3.1. Process Performance
The average COD concentration in the influent was approximately 1200 mg/L throughout the experiment. The HRT was sequentially set at 24 h, 20 h, 16 h, 12 h, and 8 h. Each time the HRT was shortened, the MLCOP reactor was not stable, and needed time to adapt to the change. During the adaption period, the COD removal rate experienced a similar trend at different stages: it first decreased, then increased quickly, and finally stabilized [
17]. The fluctuation was the largest in tank B, and the smallest in tank F. This is because the COD concentration of the influent in tank B was higher than the concentration in tanks D and F. This led to a larger impact on the microorganisms in tank B.
As the influent flow rate increased, the microorganisms in each biological tank quickly adapted to the changing environmental conditions, achieving stable operation. This demonstrated that the MLCOP reactor was highly resistant to the impact of changes in flowrate.
After the MLCOP reactor achieved stable operations at each HRT condition, the following 20 day test was conducted. The average COD concentrations of the effluent from tanks B, D, and F, and the COD removal rates, were evaluated.
Figure 4 shows the results. Each COD value was the average value of 20 days’ test results.
Figure 4 shows a significant difference in COD concentration between the biological tanks. The microorganisms in the first two biological tanks played the primary role in degrading the organic substances. The experimental test results demonstrated that the COD concentration of the final effluent remained stable below 300 mg/L.
In addition to COD concentration, other effluent indicators also reached discharge standards during the experiment. For example, through previous physicochemical treatment processes, total nitrogen and ammonia nitrogen decreased to 30 mg/L and 10 mg/L, respectively. To consistently reach the discharge standard, the MLCOP system removed these substances further, reaching levels below 10.5 mg/L and 1.4 mg/L, respectively, at different HRTs (
Figure 5).
As the HRT decreased, the total COD removal rate in the MLCOP reactor gradually decreased. Removal rates were 93.3% at a HRT of 24 h; 91.5% at a HRT of 20 h; 87.3% at a HRT of 16 h; 85.1% at a HRT of 12 h; and 83.8% at a HRT of 8 h. When the HRT was set at 4 h, the MLCOP reactor had difficulty reaching a stable state. The COD concentration of the effluent fluctuated widely, and sometimes exceeded the target of 300 mg/L. Therefore, comprehensively considering the operation cost, treatment capacity, and efficiency, 8 h was identified as a more suitable and lower HRT to remove COD in the MLCOP reactor.
Nitrogen removal in biological treatment processes requires both aerobic and anoxic environments. The results may indicate that ammonia nitrogen was first converted into nitrite or nitrate through oxidation in an aerobic environment. Then, the nitrite or nitrate was transformed into nitrogen gas in an anoxic environment, which probably existed in the inner part of the biological carriers or biofilm. The available literature also showed that under suitable conditions, the simultaneous nitrification and denitrification (SND) could occur during the aerobic treatment process, because there existed limitation to oxygen diffusion, and it could cause the formation of an anoxic environment [
22,
23].
Previous research demonstrated that the carbon source with suitable concentration control could facilitate SND, because the external carbon source could act as electron donor and stimulate the nitrogen removal by denitrification [
24,
25]. It could be inferred that the different concentration of carbon source in each biological tank may have an active effect on the nitrogen removal.
The removal efficiency of total nitrogen and ammonia nitrogen were positively correlated to the HRT. As the HRT was sequentially set at 24 h, 20 h, 16 h, 12 h, and 8 h, the removal efficiency of total nitrogen was 77.3%, 76.8%, 76.1%, 75.5%, and 65.0%, respectively; the removal efficiency of ammonia nitrogen was 96.5%, 95.7%, 94.9%, 93.6%, and 86.3% respectively. The removal efficiencies of total nitrogen were lower than ammonia nitrogen at HRTs. This can be attributed to a restrained denitrification reaction under oxic conditions. This suggests that only part of the nitrate could be transferred to the anoxic environment where the denitrification occurred [
23].
In general, using only an aerobic biological process to biodegrade the refractory organic matter is difficult and expensive. Therefore, anaerobic processes are usually used beforehand to improve wastewater biodegradation. This allows organic substances to be more easily degraded further by the aerobic biological process. Currently, hydrolysis acidification combined with aerobic biological process is relatively common when treating automobile painting wastewater to biodegrade organic substances [
8]. Nevertheless, anaerobic systems are usually larger than aerobic ones because of low biological treatment efficiency in anaerobic conditions. This requires additional construction or space costs for anaerobic systems.
Considering economic and technological limitations, the MLCOP system in this paper was used to biologically treat automobile painting wastewater to improve pollutant removal efficiency and obtain higher effluent quality. When the biological system reached stable operation with the HRT of 8 h, the variations of BOD and COD versus time are shown in
Figure 6. The BOD, COD, and SS of influent and effluent from three stages were also assessed.
Table 3 shows the average values.
The BOD/COD ratio of influent was approximately 0.15, indicating that the raw painting wastewater was very hard to biodegrade. However, the total BOD removal ratio was approximately 80.6%; the BOD/COD ratio improved during biological treatment. In particular, the BOD/COD ratio of the effluent in tanks B and D increased to 0.36 and 0.54, respectively. This is maybe because the special biological carriers, with a compact and porous fiber structure, could retain a high biomass concentration and established an anoxic microenvironment in the inner section. This facilitated the transformation of macromolecule organic substances to small molecular organic compounds through hydrolysis acidification.
Approximately 28 m
3 painting wastewater was treated during the experiment, however, less than 1 kg of excess effluent sludge was collected, and the average SS removal ratio was 82.8%. The excess sludge yield of the MLCOP system was approximately 0.03 g SS (g COD
removed)
−1, while the excess sludge yield of the conventional activated sludge treatment process was reported to be 0.2–0.4 g SS (g COD
removed)
−1 [
26,
27]. Only around 10% sludge production of conventional activated sludge treatment processes was obtained in the MLCOP system. Besides, the SS of final effluent from the biological reactor was low, and could reach the discharge requirement. There was no sedimentation tank and no sludge recycle in this system. This indicates the MLCOP system could efficiently reduce the excess sludge production compared to the conventional activated sludge treatment processes.
3.2. Microbial Community Structure Analysis
The process performance of the MLCOP system was significantly influenced by microbial behaviors. This highlights the need to analyze and demonstrate the relationship between the microbial community structure and pollutant degradation from a microbiological perspective. Therefore, further research was conducted to assess the impacts of microbial community on MLCOP performance.
The high-throughput pyrosequencing technology was applied in this study to investigate the characteristics of microbial communities in different stages of the MLCOP reactor. This allowed the identification of several sequences, and provided comprehensive information about the microbial communities. This could offer a better visualization and understanding of the variation of the complete microbial community in the MLCOP reactor.
3.2.1. Microbial Abundance and Diversity
The high-throughput pyrosequencing technology generated 119,644 high-quality sequences from the four samples, including the inoculation sludge sample (IS), and three activated sludge samples from tanks B, D, and F (BS, DS, and FS, respectively). The quantity of sequences obtained in each sample was distinct: 26,160 (IS); 34,650 (BS); 27,345 (DS); and 31,489 (FS). Sequence trimming barely had an impact on the analysis of the microbial community, and was performed for each sample to support alignment and normalization [
19]. The remaining 26,160 sequences for each sample were used for further study on the microbial community.
The Chao estimator was used to characterize species richness, with a higher value indicating greater richness. The Shannon index was used to estimate species diversity and evenness, with a higher value indicating higher diversity and evenness. High biodiversity is assumed to enhance ecosystem stability, based on the ecology principle. Coverage was used to determine the possibility that the following read would be included in the obtained OTU (Operational Taxonomic Units), with a higher value representing larger possibility [
24]. The coverage was greater than 0.995 in each sample, suggesting that the sequencing data from samples were sufficiently representative and authentic.
In
Table 4, IS stood for the inoculation sludge sample. The inoculation sludge was from a general sewage treatment plant, and the plant had operated using an anoxic/oxic process for several years. Therefore, the sequencing results of IS could represent those of the standard activated sludge.
Table 4 shows that when the MLCOP system reached stable operation, the Chao estimator and Shannon estimator of BS, DS, and FS all decreased compared with those of IS. This suggests that microbial community richness and diversity in the MLCOP system were less than those in standard activated sludge (IS). In the MLCOP system, the Chao estimator and Shannon index were highest for DS and lowest for BS. This suggests that the microbial community of the middle stage exhibited larger richness and higher diversity than the others in the MLCOP system. Given that the entire biological reactor ran under identical operating conditions, variations in microbial community structure largely depended on the influent water quality of each tank.
In the first stage, the influent had the greatest impact on the microbial community. This stage had the highest concentration of refractory poisonous organic compounds, and the lowest biodegradability. Based on the decomposition of specific microorganisms, many macromolecule organic substances were degraded into smaller molecular organic compounds, improving biodegradability. In the middle stage, influent impact was reduced, increasing the survival and abundance of the microbial community. In the last stage, influent impact on the microbial community was the lowest. This is because the easily biodegradable organics had been consumed in the previous stages. As such, the lack of organic nutrients also restrained growth and enrichment of the microbial community.
3.2.2. Microbial Community Composition
To observe the microbial community variation in different stages, the microbial community structure and composition were analyzed on the phylum and class level.
Figure 7 shows that, on the phylum level, the microorganisms in the reactor’s distinct stages were from similar dominant phyla, but there were different abundances of each phylum. In total, 10 phyla were detected. The main phyla were similar in each sample; Proteobacteria, Bacteroidetes, Firmicutes, and Chloroflexi, and represented 83% in IS, 96% in BS, 91% in DS, and 94% in FS, respectively. This indicates that the main four phyla may be enriched in painting wastewater treatment, and play a critical role in pollutant removal.
The dominant phylum in each sample was Proteobacteria; this finding is consistent with previous studies [
28]. Most of Proteobacteria belonged to facultative anaerobic heterotrophic bacteria, using organics as a carbon source during the wastewater treatment process. Proteobacteria were highly enriched and they were particularly high in BS, accounting for 71% of the total microbial composition. The relative abundance of Proteobacteria in DS and FS was decreased, in turn, compared to sample BS, which was consistent with the change trend of COD concentration, as shown in
Figure 4. As wastewater flowed sequentially through each tank, the differences in pollutant composition and concentration in each tank may have facilitated the change of microbial composition. This may suggest that microorganisms with different species and structures could efficiently decompose the refractory organic matter step by step in each tank.
Bacteroidetes decreased to 12% in BS and 18% in DS; for comparison, they were present at 21% in IS. However, Bacteroidetes were significantly enriched in FS (34%). This finding suggests that the influent wastewater with higher concentrations of COD and toxic organic substances in BS and DS may restrain Bacteroidete enrichment.
Firmicutes were more abundant in samples BS (10%), DS (12%), and FS (6%) compared to sample IS (2%). Firmicutes have a relatively thick cell wall, increasing resistance to desiccation and extreme environments [
29,
30]. Firmicutes are also reported to have hydrolytic and acidogenic abilities [
31].
Table 3 shows that the changes in the BOD/COD ratio in different stages were aligned with changes in Firmicutes. This suggests that Firmicutes were closely associated with macromolecule organic substance degradation, and could significantly contribute to improving wastewater biodegradability in the MLCOP system’s hydrolysis acidification.
Clostridium, in this study, was the main subdivision of Firmicutes at class level. Although Clostridium was not the major OTU in sample BS, Clostridium were more abundant in samples BS (8%), DS (9%), and FS (5%), compared to sample IS (1%). Besides, Clostridium was reported to enable consumption and decrease of excess sludge efficiently through fermentation [
32]; this outcome corresponds with the low excess sludge production in this study. Thus, Firmicutes may have a profound impact on sludge reduction in the MLCOP system.
Members of Chloroflexi played an ecological role in degrading cellular materials and carbohydrates.
Figure 7 shows that Chloroflexi in BS was reduced from 9% to 3%; this was mainly due to the sharp impact of the highly polluted influent wastewater. This validates the previous observation that the microbial structure in the first stage was less stable than in the subsequent stages.
The four samples had similar dominant phylum and main microbial compositions at the phylum level; however, discrepancy among samples turned out to be more evident on the class level. This level provided more information on the dynamic changes and of microbial community structure in each sample.
Figure 8 shows that the most dominant class in the initial phase was γ-Proteobacteria. However, after stable operations, the dominant classes in different stages shifted dynamically and became more distinct; these classes were β-Proteobacteria in BS, α-Proteobacteria in DS, and Sphingobacteria in FS.
Sphingobacteria, classified as a subdivision of Bacteroidetes, was present at an abundance of 16% in IS. However, it decreased to 1% in BS, and 6% in DS, and was enriched in FS at an abundance of 25%. This may be because the bacteria were inhibited by the toxicity of influent with high COD concentrations.
The distributions of the main dominant classes, including α-Proteobacteria, β-Proteobacteria, γ-Proteobacteria, and Sphingobacteria, were relatively similar, and even in IS, but became heterogenous at different proportions in the other three samples. In particular, α-Proteobacteria and β-Proteobacteria were largely enriched in BS, compared with other class compositions.
Proteobacteria mainly included four subdivisions on the class level: α-Proteobacteria, β-Proteobacteria, γ-Proteobacteria, and δ-Proteobacteria. In previous studies, β-Proteobacteria were identified as the largest class of Proteobacteria in sewage wastewater treatment systems [
19]. However, in this study, the class compositions in
Figure 9 show the discrepancies in Proteobacteria from different reactor stages.
Study results suggest that the microbial community structures in different tanks dynamically and flexibly shifted their compositions to adapt to the relevant living environments at different pollutant concentrations [
33]. During the extended acclimation to the painting wastewater, some bacteria that survived in the inoculation sludge could not adapt to the influent water quality; these bacteria were gradually eliminated in each level of the reactor.