Overall performance of Songbei WTP
After the “aeration–biofiltration” cascading treatment, the effluent iron, manganese and ammonia were all far below the permitted concentration level in the current drinking water standard. Figure 1e showed that the DO achieved 8.5 mg/L in the renovated aeration unit, which totally meet the demand of iron, manganese and ammonia oxidization in the raw groundwater. The DO concentration decreased to ~ 1.5 mg/L in the effluent of biofilter. The results of SEM and EDS analysis showed that the manganese and iron compounds attached on the surface of mature biofilter media, which could be the direct evidence for the oxidation of iron and manganese in the biofilter media by microbial catalysis. This result was also in accordance with previous studies (Du et al. 2017; Li et al. 2013). Simultaneously, the EPSs secreted by microorganisms and the sheet membrane structure composed of iron and manganese oxides provided the living space for the growth of microorganisms. With the further development of biofilms and oxides deposition, the incomplete spherical shell could gradually form to small manganese particles.
Functional bacteria distribution
As previously noted, the biological process played an important role in the biofilters for the removal of the three pollutants from groundwater. To elucidate the interactions among all of the OTUs and analyze the shared and most abundant OTUs in the three microbial consortia samples, a network representing the functional bacteria community change and linkage was constructed (Fig. 3a). Collectively, only 260 phylum-level OTUs, out of 2478 in total, were shared in these three samples. The number of OTUs shared by the UL and ML samples was 345, while 314 and 471 were shared by the UL&DL and ML&DL samples, respectively (Fig. 3a). This result was accordance with the PCoA analysis. The majority of the shared OTUs were Proteobacteria (44.2–53.0%), Acidobacteria (9.2–11.0%) and Nitrospirae (7.3–8.5%). Nitzsche et al. (2015) showed that the sand filter community was to a large extent dominated by nitrifying bacteria (Nitzsche et al. 2015). In this study, the bacteria capable of ammoxidation and nitrification, i.e., Propionibacterium, Nitrosomonas, Nitrosomonas and Candidatus Nitrotoga, accounted for 41.6% (8.9% AOB, 15.2% NOB and 17.5% NRB) of the population in the UL sample, which was 1.3 and 2.2-fold of that in the ML and DL samples (Fig. 3b). Further study on the specific composition of nitrifying bacteria showed that the NRB were dominated in the UL sample (17.5%), while NOB was in the ML sample (26.1%). Compared with the ammonia-oxidizing and nitrifying bacteria, only low numbers of OTUs related to known iron- and manganese-oxidizing bacteria (IOB and MnOB) in the three sand filter samples were quantified (< 0.2%, data not shown). Especially for the IOB, i.e., Gallionella and Leptothrix (Yang et al. 2014), there were only 6 and 7 OTUs in the UL and DL samples. Such high iron removal efficiency maybe caused by the synergistic effects of oxygen auto-catalytic and biological iron oxidation, which was in accordance with the findings of Voegelin et al. (2014) and Yang et al. (2014). By contrast, a wide range of MnOB bacterial genera was identified, including Flavobacterium (Cai et al. 2015), Hyphomicrobium (Yang et al. 2014), Planctomyces (Yang et al. 2014), Ralstonia (Yang et al. 2013) and Sphingomonas (Li et al. 2016). However, only a low relative sequence abundance of OTUs were found (Fig. 3a), the no. of all the related MnOB peaked at 102 in the ML sample, followed by the DL sample (100). The reason behind here maybe that the biological Mn oxidation was induced by microbial consortia involved in this study. Nitzsche et al. (2015) also found that only low OTUs numbers related to known IOB and none of MOB were identified in a household sand filter. They suggested that biotic Mn oxidation was most likely mediated by a phylogenetically diverse microbial community (Nitzsche et al. 2015). It is well known that the sites for the dissolved iron, manganese and ammonia removal were separated orderly along the depth of the biofilter. That is, most of the iron and ammonia were removed in the upper layer of biofilter and manganese removal was mainly concentrated in the lower layer. This study confirm the unanimous results of previous research (Li et al. 2013; Nitzsche et al. 2015; Tekerlekopoulou et al. 2013).
Correlation analysis between functional bacteria and environmental variables
To further discern the plausible correlation between biofilter samples, characteristic genera and various environmental and performance measurements, CCA analysis was performed in this study (Fig. 4). 16 characteristic bacteria and 6 environmental variables were taken into consideration. The CCA1 and CCA2 model explained 91.9 and 8.1% of total variance, which indicate that these two coordinates could represent the CCA results. The dissolved iron, manganese, ammonia and DO concentrations were positively correlated with the first canonical axis. For axis 2, only dissolved iron, layer height and temperature showed good positive correlations. The detailed information is shown in Additional file 1: Table S2. According to the length of the vector, indicating the strength of the relationship between the environmental variable and microbial community, all the six variables strongly linked to the microbial community. The characteristic genera in the UL sample had a highly positive correlation with the iron and ammonia removal. It is worthwhile to note that Propionibacterium was comparatively correlated with NH4+ removal. We also found that Mn removal had very high positive correlation with a number of functional bacteria, such as Ralstonia, Variovorax, Gallionella, Flavobacterium, Microbacterium, which were abundant in the DL sample. That is, the removal of iron and ammonia was prior to that of manganese, which occurred at the upper and deeper layers of biofilter, respectively. This was consistent with the above discussions (Functional bacteria distribution). In addition, the intersection angle between the iron and DO was slightly greater than that of ammonia, indicating that the ammonia removal was more related to DO concentration than iron. This could be verified by the coefficients of DO demand for the iron and ammonia removal, which were 0.14 and 4.57, respectively (Stumm and Morgan 2012). In this sense, the CCA results suggested that the stable biofilm on the biofilter media, created by certain microorganisms from the groundwater microflora, may play a crucial role in the simultaneous removal of iron, ammonia and manganese. Meanwhile, the relationship between the functional bacteria and the environmental variables may provide insight on possible mechanism of the “aeration–biofiltration” cascading treatment process.