Long-term operation of a pilot-scale anaerobic membrane bioreactor (AnMBR) treating 1 high salinity low loaded municipal wastewater in real environment

1 Long term operation of an anaerobic membrane bioreactor (AnMBR) treating municipal 2 wastewater was investigated in a real seawater intrusion spot in Falconara Marittima (Central 3 Italy) on the Adriatic coastline. Changes in biological conversion and system stability were 4 determined with respect to varying organic loading rate (OLR) and high salinity conditions. At 5 an OLR of 1 kgCOD . m 3-1 d -1 , biogas production was around 0.39 ± 0.2 L . d -1 . The increase of 6 the OLR to 2 kgCOD . m 3-1 d -1 resulted in the increase of biogas production to 2.8 ± 1.5 L . d -1 7 (with 33.6% ± 10.5% of CH 4 ) with methanol addition and to 4.11 ± 3.1 L . d -1 (with 29.7% ± 8 11.8% of CH 4 ) with fermented cellulosic sludge addition. COD removal by the AnMBR was 9 83% ± 1% when the effluent COD concentration was below 100 mg O 2 . L -1 . The addition of the 10 fermented sludge affected the membrane operation; significant fouling occurred after long-term 11 filtration, where the trans-membrane pressure (TMP) reached up to 500 mbar. Citric acid 12 solution was applied to remove scalants and the TMP reached the initial value. High saline 13 conditions of 1500 mgCl -. L -1 adversely affected the biogas production without deteriorating the 14 membrane operation. The treated effluent met the EU quality standards of the D.M. 185/2003 15 and the new European Commission Resolution for reuse in agriculture. 16


Introduction
1 Anaerobic treatment in high-rate bioreactors has increased in number of applications during 2 municipal wastewater treatment (MWT) in the last decade while presenting an advanced 3 technology for environmental protection and resource preservation [1,2]. The combination of 4 membrane and an anaerobic bioreactor (Anaerobic membrane bioreactor (AnMBR)) paved the 5 way for a sustainable wastewater treatment with complete biomass retention, and with 6 additional advantages such as less sludge production, high quality effluent and net energy 7 production as the organic matter is converted into high-value products (volatile fatty acids 8 (VFAs)) and energy in the form of biogas [3,4]. These advantages of anaerobic treatment 9 systems result in a decrease in operational costs compared to conventional wastewater treatment 10 plants (WWTPs) that often include aerobic processes (i.e. conventional activated sludge (CAS) 11 or aerobic membrane bioreactor (AeMBR)) [5][6][7].
The application of AnMBRs is coastal regions is also another point that lacks sufficient 1 information in the literature for a successful WWTP operation. In coastal regions, variable 2 salinity of wastewater occurs due to seawater infiltration to sewers or introduction of saline 3 water from industrial processes such as seafood and cheese production [12,17]. In general, the 4 salinity effect in anaerobic processes can cause two main critical operational problems. Firstly, 5 increased salinity results in deterioration of membrane filtration and fouling aspects due also to 6 the decrease of the biomass particle size. For instance, salinity increase from 8 20 gNa + .Lto 7 20 gNa + .L -1 was accompanied by the increase of the transmembrane pressure (TMP) up to 350 8 mbar, while a ten-fold reduction in biomass particle size resulted in a filtration resistance 9 increase [18]. Similarly, Yurtsever et al. [19] indicated that salinity induced large molecules, to 10 be detected as foulants in gel/cake layer; they may originate from biomass loosely bound 11 extracellular polymeric substances (EPS). Secondly, saline conditions can suppress microbial 12 growth and cause the disintegration of flocs and granules that further lead prominent biomass 13 wash-out affecting the sludge granulation [20].
14 The advantages of AnMBRs in MWT are evident with respect to the necessity of low-cost 15 energy technologies in WWTPs; however, long-term operational experiences in coastal regions majority of first generation MBRs was implemented in northern Europe, the presented results 1 may offer options to a critical part of the MBR market. Following the preliminary treatment (screening, degritting and oils removal), pretreated 1 influent from Falconara WWTP was sent to a pilot-scale UASB coupled with an anaerobic 2 ultrafiltration membrane. To evaluate the long-term stability period, the experimental work was 3 conducted for 480 days with real wastewater influent. As shown in Figure 2, different phases 4 were designed by increasing the influent organic loading testing the following configurations: 5 a) urban wastewater, b) co-treatment of urban wastewater and methanol, c) co-treatment of 6 urban wastewater and fermentation liquid.

7
A steady influent flow rate of about 3 L . h -1 of wastewater and the external addition of methanol 8 and fermented sludge were achieved by peristaltic pumps (Watson-Marlow, UK). The first one 9 was able to guarantee a flow rate of 15 L . h -1 , the second and the third was 3 L . h -1 -10 L . h -1 and 10 15 L . h -1 , respectively.

11
The UASB was a cylindrical Plexiglas reactor (16 L) with an internal diameter of 15 cm and a 12 total height of 136 cm. The reactor was divided into two compartments: the first was the real 13 reaction chamber at the bottom (85 cm, 12.4 L), while the second, on the top, was a tri-phase 14 separator (GLS) with 21.9 cm height and was connected to a hydraulic guard which created the 15 appropriate backpressure for the biogas release. The temperature of the UASB reactor was kept 16 constant (30˚C) by applying internal and external windings with hot water at 45˚C. The 17 produced biogas was measured by a milligas counter (Ritter, Germany). The hydraulic retention 18 time (HRT) was maintained at 5-6 h. The up-flow velocity of the UASB reactor was maintained 19 at 1 m . h -1 . The UASB reactor was inoculated with the sludge obtained from a paper mill WWTP 20 in Castelfranco Veneto (Italy).

21
The UASB effluent was collected in a mixed reactor equipped with pH and temperature probes 22 and it was partially recirculated in the UASB reactor (internal recirculation) and partially sent 23 by gravity to the second unit (anaerobic ultrafiltration tank). The performance of anaerobic 24 process was monitored throughout the experimental period with respect to OLR, indicator α, specific methanogenic activity (SMA), upflow speed and biogas production. The UASB was 1 followed by anaerobic hollow-fiber ultrafiltration membrane (PURON® Koch membrane 2 system, with 0.03 µm of nominal pore-size, a total nominal surface 0.5 m 2 , 0.25 m height) 3 installed in a Plexiglas reactor (0.29 m x 0.7 m x 0.39 m) and equipped with level and TMP 4 sensors. Different OLRs (ranging from 1 kgCOD . m 3-1 d -1 to 2 kgCOD . m 3-1 d -1 ) were studied and 7 controlled based on the influent flow. All experimental periods, applied configurations and 8 operational parameters are summarized in Table 2. In Period 1 only raw wastewater was treated 9 by UASB at 30 o C, while the co-treatment of raw wastewater and methanol was tested in Period 10 2. Period 2 was divided in 3 sub-periods: In Period 2.1, mixture of raw wastewater and methanol 11 was treated only by UASB at 30 o C. In Period 2.2, the co-treatment of raw wastewater and 12 methanol was conducted at 30°C UASB + ambient AnMBR; whereas, the same configuration 13 of Period 2.2 was applied in Period 2.3 except that the temperature of AnMBR was also kept at 14 30 o C. In Period 3, the wastewater was co-treated with the supernatant of fermented cellulosic 15 sludge in UASB + AnMBR. During Period 3, the raw wastewater was filtered by dynamic 16 rotating primary unit (SALSNES FS1000) to recover cellulosic sludge [21]. The separated 17 sludge was then sent to anaerobic fermentation reactor (1400 L) operating at 30°C in 18 uncontrolled pH. The fermented flow was dewatered (BABY 2 PIERALISI) and the liquid 19 supernatant was used to increase the OLR of UASB reactor up to 2 kgCOD . m 3-1 d -1 . Finally, 20 during the fermentation liquid co-treatment, NaCl solution was added in the UASB reactor to 21 simulate high saline conditions in Period 4. Chloride concentration was increased gradually 22 from 200 mg . L -1 to 500 mg . L -1 during first 50 days and then to 1500 mg . L -1 after 50 days and 23 the maximum of 2200 mg . L -1 at the end. The main characteristics of the influents are reported 24 in Table 3.
The performance of the system was investigated in terms of organic content removal and biogas 1 production with respect to above-listed configurations (Periods 1-2-3-4). The indicator α, that 2 is defined as the ratio of partial alkalinity over total alkalinity, was measured to verify the 3 stability of the biological process. An upflow velocity of 0.7 m . h -1 to 1 m . h -1 was maintained to 4 keep the sludge blanket in suspension. sparging method adopted in these tests, using nitrogen gas (N2) for 10 seconds off (gas off) and 17 10 seconds on (gas on), with a specific flow rate value of 2 m 3. m 2-1 h -1 [22]. Moreover, gas-18 sparging frequency was studied. The results showed that the increase of the gas sparging 19 frequency increased the percentage of degassing methane; and the degassing methane from gas-20 sparging could be recovered (see E-Supplementary material).

21
For each experimental set, the temperature was measured at the beginning and at the end of the 22 test. The temperature was normalized at 20°C using the Arrhenius equation. For each flux (J) 23 ranging from 6 LMH to 22 LMH, the average TMP (TMPave) and the slope (dTMP/dt) were 24 calculated (E-Supplementary material). The TMP reached up to 0.79 mbar . min -1 at MLSS concentration of 300 mgMLSS . L -1 without gas sparging. Differently, for the same MLSS 1 concentration, the TMP was maintained below 0.1 mbar . min -1 by switching on the gas sparging.
2 Therefore, the gas sparging decreasing the fouling rate independently from the MLSS 3 concentrations and the preliminary tested gas sparging durations were adopted for the long 4 operating periods.

5
Granulometric characterizations of the influent of AnMBR was conducted to characterize the 6 particle size of influent solids. Accumulative volume percent was calculated. Diameters d50 and 7 d90 of particles were found to be 14 µm and 58 µm, respectively. every 45 days to remove organic fouling of the membrane; while NaOCl at a concentration of 10 1000 mg . L -1 or citric acid (C6H8O7) at 1000 mg . L -1 was used to restore the initial permeability 11 of the membrane. Permeability tests were carried out with tap water and TMP values were 12 measured at different steps of permeation flow rate. Following the preliminary tests, the 13 AnMBR reactor was fed with UASB effluent in different Periods ( Table 2) and coupled with 14 UASB as mentioned earlier. Moreover, in each period, anaerobic biomass was sampled from UASB reactor to investigate 1 SMA (data not shown) at different OLRs according to the experimental method reported by 2 Hussain et al. [24]. Acetate (solution of 2 gCOD . L -1 , with a ratio VS/COD of 2) was used and 3 its degree of conversion into methane was normalized considering the volatile solids (VS) and 4 expressed in m 3 CH4 . kgVS -1 d -1 . The CH4 content of the biogas was analyzed by a Brüel and   The main characteristics of the influent for each operational period is given in Table 3. pH of 20 the influent remained almost stable (7.5 -7.8). The total alkalinity of the influent ranged from 21 281 ± 92 mgCaCO3 . L -1 to 526 ± 110 mgCaCO3 . L -1 , with the highest concentration observed in 22 Period 3. EC was 1464 ± 153 ms/cm and 1467 ± 380 ms/cm in Period 1 and Period 2 23 respectively, while the addition the fermentation liquid increased the EC to 1817 ± 350 ms . cm -1 . Clconcentration of the influent was 282 ± 126 mg . L -1 , 304 ± 300 mg . L -1 , 393 ± 377 mg . L -1 1 and 1100 ± 618 mg . L -1 during Periods 1, 2, 3 and 4, respectively. COD concentrations were 207 2 ± 73 mg . L -1 , 388 ± 80 mg . L -1 , 375 ± 148 mg . L -1 and 550±330 mg . L -1 in Periods 1,2,3 and 4,  Castelfranco Veneto, Italy. The temperature of the UASB was increased gradually to 30 o C after 17 10 days of operation period. The reactor operated for 5 months at 30 o C with OLR value of 1.05 18 ± 0.4 kgCOD . m 3-1 d -1 . The flow rate was maintained between at 3.38 ± 0.6 L . h -1 . Meanwhile,  19 was between 0.12 and 0.43, which indicated a stable biological process. The average biogas 20 production was 0.39 ± 0.2 Lbiogas . d -1 . The COD and TSS removal efficiencies were 63% and 21 84%, respectively. In addition, 86% of P and N were released during the start-up period of 22 UASB.

Effect of OLR in the system performance
Following the start-up period, the UASB was first fed with the mixture of municipal wastewater 1 and methanol (as the external C source) and the OLR was increased to 2.1 ± 0.6 kgCOD . m 3-1 d -2 1 (Period 2). The flow rate was maintained at 2.98 ± 0.3 L . h -1 . The variations in OLR and α 3 value throughout the operation period are given in Figure 3a and Figure 3b, respectively. α 4 value remained almost stable in the beginning of Period 2, and then tended to increase up to 5 0.65. while 86% P and 88% N were released. The average biogas production increased to 2.8 ± 6 1.5 Lbiogas . d -1 with 33.6% ± 10.5% of CH4 in Period 2. The COD and TSS removal efficiencies 7 of the UASB was 70% and 48% respectively while the application of the AnMBR increased 8 the average COD and TSS removal to 85%, and > 99.99%. On the other hand, the release of P 9 and N in the UASB was 86% P and 88% N and then slightly decreased to 76% and 83% in the 10 integrated UASB+AnMBR configuration, respectively.

11
In Period 3, α value was between 0.16 and 0.52 with an average of 0.29 ± 0.1. The addition of 12 the fermentation liquid in the influent resulted in a peak in the biogas production; while only 13 9% CH4 was measured indicating excess CO2 production via fermentation (Figure 4). The 14 biogas production started to increase gradually together with the CH4 content of the biogas.

15
Hence, up to 10.25 Lbiogas . d -1 was generated with 51.9% of CH4 (average of 4.11 ± 3.1 16 Lbiogas . d -1 with 29.7% ± 11.8% of CH4). The addition of fermentation liquid as the external 17 carbon source increased the biogas production without affecting the overall CH4 content of the 18 biogas. In Period 3, the COD removal efficiencies were 42% and 83% in UASB and 19 UASB+AnMBR, respectively. The TSS removal efficiencies in UASB and UASB+AnMBR 20 were 38% and 100%, respectively, while P and N releases were 85% and 75%, respectively.

21
The application of the AnMBR in the study of Gouveia et al. [2] for the treatment of municipal wastewater under psychrophilic conditions and loading rate of 2 and 2.5 kgtCOD . m 3-1 d -1 23 resulted in effluent tCOD concentrations of 100 mg . L -1 -120 mg . L -1 . In another study by Wei 24 et al. [28], a wide range of volumetric OLR (0.8-10 gCOD . L -1 d -1 ) was tested in AnMBR to treat synthetic municipal wastewater. The results showed that at steady conditions, 98% COD 1 removal was achieved while the application of high sludge OLR led to high methane production 2 of over 300 mL . gCOD -1 . Wijekoon et al. [29] tested the performance of a thermophilic AnMBR 3 at different OLRs ranging from 5 12 kgCOD . m 3-1 d -1 to 12 kgCOD . m 3-1 d -1 . The authors reported 4 an average biogas production of 15 L . d -1 , 20 L . d -1 and 35 L . d -1 at OLRs of 5.1 ± 0.1 kgCOD . m 3-5 1 d -1 , 8.1 ± 0.3 kgCOD . m 3-1 d -1 and 12.0 ± 0.2 kgCOD . m 3-1 d -1 , respectively, with CH4 content of 6 about 55%-65%. In addition, the reactor showed optimum COD removal efficiencies at 8 ± 0.3 7 kgCOD . m 3-1 d -1 OLR. In a recent study, the highest VFA yield (48.20 ± 1.21 mgVFA . 10 8 mgCODfeed) was observed at OLR of 550 mgCOD . L -1 ; however, the authors achieved less VFA 9 yield at the examined maximum OLR (715 mgCOD . L -1 ), indicating that elevated OLRs can 10 lead to high VFA production but it is also crucial to optimize operating OLR during the 11 treatment of low strength wastewater in AnMBR [30]. In high-rate bioreactors such as UASB 12 and AnMBR, VFAs may not be efficiently converted to methane due to low-retention times 13 and can accumulate in the reactor and thus can be detected in the effluent [1,31]. In operating 14 conditions at elevated OLRs, VFAs should be therefore monitored to meet the local standards 15 for discharge or reuse. the reported values under low saline conditions. The CH4 content of the biogas was adversely affected (10%-20% in the beginning of Period 4 and 5% at the end) by high saline conditions.

1
The system almost failed to operate at the maximum examined Clconcentration (app. 2200 2 mg . L -1 ), since the biogas production was 0.08 L . d -1 with 3% CH4. is required with high biomass concentrations for the system to regain its stability.

Effect of OLR and high salinity on membrane operation
The variations in the TMP value is shown in Figure 6a. The red points mark the membrane 1 cleaning days. The TMP of the membrane was stable at around 50 mbar when the system was 2 operated with methanol addition (Period 2, OLR of 2 kgCOD . m 3-1 d -1 ), with gas-sparging 3 condition of 10 seconds on and 120 seconds off. The specific flux normalized at 20°C was 175 4 L . h -1 m 2-1 bar -1 and only NaOCl cleaning was necessary after 50 days of operation. (Figure 6b) 5 During the fermentation liquid addition (Period 3, OLR of 2 kgCOD . m 3-1 d -1 ) the behavior was 6 first similar to the methanol co-treatment; then the TMP increased gradually after 50 days of 7 operation and reached to 500 mbar after 100 days of operation. Thus, a more intense chemical 8 cleaning was applied to restore the initial permeability (citric acid at 1000 mg . L -1 ) on day 315. 9 A significant EPS production was observed when the fermentation liquid was used, leading to 10 larger formation of the "cake" on the surface of the membrane. This was mainly due to the 11 fluctuation of the characteristics of the production of fermentation liquid. There was an increase 12 in EPS concentration from 52.8 mgEPS . L -1 in the Period 2 to 70.8 mgEPS . L -1 in the Period 3.

13
The latter decreased the membrane permeability a short time, higher rate pore obstruction and 14 therefore more intense and frequent cleaning was required.

15
In Period 4, the TMP remained stable at 12 mbar when the reactors were fed with the 16 fermentation liquid together with additional NaCl to increase salinity in the system, at the 17 concentration of 500 mgCl -. L -1 . The average TMP was around 50 mbar, at Clconcentration of 18 1500 mgCl -. L -1 , caused by the increased filtration resistance following the increased TMP due 19 to the increased EPS concentration in Period 3. The saline conditions therefore only affected 20 the initial TMP value and remained constant during the system operation.

21
The TMP of AnMBRs is highly dependent on the critical flux and the sparging rate together 22 with other environmental and operating conditions [35]. Furthermore, an initial flux below the 23 critical flux, prior to the introduction of peak flow, is reported to be advantageous to 24 permeability recovery [36]. In the study of Muñoz Sierra et al. [18], increased salt concentration was found to affect the TMP negatively (350 mbar at a flux of 4.0 Lm 2. h -1 ). The deterioration 1 of membrane filtration performance was attributed to the decrease of biomass particle size when 2 salinity was increased. The small particle size had a significant influence on the cake layer 3 compaction that increased the operational values of the filtration resistance. Furthermore, 4 higher stability of process performances of AnMBR over UASB was reported to overcome high 5 salinity [34]. Elevated TMP values were also reported by Yurtsever et al. [19] with respect to 6 high salinity conditions. The salinity induced large molecules as foulants in gel/cake layer, that 7 may originate from biomass loosely bound EPS. The EPS properties are highly dependent on 8 the operating OLR [37]; and the OLR increase is often accompanied with high EPS production. 9 3.6. Relationship between process parameters and system performance 10 PCA was carried out to reveal the relationships between the applied operating conditions (i.e.

11
OLR, salinity) and system performance in terms of biogas production, CH4 content of the 12 biogas, α value, TMP and Js. The PCA supported our previous discussion regarding the 13 performance of the UASB+AnMBR system at different periods. OLR was closely grouped 14 together with the biogas production, CH4 content and α value (Figure 7). This cluster showed 15 the close relationship between these parameters mostly in Period 1 which included only the 16 UASB operation. The data-points of Period 2 was comparatively more equally distributed 17 between the parameters; where the negative impacts of OLR on TMP were clearly reflected in 18 our data especially in Period 3. Furthermore, data points of Period 4 were characterized by high 19 chloride concentrations. The displayed negative correlation between the salinity and TMP 20 and/or Js in Period 4 was due the citric acid cleaning of the membrane following the fouling 21 occurred in Period 3 that was previously mentioned. TMP value was close to its initial value 22 right after the citric acid cleaning and although TMP also slightly increased Period 4 at high  The obtained effluent quality was compared against the values reported in Table 4 (minimum 20 requirements for reuse). Analyzing the results obtained from this study, the UASB effluent is

11
Regarding the change of paradigm in the context of circular economy in terms of water, 12 nutrients and energy; site-specific optimization of an AnMBR is crucial especially in coastal 13 regions. In this particular study, the addition of fermented cellulosic sludge to raw wastewater 14 increased the biogas production without affecting the overall CH4 content of the biogas. In case 15 of the application of the process in coastal areas, the biogas production decreased by 27% due 16 to the saline conditions when Cl concentration was up to 1500 mg . L -1 . Moreover, the CH4 17 content of the biogas was also adversely related to the high saline conditions up to almost null 18 value for chloride higher than 2000 mg . L -1 . At Clconcentration less than 1500 mgCl . L -1 , long 19 term adaptation of microbial community (i.e. halotolerant or even halophilic microorganisms) 20 may be required with high biomass concentrations for the system to regain its stability and 21 recover the bioreactor performance in UASB+SuMBR. Concerning the membrane operation,  Validated innovative approaches for energy-efficient resource recovery and re-use        NOTES: 1) guide values, regions may authorize different limits, 2) up to 10 mg P/L and up to 35 mg N/L for irrigation use in non-vulnerable areas, 3) guide values, but 4000 is the maximum permissible, 4) for 80% of samples, 100 UFC/ml 100 is the maximum value and for water impoundment or phytoremediation apply the limits of 50 (80% of the samples) and 200 UFC /ml 100 (punctual maximum value), 5) for 100% of samples, 6) limit value not provided. Tel: +39 071 2204911 Fax: +39 071 2204729