Experimental results of the RO pilot test
Figure 2 shows the flux behavior comparison during the pilot-scale RO process under intermittent and continuous operating modes by feeding anoxic and oxic influent water. Flux decreased continuously in the intermittent operation mode, even if there was fluctuation during the operation time. In the anoxic-continuous mode, flux was decreased about 20% from 30 h after fluctuated flux pattern, and a sharp decrease in flux in the oxic-continuous operation mode was observed around 25 h after a constant flux pattern. In the anoxic condition (Fig. 2 (a)), normalized flux variated from 1.1 to 0.6 (average flux: 16.1 LMH, average salt rejection: 90%) in the intermittent mode and from 1.1 to 0.7 in the continuous mode (average flux: 14.6 LMH, average salt rejection: 92%), respectively. In the oxic condition (Fig. 2 (b)), normalized flux fluctuated from 1.0 to 0.4 (average flux: 12.8 LMH, average salt rejection: 89%) in the intermittent mode and from 1.0 to 0.5 (average flux: 13.0 LMH, average salt rejection: 89%) in the continuous mode. Although the initial flux was similar under oxic and anoxic conditions, a decrease in flux about 25% within the first hour was observed under oxic conditions, indicating that a higher flux was maintained under anoxic conditions. The most stable flux tendency was observed under intermittent operation under anoxic conditions. The flux decline behavior was strongly influenced by the fouling layer on the membrane surface.
Comprehensive investigation of biofilm community structures
The major OTUs (i.e., the top ten abundant OTUs in each sample) were investigated to understand the biofilm communities in the pilot-scale BWRO process. As a result, 25 bacterial OTUs were selected as the major (Fig. 7, Table 3, and Table S2). The results of microbial identification are summarized in Table S2.
Table 3
Taxonomic information of the major bacterial OTUs obtained from the influent and biofilm samples.
No. | Closet taxon | Accession No. | Similarity (%) |
OTU 1 | Pseudomonas qingdaonensis | NR_169411.1 | 99.4 |
OTU 2 | Aeromonas finlandiensis | NR_136830.1 | 99.4 |
OTU 3 | Acinetobacter gandensis | NR_133953.1 | 99.1 |
OTU 4 | Acinetobacter bohemicus | MK606064.1 | 99.6 |
OTU 5 | Shewanella profunda | NR_104770.1 | 99.8 |
OTU 6 | Paraclostridium benzoelyticum | NR_148815.1 | 99.5 |
OTU 7 | Parabacteroides chartae | NR_109439.1 | 99.1 |
OTU 8 | Bacillus cereus | NR_074540.1 | 100.0 |
OTU 9 | Comamonas testosteroni | NR_029161.2 | 99.8 |
OTU 10 | Uncultured bacterium clone WB11 | KP241972.1 | 99.5 |
OTU 11 | Citrobacter freundii | NR_113596.1 | 99.5 |
OTU 12 | Clostridium collagenovorans | NR_029246.1 | 99.1 |
OTU 13 | Bacteroides luti | NR_125463.1 | 98.9 |
OTU 14 | Azospira oryzae | NR_074103.1 | 100.0 |
OTU 15 | Sedimentibacter saalensis | NR_025498.1 | 97.0 |
OTU 17 | Asaccharospora irregularis | NR_119034.1 | 99.5 |
OTU 18 | Acinetobacter bouvetii | NR_117628.1 | 98.9 |
OTU 19 | Uncultured bacterium clone SuLe1_E13 | KY696946.1 | 100.0 |
OTU 20 | Exiguobacterium undae | NR_043477.1 | 99.5 |
OTU 21 | Uncultured clone 018MICCbiofilm | JF341918.1 | 99.8 |
OTU 22 | Desulfovibrio vulgaris | NR_041855.1 | 99.3 |
OTU 23 | Acinetobacter johnsonii | NR_164627.1 | 99.1 |
OTU 24 | Sedimentibacter acidaminivorans | NR_148817.1 | 98.4 |
OTU 25 | Uncultured bacterium clone: C168A-31 | AB717123.1 | 99.8 |
OTU 29 | Pseudomonas vancouverensis | NR_041953.1 | 99.8 |
The predominant microbial communities differed with the input water source (i.e., oxic vs. anoxic), although several major OTUs were observed (Fig. 7, Table 3, and Table S2). For the aerobic influent (i.e., Ox-CON and Ox-INT), OTUs 2, 3, and 4 were abundant with > 10% relative abundance. OTU 2 was identified as Aeromonas finlandiensis, a facultative anaerobe that can utilize various organic matter as a carbon source36. OTUs 3 and 4 had the highest similarity with strict aerobic microbes, Acinetobacter gandensis and Acinetobacter bohemicus37,38. Shewanella profunda (OTU 5) and Bacillus cereus (OTU 8) were also predominant and are known as facultative anaerobes that can grow under both aerobic and anaerobic conditions39,40. OTUs 1, 2, 6, 7, and 9 were predominant in the anoxic influent (Ano-CON and Ano-INT). OTU 1 was identified as an aerobe, Pseudomonas qingdaonensis41; which can grow under low DO conditions. In contrast to the oxic influent, obligate anaerobes, such as Paraclostridium benzoelyticum (OTU 6) and Parabacteroides chartae (OTU 7), were predominant42,43 in the anoxic influent. Comamonas testosterone (OTU 9), which can be microaerophilic44, was also abundant in the anoxic influent. In other words, different bacterial communities were formed by different input water sources during the pilot-scale BWRO process.
In addition, the biofilm community structures differed with the type of input water. For the biofilms originating from the BWRO processes treating oxic influent (i.e., Ox-CON-Bio and Ox-INT-Bio), OTUs 2, 3, and 4 were determined to be predominant communities because of the abundance of these microbes in the influents accumulated on the RO membrane surfaces. However, OTU 7 became predominant in the biofilm samples and P. chartae was also found in the anode biofilm of microbial fuel cells (MFC)45. Similarly, several microbes, such as OTUs 9, 15, 17, 21, 22, and 24, increased compared to the input water sources (Fig. 7, Table 3, and Table S2) and some of the major OTUs were discovered in various environmental biofilms. For example, OTU 9 was identified as C. testosterone, which has a strong propensity for biofilm formation46. OTU 15, Sedimentibacter saalensis, was found in the anode-attached biofilm of the sediment MFC system47. OTU 21 has high similarity to uncultured bacterium clone 018MICCbiofilm (99.8%) discovered in concrete sewer biofilms48. OTU 22 was closely related to Desulfovibrio vulgaris, a sulfate-reducing bacterium which are often found in biofilms49. Some anaerobic microbes, such as P. chartae and S. saalensis, increased in the biofilms of the BRWO process treating oxic influent, probably because the biofilm provided various oxygen conditions that facilitated both anaerobic and anaerobic environments for microbes50,51.
For the biofilms formed in the BWRO processes treating anoxic influent (i.e., Ano-CON-Bio and Ano-INT-Bio), OTUs 1, 2, 3, and 7 were predominant. A. gandensis (OTU 3) and P. chartae (OTU 7) increased in the biofilm samples compared to the influent, indicating that these microbes played an important role in biofouling in the BWRO process. It has been reported that Acinetobacter spp. exhibit a variety of physiological characteristics, including biofilm formation52. P. chartae can form biofilms during environmental processes45. In addition, other microbes, such as OTUs 11, 14, and 19, also increased in the anoxic biofilms during the BWRO process (Fig. 7, Table 3, and Table S2). OTU 11 was identified as Citrobacter freundii, a facultative anaerobic bacterium that was also isolated from the environmental biofilm of an MFC system inoculated with aerobic sludge53. OTU 14 has the highest 16S rRNA gene similarity to Azospira oryzae, a bacteria that can grow under microaerophilic conditions by utilizing various organic compounds54 and that has been found in biofilms formed on hollow-fiber membrane surfaces installed in membrane biofilm reactors55. OTU 19 was closely related to the uncultured bacterium clone SuLe1_E13, which has been discovered in biofilms formed on anodic carbon cloths56. As described in previous chapter, dead cells were observed under anoxic conditions (Fig. 4), implying that some major OTUs might exist as dead cells. Taken together, various biofilm-forming bacteria can be formed on the RO membrane surface after BWRO. The biofilm communities differed with influent water source, suggesting that different microbial consortia can cause biofouling.
Evaluation of cleaning
The permeate flux behavior during the lab-scale cross-flow biofouling experiment using intermittent operation is shown in Fig. 8(a). Permeate flux decreased during intermittent operation for 5 h; from 23.4 to 9.6 LMH at 1st intermittent operation and from 18.7 to 10.2 LMH at 2nd intermittent operation. The data are shown as the average permeate flux for three repeated experiments at each measured time point. The severe permeate flux decline was caused by the high concentration of microorganisms.
Figure 8(b) shows the flux recovery by physical or chemical cleaning after intermittent operation. The flux recoveries after the 1st and 2nd intermittent operations were 80.0 ± 10.3% and 69.0 ± 7.6%, respectively. The lower flux recovery at 2nd intermittent operation could be due to the formation of more mature biofilms on the RO membranes. The permeate flux after 2nd intermittent operation at 48 h of operation time was approximately 16.1 LMH.
Physical cleaning was performed using deionized water at various conditions, such as 0.6–1.2 L/min of flow rate and 15–60 min of cleaning time, achieving similar results for flux recovery, attached microorganisms, and EPS concentration under all conditions (Table S3). The highest cleaning efficiency was observed for chemical cleaning, achieving 88.3 ± 2.8% flux recovery by applying a 0.5 wt% NaOCl solution at 0.6 L/min for 15 min. Physical cleaning achieved a flux recovery of 72.2 ± 3.3% and a similar flux recovery was achieved after the 2nd intermittent operation (69.0 ± 7.6%). A similar flux recovery indicated insignificant physical cleaning during intermittent operation.
Quantitative analysis of the attached microorganisms and EPS concentrations on the membranes after intermittent operation and physical or chemical cleaning are summarized in Table 4. The microorganism concentration after physical cleaning was 0.3 × 107 CFU/cm2, slightly less than that of the control (sample obtained after 2nd filtration at 29 h of operation time) and after the 2nd intermittent operation (sample obtained at 48 h of operation time); both concentrations were approximately 1.0 × 107 CFU/cm2. In contrast, no microorganisms were observed after chemical cleaning. Similar trends were observed in attached protein concentrations; 6.5–6.9 µg/cm2 for control after 2nd intermittent and physical cleaning and 0.8 µg/cm2 for chemical cleaning. There were no significant differences in the polysaccharide concentration; however, a slightly lower concentration (34.7 µg/cm2) was measured after chemical cleaning compared to the physical cleanings (37.7–40.9 µg/cm2). This means that residual polysaccharides contributed to the permeate flux decline (~ 12%) after chemical cleaning. In addition, Pearson’s correlation analysis (r) was carried out between flux recovery and the concentrations of attached microorganisms, proteins, and polysaccharides as shown in Table S3. The results indicate strong relationships between flux recovery and protein concentration (r = -0.90, p = 0.01) and polysaccharide concentration (r = 0.85, p = 0.03) but no significant relationship was observed with attached microorganism concentration (r = 0.21 and p = 0.69). Note that an r value closer to ± 1 indicates a linear relationship.
Table 4
Quantitative analysis for attached microorganism concentration and EPS (e.g. protein and polysaccharide) concentration onto the membrane after physical and chemical cleaning processes.
Condition | Microorganism conc. (× 107 CFU/cm2) | Protein conc. (µg/cm2) | Polysacchride conc. (µg/cm2) |
Control a | 1.0 | 6.9 | 40.9 |
After 2nd intermittent b | 1.1 | 6.5 | 37.7 |
Physical cleaning | 0.3 | 6.7 | 38.2 |
Chemical cleaning | n.a. | 0.8 | 34.7 |
a Control data was obtained after 2nd filtration (29 h operation) |
b Sample was obtained after 2nd resting (48 h operation). |