Pilot study on the treatment of lake water with algae by ultrafiltration–ozone–biologically activated carbon

Eutrophication has been affecting the water ecological health of many freshwater lakes and reservoirs in the world since the 1960s (de Oliveira et al. 2020). The sudden excessive growth of seasonal algae in lakes and reservoirs excessively burdens drinking water treatment plants (DWTPs) and worsens the water quality (Lin et al. 2018). However, conventional water treatment technologies, including coagulation, flocculation and filtration, are inefficient for removing cyanobacterial blooms and some dissolved organic matter (Bai et al. 2019), and even excessive algae will clog filtration tanks during the operation of waterworks (He et al. 2021). During the growth and death of algae, undesirable organic matter is released into the water, such as algae organic matter (AOM) (Dong et al. 2021). As AOM contains numerous carbohydrates, amino acids and proteinaceous compounds, all these are potential precursors of carbonaceous disinfection by-products (C-DBPs) and nitrogenous disinfection by-products (N-DBPs) (Zhu et al. 2015), which will cause a negative impact on subsequent treatment units and drinking water quality. According to Lui et al. (2011), some algal-derived organic matters with hydrophilicity and low molecular weight (MW) cannot be easily removed by flocculation/coagulation, and eventually contribute substantially to DBPs formation. Previous toxicological studies have reported that DBPs could cause public health and safety risks, so it is necessary to control them. There are a number of strategies to limit the concentrations of DBPs, such as removing the DBPs precursors before disinfection, removal of DBPs after formation or by using substitute disinfectants (Mazhar et al. 2020). Some studies believed that the application of reverse osmosis membranes or improved disinfection methods were effective options for controlling the formation of DBPs (Yang et al. 2017; Thostenson et al. 2018). However, costs and technologies usually limit the choice of water treatment methods. At present, an effective and economical strategy to control the formation of DBPs in DWTPs is to remove the precursors before they react with disinfectants (Chaukura et al. 2020).

There are quite a few strategies to solve the algae problem. Many previous studies had shown that for the treatment of water with algae, chemical pre-oxidation could change the morphology and structure of algae and reduce algal cells stability, which was very efficient for the removal of algae and also conducive to the saving of coagulants in the follow-up coagulation process (Ma & Liu 2002; Chen et al. 2009; Lin et al. 2021). Therefore, chemical pre-oxidation was once considered a promising strategy. However, excessive oxidants will also cause the release of AOM, especially intracellular organic matter (IOM) (Qi et al. 2016; Wang et al. 2021). Some references have demonstrated that IOM is an important C-DBPs and N-DBPs precursors compared with other AOM and/or natural organic matter (NOM) (Wang et al. 2021). In particular, the release of IOM during pre-oxidation prior to coagulation will result in the formation of more trihalomethanes (THMs) (Zhao et al. 2020). In terms of pre-ozonation, Dong et al. and Lin et al. reported that pre-ozonation increased the formation of haloacetic acids (HAA₅) and THMs from algae (Dong et al. 2021; Lin et al. 2021). In addition, the strong oxidizability of ozone will also destroy algal cells and make them release substances such as algal toxins and odors (e.g., 2-methylisocamphanol, 2-MIB). Although some studies have suggested that ozonation and activated carbon adsorption can effectively reduce the concentration of 2-MIB in water, this approach is not always sufficient to guarantee that the treated water concentration achieves the China potable water standard for the odor threshold (10 ng/L) (Li et al. 2019). Ozonation is commonly considered to be an efficient approach to remove algal toxins. However, due to the relatively complicated structure of toxins and the interference of coexisting organic compounds, it is unrealistic to thoroughly mineralize algal toxins by ozonation (Chang et al. 2014). Moreover, the increase of algal toxins and other substances also increased ozone consumption, which inevitably increased economic costs and resulted in the formation of DBPs (Rangesh & Sorial 2011). It is unacceptable due to economic and health concerns. Therefore, it is crucial to remove algal cells and reduce the degree of algal cell rupture in the pretreatment stage to ensure the safety of drinking water.

The coagulation–ultrafiltration process (CUF) is an efficient water treatment technology with the advantages of low cost and easy operation (Li et al. 2020). It can remove particles, colloids, microorganisms, pathogens and reduce transmembrane pressure (TMP) development and the extent of irreversible fouling in the UF system (Tabatabai et al. 2014; Li et al. 2020). Hence, the CUF unit is used as the pretreatment stage to remove high MW fractions and algae cells by coagulation and UF. This not only reduces the burdens of follow-up treatment but also reduces the release of IOM and other substances to the greatest extent, which plays an important role in ensuring drinking water quality. The ozone–biologically activated carbon (O₃–BAC) advanced treatment process is an advanced water purification technology because it can remove DBPs precursors, toxic pollutants and improve biostability (Wang et al. 2019). It has been broadly applied in DWTPs in various countries (Kim et al. 2005). Therefore, the coagulation–ultrafiltration–ozone–biologically activated carbon (CUF–O₃–BAC) integrated process was first used to treat East Taihu Lake water in China, aiming at evaluating the removal efficiencies of algae, permanganate index (COD_Mn), UV₂₅₄, NH₃-N and DBPs precursors, and observing the variation of TMP under various fluxes. Finally, the ultimate purpose of this study is to provide a meaningful reference for the process improvement and research in the future from the perspectives of theory and practice.

East Taihu Lake in China serves as the water source for the pilot device, with a scale of about 510 L/h. The raw water quality indicators are given in Table 1. The experimental parameters were previously obtained by Yin et al. (2019). The coagulant used in the experiment was polyaluminum chloride (PAC), and the dosage was 4 mg/L as aluminum. The membrane material was a multibore polyethersulfone hollow fiber UF membrane, using internal pressure filtration, with an effective surface area of 6.5 m² and a cut-off MW of 150,000 Da. The dosage of ozone was 1 mg/L, and the exposure time was 15 min. The coal-based granular activated carbon (GAC) was used as the carbon pool filter material. The depth of the carbon layer was 1 m, the filtration rate was 6.5 m/h and the empty bed exposure time was 10 min.

As shown in Figure 1, the raw water first passed through 200 μm butterfly filter to remove large-sized suspended and floating matter, and then directly entered the UF tank after coagulation and finally entered the O₃–BAC tank. The removal efficiencies of algae, COD_Mn, UV₂₅₄, NH₃-N and DBPs precursors in the CUF–O₃–BAC integrated process were investigated. Turbidity was measured by a desktop turbidity meter (2100N, Hach, USA). The pH was measured using a pH meter (PB-21, Sartorius, Germany). COD_Mn was determined by the acidic potassium permanganate method. UV₂₅₄ was measured at a wavelength of 254 nm (Evolution 200, Thermo Fisher, Germany) using a 1 cm quartz cell. The concentration of NH₃-N was determined by Nessler's reagent spectrophotometry. The instrument was a dual-beam ultraviolet-visible ultraviolet spectrophotometer (Evolution 300, Thermo Fisher, Germany).

All water samples were buffered with pH 7 phosphate buffer before chlorination. If necessary, the water samples were first adjusted to pH 7 with dilute sulfuric acid and/or dilute sodium hydroxide, and then buffered with pH 7 phosphate buffer. Then each sample was disinfected with sodium hypochlorite, and the dosage was 2 mg/L as Cl₂. After chlorination, the samples were incubated at constant temperature (25 ± 1 °C) for 24 h in the dark to simulate water chlorination and investigate the formation of DBPs in water samples (Prasert et al. 2021). The reaction was stopped after 24 h by adding excessive ascorbic acid (Watson et al. 2018). THMs and HAA₅ were tested using a gas chromatograph with an electron capture detector (Agilent, 7890A, USA). For THMs testing, 5 ml samples were directly taken and then placed in a gas chromatograph for determination. According to the water quality determination of HAA₅-gas chromatography (Ministry of Environmental Protection 2015), HAA₅ was extracted by the liquid/liquid technique with methyl tertiary-butyl ether followed by derivatization with acidic methanol.

Four different fluxes (60, 70, 80 and 90 L/(m²·h)) were used to investigate the performance of the membrane system during long-term operation, and each operating mode was continuously operated for 30 days. Constant flux operation mode was adopted for UF. One-minute hydraulic positive flushing and one-minute hydraulic backwashing were carried out every 30 min of operation, and online chemical cleaning was used every 24 h. The changes of TMP during the experiment were automatically recorded by the PLC (Programmable Logic Controller) system.

The average removal efficiencies of phycocyanin, COD_Mn, UV₂₅₄ and NH₃-N via the CUF unit and the integrated process are shown in Figure 2.

Phycocyanin is a phycobilin pigment, mainly found in cyanobacteria (Thomson-Laing et al. 2020). Published literature has shown that phycocyanin is an important indicator of cyanobacteria, and cyanobacterial blooms can be quickly and easily monitored by the phycocyanin concentration (Lzydorczyk et al. 2009). Therefore, Phycocyanin was used to characterize the concentration of cyanobacteria in the experiment. Figure 2 shows that the removal rates of phycocyanin via the CUF unit and the integrated process reached 94.39 ± 2.03 and 95.89 ± 1.52%, respectively. It indicated that both the CUF unit and the integrated process could effectively reduce the concentration of cyanobacteria, but the CUF unit played a major role. Previous studies have shown that the algae in Taihu Lake are mainly cyanobacteria with a diameter between 2 and 15 μm, which are larger than the pore size of the UF membrane. Therefore, the UF has a good removal effect for cyanobacteria. The stimulation of ozone to algal cells was significantly reduced by this physical interception, thus reducing the release of AOM, which also indicated that the formation of DBPs could be reduced during subsequent chlorination.

The average concentrations of COD_Mn in the effluent from the CUF unit and the integrated process were 2.01 ± 0.16 and 0.86 ± 0.10 mg/L, respectively, and the removal rates were 44.51 ± 7.33 and 76.18 ± 4.38%, respectively (Figure 2). It can be seen that the COD_Mn value of the effluent from the CUF unit has met the requirement of China's ‘Standards for drinking water quality’ (GB 5749-2006) of no more than 3 mg/L, indicating that the CUF unit has positive impaction of COD_Mn removal. The removal of COD_Mn via the CUF process mainly depends on the coagulation process. Because the coagulant first adsorbed the −OH group on the surface of the particulates, resulting in electric neutralization and bridging effect, and then reacted with the solution hydrolysis to produce gel precipitates, which further played the role of bonding and coagulation (Han & Huang 2013), thereby increasing the size of the organic colloid in water and making it easy to be removed via UF. Because COD_Mn reflects organic matter with lower MW fractions and the aperture of the UF membrane is large, the removal of COD_Mn via the UF membrane mainly relies on the adsorption and retention of the filter cake layers on the membrane surface and the membrane itself. The integrated process further improved the removal rate of COD_Mn. First, ozonation caused breakage of carbon–carbon double bonds of some organic matter, which resulted in the production of more bioavailable organic matter (Lundqvist et al. 2019). Therefore, the concentration of COD_Mn can be reduced via the role of subsequent biological activated carbon tanks.

UV₂₅₄ stands for aromatic structures of organics that are primarily humics and is sensitive to the formation of THMs (Xu et al. 2011, 2019), so it is necessary to control it. During the experiment, the UV₂₅₄ value of raw water was relatively low, with an average concentration of 0.06 ± 0.01 cm⁻¹, which was mainly related to the ecological environment of the lake region. Due to the dense aquatic plants in East Taihu Lake, the hydrophilic substances after decomposition of plants, which were easily decomposed and utilized by microorganisms, were released into the aquatic environment, so the UV₂₅₄ value of the hydrophobic substance was relatively low (Feng et al. 2016). As shown in Figure 2, the average removal efficiencies of UV₂₅₄ via the CUF unit and the integrated process reached 20.94 ± 7.42 and 72.06 ± 4.72%, respectively. It indicated that the integrated process could effectively reduce the risk of DBPs formation during disinfection. Since coagulation had a good effect on the removal of macromolecules and hydrophobic organic compounds, the removal of UV₂₅₄ via the CUF unit was mainly completed via the coagulation process. The removal rate of UV₂₅₄ was significantly improved by the integrated process. It was because post-ozone had strong oxidizability, and it was expert in reducing humic and fulvic substances via converting them into simple non-humic forms with small molecules that were easier to biodegradation (Sun et al. 2018).

Because East Taihu Lake is a typical grass-type lake, NH₃-N is mainly produced via the decomposition of dead aquatic plants and other residues (Fan et al. 2004). The experiment was carried out under the conditions of water temperature around 28 °C and pH around 8.0. Therefore, the NH₄⁺ and NH₃ accounted for about 90 and 10% of NH₃-N (Zhang 2000). As shown in Figure 2, the average removal efficiencies of NH₃-N via the CUF unit and the integrated process reached 54.32 ± 8.60 and 81.31 ± 6.71%, respectively. It showed that both the CUF unit and O₃–BAC process had positive impaction of NH₃-N removal. This phenomenon seems very interesting, and the removal mechanism of NH₃-N will be discussed below. The removal mechanism of NH₃-N by the CUF unit is shown in Figure 3.

The removal efficiency of NH₃-N by coagulant was poor. This was due to the fact that metal hydroxide sols formed during coagulation carried the like charge as NH₄⁺ (Wongcharee et al. 2020), and NH₃-N reflected small MW fractions. Therefore, the removal mechanism of NH₃-N in this experiment was mainly based on the viewpoints of Yu et al., Liu et al. and Chen et al.Yu et al. (2020) and Liu et al. (2017) believed that the formation of filter cake layer on the membrane surface and the accumulation and growth of nitrifying bacteria could induce an effective removal of NH₄⁺-N by adsorption and oxidation. Chen et al. (2020) considered that electrostatic attraction was the main mechanism for NH₃-N removal during the CUF, through which NH₃-N adhered to the surface of the negatively charged colloidal particles and was removed through particles discharging. In this research, these colloidal particles were more inclined to humic acid in the aquatic environment.

Some studies have shown that (Wongcharee et al. 2020) the NOM in water contributes to the removal of NH₃-N, among which humic acid may play a crucial role. Humic acid is a complex compound with amino acids and polysaccharides formed via lignin through a series of oxidation and demethylation processes. It contains both hydrophobic and hydrophilic parts and many chemical functional groups, such as carboxyl, phenolic, carbonyl and hydroxyl (Zhang & Bai 2003). The existence of these groups causes the acidity and exchangeability of humic acid, which can carry out ion exchange and complexation with NH₄⁺ in aquatic environment, thereby promoting the adsorption of NH₄⁺ (Meng et al. 2011). Furthermore, NH₃ can also combine with humic substances in water. Some researchers argued that a relatively small amount of quinones or phenols in humic substance would fix a large amount of NH₃, and the reaction model is shown in Figure 4 (Stevenson & Xia 1994). However, the BAC also played a positive role in the removal of NH₃-N. It is because the decomposition of ozone can provide an abundance of dissolved oxygen. Under such conditions, NH₃/NH₄⁺ can be converted into NO₂⁻ and NO₃⁻ by ammonia-oxidizing microorganisms, thus further reducing the value of NH₃-N.

To guarantee the health and safety of potable water, chlorine disinfection is commonly used to kill pathogens in potable water (Zhang et al. 2020). However, the existence of toxic and harmful DBPs such as THMs and HAAs in water during chlorination have aroused great concern to human health (Farré et al. 2011). Epidemiological studies have shown that drinking chlorinated water is associated with bladder cancer, colon cancer and rectal cancer (Pan et al. 2014; Chen et al. 2021). Therefore, it is critical to evaluate the process for the removal of DBPs. The concentrations of DBPs in influent and effluent are given in Table 2.

Figure 5 shows the removal efficiencies of THMs and HAA₅ precursors by the integrated process. Figure 5(a) shows that the CUF process could only reduce the formation potential of TCM, but only about 0.76%. The occurrence of this phenomenon may be related to the characteristics of organic matter in water. It is because UF usually exhibited better removal efficiency for high MW fractions, which is relatively poor for low and medium MW fractions. Some studies have reported that the small and medium MW fractions are major DBPs precursors (Yan et al. 2010; Xu et al. 2019). Generally, the higher MW fractions of NOM are more aliphatic, while the lower MW fractions contain more aromatic and carboxyl groups, which are more reactive to the formation of THMs and HAA₅ during chlorination (Xu et al. 2011). UF can effectively remove high MW fractions, which is propitious to the reaction of small and medium MW fractions with adequate chlorine, thus resulting in more DBPs. This also indicates that the CUF process has a tendency to increase the formation potentials of DBPs.

Figure 5(b) and Table 2 show that after CUF–O₃ treatment, the formation potentials of brominated DBPs increased significantly and those of chlorinated THMs decreased, while those of chlorinated HAA₅ did not change significantly. The removal efficiencies of BDCM and TCM reached 18.54 and 31.13%, respectively. The results indicated that ozone could reduce the formation risk of chlorinated THMs to a certain extent, but it was ineffective to reduce the formation risks of brominated THMs and HAA₅. This may be because ozonation transforms NOM from hydrophobic character to hydrophilic character and/or higher MW fractions to lower MW fractions, which are more likely to generate brominated DBPs during subsequent chlorination (Hua & Reckhow 2013; Gao et al. 2020). Figure 5(c) shows that the removal efficiencies of TBM, BDCM, DBCM, TCM, DCAA, TCAA and DBAA are −43.51, 68.31, 36.43, 76.86, 56.61, 31.42 and 8.00%, respectively, whereas MBAA was not detected. Although the removal efficiency of TBM formation potential in BAC effluent was negative, the removal efficiency of TBM via the CUF–O₃–BAC process was still significantly improved compared to the CUF–O₃ unit. Therefore, the CUF–O₃–BAC process is an effective technology to remove THMs and HAA₅ precursors in drinking water treatment. Some studies have shown that GAC exhibits very high efficiency to adsorb small, neutral and hydrophobic molecules but relatively lower adsorption capacity for hydrophilic molecules (Fu et al. 2017). Therefore, hydrophilicity and low and medium MW fractions may be the major reason for the increase of brominated DBPs based on these theories.

In the membrane treatment process, membrane flux and TMP are the most basic indicators of the membrane's operational stability and membrane fouling (Wang et al. 2020). Therefore, the long-term continuous operation of the membrane system was investigated by changing the membrane fluxes and recording the corresponding TMP values. During the experiment, the PLC system automatically recorded a TMP value every 5 min. To facilitate experimental analysis, the average value of data per 8 h was used to represent the variation of TMP. The variation of TMP under various membrane fluxes is shown in Figure 6(a). Figure 6(a) shows that the initial TMP values are 9.6, 12.3, 14.3 and 16 kPa at the fluxes of 60, 70, 80 and 90 L/(m²·h), respectively. After 30 days of operation, the TMP increased to 15.2, 26.3, 30.3 and 37.5 kPa, respectively (Figure 6(a)). Now, the kinetic simulation of the relationship between TMP and time under various fluxes is carried out.

According to the above linear fitting results, the correlation coefficients (R²) of first-order dynamics and second-order dynamics were all greater than 0.9 under the four fluxes of 60, 70, 80 and 90 L/(m²·h). The results indicated that the relationship of TMP with time accorded with both the first-order and second-order kinetics models. According to the fitting results in Figure 6(b) and Table 3, the rising rate of TMP did not increase with the increase of flux. On the contrary, when the flux was 70 L/(m²·h), the growth rate of TMP was the slowest and the correlation coefficient was the smallest (R² = 0.93), which was also reflected in the second-order kinetics (R² = 0.91). It preliminarily indicated that the fouling extent of the membrane system was relatively the lowest under the flux of 70 L/(m²·h).

The average TMP (TMP_avg) vs. flux plots are frequently used to determine critical and threshold fluxes from flux stepping experiments (Miller et al. 2014). To further determine the optimal flux, the relationship between TMP_avg and flux is plotted, and the results are shown in Figure 6(d). The curve of Figure 6(d) was divided into three sections and analyzed by linear regression, respectively. Figure 6(d) shows that the development of TMP_avg is particularly rapid with the increase of flux when the flux was >80 or <70 L/(m²·h). On the contrary, the development of TMP_avg was relatively moderate under the condition of 70–80 L/(m²·h). Therefore, we can draw a conclusion that the membrane system has relatively strong anti-fouling ability under the flux of 70 L/(m²·h).

For the treatment of lake water with algae, this research adopted the CUF–O₃–BAC integrated process to treat East Taihu Lake water and first investigated the removal efficiencies of algae, COD_Mn, UV₂₅₄, NH₃-N and DBPs precursors. We found that the integrated process had positive impaction of NH₃-N removal, which might be mainly determined by the formation of cake layer on the membrane surface, the nitrification of nitrifying bacteria and the influence of humic acid in raw water. However, further research is necessary to determine which factor dominates. The DBPs results indicated that the CUF unit tended to increase the formation potentials of THMs and HAA₅. This may be because the UF can effectively remove high MW fractions, which is more conducive to the reaction of small and medium MW fractions with sufficient chlorine, thus resulting in more DBPs. The formation potentials of brominated DBPs after CUF–O₃ effluent were significantly increased. This is attributed to ozone transforming NOM from hydrophobic character to hydrophilic character and/or higher MW fractions to lower MW fractions, which are considered brominated DBPs precursors. However, the CUF–O₃–BAC integrated process was an effective technology for the removal of THMs and HAA₅ precursors in drinking water treatment. Finally, we established the kinetic models of TMP and operation time and determined that the membrane system had relatively strong anti-fouling ability under the flux of 70 L/(m²·h). However, we did not find out exactly what factors caused the relatively strong anti-fouling ability of the membrane system under the flux of 70 L/(m²·h), which would continue to be observed and analyzed in the follow-up experiments.

This research has been supported by the ‘Integration and comprehensive application of drinking water security technology in Suzhou’ (No. 2017ZX07201001). The authors wish to gratefully acknowledge Lanzhou Jiaotong University and Tongji University for their support and assistance in this research.

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