start-up experiments to yield a comparison with the same water parameters. Figure 2 presents ammonium and nitrite concentrations measured at the outlet of the filter for RefA1 (Figure 2(a)) and RefA2 (Figure 2(b)). During RefA1, complete ammonia removal took 40 days. The outlet nitrite concentration first rose until a maximum at day 42, then decreased to values under the detection limit at day 92. During this experiment, disrupting events occurred (particularly the water feeding pump stopped at day 60), resulting in a temporary loss of ammonium oxidation activity (around 80% of the nitritation activity), and a decrease in the accumulation of nitrite. However, complete nitritation was restored in 5 days. For the RefA2 experiment (Figure 2(b)), nitritation was complete in 32 days, and nitratation in 63 days. Thus, during the natural start-up period, the nitrification process proceeded in two steps as is commonly described (Niu et al. 2013) and a long duration for start-up was confirmed. The start-up duration significantly varied (complete nitritation needed 30–40 days and complete nitrification, 63–92 days). The range in variability could be partly explained, especially for nitritation, by an incident causing a technical stop for a few days at day 60 for experiment RefA1. These durations are high compared to the ones reported in the literature, probably due to the low temperature (∼12 °C) and the very low concentration of ammonium (∼80 μgN-NH₄⁺/L) in borehole A. The duration variation could be also explained by the sensitivity of microbial activities to changes in environmental conditions, including the concentration of microelements, and the initial concentration of the active microbial population (which is the only origin responsible for the seeding of the filter).

Figure 3 presents the outlet manganese concentrations for RefA1 (Figure 3(a)) and RefA2 (Figure 3(b)). For RefA1, manganese concentration showed no significant evolution for 32 days, but then started to decrease below detection limit at day 41. For RefA2 (Figure 3(b)), high variations in inlet manganese concentrations during the first 15 days were observed. 80% of manganese concentration was removed at day 25, which was 16 days faster than in RefA1. Difference in the inlet soluble manganese concentration is the main difference between the two experiments (∼120 μMn/L for RefA and ∼ 55 μgMn/L for RefA2). As for nitrification, the duration required to complete manganese oxidation is variable, with a higher activity for RefA2. This could be explained by the high variations in inlet concentration. Finally, it can be concluded that the durations required to complete manganese oxidation (between 25 and 40 days) are in the range of values found in the literature (15–75 days (Bruins et al. 2015b; Cheng et al. 2018; Breda et al. 2019)).

Experiments RefA1 and RefA2 demonstrate the feasibility of ammonium, nitrite and manganese co-elimination with unreactive sand, confirming previous observations by various authors (Bruins et al. 2014; Cai et al. 2015; Zeng et al. 2020). To go deeper into our understanding of the removal processes, concentration profiles were realized along the depth of the filter for RefA2 at different times (day 4 and day 13, no measurable biological activity; day 27, start of the nitritation and the manganese oxidation; day 40 complete nitritation and complete manganese oxidation; and day 65, complete nitritation, end of nitratation and complete manganese oxidation (Figure 4)). From these profiles, it can be observed that biological activities are initially distributed along all the filter depth (see day 27 for ammonium and manganese oxidation, with a regular consumption of the substrate across the full depth of the filter). However, over time, the biological activities concentrate at the top of the filter, confirming the observations of Breda et al. (2019). Additionally, the concentration profiles for nitrite and nitrate confirmed a nitrification in two steps during start up.

Considering a maximal growth yield for nitrification of 0.18 gVSS.gN⁻¹ (Blackburne et al. 2007), and a mean manganese content in bacteria of 5e⁻³ gMn/gVSS (Novoselov et al. 2013), the manganese consumed by the complete nitrification correspond to 0.4 μgMn²⁺/L, thus between 0.3 and 0.7% of the manganese concentration in the raw water. Moreover, the virgin quartz sand showed no interaction with manganese. Therefore, specific biological activity should be responsible of the initiation of manganese oxidation. In the present study, manganese oxidation occurred simultaneously with nitritation, which is not in agreement with results of various authors (Vandenabeele et al. 1995; Han et al. 2013; Cheng et al. 2016). It can be seen in Figure 4, at day 40, that nitrite accumulation occurred across the full depth of the filter without any effect on manganese oxidation. Thus, no inhibiting effect of nitrite on manganese oxidation could be observed in these experimental conditions. However, it should be underlined that ammonia concentrations in raw water (max 90 μgN-NH₄⁺/L) were low compared to the concentrations reported in the literature (for example Cheng et al. 2016 described concentrations up to 930 μgN-NH₄⁺/L). These observations tend to be in line with a recent study showing no influence of ammonium pulses on manganese oxidation under 0.9 mg/L of ammonia (Tang et al. 2021).

Start-Up strategies were applied to borehole A with temperature increase and substrate dosing as presented in Table 2. Four experiments were carried out (SUA1, SUA2, SUA3 and SUA4). Figure 5 presents the results for ammonium, nitrite, and manganese concentrations at the outlet of the filter during and after the start-up, when the filter was operated back into reference conditions (raw water temperature and concentrations, no substrate added). For experiments SUA1, SUA2 and SUA3, temperature increases were applied with only nitrogen dosing (ammonium and nitrite). Nitritation and nitratation periods were reduced to 5–17 days (compared to 30–40 days in reference conditions), and to 10–20 days (compared to 63–92 days in reference conditions), respectively, thus yielding a gain of 43–72 days to obtain the same removal efficiency for nitrification. For manganese oxidation, experiments SUA1, SUA2 and SUA3 (Figure 5), with only temperature increase, showed a complete removal in 20–40 days (compared to 25–40 days in reference conditions). Hence, no clear impact of a temperature increase on the manganese oxidation was highlighted. Experiment SUA4 was operated with temperature increase and with manganese dosing only. During manganese dosing (500 μgMn²⁺/L during the 50 first days), no manganese conversion was measured. After 50 days, only temperature increase was applied and manganese dosing was stopped, then manganese oxidation started, highlighting a potential inhibiting effect of the inlet manganese concentration on the biological oxidation (self-substrate inhibition existing for manganese oxidizers but at higher manganese levels was reported by Therdkiattikul et al. 2020). For nitrification, with no nitrogen dosing, experiment SUA4 showed (Figure 5) an acceleration of nitritation (17 days) and nitratation (35 days), due to the temperature increase. Nonetheless, acceleration factors were lower than with both temperature increase and substrate dosing.

After the start-up period, the filter was brought back into reference operating conditions (for all the experiments), to evaluate the potential of the accelerated start-ups to develop efficient biological activities at lower temperatures. For all the start-up experiments on borehole A, nitrogen and manganese removals were at the same level in natural conditions than at the end of the start-up period. Start-up strategy with temperature increase and substrate dosing showed a clear positive effect on nitrification. No gain was obtained for manganese oxidation in the tested experimental conditions. However, in reference conditions, manganese oxidation was achieved before complete nitrification, thus acceleration of the nitrification may allow a reduction in the time required to install the biological activities for the co-removal of nitrogen and manganese. Moreover, in these conditions, temperature increase, without substrate dosing, enables accelerating the start-up period for nitrification (SUA4). Compared to experiment RefA2 (63 days to complete nitrification), 40 days to install all the biological activities (highest duration recorded for manganese oxidation without dosing), enables 36% of water saving, with only a temperature increase (30 °C).

For all experiments except SUA4, the nitrite peak was around 500 μgN-NO₂⁻/L and manganese oxidation always occurred after complete nitrification, thus indicating a possible interaction of manganese oxidation with nitrification as described in the literature (Vandenabeele et al. 1995; Cheng et al. 2016). However, Han et al. (2013) showed that ammonium pulses logically increase the nitrification activity, with no negative influence on the simultaneous removal of ammonium and manganese. In our experiments, manganese oxidation does not appear to be influenced by the chosen parameters for the start-up strategy, unlike nitrification. Therefore, the results could only be linked to a different behavior of the involved specific microorganisms towards environmental conditions. Further investigation is needed to improve knowledge on the interactions between these biological activities.

During the initial start-up phase, we confirmed that manganese oxidation is biologically supported, and observed that manganese oxidation always occurred after nitritation. To further investigate the influence of nitrification on manganese oxidation, nitrite dosing experiments were conducted. Figure 6 presents manganese concentrations in influent, effluent, and across the depth of the filter during experiment SUA2. In this experiment, successive pulses of nitrite at the pilot inlet were applied during the installation of biological manganese oxidation, with the aim of detecting potential negative interactions between the biological processes, at day 29, day 30 and day 35. Data clearly refute any influence of nitrite on biological manganese removal during the start-up phase of manganese oxidation, in our experimental conditions.

Manganese oxidation in rapid sand filters is a complex phenomenon including biological oxidation and chemical heterogenous catalysis due to the production during biological reaction of solid manganese oxides with catalytic properties for Mn²⁺ oxidation (Tekerlekopoulou et al. 2013; Bruins et al. 2014; Breda et al. 2019). To investigate the influence of nitrite on the global manganese oxidation process, two pulses of nitrite were realized in the influent water of the biofilter during experiment SUA3 at day 119 and day 122 (no modification of the raw water conditions). Results in Figure 7 show that concentration profiles of manganese before and after the injections did not significantly differ at both concentrations. In our experimental conditions, nitrite seems to have no significant impact on global manganese oxidation, in contrast with some results that are often cited in the literature (Vandenabeele et al. 1995; Cheng et al. 2016). Interestingly, more recent data have similarly shown no influence of ammonium pulses on manganese oxidation under 0.9 mg/L of ammonia (Tang et al. 2021), which brings into question the role of nitrogen compounds on the sequence of nitrification followed by manganese oxidation during the start-up phase, in these experimental conditions.

Experiments on borehole A showed no clear gain for the chosen start-up strategy on manganese oxidation but showed interesting trends in nitrification activity. Thus, to improve the start-up strategy and challenge it, the pilot was used on another borehole, in another industrial location (borehole B), using MnO₂ reactive sand to focus on nitrification. The experimental plan was then completed with other temperatures (with or without substrates dosing) and with a new parameter for the start-up strategy: partial recirculation of the filtered water (Table 2). Experiments SUB1 and SUB2 were set without recirculation of the filtered water, and experiments SUB3 and SUB4 were set with 75% recirculation of filtered water. Figure 8 presents ammonium and nitrite time series obtained for the four experiments. Firstly, for all experiments, nitrification was achieved, with transitional nitrite accumulation (indicating, like for borehole A, a nitrification in two steps during the start-up periods). Regarding experiments SUB1 and SUB2 operated in the same conditions, nitritation was achieved in 15 and 20 days respectively, and nitratation in 26 and 27 days respectively, indicating a good repeatability of the start-up strategy. Compared with experiments SUA1, SUA2 and SUA3, differences were only the composition of the source water, and the temperature used for the accelerated start-up strategy (∼30 °C for SUA experiments and ∼20 °C for SUB experiments). For the SUA experiments, nitritation was achieved in 11.3 days compared to 17.5 days for SUB experiments (mean values), and nitratation was achieved for SUA in 15.7 days compared to 26.5 days for SUB (mean values), indicating a higher nitrification rate at 30 °C than at 20 °C, as expected for common nitrifiers (Antoniou et al. 1990; Kaelin et al. 2009). This suggests the possibility of different start-up strategies adapted for different sites. However, as water is lost, thermal energy is consumed, and chemicals are used during this start-up. Two experiments were operated to try to reduce the costs of the start-up strategy, while maintaining the reduced durations. Experiment SUB3 was carried out in the same conditions as SUB1 and SUB2 but with 75% recirculation of outlet water, and experiment SUB4, with the same recirculation rate, but with no temperature increase (with the aim of limiting energy consumption). For experiment SUB3, nitritation was achieved in 15 days and nitratation in 25 days, corresponding to the values recorded without recirculation. Henceforth, 75% of the water could be saved in this configuration with no impact on the time gained. Regarding experiment SUB4, nitritation was achieved in 48 days and nitratation in 57 days, indicating that recirculation without temperature was not an interesting strategy for start-up acceleration despite the nutrient supplementation. Thus, temperature remains the key parameter to accelerate nitrification.

The results obtained in this study have shown that for specific groundwaters where no exogenous biomass inoculation is possible, different start-up strategies are feasible. The time to obtain complete biological activities was 63 days in experiment RefA2, leading to 151.2 m³ water consumed before production. Different strategies were investigated: (1) with no substrates dosing and no temperature increase, the recirculation of 75% of the water enabled completing the start-up phase in 63 days but with only 37.8 m³ of water consumed; (2) with temperature and substrates dosing and without water recirculation, complete start-up phase was reached in 40 days, i.e. 96 m³ of water consumed; (3) combining recirculation of 75% of the filtered water, temperature increase and substrates dosing, the production period was obtained in 40 days, but with only 24 m³ of water consumed, thus 84% of water saved.

This study investigated start-up strategies to produce drinking water from groundwaters with temperature increase, substrates dosing, and without any addition of exogenous biomass. Focus on nitrification and manganese oxidation highlighted a different impact of the chosen parameters on both activities, with an acceleration of nitrification and no influence on manganese oxidation. In all start-up experiments, manganese oxidation always occurred after nitritation. However, nitrite pulses have shown no influence of nitrite on manganese oxidation in short and long periods after start-up, in the experimental conditions of this study.

The start-up strategy for nitrification was challenged in another borehole, in another location. Table 3 summarizes the results obtain for the applied experimental conditions. Combining recirculation, ammonium, nitrite and phosphorus dosing, and increased temperature could save up to 90% of water for an installation of efficient biological activities for nitrogen and manganese removal. In addition to these savings, start-up time reduction is also paramount from the business perspective, as each day of the start-up phase is a day without production. This study proposed a strategy to limit these constraints in an industrial situation where exogenous biomass could not be added to start a new biofilter, such as with natural mineral waters.

The authors want to thank all the people who participated in this study from DANONE industrial sites, as well as from INSA-Toulouse. Thomas Etcheberry would like to thank ANRT (French Association for National Research and Technology) for their financial support of his PhD (CIFRE convention n° 2017/0429).

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