Implementation of an anaerobic selector step for the densification of activated sludge treating high-salinity petrochemical wastewater

A primary and long-standing problem in industrial biological wastewater treatment is bulking sludge. This phenomenon commonly leads to bad sludge settling characteristics, sludge washout and effluent discharge problems. The extensive growth of filamentous organisms is seen as one of the main causes of bulking sludge, leading to the specific terminology of filamentous bulking sludge. As described by Jenkins et al. (2003) and Martins et al. (2004), different aspects such as low nutrient, substrate or oxygen concentrations, can provoke the extensive growth of filaments, with the organisms forming a net-like structure hindering the settling of sludge. Bulking sludge is a problem that widely occurs in many industries, as a study by Cornelissen et al. (2018), conducted in Flanders (Belgium), showed. In this study, 36% of the treatment plants investigated experienced bulking sludge, including plants in the (petro)chemical industry.

Many strategies have been applied to overcome this problem, one being the anaerobic feast/aerobic famine feeding strategy, thus implementing an anaerobic selector. This strategy relies on the selection of slow-growing organisms such as polyphosphate accumulating organisms (PAOs) and glycogen-accumulating organisms (GAOs), another key for the successful formation of aerobic granular sludge (AGS) (de Kreuk & van Loosdrecht 2004; Caluwé et al. 2022). These storage organisms are able to store carbon (e.g. volatile fatty acids (VFA)) as storage polymers under anaerobic/anoxic conditions, giving them a clear advantage over filamentous organisms. Ultimately, filaments will be out selected, giving way to the formation of more dense and settleable sludge flocs (and even granules), and hereby improving sludge settling and effluent discharge. The feast/famine strategy has been applied successfully before with various industrial wastewaters. Caluwé et al. (2017) were able to improve the sludge characteristics of sludge originating from a petrochemical company. Through the implementation of an anaerobic feast/famine regime for the treatment of petrochemical wastewater, the sludge volume index (SVI) was decreased from 285 to 56 mL·g⁻¹ after 30 days. Other industries where the same strategy has been successfully implemented include the tank truck cleaning industry (Caluwé et al. 2018, 2022), the potato processing industry (Dobbeleers et al. 2017) and breweries (Corsino et al. 2017).

Salinity effects on the formation and stability of AGS or dense sludge flocs in general have been widely described, including the effects on the metabolism of PAOs and GAOs, mostly on the basis of NaCl salinity. Pronk et al. (2014) investigated the influence of increasing salinity on AGS in a sequencing batch reactor (SBR), where starting at concentrations of 20 g Cl⁻·L⁻¹, the granule size decreased and inhibition of nitrite oxidation occurred which ultimately led to the inhibition of phosphate removal by PAOs. Welles et al. (2014) described the short-term influence of Cl⁻ salinity on the anaerobic metabolism of PAOs and GAOs, showing a 71 and 41% decrease, respectively, in the acetate uptake rate when increasing from 0 to 10 g Cl⁻·L⁻¹. While the anaerobic energy requirements increased by 4%, the GAOs seemed to be less influenced by higher NaCl salinity.

The impact of salinity as NaCl has been investigated in great detail, but only a little research is available on the effect of a real salt matrix (e.g. seawater) on PAOs, GAOs and in general the AGS formation processes. In comparison to Pronk et al. (2014), de Graaff et al. (2020a) showed stable PAO activity when cultivating AGS using artificial seawater (salinity of 35 g·L⁻¹), a high enrichment of the PAO ‘Ca. Accumulibacter’ and a complete absence of GAOs. In the research of Ramos et al. (2015) on the other hand, where saline industrial wastewater was mimicked through the addition of a salt mixture, increasing salt concentrations up to 29 g·L⁻¹ were applied to treat aromatics-based wastewater with AGS. An initial acclimation to the salinity and complete biodegradation was observed, but at the highest salt concentrations, a complete deterioration of granules was observed.

While there are a few studies where artificial seawater was used, only a minor amount of research is available on the treatment of real industrial saline wastewaters with AGS, and operated using a feast/famine regime. Carrera et al. (2019) investigated the SBR treatment of fish-canning wastewater (salinity up to 13.45 g NaCl·L⁻¹), where fast and relatively stable granulation was achieved, with GAOs being dominant in the granules. Pilot-scale experiments with saline (up to 14.35 g NaCl·L⁻¹) fish-canning wastewater were examined in follow-up research (Carrera et al. 2021), where granulation and good settleability were achieved after 30 days using an anaerobic feast/aerobic famine strategy.

Although knowledge is available on using feast/famine strategies to treat certain high-salinity complex wastewaters, it is still missing on the use of this strategy for high-saline petrochemical wastewater. Therefore, the present research is aimed at the treatment of high-salinity petrochemical wastewater (average salinity of 35 g salts·L⁻¹) with an anaerobic feast/aerobic famine process in order to promote sludge densification combined with high removal rates of chemical oxygen demand (COD) and nutrients. Laboratory-scale SBR experiments were conducted with analysis of the activated sludge (settling, morphology, biomass) and a subsequent determination of the microbial community through 16S rRNA amplicon sequencing to address the influence of salinity on the growth of storage-capable organisms.

A laboratory-scale SBR with a working volume (before feeding) of 11 L was used in this study. The SBR was equipped with a mixer (Heidolph, Germany) to keep the sludge in suspension, an influent feeding pump (Iwaki, Germany) and a discharge valve. Oxygen was supplied to the mixed liquor by an AquaForte aeration pump, connected to a ceramic disk diffuser. The aeration was done between two setpoints, with a subsequent calculation of the oxygen uptake rates (OUR) within each decrease in oxygen concentration. The oxygen, pH and redox potential (ORP) were constantly monitored through an optical dissolved oxygen sensor (Jumo ecoLine O-DO B202613, Jumo, Germany), pH sensor (Jumo tecLine 201021, Jumo, Germany) and ORP sensor (Jumo tecLine 201026, Jumo, Germany), respectively. Data logging of the sensors was done using an Aquis Touch S system (Jumo, Germany). General communication and operation with/of the reactor system was done using LabView™-software (National Instruments, USA) through a data acquisition card (National Instruments, USA).

The SBR was inoculated with seed sludge originating from a petrochemical company and subsequently fed with the wastewater from the same company. The reactor was operated for 150 days with an SBR cycle of 12 h. A standard cycle consisted of several steps, including an anaerobic feeding step, applying the anaerobic feast/aerobic famine method to induce the growth of slow-growing organisms such as GAO. The anaerobic feeding step was split into a non-mixed and mixed phase, then followed by an extended anaerobic mixing step, an aerobic period with aeration between 1 and 3 mg O₂·L⁻¹, and ultimately settling and effluent discharge steps. Throughout the operational period of the SBR, three periods or stages can be distinguished where changes to the cycle were made. During stage I (days 1–21), 1 L of influent was fed per cycle; during stage II (days 21–73), the influent volume was increased to 1.5 L per cycle; and during stage III (days 73–150) the length of the anaerobic mixing period, which follows the feeding step (1.5 L influent/cycle), was increased by 30 min to promote substrate uptake. The different times of each cycle step can be found in Table 1 for all three stages.

A constant temperature of 25 °C was used throughout the reactor operations. Additionally, during stage III, excess sludge was periodically withdrawn to maintain a constant sludge retention time (SRT) of 50 days. A schematic overview of the SBR cycle and set-up can be found in the Supplementary Information S1.

Nutrient-limited wastewater (COD:N:P = 100:2.35:0.01) from a petrochemical company was used, with different batches (collected at the company at different timepoints) being fed to the reactor during the operational period of 150 days. The average concentration of the wastewater is shown in Table 2, showing the variation of the influent. Concentrations of Cl⁻, and Na⁺ were also determined, being 18, 5 and 10 g·L⁻¹, respectively. Combined with high electrical conductivity (EC) values, the wastewater can therefore be classified in the category of seawater (30–35 g salts·L⁻¹). Nutrients were dosed (as NH₄Cl and K₂HPO₄ solutions) in the wastewater when nutrient requirements for optimal growth were not met (to a COD:N:P ratio of 100:3.5:0.4) based on Hamza et al. (2019) and preliminary research.

The influent and effluent waters were characterized using the following parameters: total COD, soluble COD (sCOD), DOC, -N, -N, -N, total N, -P, pH and electrical conductivity (EC). VFA, chloride (Cl⁻), sulfate (⁠⁠) and sodium (Na⁺) concentrations were only determined for influent. COD and sCOD measurements of influent samples were conducted with a Hanna Instruments (Temse, Belgium) COD Medium Range test kit (HI93754B, accuracy of ± 15 mg O₂·L⁻¹) with sufficient dilution to avoid chloride interference. The COD and sCOD of the effluent samples were measured with a Nanocolor COD 60 (Machery-Nagel, accuracy of ± 3.87 mg O₂·L⁻¹) test kit using a Nanocolor chloride complexing agent (Machery-Nagel) to increase the chloride interference range and to minimize the dilution for lower COD-values. -N, -N, -N and -P were analyzed using Hanna Instruments test kits HI93715, HI93707, HI93766 and HI93717, respectively. Accuracies for the used test kits were ± 0.05 mg N·L⁻¹ for -N, ± 20 μg N·L⁻¹ for -N, ± 1,0 mg N·L⁻¹ for -N, ± 1.0 mg P·L⁻¹ for -P. The DOC was measured with a Sievers InnoVox Laboratory Total Organic Carbon Analyser. The pH and conductivity were measured using a HI99141 portable meter and a HI99301 N portable meter (Hanna Instruments), respectively. VFA, Cl⁻ and were measured with the following test kits; LCK-365 (Hach), HI3815 (Hanna Instruments) and SulfaVer 4 powder pillows (Hach), respectively. Na⁺ concentrations were provided by the petrochemical company. For all the above-mentioned tests, except for the total COD, filtered samples (through 1.2 μm glass microfiber filters) were used.

Activated sludge was characterized through biomass measurements as mixed liquor suspended solids (MLSS) and mixed liquor volatile suspended solids (MLVSS), and sludge settling using the SVI, including a diluted (4×) SVI, shown as dSVI. The sludge morphology was followed through brightfield microscopy using an Olympus CX43 Biological Microscope (Olympus, Japan). (d)SVI, MLSS and MLVSS were measured according to the APHA Standard Methods for Examination of Water and Wastewater (APHA 2017).

Biomass samples were collected weekly, typically at the end of each cycle, then centrifuged with the resulting pellet being stored at −20 °C prior to DNA extraction. DNA was isolated using the FastDNA^® SPIN kit following the van Loosdrecht et al. (2016) protocol. The DNA concentration was then measured using the Qubit dsDNA Assay kit (Invitrogen, USA), following the manufacturer's protocol.

The bacterial 16S rRNA gene hypervariable region, targeting V1–3, was amplified through a polymerase chain reaction (PCR). Barcoded primers were used 27F (5′-AGAGTTTGATCCTGGCTCAG-3′), 534R (5′-ATTACCGCGGCTGCTGG-3′), and KAPA HiFi HotStart PCR kit (Boston, USA). PCR products were purified using the Agent AMPure XP Beads kit, following the manufacturer's protocol, and then pooled into one library. The resulting library had a final equimolar concentration of 4 nM. The sequencing of the amplified libraries was carried out on an Illumina MiSeq platform at the Center for Medical Genetics (Edegem, Belgium), using a MiSeq Reagent kit v (Illumina), following standard guidelines. The reads were clustered using the USEARCH v.11.0.667 (Edgar 2013) toolbox and taxonomically classified with the MiDAS 4 database (Dueholm et al. 2022). Amplicon sequencing data analysis was performed using the ampvis2 package v. 1.24.0.

Figure 1 also shows the food to microorganisms ratio (F/M ratio) over time for the working period of the reactor. The F/M ratio shows the amount of food available (as COD) for the amount of microorganisms (as MLVSS) present, giving a glance at COD loading of the activated sludge. During stage I, an average F/M of 0.18 ± 0.02 gCOD·gVSS⁻¹·d⁻¹ was achieved, with a corresponding organic loading rate (OLR) of 0.41 ± 0.01 kgCOD·m⁻³·d⁻¹. The stable OLR was achieved due to the same influent being used during the first 20 days of operation. With higher influent variability and an increased feeding volume during stage II, the average F/M increased to 0.20 ± 0.03 gCOD·gVSS⁻¹·d⁻¹, with a respective average OLR of 0.50 ± 0.06 kgCOD·m⁻³·d⁻¹. During stage III, with an increase in the length of the anaerobic mixing phase, a much lower average F/M of 0.15 ± 0.02 gCOD·gVSS⁻¹·d⁻¹ was observed with a clear decrease during the period overall (Figure 1). The respective OLR was higher compared to stages I and II, at 0.57 ± 0.08 kgCOD·m⁻³·d⁻¹, with a higher biomass concentration (section 3.2) causing the lower F/M ratio. Overall, the OLR in this study was lower than that used by Carrera et al. (2019) who used saline fish-canning wastewater, with the lowest used OLR at 1.8 kgCOD·m⁻³·d⁻¹. van den Akker et al. (2015) on the other hand, used a more comparable OLR between 0.98 and 1.55 kgCOD·m⁻³·d⁻¹, achieving 98% BOD₅ removal treating saline municipal sewage (5.3–6.1 g NaCl·L⁻¹) with AGS.

Sludge settling (as dSVI₃₀ in Figure 2) started on the higher end, with dSVI₃₀ between 400 and 500 mL·gMLSS⁻¹, being categorized as poorly settling sludge. During stage II, a small increase in the dSVI₃₀ was observed, shifting to a steady decrease starting around day 30, and ultimately reaching 280 mL·g⁻¹ after 70 days of operation. This suggests that the higher feeding volume had an influence on sludge densification. The decreasing SVI trend continued in stage III, where the prolonged anaerobic mixing period further induced the formation of dense sludge flocs with dSVI₃₀ reaching its lowest value of 88 mL·g⁻¹ at day 147. By reaching the threshold of 100 mL·g⁻¹ towards the end of the experiment, the sludge could be classified as well-settling, showing the feasibility of sludge densification by using anaerobic feast/aerobic famine in the presence of high salinity. Looking at the fish-canning industry, Carrera et al. (2019) with a starting SVI₃₀ of 300 mL·g⁻¹, were able to achieve a decrease in SVI₃₀ ending up below 50 mL·g⁻¹ and below 100 mL·g⁻¹ after 150 days using a short anaerobic feeding with aerobic reaction and plug-flow anaerobic feeding, respectively. This is comparable to the settling properties achieved in the present study. Similar results were achieved by Carrera et al. (2021) at a pilot-scale. This clearly indicates that problems caused by salinity can be overcome to achieve good sludge settling, as similar results were also achieved in different studies (Ramos et al. 2015; van den Akker et al. 2015; Ou et al. 2018).

Figure 4(b) shows an overview of two in situ cycle measurements with the changing DOC concentrations for a cycle during stage II and a cycle during stage III (with additional anaerobic mixing time). Anaerobic DOC uptake percentages were 18 (stage II, day 44) and 40 (stage III, day 128) with %VFA being 30 and 36, respectively. Both cycle measurements showed a steady DOC increase during feeding. During the anaerobic mixing periods, both in stage II and stage III, a decrease in DOC was observed, which can mainly be attributed to anaerobic uptake. The additional 30 min mixing time did not show a clear additional decrease in DOC during measurements, although a higher anaerobic uptake was observed. Most of the anaerobic uptake already happened during the (mixed) feeding step. This could be concluded by taking into account the theoretical DOC value that was expected at the end of feeding (Figure 4(b)). As this is mainly through the uptake of VFAs, the DOC decrease during the anaerobic mixing phase was less steep. Since not all DOC was taken up anaerobically, a part of the DOC was still available for filamentous (and other ordinary heterotrophic) organisms to grow during the aerobic phase, probably explaining the presence of filaments even when dense sludge structures were achieved (Figure 3).

Successful densification of activated sludge treating high-salinity petrochemical wastewater could be achieved in this study. High COD removal rates (95%) were achieved, with effluent discharge limits being met. In terms of settling characteristics, well-settling sludge (dSVI₃₀ < 100 mL·g⁻¹) was observed at the end of the operational period with clear densification of sludge flocs observed by microscopy analysis. The presence of the GAO ‘Ca. Competibacter’ was confirmed through 16S rRNA amplicon sequencing, showing the resistance of these organisms to elevated salt concentrations in the bulk liquid. Even though stable densification was achieved in this case, future research should be conducted focusing on pilot-scale testing. This can lead to insights of system stability, as a higher variation of wastewater composition and salt concentrations will likely occur. Investigation of the microbial community in these saline environments should also be optimized in the future. In conclusion, the implementation of an anaerobic feast/aerobic famine strategy provides a useful and robust solution to overcoming filamentous bulking problems, also in high-salinity petrochemical environments.

This research was funded by the Industrial Research Fund (IOF) of the University of Antwerp through a Proof of Concept (POC) project.

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

The authors declare there is no conflict.