- Email: [email protected]
Anaerobic dynamic membrane bioreactor for wastewater treatment at ambient temperature
Accepted Manuscript Anaerobic dynamic membrane bioreactor for wastewater treatment at ambient temperature Luca Alibardi, Nicoletta Bernava, Raffaello Cossu, Alessandro Spagni PII: DOI: Reference:
S1385-8947(15)01194-8 http://dx.doi.org/10.1016/j.cej.2015.08.111 CEJ 14104
To appear in:
Chemical Engineering Journal
Received Date: Revised Date: Accepted Date:
26 June 2015 25 August 2015 26 August 2015
Please cite this article as: L. Alibardi, N. Bernava, R. Cossu, A. Spagni, Anaerobic dynamic membrane bioreactor for wastewater treatment at ambient temperature, Chemical Engineering Journal (2015), doi: http://dx.doi.org/ 10.1016/j.cej.2015.08.111
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Anaerobic dynamic membrane bioreactor for wastewater treatment at ambient temperature
Luca Alibardi a, Nicoletta Bernava b, Raffaello Cossu a, Alessandro Spagni c,*
a
Department of Industrial Engineering, University of Padova, via Marzolo 9, 35131
Padova, Italy b
Department of Civil, Environmental and Architectural Engineering, University of
Padova, via Marzolo 9, 35131 Padova, Italy c
Water Resources Management Laboratory, Italian National Agency for New Technology,
Energy and Sustainable Economic Development (ENEA), via M.M. Sole 4, 40129 Bologna, Italy
* Corresponding author: Alessandro Spagni, ENEA, via M.M. Sole 4, 40129 Bologna, Italy. Email: [email protected] Phone: +39 051 6098779; fax: +39 051 6098309
Abstract A bench-scale dynamic membrane (DM) bioreactor was operated to evaluate the anaerobic treatment of a synthetic municipal wastewater at ambient temperature. The DM was developed over a large pore size (200 µm) mesh in order to improve sludge filterability and reduce energy consumption. The system achieved average organic removals higher than 80 and 90% for total COD and filtered COD, respectively. Results also demonstrated that the biofilm, which the DM is made of, played a significant part in obtaining the overall
1
organic removal efficiency. The large pore size of the mesh allowed for high membrane fluxes (approximately 15-20 L m-2 h-1) applying low TMP (usually lower than 50-100 mbar). Fluxes higher than 20 L m-2 h-1 produced low solid removal efficiency indicating deterioration of the DM. COD mass balance suggests that the low hydraulic retention times applied to the system caused methane loss through the effluent due to oversaturation.
Keywords: Ambient temperature; Anaerobic processes; Dynamic membrane; Membrane bioreactor; Mesh filtration;.
1. Introduction Municipal wastewaters are, currently, mainly treated by the use of activated sludge systems which, although effective, require a great deal of energy [1]. As a result, anaerobic technologies have been widely investigated for the treatment of municipal wastewater. Their benefits are the production of biogas as a renewable energy source and reduced energy consumption if compared to conventional aerobic treatment [1-4]. The advantages of anaerobic treatments will also be emphasised in future water and wastewater management scenarios [5]. Anaerobic processes can in fact improve the recovery of energy, materials and water from concentrated and diversified wastewater streams both in centralised and decentralised systems [5]. Mesophilic conditions are the preferred option for anaerobic wastewater treatment. These conditions, however, make anaerobic processes of medium to low strength streams such as municipal wastewater non cost-effective. The low energy recovery per unit of volume is not in fact sufficient to satisfy the heat and power requirements of reactors [1, 6]. Solutions allowing for the application of anaerobic processes at ambient or psychrophilic temperatures for low strength wastewater are, therefore, very attractive [6, 7].
2
However, several aspects related to the application of psychrophilic conditions are still under investigation. For example psychrophilic conditions do not only reduce microbial kinetic rates but also increase gas solubility, thus leading to lower methane recovery and methane losses with the effluent [1, 8]. Anaerobic processes under psychrophilic conditions have been successfully applied to municipal wastewater treatment by high rate systems such as the up-flow anaerobic sludge blanket (UASB) and expanded granular sludge bed (EGSB) reactors [6, 7]. However, serious problems arise when treating low strength wastewater under psychrophilic conditions using these technologies. In fact, low biomass growth rates at low temperatures and high hydraulic loading rates increase sludge wash out. In addition, applications of UASB and EGSB reactors for municipal wastewater treatment are limited by the influent high concentrations of suspended solids and inert substances [6]. Over the last decade, several studies have investigated the application of membrane bioreactors (MBRs) in anaerobic conditions [2, 9-12]. Microfiltration (MF) and ultrafiltration (UF) membranes, in fact, allow complete biomass retention so that a high concentration of slow-growing anaerobic bacteria in the reactor can be reached, even at low hydraulic retention time (HRT) [2, 9, 12]. Only a few studies have evaluated the performance of anaerobic MBR (AMBR) at psychrophilic or ambient temperature, although, very promising results have been obtained [2, 12, 13]. The first studies on sludge filterability in AMBRs [12, 14] have showed, although not conclusively [10], more severe fouling phenomena under anaerobic rather than aerobic conditions. Consequently, mesh filtration has recently been proposed as an alternative to the use of MF/UF-MBRs in order to improve sludge filterability and reduce capital and management costs [15-19].
3
When mesh filtration is applied, a dynamic membrane (DM) develops on the support material (the mesh). DM is a cake layer or biofilm obtained through the deposition over the mesh of organic substances and bacteria present in the reactor. Once DM is formed, solids rejection is carried out by this regenerative biological layer while the mesh only acts as a support [17, 20]. The layer can be formed and re-formed as a self-forming DM and the permeability can be affected by controlling its thickness while the support (mesh) can be characterised by large pores. The cake layer that develops on the mesh plays, thus, a central role during DM filtration [17] while it is considered the main drawback for the widely adopted conventional MF/UF-MBRs. DMs cannot, it seems, achieve the high water quality obtained by MF and UF due to the different membrane cut-off. However, DMs could represent a worthwhile compromise between water quality and plant costs in anaerobic processes, since effluent posttreatments are usually considered prior to final water discharge (e.g. for nutrient removal). DM technology can offer benefits over traditional filtration in biological treatments by precluding the need of costly membrane modules and by providing a low-cost, regenerative, self-forming filtration surface with adaptable permeability and high fluxes [16, 17]. The link between mesh pore size and dynamic layer formation is still not clear. The use of large pore size mesh can reduce the overall filtration resistance and the cost of the filtration module; however, the maximum size allowing the development of a reliable dynamic filtration layer has not been defined yet. Experiments on DM development have been carried out with mesh opening between 30 and 90 μm [20-23]. Preliminary studies indicated that an effective cake layer does not develop over meshes with pore sizes larger than 60-70µm [23]. On the contrary, Alibardi et al. [15] developed a DM on a large poresized mesh (200 μm) in mesophilic conditions by properly managing hydrodynamic
4
conditions. Moreover, Kiso et al. [24] suggested that DMs not only act as filters but may also improve overall pollutant removal efficiency through the biochemical reactions that occur within the cake layer. This study aimed to evaluate the application of an anaerobic dynamic membrane bioreactor (ADMBR) equipped with a coarse filtration mesh (200 m) for the treatment of highstrength municipal wastewater at ambient temperature. A large pore size mesh was used in order to improve filterability and reduce energy consumption. The study also aimed to evaluate the contribution of the DM biofilm on the overall organic removal efficiency of the reactor.
2. Material and methods 2.1. Experimental setup The study was performed by using a bench-scale ADMBR equipped with an external cross-flow filtration unit (Fig. 1). The reactor had a total volume of 898 mL (W x H x D: 9.5 x 10.5 x 9 cm) and a working volume of 684 mL. The level of the mixed liquor was kept constant by using a level sensor connected to the influent pump. The filtration support had an inner volume of 60 mL (W x H x D: 20 x 1.5 x 2 cm). A monofilament woven mesh made of polyamide/nylon (SaatiMil PA 7, Saati s.p.a., Italy) with openings of 200 µm, thread diameter of 120 µm, mesh count of 31/cm and 39% opening area (data from the supplier) was inserted into the central longitudinal part of a filtration support. The filtration area was 40 cm2 (L x W: 20 x 2 cm). The overall working volume composed by reactor, membrane support and pipe connections, was 745 mL.
5
The feed was provided by a peristaltic pump (Watson Marlow 401U/D1) controlled by the level sensor inside the reactor. Sludge mixing was provided by a magnetic stirrer (Heidolph, Hei-Standard). The circulation of the mixed liquor along the mesh surface and the return of the retentate into the reactor were provided by a second peristaltic pump (Watson Marlow 505S) applying a cross flow velocity of about 10 m h-1. A third peristaltic pump (Watson Marlow 401U/DM3) was installed to extract the effluent out of the system and to provide the necessary transmembrane pressure (TMP) for filtration. An airtight vessel of approximately 100 mL was installed down-flow from the third pump to assess the presence of oversaturated biogas in the effluent. The effluent flowed out this vessel into a collection tank. TMP was measured by means of two U- tube pressure gauges filled with water and placed up- and down- flow from the filtration support, respectively. Biogas production was monitored by using three homemade wet-tip gas meters connected to the treatment system. The first gas meter was connected to the main vessel while the second to the mesh support in order to estimate the biogas production by the reactor and by the DM biofilm, respectively. The third gas meter was connected to the effluent collection vessel in an attempt to collect (at least) part of the biogas that could escape with the effluent. Effluent volume, TMP and biogas production were recorded on a daily basis from Monday to Saturday, except during the holiday closure period (from day 38 to day 53). Mixed liquor samples were collected from reactor once to twice per week (to avoid excessive sludge removal) and analysed for the following parameters: total suspended solids (TSS), volatile suspended solids (VSS), total chemical oxygen demand (CODt) and filtrable COD (CODf). Effluent samples were collected three to five times per week and analysed for the following parameters: TSS, VSS, CODt, CODf.
6
2.2. Inoculum The reactor was inoculated with anaerobic sludge taken from a full-scale mesophilic anaerobic digester treating the excess sludge of a municipal wastewater treatment plant located in Padova, Italy. The inoculum had a concentration of total (TSS) and volatile suspended solids (VSS) of 12 gTSS L-1 and 6 gVSS L-1, respectively. Sludge was added on three further occasions during the experiment. Sludge had to be topped up to compensate for biomass loss caused by cleaning of the filtration support due to clogging and by other maintenance procedures.
2.3. Feed The reactor was fed with synthetic wastewater, which simulated high strength municipal wastewater. The feed had a concentration of 900 mgCOD L-1 composed of sucrose (45% of total COD), powder milk (10%) and corn starch (45%) as organic matter. The followings chemicals were added to ensure alkalinity, macro- and micronutrients: NaHCO3 (830 mg L-1feed), NH4Cl (50 mgN L-1feed), KH2PO4 (10 mgP L-1 feed), FeCl3·6H2O (2.1 mgFe L-1 feed), CaCl2·2H2O (8.2 mgCa L-1 feed), MgCl2·6H2O (2.4 mgMg L-1 feed), Na2MoO4·2H2O (0.22 mgMo L-1feed), ZnSO4·7H2O (0.23 mgZn L-1 feed), CuSO4·5H2O (0.128 mgCu L-1 feed), NiCl2·6H2O (0.1 mgNi L-1 feed), H3BO4 (0.007 mgB L-1feed). All chemicals were dissolved in tap water.
2.4. Operating parameters The feed concentration and the filtration area were kept constant during the entire study. The HRT was intentionally varied between 5.7 and 0.25 d to evaluate the performance of the systems under different operating conditions. Consequently, the organic load rate
7
(OLR) and the flux applied to the membrane varied in the range 0.16-3.3 kgCOD m-3 d-1 and 1.4-28 L m-2 h-1, respectively. Any condition was maintained for a period of time longer than 3 HRTs to observe steady state operation. Since the temperature was not controlled, the system worked at ambient temperature (in the laboratory) in the range of 20 – 24 °C.
2.5. COD mass balance COD mass balance of the system was performed for three phases characterised by the following hydraulic retention times (HRTs): 2 d (experimental days 63 – 83), 1 d (days 84 – 97), 0.5 d (days 98 - 121). The following equation was used:
CODin ·Qin CODeff ·Qeff CODCH4(g)·QCH4 CODCH4(d)·Qeff CODW ·QW CODML (1)
Where: - CODin, CODeff, and CODw are the average COD concentration of the influent, effluent and waste (i.e. the sludge extracted from the reactor for sampling) streams, respectively; - CODCH4(g) is the average COD of the methane in the biogas; - CODCH4(d) is the average COD of the methane dissolved in the effluent stream; - ΔCODML is the variation of the COD due to the change of the mixed liquor suspended solids concentration; - Qin, Qeff, Qw are the influent, effluent, waste (i.e. the volume of the mixed liquor sampled for analysis) flow; - QCH4 is the flow of the methane production in the biogas.
8
The COD of the mixed liquor volatile suspended solids was estimated assuming the ratio of 1.42 gCOD gVSS-1. CODCH4(g) was computed by considering the daily biogas production, the methane concentration in the biogas, the molar volume of the gas (according to the experimental temperature) and the COD equivalence of 64 gCOD molCH4-1 [25]. The methane dissolved in the liquid phase was estimated using Henry’s law assuming equilibrium with the gas phase and using a Henry’s constant (KH) value at 25°C of 0.0014 Mliq bargas-1 (i.e. molarity in the liquid phase over partial pressure, P, in the gas phase) [25]:
(2)
The values of KH at different temperatures (KH(T)) were estimated according to the following equation [25]:
(3)
where θ is the temperature dependence constant assumed equal to -0.01929 T-1 [25]. The amount of dissolved COD (methane and soluble organic matter) lost via the sludge taken for sampling (Qw) was neglected since the volume sampled was very low (approximately 5-10 mL twice or three times per week).
2.6. Short-term experiment for evaluating COD removal efficiency by the DM A short-term experiment was performed at the end of the study to evaluate the contribution of the well-formed DM on overall COD removal efficiency. The experiment was carried out by means of the same setup described in Section 2.1.
9
The reactor was emptied of the mixed liquor and filled with the feed solution. During the first 72 hours the HRT was set to approximately 1.5 d. Thereafter, HRT was reduced to 0.6 d and kept constant until the end of the experiment. The entire experiment lasted for almost 90 h.
2.7. Analytical methods Total chemical oxygen demand (CODt), filtered COD (CODf), TSS and VSS concentrations were measured according to Standard Methods [26]. pH was measured using a Crison instrument (model micropH 2001) and probe (model 5011). Sample filtration for TSS and VSS analysis was performed using Whatman GF/C filters. Sample filtration for CODf analysis was performed using 0.20 µm syringe filters. Biogas composition was analysed using a micro-GC (Varian 490-GC) equipped with a 10meter MS5A column (to analyse CH4) and a 10-meter PPU column (to analyse CO2) and two Thermal Conductivity Detectors (TCDs). Argon was used as carrier gas at a pressure of 60 kPa in columns. Injector and column temperatures were both set to 80 °C. All analyses were performed in triplicate and data reported are average values. Measurements variability resulted within analytical error reported by Standard Methods [26].
3. Results and discussion 3.1. Transmembrane Pressure and Fluxes The system was started up with a HRT of approximately 5 d in order to enhance biomass acclimation and development of the DM on the mesh. On day 15, the HRT was halved to approximately 2.5 d for three weeks until day 38 when the HRT was returned to values of about 5 d due to the summer holidays. Thereafter, the HRT was stepwise reduced to the
10
final value of 0.27 d over a period of approximately two months (Fig. 2) in order to assess system performance. Owing to the HRT variations, the fluxes applied across the DM resulted in the range of between 1.4 and 28 L m-2 h-2 (Fig. 2). During the first three months of the ADMBR operation, TMP experienced a fluctuating up-and-down trend with values, anyway, constantly below 50 mbar (Fig. 2). The up-anddown trend may have been due to the slow development of the DM. TMP started to steadily increase when the HRT was decreased to 0.5 d, corresponding to membrane fluxes of approximately 15 L m-2 h-1. TMP stabilised at 110-120 mbar in two weeks and then it increased to almost 200 mbar when the membrane flux was set to 28 L m-2 h-2 by reducing HRT to 0.25 d (Fig. 2). The results show that during the first three months a stable DM was not formed as proportionality between the applied membrane flux and the measured TMP (according to Darcy’s law) was not observed. At fluxes lower than 10 L m-2 h-1 the step by step increase in the membrane flux caused an immediate decrease in pressure (see for instance on days 15, 65 and 85 in Fig. 2). Thereafter, the cake layer showed a rapid regrowth process indicated by a fast increase of TMP. This fast increase is due to a high level of deposited solids produced by increased fluxes and because of the presence of a previously developed fouling layer. The rough proportionality between the flux and TMP observed after approximately three months of operation is in accordance with Darcy’s law and confirms the development of a stable DM. The evolution of TMP during the experiment agrees with the hypothesis proposed by Zhang et al. [20] and observed by Alibardi et al. [15] of the three stages DM development. The initial instability of TMP can be explained as a consequence of the process of formation of the separation layer. Then, particulate deposition and biofilm growth
11
produced the formation of the DM (DM growth stage) and the flux increase led to the development of a thick cake layer (fouling stage). The TMP values observed in the present study were, however, somewhat different from those previously reported by Alibardi et al. [15] which used the same setup and mesh porosity but applying mesophilic conditions and an OLR that five times higher. In fact, the higher fluxes and lower OLRs applied in the present study than those of Alibardi et al. [15] produced much lower TMP values than those measured in the previous study. These results, thus, highlight the dynamic state of the filtration layer and suggest that operating conditions greatly affect DM formation. The small OLR applied in this study, even at low HRTs, slowed down the formation of the DM, maintaining a limited filtration resistance for a longer time than the previous study [15]. Wastewater characteristics, reactor configuration and conditions applied to the filtration module, thus, influence the initial formation, consolidation and filtration resistance of a DM. It is important to state that the use of a support mesh characterised by large pore opening (200 µm) allowed the development of a DM filtration layer characterised by stable fluxes of 15 L m-2 h-1 at TMP values lower than 100 mbar. These fluxes are comparable to those applied at full scale aerobic submerged MBRs while those obtained by anaerobic MBRs are much lower. For example, Spagni et al. [14] obtained sustainable fluxes between 2 and 5 L m-2 h-1 operating a submerged anaerobic MBR equipped with a conventional flat sheet membrane module characterised by a nominal pore size was of 0.4 µm. Ho and Sung [27] obtained fluxes from 5 to 10 L m-2 h-1 in an anaerobic MBR equipped with an external non-woven filter with pore size of 10 µm and operated at temperature and OLR values similar to those of this study. Shin et al. [3] and Smith et al. [2] recently obtained fluxes below 10 L m-2 h-1 during municipal wastewater treatment under psychrophilic conditions by using AMBRs equipped with hollow fibre polyvinylidene fluoride membranes with
12
pore size of 0.03 μm and flat-sheet microfiltration polyethersulfone membranes with a pore size of 0.2 µm, respectively. A very low flux of 2.6 L m-2 h-1 was also reported by Ersahin et al., [28] when operation of a laboratory scale ADMBR treating high-strength synthetic wastewater by using as support material a monofilament woven fabric made of polypropylene material with an average pore size of 10 µm. Zhang et al. [20], on the contrary, obtained high fluxes using a small-size Dacron mesh (61 µm). The mesh was however installed in the top part of a UASB reactor where the solids load on the filtration unit was limited. The use of a large pore size mesh support also allowed fluxes to be maintained at a level comparable to those reported in literature for DM but without the need for any cleaning procedure. This observation can indicate a potential advantage on the use of a large pore size mesh since membrane cleaning procedures are often used also for DM [20, 22, 23, 28].
3.2. Wastewater treatment The concentrations of CODt and CODf in the effluent remained below 100 mg L-1 during the first month of operation at 2 d HRT, resulting in a COD removal efficiency of over 90% (Fig. 3a). In the second month effluent COD concentrations slightly increased even though HRT was raised to 4 days. After two months of operation, COD removal increased again when HRT was reduced to 2 days and then further decreased to approximately 1 d. These results suggest that the DM contributes to the organic matter degradation as better COD removals were obtained at lower HRTs and consequently increased membrane fluxes (Fig. 2). The increasing organic load on the filtration membrane applied by lowering HRTs can stimulate and support the growth of a thicker biofilm, capable not only of solids rejection but also of organic biodegradation (Fig. 3b).
13
When the HRT was reduced to 0.5 d on day 98, the effluent COD began to gradually increase to nearly 200 mg L-1 (day 101) and also CODf concentration increased from values lower than 10 mg L-1 to about 100 mg L-1. This increase lasted for a few days, after which CODt removal returned to values close to 90%. Further HRT reduction to 0.25 d (Fig. 2), contributed to the decrease in COD removal (Fig. 3). These results highlight the process limits for the bench scale ADMBR used in this study. Low HRTs (0.5 or 0.25 d) and the resulting high fluxes applied to the dynamic membrane produced a great accumulation of solids in the filtration layer causing the TMP increase. The ambient temperature and the high wastewater flow through the system negatively affected the anaerobic digestion process. Considering the system during the first 122 d of the experimental run (i.e. before the deterioration of treatment processes), the reactor achieved the average removal efficiencies of 87 ± 4 % and 94 ± 4 % (mean ± standard deviation) for total and filtered COD, respectively. These performances are comparable with those of conventional AMBRs, whose COD removal efficiency usually ranges between 75% and 99% [16]. If data are compared with other studies on ADMBR, COD removal obtained in this study is generally higher. Alibardi et al. [15] operated an identical ADMBR at mesophilic temperatures and HRTs between 2 and 4 days for treatment of synthetic industrial wastewater. COD removals ranged from 65% to 92% (average of 75%) from an influent COD of 5 g L-1. Also Ma et al. [22] reported an average COD removal of 79% during treatment of real municipal wastewater characterised by a COD concentration that varied between 300 and 800 mg L-1. Ersahin et al. [28] achieved higher COD removals (99%) but during the treatment of high-strength synthetic industrial wastewater (COD > 20 g L-1) at 35°C and using a support mesh with average pore size of 10 µm. Results demonstrate that anaerobic processes coupled with DMs developed over large pore size
14
mesh can consistently achieve high treatment efficiencies at ambient temperature in a perspective of lower capital and management costs. The reactor was started up with approximately 5 and 3.5 g L-1 of TSS and VSS, respectively. During the first month, solids rejection by the DM was very efficient producing an effluent TSS concentration below 25 mg L-1 (Fig. 4a). However, after approximately three weeks of operation, the membrane support was opened due to clogging problems that occurred in the pipes. Due to the partial loss of the mixed liquor (ML) sludge a small amount of sludge was added, and this justifies the mixed liquor TSS (MLTSS) and VSS (MLVSS) peaks of 9.5 and 5.3 g L-1 observed on day 29 (Fig. 4b). During the second month, fluctuations in effluent solids content were noticed as decreasing concentration of MLTSS in the reactor (Fig. 4). Even though high effluent TSS concentrations were measured, the solids lost in the exiting flow does not justify the significant decrease in the mixed liquor suspended solids (MLSS) (Fig. 4b) in the reactor during the second month. The addition of new inoculum performed on day 58 produced only a temporary increase of biomass concentration in the reactor as a rapid decrease of MLSS concentration was observed immediately after. The simultaneous low TSS content of the effluent suggests that MLSS were accumulated on the mesh forming a thick cake layer. The TSS effluent concentration was for most of the study below 100 mg L -1 with only some exceptions. This value is comparable to effluent TSS concentrations reported for other high rate anaerobic treatment processes [6]. This result demonstrates that DM supported by large pore size mesh can obtain efficient solid rejections at ambient temperatures. Filtration efficiency, therefore, is not related to the pore size of the support mesh but it only depends by the formation and consolidation state of the dynamic cake layer formed over the mesh.
15
After day 110, with the further decrease of HRT to 0.5 and then to 0.27 d, the effluent TSS and VSS concentrations increased (Fig. 4a) causing almost complete depletion of the MLSS (Fig. 4b). A third biomass addition was not able to maintain the high MLSS concentration. These results indicate that such low HRTs cannot sustain sufficient biomass growth rate at ambient temperature when treating municipal wastewater. Moreover, the high membrane fluxes (15-28 L m-2 h-1) created by the low HRTs, caused deterioration in the structure of the DM allowing solids to escape into the effluent. Both phenomena produced the observed reduction in COD removal (Fig. 3). However, although MLTSS was lower than 0.8 g L-1 and relatively low HRTs were applied (0.27 - 0.5 d), COD removal remained higher than 50% and 60% for CODt and CODf, respectively (Fig. 3). This indicates that deterioration of the DM was only partial as it was still able to operate partial solids retention and also consistent COD removal. This strongly supports the hypothesis that the biofilm of the cake layer forming the DM not only retains solids but also consistently concurs in the COD removal processes [24]. The SRT was estimated as the ratio between the biomass in the reactor and the VSS lost through both the effluent and the sampled mixed liquor per day. SRT remained above 110 d during system operating HRT of 2 d or higher while it gradually decreased to approximately 40 d during operation at 1 d of HRT. When HRT was reduced to 0.5 d or lower, SRT dropped to about 5 d. This value is not sufficient to support the growth of methanogenic biomass as confirmed by the reduction of COD removal observed at the end of the experiment (Fig. 3).
3.3. Biogas production Low biogas production was observed (< 50 mL d-1) during the first two months of operation (Fig. 5a) due to both the low OLR applied to the system and the characteristics
16
of the mesophilic anaerobic sludge used as seed which was not previously acclimated to operate at ambient temperature. The decrease of the HRT to 2 d after two months of operation led to a continuous increase of the biogas production rate from approximately 25 to 175 mL d-1. The increase of biogas production can be attributed to the higher OLRs and to acclimation of the biomass to ambient temperatures. During the first two months, the vast majority of biogas was emitted from the reactor, which, therefore, accounted for nearly 100% of the total volume measured (Fig. 5b). After two months of operation, biogas began to be emitted from the membrane support too, probably due to the further decrease of HRT, higher OLR and the stable development of the DM. The volume of biogas emitted from the membrane support reached up to 60% of total biogas production from the second month until the end of the experiment. The smaller volume and specific surface area of the membrane module, if compared to the reactor, and the significant biogas production measured from the membrane, indicate the high biological activity of the biofilm forming the DM. This result is also confirmed by the fact that total biogas production was only slightly affected by the decreasing trend of MLSS concentration inside the reactor, thus, supporting the idea, once more, that the biofilm forming the DM could substantially contribute to the biochemical processes of the treatment system. During the last two weeks of the experiment, the very low HRT caused a high quantity of biogas (up to 55%) to be released into the vessel collecting the effluent (Fig. 5b). This observation agrees with other studies indicating that a significant fraction of biogas can be lost in the effluent stream in MBR configurations operated at low HRTs [1, 4, 8, 29]. This phenomenon could also be enhanced in ADMBR as consistent COD removal is operated by the biofilm forming the DM (see Section 3.4). Even though DM can support high fluxes as demonstrated in this study, this configuration could represent an overall disadvantage
17
for the sustainability of the process. Methane can in fact be forced to escape from the biofilm through the effluent by the water flow producing a reduction of energy recovery. In the first two months, the methane content in the biogas extracted from the reactor was between 55% and 65% (Fig. 5c). The decrease of HRT led to an increase of the methane content in the biogas (up to 80%). This effect can be related to the different solubility of CH4 if compared to CO2 the latter being about 25 times more soluble in water than the former. At low HRTs, CO2 release to the headspace is limited and most of it is removed with the effluent stream thus increasing the methane content of the biogas produced [12]. Mixed liquor pH ranged between 6.8 and 7.5, while effluent pH resulted between 7.1 and 8.7 (data not shown). Similar results were reported by Ho and Sung [27] during treatment of a synthetic municipal wastewater with AMBRs at ambient temperatures. The different pH values could be due to the transformation of organic acids into methane by the biofilm developed on the DM membrane and also by a generation of alkalinity by degradation organics containing nitrogen accumulated on the cake layer. The different pH values also support the slightly higher CH4 content of the biogas collected above the membrane than that of the reactor (Fig. 5c). Continuous recirculation of mixed liquor was carried out between the membrane support and the anaerobic reactor, so dissolved biogas could escape from both compartments. The biogas collected from the membrane support is therefore a very rough estimation of the methanogenic activity of the DM.
3.4. Specific contribution of the Dynamic Membrane to treatment process The trends of the CODf concentrations measured in the reactor and in the effluent (Fig. 3b) highlight a contribution towards COD removal by the DM. CODf in the effluent was, in fact, most of the time lower than that of the anaerobic reactor. Approximately 20% of
18
overall CODf removal can be ascribed to the DM (Fig. 3b). The effect of the DM on CODf removal began after approximately 3 weeks from start-up. A significant contribution to COD removal by membranes has already been observed in other studies [e.g. 2, 27] but to the best of author's knowledge this effect has not previously been reported for DM developed over a large pore size mesh.
3.4.1. Short-term experiment on membrane filterable COD removal To confirm the hypothesis suggested by the difference in CODf concentrations, a shortterm experiment was performed at the end of the study once the DM was well-formed. The anaerobic reactor was emptied of the mixed liquor and filled with the synthetic feed. The only active biomass in this configuration was, thus, the biofilm forming the DM. Surprisingly, the DM demonstrated very high COD removal (Fig. 6). At an HRT of 1.5 d (corresponding to a flux of 5 L m-2 h-1) the DM achieved COD removal of 80-85 and 9095% for CODt and CODf, respectively. When HRT was reduced to 0.6 d and flux increased to approximately 15 L m-2 h-1, COD removals decreased as well but remained anyway above 70 and 80 % for CODt and CODf, respectively (Fig. 6). Although the mixed liquor was completely removed from the reactor, TSS and VSS concentrations of 40-80 and 20-60 mg L-1, respectively, were measured in the effluent during the phase at an HRT of 1.6 d. When HRT was decreased to 0.6 d, TSS and VSS concentration increased to 130 and 110 mg L -1, respectively (data not shown). The significant amount of suspended solids measured in the effluent during the short-term experiment suggests that the DM establishes an equilibrium between the solids deposited on the inner part of the membrane and the loss of solids with the water flow across the membrane from its external part. Solids were also removed from the DM by cross flow and transported into the reactor establishing a MLTSS concentration of 0.5 g L-1 at the end of
19
the short-term experiment. The solids present in the effluent were probably removed from the external part of the DM; the visual inspection carried out after system shutdown, in fact revealed that a dense biofilm was attached to both surfaces (internal and external) of the mesh. Approximately 400 mL of biogas were released by the system during the short-term test, with methane concentrations of 80-85%. However, most of the biogas (62%) was measured in the effluent collection vessel, confirming the importance of monitoring the dissolved methane in the liquid phase when working at low HRT and temperature. Surprisingly, the biogas released from the membrane support (i.e. collected by the gas meter located above the membrane support) was only 8% of total production, demonstrating that most of the dissolved biogas produced in the DM is transported by the liquid recirculation flow and released into the reactor. Over the short-term experiment the TMP remained around 50 mbar during phase at an HRT of 1.5 d, while it gradually increased up to 120 mbar when HRT was decreased to 0.6 d (data not shown).
3.5. COD Mass Balance COD mass balance was performed during three phases of the experiment at an HRT of 2 d, 1 d and 0.5 d. The results are expressed as a percentage of the total COD entering the system through the influent (Fig. 7). COD removed as waste sludge for sampling (CODw*Qw) and COD variation in the sludge (ΔCODML) are merged together and reported as “waste” in Figure 7. The results highlight that a large fraction of the COD that enters is missing in the mass balance (53-59%) for all the phases considered. Other authors have also reported significant amounts of unaccounted for COD in mass balances during operation of AMBRs
20
at ambient temperatures [3, 8]. The values obtained in this study are, however, higher than those previously reported (9-18%). The large amount of unaccounted for COD can be due to several reasons. Contrary to other studies [3, 8], COD removal due to other electron acceptors [3] can be neglected since sulphate and nitrate concentrations in the influent were both below 15 mg L-1 and therefore they can account for up to 2 % of total COD input. The sampling procedure and frequency adopted for the monitoring of the system might have underestimated biomass loss through the effluent or biomass growth and accumulation on the DM due to the small volume of the system. Also, biofilm formation inside the reactor and the piping might have contributed to the missing fraction of the COD mass balance. Considering the low yield value of anaerobic biomass it can, however, be presumed that these underestimated COD fractions do not represent the entire missing COD. It is believed that the missing fraction COD is largely due to an underestimated dissolved methane concentration in the liquid phase. Dissolved methane was calculated assuming thermodynamic equilibrium (eq. 2) and accounted for 8% of the influent COD for all HRTs while CH4 collected in the gas phase accounted for 13-24 % of the influent COD (Fig. 7). The methane yield calculated comprising both the gaseous and dissolved phase, was recorded at between 85 and 133 LCH4 kgCOD-1. These yields are much lower than the theoretical value of 350
kgCOD-1 [25], although some authors also reported lower
methane yields (210-270 LCH4 kgCOD-1) than the theoretical values during operation of AMBRs [13]. The low values obtained in this study support the hypothesis that the underestimated methane production is due to an actual dissolved methane concentration in the liquid phase that is considerably higher than the amount calculated under thermodynamic equilibrium. The oversaturation conditions produced a loss of dissolved
21
methane through the reactor effluent that was not detected, explaining the unaccounted COD fraction. Methane oversaturation in the liquid phase is likely to occur in anaerobic systems working at ambient temperature as reported by several authors [3, 4, 30]. Pauss et al. [30] stated that the liquid-to-gas mass transfer coefficient significantly changes according to reactor configuration and operating conditions and can lead to methane concentrations in the liquid phase that are up to 12 times higher than the equilibrium values. Indeed, if dissolved methane concentration at equilibrium was increased to values indicated by Pauss et al. [30], the missing COD fraction in the mass balances was reduced to less than 5% (data not shown) highlighting the need for proper recovery strategies of methane losses in effluents from low temperature AMBRs [1].
4. Conclusions The results demonstrated that an anaerobic DM can be achieved on a mesh with large pore size (200 µm) treating municipal wastewater at ambient temperature. The large pore size allows for high membrane fluxes (15-20 L m-2 h-1) applying low TMP (lower than 50-100 mbar). The treatment system achieved an average COD removal of higher than 80 % for HRTs higher than 0.5 d. On the contrary at HRTs lower than 0.5 d the average COD removal decreased to 50 - 60%. The results also demonstrated that the biofilm composing the cake layer of the DM significantly concurs in obtaining the high treatment efficiency. COD mass balance suggested that the low HRTs applied to the ADMBR produced dissolved methane concentrations higher than those that are estimated by Henry’s law, thus causing important methane loss through the reactor effluent.
22
5. References [1] A.L. Smith, L.B. Stadler, N.G. Love, S.J. Skerlos, L. Raskin, Perspectives on anaerobic membrane bioreactor treatment of domestic wastewater: a critical review, Bioresource Technol. 122 (2012) 149-159.
[2] A.L. Smith, S.J. Skerlos, L. Raskin, Psychrophilic anaerobic membrane bioreactor treatment of domestic wastewater, Water Res. 47 (2013) 1655-1665.
[3] C. Shin, P.L. McCarty, J. Kim, J. Bae, Pilot-scale temperate-climate treatment of domestic wastewater with a staged anaerobic fuidized membrane bioreactor (SAF-MBR), Bioresource Technol. 159 (2014) 95-103.
[4] R. Yoo, J. Kim, P.L. McCarty, J. Bae, Anaerobic treatment of municipal wastewater with a staged anaerobic fluidized membrane bioreactor (SAF-MBR) system, Bioresource Technol. 120 (2012) 133-139.
[5] W. Verstraete, P. Van de Caveye, V. Diamantis, Maximum use of resources present in domestic “used water", Bioresource Technol. 100 (2009) 5537-5545.
[6] G. Lettinga, S. Rebac, G. Zeeman, Challenge of psychrophilic anaerobic wastewater treatment, Trends Biotechnol. 19 (2001) 363-370. [7] V. O’Flaherty, G. Collins, T. Mahony, The microbiology and biochemistry of anaerobic bioreactors with relevance to domestic sewage treatment, Rev. Environ. Sci. Biotechnol. 5 (2006) 39–55.
[8] J.B. Giménez, N. Martí, J. Ferrer, A. Seco, Methane recovery efficiency in a submerged anaerobic membrane bioreactor (SAnMBR) treating sulphate-rich urban wastewater: evaluation of methane losses with the effluent, Bioresource Technol. 118 (2012) 67-72.
[9] S. Casu, N.A. Crispino, R. Farina, D. Mattioli, M. Ferraris, A. Spagni, Wastewater treatment in a submerged anaerobic membrane bioreactor. J. Environ. Sci. Heal. A. 47
23
(2012) 204-209.
[10] A. Robles, M.V. Ruano, J. Ribes, J. Ferrer, Factors that affect the permeability of commercial hollow-fibre membranes in a submerged anaerobic MBR (HF-SAnBR) system, Water Res. 47 (2013) 1277-1288.
[11] G. Skouteris, D. Hermosilla, P. López, C. Negro, Á. Blanco, Anaerobic membrane bioreactors for wastewater treatment: a review, Chem. Eng. J. 198 (2012) 138-148.
[12] E.R. Hall, P.R. Berube, Membrane bioreactors for anaerobic treatment of wastewaters: phase II, WERF report (Alexandria) and IWA Publishing (London), 2006.
[13] D. Martinez-Sosa, B. Helmreich, T. Netter, S. Paris, F. Bischof, H. Horn, Anaerobic submerged membrane bioreactor (AnSMBR) for municipal wastewater treatment under mesophilic and psychrophilic temperature conditions, Bioresource Technol. 102 (2011) 10377-10385.
[14] A. Spagni, S. Casu, N.A. Crispino, R. Farina, D. Mattioli, Filterability in a submerged anaerobic membrane bioreactor, Desalination, 250 (2010) 787-792.
[15] L. Alibardi, R. Cossu, M. Saleem, A. Spagni, Development and permeability of a dynamic membrane for anaerobic wastewater treatment, Bioresource Technol. 161 (2014) 236-244.
[16] H. Chu, Y. Zhang, X. Zhou, Y. Zhao, B. Dong, H. Zhang, Dynamic membrane bioreactor for wastewater treatment: Operation, critical flux, and dynamic membrane structure, J. Membrane Sci. 450 (2014) 265-271.
[17] M.E. Ersahin, H. Ozgun, R.K. Dereli, I. Ozturk, K. Roest, J.B. van Lier, A review on dynamic membrane filtration: Materials, applications and future perspectives, Bioresource Technol. 122 (2012) 196-206.
[18] H. Liu, C. Yang, W. Pu, J. Zhang, Formation mechanism and structure of dynamic membrane in the dynamic membrane bioreactor, Chem. Eng. J. 148 (2009) 290-295. 24
[19] C. Wang, W-N. Chen, Q-Y. Hu, M. Ji, X. Gao, Dynamic fouling behavior and cake layer structure changes in nonwoven membrane bioreactor for bath wastewater treatment, Chem. Eng. J. 264 (2015) 462-469.
[20] X. Zhang, Z. Wang, Z. Wu, F. Lu, J. Tong, L. Zang, Formation of dynamic membrane in an anaerobic membrane bioreactor for municipal wastewater treatment. Chem. Eng. J. 165 (2010) 175-183. [21] C. Loderer, B. Gahleitner, K. Steinbacher, C. Stelzer, W. Fuchs, Dynamic filtration – A novel approach for critical flux determination using different textiles, Sep. Purif. Technol. 120 (2013) 410-414.
[22] J. Ma, Z. Wang, X. Zou, J. Feng, Z. Wu, Microbial communities in an anaerobic dynamic membrane bioreactor (AnDMBR) for municipal wastewater treatment: Comparison of bulk sludge and cake layer, Process Biochem. 48 (2013) 510-516.
[23] J. Ma, Z. Wang, Y. Xu, Q. Wang, Z. Wu, A. Grasmick, Organic matter recovery from municipal wastewater by using dynamic membrane separation process, Chem. Eng. J. 219 (2013) 190-199.
[24] Y. Kiso, Y.J. Jung, T. Ichinari, M. Park, T. Kitao, K. Nishimura, K.S. Min, Wastewater treatment performance of a filtration bio-reactor equipped with a mesh as a filter material, Water Res. 34 (2000) 4143-4150.
[25] D. Batstone, J. Keller, R.I. Angelidaki, S.V. Kalyuzhnyi, S.G. Pavlostathis, A. Rozzi, W.T.M. Sanders, H. Siegrist, V.A. Vavilin, Anaerobic Digestion Model No.1 (ADM1), IWA Publishing, London, 2002.
[26] APHA, AWWA, WEF, Standard Methods for the examination of water and wastewater, 21st ed., American Public Health Association/ American Water Works Association/ Water Environmental Federation. Washington, DC., 2005.
25
[27] J. Ho, S. Sung, Methanogenic activities in anaerobic membrane bioreactors (AnMBR) treating synthetic municipal wastewater, Bioresource Technol. 101 (2010) 2191-2196.
[28] M.E. Ersahin, H. Ozgun, Y. Tao, J.B. van Lier, Applicability of dynamic membrane technology in anaerobic membrane bioreactors, Water Res. 48 (2014) 420-429.
[29] H. Yeo, H.S. Lee, The effect of solids retention time on dissolved methane concentration in anaerobic membrane bioreactors, Environ. Technol. 34 (2013) 2105-2112.
[30] A. Pauss, G. Andre, M. Perrier, S. Guiot, Liquid-to-gas mass transfer in anaerobic processes: inevitable transfer limitations of methane and hydrogen in the biomethanation process, Appl. Environ. Microb. 56 (1990) 1636−1644.
26
Figure captions
Figure 1. Schematic diagram of the bench-scale ADMBR.
Figure 2. a) TMP variations, applied HRT and b) membrane fluxes for the ADMBR.
Figure 3. a) Trends of the effluent total COD (Eff. CODt), effluent filterable COD (Eff. CODf), total COD removal (CODt Rem.) and filterable COD removal (CODf Rem.) in the ADMBR; b) variations of mixed liquor filtered COD (CODf ML), effluent filtered COD (CODf Effl.), and filtered COD removal due to the DM (CODf Rem.) in the ADMBR.
Figure 4. a) Effluent TSS and VSS variations in the ADMBR. b) MLTSS and MLVSS variations in the ADMBR. Arrows indicate the further seeds addition during the experiments.
Figure 5. Biogas production (a), biogas production fractions emitted from the reactor, membrane module and effluent vessel (b) and CH4 content of the biogas emitted from the reactor and from the membrane module (c).
Figure 6. Variations of the total (CODt) and filtered COD (CODf) concentration and removal (Rem.) in the short-term experiment.
Figure 7. Mass balance COD distributions (%) for a) HRT = 2 d; b) HRT= 1 d; c) HRT=0.5 d.
27
Figure 1. Schematic diagram of the bench-scale ADMBR.
28
6
a)
200 HRT TMP
5
HRT (d)
4 3
100
2
TMP (mbar)
150
50 1 0
Flux (L m-2 h-1)
b)
0
30 25 20 15 10 5 0 0
20
40
60
80
100
120
140
Time (d)
Figure 2. a) TMP variations, applied HRT and b) membrane fluxes for the ADMBR.
29
900
100
800
90
700
80 70
600 Eff. CODt Eff. CODf CODt Rem. CODf Rem.
500 400
60 50 40
300
30
200
20
100
10
0
COD removal (%)
Effluent COD (mg L-1)
a)
0 0
20
40
60 80 Time (d)
100
120
400
140
25 ML CODf Eff. CODf CODf Rem.
b)
20
15 200 10
CODf removal (%)
CODf (mg L-1)
300
100 5
0
0 0
20
40
60
80
100
120
140
Time (d)
Figure 3. a) Trends of the effluent total COD (Eff. CODt), effluent filterable COD (Eff. CODf), total COD removal (CODt Rem.) and filterable COD removal (CODf Rem.) in the ADMBR; b) variations of mixed liquor filtered COD (CODf ML), effluent filtered COD (CODf Effl.), and filtered COD removal due to the DM (CODf Rem.) in the ADMBR.
30
b)
Effluent TSS and VSS (mg L -1)
1000
MLTSS and MLVSS (g L -1)
a)
800
TSS VSS
600 400 200 0 12
MLTSS MLVSS
10 8 6 4 2 0 0
20
40
60
80
100
120
140
Time (d)
Figure 4. a) Effluent TSS and VSS variations in the ADMBR. b) MLTSS and MLVSS variations in the ADMBR. Arrows indicate the further seeds addition during the experiments.
31
a)
150
100
50
0
100 80
b)
60
Reactor Membrane Effluent vessel
40 20
100
CH4 Content (%)
c)
0 Reactor Membrane
80
Biogas production fraction (%)
Biogas production (mL d-1)
200
60 40 20 0 0
20
40
60
80
100
120
140
Time (d)
Figure 5. Biogas production (a), biogas production fractions emitted from the reactor, membrane module and effluent vessel (b) and CH4 content of the biogas emitted from the reactor and from the membrane module (c).
32
CODt 400
CODf
CODf Rem.
CODt Rem.
HRT = 1.5 d
HRT = 0.6 d
100
80 200
70 60
100
COD removal (%)
-1 COD (mg L )
90 300
50 0
40 0
20
40
60
80
100
Time (h)
Figure 6. Variations of the total (CODt) and filtered COD (CODf) concentration and removal (Rem.) in the short-term experiment.
33
Waste 3% a)
Effluent 11%
CH4 (g) 19% Unknown 59%
CH4 (d) 8% b)
Waste 2%
Effluent 13%
Unknown 53%
CH4 (g) 24% CH4 (d) 8%
Waste 1% c)
Effluent 25% Unknown 53% CH4 (g) 13% CH4(d) 8%
Figure 7. Mass balance COD distributions (%) for a) HRT = 2 d; b) HRT= 1 d; c) HRT=0.5 d.
34
Highlights
Anaerobic dynamic membranes can be developed at ambient temperature Dynamic membrane over large pore size mesh can obtain efficient solid rejections Biofilm forming the dynamic membrane significantly contributed to organics removal Methane oversaturation occured due to ambient temperature and low HRTs
35
S1385-8947(15)01194-8 http://dx.doi.org/10.1016/j.cej.2015.08.111 CEJ 14104
To appear in:
Chemical Engineering Journal
Received Date: Revised Date: Accepted Date:
26 June 2015 25 August 2015 26 August 2015
Please cite this article as: L. Alibardi, N. Bernava, R. Cossu, A. Spagni, Anaerobic dynamic membrane bioreactor for wastewater treatment at ambient temperature, Chemical Engineering Journal (2015), doi: http://dx.doi.org/ 10.1016/j.cej.2015.08.111
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Anaerobic dynamic membrane bioreactor for wastewater treatment at ambient temperature
Luca Alibardi a, Nicoletta Bernava b, Raffaello Cossu a, Alessandro Spagni c,*
a
Department of Industrial Engineering, University of Padova, via Marzolo 9, 35131
Padova, Italy b
Department of Civil, Environmental and Architectural Engineering, University of
Padova, via Marzolo 9, 35131 Padova, Italy c
Water Resources Management Laboratory, Italian National Agency for New Technology,
Energy and Sustainable Economic Development (ENEA), via M.M. Sole 4, 40129 Bologna, Italy
* Corresponding author: Alessandro Spagni, ENEA, via M.M. Sole 4, 40129 Bologna, Italy. Email: [email protected] Phone: +39 051 6098779; fax: +39 051 6098309
Abstract A bench-scale dynamic membrane (DM) bioreactor was operated to evaluate the anaerobic treatment of a synthetic municipal wastewater at ambient temperature. The DM was developed over a large pore size (200 µm) mesh in order to improve sludge filterability and reduce energy consumption. The system achieved average organic removals higher than 80 and 90% for total COD and filtered COD, respectively. Results also demonstrated that the biofilm, which the DM is made of, played a significant part in obtaining the overall
1
organic removal efficiency. The large pore size of the mesh allowed for high membrane fluxes (approximately 15-20 L m-2 h-1) applying low TMP (usually lower than 50-100 mbar). Fluxes higher than 20 L m-2 h-1 produced low solid removal efficiency indicating deterioration of the DM. COD mass balance suggests that the low hydraulic retention times applied to the system caused methane loss through the effluent due to oversaturation.
Keywords: Ambient temperature; Anaerobic processes; Dynamic membrane; Membrane bioreactor; Mesh filtration;.
1. Introduction Municipal wastewaters are, currently, mainly treated by the use of activated sludge systems which, although effective, require a great deal of energy [1]. As a result, anaerobic technologies have been widely investigated for the treatment of municipal wastewater. Their benefits are the production of biogas as a renewable energy source and reduced energy consumption if compared to conventional aerobic treatment [1-4]. The advantages of anaerobic treatments will also be emphasised in future water and wastewater management scenarios [5]. Anaerobic processes can in fact improve the recovery of energy, materials and water from concentrated and diversified wastewater streams both in centralised and decentralised systems [5]. Mesophilic conditions are the preferred option for anaerobic wastewater treatment. These conditions, however, make anaerobic processes of medium to low strength streams such as municipal wastewater non cost-effective. The low energy recovery per unit of volume is not in fact sufficient to satisfy the heat and power requirements of reactors [1, 6]. Solutions allowing for the application of anaerobic processes at ambient or psychrophilic temperatures for low strength wastewater are, therefore, very attractive [6, 7].
2
However, several aspects related to the application of psychrophilic conditions are still under investigation. For example psychrophilic conditions do not only reduce microbial kinetic rates but also increase gas solubility, thus leading to lower methane recovery and methane losses with the effluent [1, 8]. Anaerobic processes under psychrophilic conditions have been successfully applied to municipal wastewater treatment by high rate systems such as the up-flow anaerobic sludge blanket (UASB) and expanded granular sludge bed (EGSB) reactors [6, 7]. However, serious problems arise when treating low strength wastewater under psychrophilic conditions using these technologies. In fact, low biomass growth rates at low temperatures and high hydraulic loading rates increase sludge wash out. In addition, applications of UASB and EGSB reactors for municipal wastewater treatment are limited by the influent high concentrations of suspended solids and inert substances [6]. Over the last decade, several studies have investigated the application of membrane bioreactors (MBRs) in anaerobic conditions [2, 9-12]. Microfiltration (MF) and ultrafiltration (UF) membranes, in fact, allow complete biomass retention so that a high concentration of slow-growing anaerobic bacteria in the reactor can be reached, even at low hydraulic retention time (HRT) [2, 9, 12]. Only a few studies have evaluated the performance of anaerobic MBR (AMBR) at psychrophilic or ambient temperature, although, very promising results have been obtained [2, 12, 13]. The first studies on sludge filterability in AMBRs [12, 14] have showed, although not conclusively [10], more severe fouling phenomena under anaerobic rather than aerobic conditions. Consequently, mesh filtration has recently been proposed as an alternative to the use of MF/UF-MBRs in order to improve sludge filterability and reduce capital and management costs [15-19].
3
When mesh filtration is applied, a dynamic membrane (DM) develops on the support material (the mesh). DM is a cake layer or biofilm obtained through the deposition over the mesh of organic substances and bacteria present in the reactor. Once DM is formed, solids rejection is carried out by this regenerative biological layer while the mesh only acts as a support [17, 20]. The layer can be formed and re-formed as a self-forming DM and the permeability can be affected by controlling its thickness while the support (mesh) can be characterised by large pores. The cake layer that develops on the mesh plays, thus, a central role during DM filtration [17] while it is considered the main drawback for the widely adopted conventional MF/UF-MBRs. DMs cannot, it seems, achieve the high water quality obtained by MF and UF due to the different membrane cut-off. However, DMs could represent a worthwhile compromise between water quality and plant costs in anaerobic processes, since effluent posttreatments are usually considered prior to final water discharge (e.g. for nutrient removal). DM technology can offer benefits over traditional filtration in biological treatments by precluding the need of costly membrane modules and by providing a low-cost, regenerative, self-forming filtration surface with adaptable permeability and high fluxes [16, 17]. The link between mesh pore size and dynamic layer formation is still not clear. The use of large pore size mesh can reduce the overall filtration resistance and the cost of the filtration module; however, the maximum size allowing the development of a reliable dynamic filtration layer has not been defined yet. Experiments on DM development have been carried out with mesh opening between 30 and 90 μm [20-23]. Preliminary studies indicated that an effective cake layer does not develop over meshes with pore sizes larger than 60-70µm [23]. On the contrary, Alibardi et al. [15] developed a DM on a large poresized mesh (200 μm) in mesophilic conditions by properly managing hydrodynamic
4
conditions. Moreover, Kiso et al. [24] suggested that DMs not only act as filters but may also improve overall pollutant removal efficiency through the biochemical reactions that occur within the cake layer. This study aimed to evaluate the application of an anaerobic dynamic membrane bioreactor (ADMBR) equipped with a coarse filtration mesh (200 m) for the treatment of highstrength municipal wastewater at ambient temperature. A large pore size mesh was used in order to improve filterability and reduce energy consumption. The study also aimed to evaluate the contribution of the DM biofilm on the overall organic removal efficiency of the reactor.
2. Material and methods 2.1. Experimental setup The study was performed by using a bench-scale ADMBR equipped with an external cross-flow filtration unit (Fig. 1). The reactor had a total volume of 898 mL (W x H x D: 9.5 x 10.5 x 9 cm) and a working volume of 684 mL. The level of the mixed liquor was kept constant by using a level sensor connected to the influent pump. The filtration support had an inner volume of 60 mL (W x H x D: 20 x 1.5 x 2 cm). A monofilament woven mesh made of polyamide/nylon (SaatiMil PA 7, Saati s.p.a., Italy) with openings of 200 µm, thread diameter of 120 µm, mesh count of 31/cm and 39% opening area (data from the supplier) was inserted into the central longitudinal part of a filtration support. The filtration area was 40 cm2 (L x W: 20 x 2 cm). The overall working volume composed by reactor, membrane support and pipe connections, was 745 mL.
5
The feed was provided by a peristaltic pump (Watson Marlow 401U/D1) controlled by the level sensor inside the reactor. Sludge mixing was provided by a magnetic stirrer (Heidolph, Hei-Standard). The circulation of the mixed liquor along the mesh surface and the return of the retentate into the reactor were provided by a second peristaltic pump (Watson Marlow 505S) applying a cross flow velocity of about 10 m h-1. A third peristaltic pump (Watson Marlow 401U/DM3) was installed to extract the effluent out of the system and to provide the necessary transmembrane pressure (TMP) for filtration. An airtight vessel of approximately 100 mL was installed down-flow from the third pump to assess the presence of oversaturated biogas in the effluent. The effluent flowed out this vessel into a collection tank. TMP was measured by means of two U- tube pressure gauges filled with water and placed up- and down- flow from the filtration support, respectively. Biogas production was monitored by using three homemade wet-tip gas meters connected to the treatment system. The first gas meter was connected to the main vessel while the second to the mesh support in order to estimate the biogas production by the reactor and by the DM biofilm, respectively. The third gas meter was connected to the effluent collection vessel in an attempt to collect (at least) part of the biogas that could escape with the effluent. Effluent volume, TMP and biogas production were recorded on a daily basis from Monday to Saturday, except during the holiday closure period (from day 38 to day 53). Mixed liquor samples were collected from reactor once to twice per week (to avoid excessive sludge removal) and analysed for the following parameters: total suspended solids (TSS), volatile suspended solids (VSS), total chemical oxygen demand (CODt) and filtrable COD (CODf). Effluent samples were collected three to five times per week and analysed for the following parameters: TSS, VSS, CODt, CODf.
6
2.2. Inoculum The reactor was inoculated with anaerobic sludge taken from a full-scale mesophilic anaerobic digester treating the excess sludge of a municipal wastewater treatment plant located in Padova, Italy. The inoculum had a concentration of total (TSS) and volatile suspended solids (VSS) of 12 gTSS L-1 and 6 gVSS L-1, respectively. Sludge was added on three further occasions during the experiment. Sludge had to be topped up to compensate for biomass loss caused by cleaning of the filtration support due to clogging and by other maintenance procedures.
2.3. Feed The reactor was fed with synthetic wastewater, which simulated high strength municipal wastewater. The feed had a concentration of 900 mgCOD L-1 composed of sucrose (45% of total COD), powder milk (10%) and corn starch (45%) as organic matter. The followings chemicals were added to ensure alkalinity, macro- and micronutrients: NaHCO3 (830 mg L-1feed), NH4Cl (50 mgN L-1feed), KH2PO4 (10 mgP L-1 feed), FeCl3·6H2O (2.1 mgFe L-1 feed), CaCl2·2H2O (8.2 mgCa L-1 feed), MgCl2·6H2O (2.4 mgMg L-1 feed), Na2MoO4·2H2O (0.22 mgMo L-1feed), ZnSO4·7H2O (0.23 mgZn L-1 feed), CuSO4·5H2O (0.128 mgCu L-1 feed), NiCl2·6H2O (0.1 mgNi L-1 feed), H3BO4 (0.007 mgB L-1feed). All chemicals were dissolved in tap water.
2.4. Operating parameters The feed concentration and the filtration area were kept constant during the entire study. The HRT was intentionally varied between 5.7 and 0.25 d to evaluate the performance of the systems under different operating conditions. Consequently, the organic load rate
7
(OLR) and the flux applied to the membrane varied in the range 0.16-3.3 kgCOD m-3 d-1 and 1.4-28 L m-2 h-1, respectively. Any condition was maintained for a period of time longer than 3 HRTs to observe steady state operation. Since the temperature was not controlled, the system worked at ambient temperature (in the laboratory) in the range of 20 – 24 °C.
2.5. COD mass balance COD mass balance of the system was performed for three phases characterised by the following hydraulic retention times (HRTs): 2 d (experimental days 63 – 83), 1 d (days 84 – 97), 0.5 d (days 98 - 121). The following equation was used:
CODin ·Qin CODeff ·Qeff CODCH4(g)·QCH4 CODCH4(d)·Qeff CODW ·QW CODML (1)
Where: - CODin, CODeff, and CODw are the average COD concentration of the influent, effluent and waste (i.e. the sludge extracted from the reactor for sampling) streams, respectively; - CODCH4(g) is the average COD of the methane in the biogas; - CODCH4(d) is the average COD of the methane dissolved in the effluent stream; - ΔCODML is the variation of the COD due to the change of the mixed liquor suspended solids concentration; - Qin, Qeff, Qw are the influent, effluent, waste (i.e. the volume of the mixed liquor sampled for analysis) flow; - QCH4 is the flow of the methane production in the biogas.
8
The COD of the mixed liquor volatile suspended solids was estimated assuming the ratio of 1.42 gCOD gVSS-1. CODCH4(g) was computed by considering the daily biogas production, the methane concentration in the biogas, the molar volume of the gas (according to the experimental temperature) and the COD equivalence of 64 gCOD molCH4-1 [25]. The methane dissolved in the liquid phase was estimated using Henry’s law assuming equilibrium with the gas phase and using a Henry’s constant (KH) value at 25°C of 0.0014 Mliq bargas-1 (i.e. molarity in the liquid phase over partial pressure, P, in the gas phase) [25]:
(2)
The values of KH at different temperatures (KH(T)) were estimated according to the following equation [25]:
(3)
where θ is the temperature dependence constant assumed equal to -0.01929 T-1 [25]. The amount of dissolved COD (methane and soluble organic matter) lost via the sludge taken for sampling (Qw) was neglected since the volume sampled was very low (approximately 5-10 mL twice or three times per week).
2.6. Short-term experiment for evaluating COD removal efficiency by the DM A short-term experiment was performed at the end of the study to evaluate the contribution of the well-formed DM on overall COD removal efficiency. The experiment was carried out by means of the same setup described in Section 2.1.
9
The reactor was emptied of the mixed liquor and filled with the feed solution. During the first 72 hours the HRT was set to approximately 1.5 d. Thereafter, HRT was reduced to 0.6 d and kept constant until the end of the experiment. The entire experiment lasted for almost 90 h.
2.7. Analytical methods Total chemical oxygen demand (CODt), filtered COD (CODf), TSS and VSS concentrations were measured according to Standard Methods [26]. pH was measured using a Crison instrument (model micropH 2001) and probe (model 5011). Sample filtration for TSS and VSS analysis was performed using Whatman GF/C filters. Sample filtration for CODf analysis was performed using 0.20 µm syringe filters. Biogas composition was analysed using a micro-GC (Varian 490-GC) equipped with a 10meter MS5A column (to analyse CH4) and a 10-meter PPU column (to analyse CO2) and two Thermal Conductivity Detectors (TCDs). Argon was used as carrier gas at a pressure of 60 kPa in columns. Injector and column temperatures were both set to 80 °C. All analyses were performed in triplicate and data reported are average values. Measurements variability resulted within analytical error reported by Standard Methods [26].
3. Results and discussion 3.1. Transmembrane Pressure and Fluxes The system was started up with a HRT of approximately 5 d in order to enhance biomass acclimation and development of the DM on the mesh. On day 15, the HRT was halved to approximately 2.5 d for three weeks until day 38 when the HRT was returned to values of about 5 d due to the summer holidays. Thereafter, the HRT was stepwise reduced to the
10
final value of 0.27 d over a period of approximately two months (Fig. 2) in order to assess system performance. Owing to the HRT variations, the fluxes applied across the DM resulted in the range of between 1.4 and 28 L m-2 h-2 (Fig. 2). During the first three months of the ADMBR operation, TMP experienced a fluctuating up-and-down trend with values, anyway, constantly below 50 mbar (Fig. 2). The up-anddown trend may have been due to the slow development of the DM. TMP started to steadily increase when the HRT was decreased to 0.5 d, corresponding to membrane fluxes of approximately 15 L m-2 h-1. TMP stabilised at 110-120 mbar in two weeks and then it increased to almost 200 mbar when the membrane flux was set to 28 L m-2 h-2 by reducing HRT to 0.25 d (Fig. 2). The results show that during the first three months a stable DM was not formed as proportionality between the applied membrane flux and the measured TMP (according to Darcy’s law) was not observed. At fluxes lower than 10 L m-2 h-1 the step by step increase in the membrane flux caused an immediate decrease in pressure (see for instance on days 15, 65 and 85 in Fig. 2). Thereafter, the cake layer showed a rapid regrowth process indicated by a fast increase of TMP. This fast increase is due to a high level of deposited solids produced by increased fluxes and because of the presence of a previously developed fouling layer. The rough proportionality between the flux and TMP observed after approximately three months of operation is in accordance with Darcy’s law and confirms the development of a stable DM. The evolution of TMP during the experiment agrees with the hypothesis proposed by Zhang et al. [20] and observed by Alibardi et al. [15] of the three stages DM development. The initial instability of TMP can be explained as a consequence of the process of formation of the separation layer. Then, particulate deposition and biofilm growth
11
produced the formation of the DM (DM growth stage) and the flux increase led to the development of a thick cake layer (fouling stage). The TMP values observed in the present study were, however, somewhat different from those previously reported by Alibardi et al. [15] which used the same setup and mesh porosity but applying mesophilic conditions and an OLR that five times higher. In fact, the higher fluxes and lower OLRs applied in the present study than those of Alibardi et al. [15] produced much lower TMP values than those measured in the previous study. These results, thus, highlight the dynamic state of the filtration layer and suggest that operating conditions greatly affect DM formation. The small OLR applied in this study, even at low HRTs, slowed down the formation of the DM, maintaining a limited filtration resistance for a longer time than the previous study [15]. Wastewater characteristics, reactor configuration and conditions applied to the filtration module, thus, influence the initial formation, consolidation and filtration resistance of a DM. It is important to state that the use of a support mesh characterised by large pore opening (200 µm) allowed the development of a DM filtration layer characterised by stable fluxes of 15 L m-2 h-1 at TMP values lower than 100 mbar. These fluxes are comparable to those applied at full scale aerobic submerged MBRs while those obtained by anaerobic MBRs are much lower. For example, Spagni et al. [14] obtained sustainable fluxes between 2 and 5 L m-2 h-1 operating a submerged anaerobic MBR equipped with a conventional flat sheet membrane module characterised by a nominal pore size was of 0.4 µm. Ho and Sung [27] obtained fluxes from 5 to 10 L m-2 h-1 in an anaerobic MBR equipped with an external non-woven filter with pore size of 10 µm and operated at temperature and OLR values similar to those of this study. Shin et al. [3] and Smith et al. [2] recently obtained fluxes below 10 L m-2 h-1 during municipal wastewater treatment under psychrophilic conditions by using AMBRs equipped with hollow fibre polyvinylidene fluoride membranes with
12
pore size of 0.03 μm and flat-sheet microfiltration polyethersulfone membranes with a pore size of 0.2 µm, respectively. A very low flux of 2.6 L m-2 h-1 was also reported by Ersahin et al., [28] when operation of a laboratory scale ADMBR treating high-strength synthetic wastewater by using as support material a monofilament woven fabric made of polypropylene material with an average pore size of 10 µm. Zhang et al. [20], on the contrary, obtained high fluxes using a small-size Dacron mesh (61 µm). The mesh was however installed in the top part of a UASB reactor where the solids load on the filtration unit was limited. The use of a large pore size mesh support also allowed fluxes to be maintained at a level comparable to those reported in literature for DM but without the need for any cleaning procedure. This observation can indicate a potential advantage on the use of a large pore size mesh since membrane cleaning procedures are often used also for DM [20, 22, 23, 28].
3.2. Wastewater treatment The concentrations of CODt and CODf in the effluent remained below 100 mg L-1 during the first month of operation at 2 d HRT, resulting in a COD removal efficiency of over 90% (Fig. 3a). In the second month effluent COD concentrations slightly increased even though HRT was raised to 4 days. After two months of operation, COD removal increased again when HRT was reduced to 2 days and then further decreased to approximately 1 d. These results suggest that the DM contributes to the organic matter degradation as better COD removals were obtained at lower HRTs and consequently increased membrane fluxes (Fig. 2). The increasing organic load on the filtration membrane applied by lowering HRTs can stimulate and support the growth of a thicker biofilm, capable not only of solids rejection but also of organic biodegradation (Fig. 3b).
13
When the HRT was reduced to 0.5 d on day 98, the effluent COD began to gradually increase to nearly 200 mg L-1 (day 101) and also CODf concentration increased from values lower than 10 mg L-1 to about 100 mg L-1. This increase lasted for a few days, after which CODt removal returned to values close to 90%. Further HRT reduction to 0.25 d (Fig. 2), contributed to the decrease in COD removal (Fig. 3). These results highlight the process limits for the bench scale ADMBR used in this study. Low HRTs (0.5 or 0.25 d) and the resulting high fluxes applied to the dynamic membrane produced a great accumulation of solids in the filtration layer causing the TMP increase. The ambient temperature and the high wastewater flow through the system negatively affected the anaerobic digestion process. Considering the system during the first 122 d of the experimental run (i.e. before the deterioration of treatment processes), the reactor achieved the average removal efficiencies of 87 ± 4 % and 94 ± 4 % (mean ± standard deviation) for total and filtered COD, respectively. These performances are comparable with those of conventional AMBRs, whose COD removal efficiency usually ranges between 75% and 99% [16]. If data are compared with other studies on ADMBR, COD removal obtained in this study is generally higher. Alibardi et al. [15] operated an identical ADMBR at mesophilic temperatures and HRTs between 2 and 4 days for treatment of synthetic industrial wastewater. COD removals ranged from 65% to 92% (average of 75%) from an influent COD of 5 g L-1. Also Ma et al. [22] reported an average COD removal of 79% during treatment of real municipal wastewater characterised by a COD concentration that varied between 300 and 800 mg L-1. Ersahin et al. [28] achieved higher COD removals (99%) but during the treatment of high-strength synthetic industrial wastewater (COD > 20 g L-1) at 35°C and using a support mesh with average pore size of 10 µm. Results demonstrate that anaerobic processes coupled with DMs developed over large pore size
14
mesh can consistently achieve high treatment efficiencies at ambient temperature in a perspective of lower capital and management costs. The reactor was started up with approximately 5 and 3.5 g L-1 of TSS and VSS, respectively. During the first month, solids rejection by the DM was very efficient producing an effluent TSS concentration below 25 mg L-1 (Fig. 4a). However, after approximately three weeks of operation, the membrane support was opened due to clogging problems that occurred in the pipes. Due to the partial loss of the mixed liquor (ML) sludge a small amount of sludge was added, and this justifies the mixed liquor TSS (MLTSS) and VSS (MLVSS) peaks of 9.5 and 5.3 g L-1 observed on day 29 (Fig. 4b). During the second month, fluctuations in effluent solids content were noticed as decreasing concentration of MLTSS in the reactor (Fig. 4). Even though high effluent TSS concentrations were measured, the solids lost in the exiting flow does not justify the significant decrease in the mixed liquor suspended solids (MLSS) (Fig. 4b) in the reactor during the second month. The addition of new inoculum performed on day 58 produced only a temporary increase of biomass concentration in the reactor as a rapid decrease of MLSS concentration was observed immediately after. The simultaneous low TSS content of the effluent suggests that MLSS were accumulated on the mesh forming a thick cake layer. The TSS effluent concentration was for most of the study below 100 mg L -1 with only some exceptions. This value is comparable to effluent TSS concentrations reported for other high rate anaerobic treatment processes [6]. This result demonstrates that DM supported by large pore size mesh can obtain efficient solid rejections at ambient temperatures. Filtration efficiency, therefore, is not related to the pore size of the support mesh but it only depends by the formation and consolidation state of the dynamic cake layer formed over the mesh.
15
After day 110, with the further decrease of HRT to 0.5 and then to 0.27 d, the effluent TSS and VSS concentrations increased (Fig. 4a) causing almost complete depletion of the MLSS (Fig. 4b). A third biomass addition was not able to maintain the high MLSS concentration. These results indicate that such low HRTs cannot sustain sufficient biomass growth rate at ambient temperature when treating municipal wastewater. Moreover, the high membrane fluxes (15-28 L m-2 h-1) created by the low HRTs, caused deterioration in the structure of the DM allowing solids to escape into the effluent. Both phenomena produced the observed reduction in COD removal (Fig. 3). However, although MLTSS was lower than 0.8 g L-1 and relatively low HRTs were applied (0.27 - 0.5 d), COD removal remained higher than 50% and 60% for CODt and CODf, respectively (Fig. 3). This indicates that deterioration of the DM was only partial as it was still able to operate partial solids retention and also consistent COD removal. This strongly supports the hypothesis that the biofilm of the cake layer forming the DM not only retains solids but also consistently concurs in the COD removal processes [24]. The SRT was estimated as the ratio between the biomass in the reactor and the VSS lost through both the effluent and the sampled mixed liquor per day. SRT remained above 110 d during system operating HRT of 2 d or higher while it gradually decreased to approximately 40 d during operation at 1 d of HRT. When HRT was reduced to 0.5 d or lower, SRT dropped to about 5 d. This value is not sufficient to support the growth of methanogenic biomass as confirmed by the reduction of COD removal observed at the end of the experiment (Fig. 3).
3.3. Biogas production Low biogas production was observed (< 50 mL d-1) during the first two months of operation (Fig. 5a) due to both the low OLR applied to the system and the characteristics
16
of the mesophilic anaerobic sludge used as seed which was not previously acclimated to operate at ambient temperature. The decrease of the HRT to 2 d after two months of operation led to a continuous increase of the biogas production rate from approximately 25 to 175 mL d-1. The increase of biogas production can be attributed to the higher OLRs and to acclimation of the biomass to ambient temperatures. During the first two months, the vast majority of biogas was emitted from the reactor, which, therefore, accounted for nearly 100% of the total volume measured (Fig. 5b). After two months of operation, biogas began to be emitted from the membrane support too, probably due to the further decrease of HRT, higher OLR and the stable development of the DM. The volume of biogas emitted from the membrane support reached up to 60% of total biogas production from the second month until the end of the experiment. The smaller volume and specific surface area of the membrane module, if compared to the reactor, and the significant biogas production measured from the membrane, indicate the high biological activity of the biofilm forming the DM. This result is also confirmed by the fact that total biogas production was only slightly affected by the decreasing trend of MLSS concentration inside the reactor, thus, supporting the idea, once more, that the biofilm forming the DM could substantially contribute to the biochemical processes of the treatment system. During the last two weeks of the experiment, the very low HRT caused a high quantity of biogas (up to 55%) to be released into the vessel collecting the effluent (Fig. 5b). This observation agrees with other studies indicating that a significant fraction of biogas can be lost in the effluent stream in MBR configurations operated at low HRTs [1, 4, 8, 29]. This phenomenon could also be enhanced in ADMBR as consistent COD removal is operated by the biofilm forming the DM (see Section 3.4). Even though DM can support high fluxes as demonstrated in this study, this configuration could represent an overall disadvantage
17
for the sustainability of the process. Methane can in fact be forced to escape from the biofilm through the effluent by the water flow producing a reduction of energy recovery. In the first two months, the methane content in the biogas extracted from the reactor was between 55% and 65% (Fig. 5c). The decrease of HRT led to an increase of the methane content in the biogas (up to 80%). This effect can be related to the different solubility of CH4 if compared to CO2 the latter being about 25 times more soluble in water than the former. At low HRTs, CO2 release to the headspace is limited and most of it is removed with the effluent stream thus increasing the methane content of the biogas produced [12]. Mixed liquor pH ranged between 6.8 and 7.5, while effluent pH resulted between 7.1 and 8.7 (data not shown). Similar results were reported by Ho and Sung [27] during treatment of a synthetic municipal wastewater with AMBRs at ambient temperatures. The different pH values could be due to the transformation of organic acids into methane by the biofilm developed on the DM membrane and also by a generation of alkalinity by degradation organics containing nitrogen accumulated on the cake layer. The different pH values also support the slightly higher CH4 content of the biogas collected above the membrane than that of the reactor (Fig. 5c). Continuous recirculation of mixed liquor was carried out between the membrane support and the anaerobic reactor, so dissolved biogas could escape from both compartments. The biogas collected from the membrane support is therefore a very rough estimation of the methanogenic activity of the DM.
3.4. Specific contribution of the Dynamic Membrane to treatment process The trends of the CODf concentrations measured in the reactor and in the effluent (Fig. 3b) highlight a contribution towards COD removal by the DM. CODf in the effluent was, in fact, most of the time lower than that of the anaerobic reactor. Approximately 20% of
18
overall CODf removal can be ascribed to the DM (Fig. 3b). The effect of the DM on CODf removal began after approximately 3 weeks from start-up. A significant contribution to COD removal by membranes has already been observed in other studies [e.g. 2, 27] but to the best of author's knowledge this effect has not previously been reported for DM developed over a large pore size mesh.
3.4.1. Short-term experiment on membrane filterable COD removal To confirm the hypothesis suggested by the difference in CODf concentrations, a shortterm experiment was performed at the end of the study once the DM was well-formed. The anaerobic reactor was emptied of the mixed liquor and filled with the synthetic feed. The only active biomass in this configuration was, thus, the biofilm forming the DM. Surprisingly, the DM demonstrated very high COD removal (Fig. 6). At an HRT of 1.5 d (corresponding to a flux of 5 L m-2 h-1) the DM achieved COD removal of 80-85 and 9095% for CODt and CODf, respectively. When HRT was reduced to 0.6 d and flux increased to approximately 15 L m-2 h-1, COD removals decreased as well but remained anyway above 70 and 80 % for CODt and CODf, respectively (Fig. 6). Although the mixed liquor was completely removed from the reactor, TSS and VSS concentrations of 40-80 and 20-60 mg L-1, respectively, were measured in the effluent during the phase at an HRT of 1.6 d. When HRT was decreased to 0.6 d, TSS and VSS concentration increased to 130 and 110 mg L -1, respectively (data not shown). The significant amount of suspended solids measured in the effluent during the short-term experiment suggests that the DM establishes an equilibrium between the solids deposited on the inner part of the membrane and the loss of solids with the water flow across the membrane from its external part. Solids were also removed from the DM by cross flow and transported into the reactor establishing a MLTSS concentration of 0.5 g L-1 at the end of
19
the short-term experiment. The solids present in the effluent were probably removed from the external part of the DM; the visual inspection carried out after system shutdown, in fact revealed that a dense biofilm was attached to both surfaces (internal and external) of the mesh. Approximately 400 mL of biogas were released by the system during the short-term test, with methane concentrations of 80-85%. However, most of the biogas (62%) was measured in the effluent collection vessel, confirming the importance of monitoring the dissolved methane in the liquid phase when working at low HRT and temperature. Surprisingly, the biogas released from the membrane support (i.e. collected by the gas meter located above the membrane support) was only 8% of total production, demonstrating that most of the dissolved biogas produced in the DM is transported by the liquid recirculation flow and released into the reactor. Over the short-term experiment the TMP remained around 50 mbar during phase at an HRT of 1.5 d, while it gradually increased up to 120 mbar when HRT was decreased to 0.6 d (data not shown).
3.5. COD Mass Balance COD mass balance was performed during three phases of the experiment at an HRT of 2 d, 1 d and 0.5 d. The results are expressed as a percentage of the total COD entering the system through the influent (Fig. 7). COD removed as waste sludge for sampling (CODw*Qw) and COD variation in the sludge (ΔCODML) are merged together and reported as “waste” in Figure 7. The results highlight that a large fraction of the COD that enters is missing in the mass balance (53-59%) for all the phases considered. Other authors have also reported significant amounts of unaccounted for COD in mass balances during operation of AMBRs
20
at ambient temperatures [3, 8]. The values obtained in this study are, however, higher than those previously reported (9-18%). The large amount of unaccounted for COD can be due to several reasons. Contrary to other studies [3, 8], COD removal due to other electron acceptors [3] can be neglected since sulphate and nitrate concentrations in the influent were both below 15 mg L-1 and therefore they can account for up to 2 % of total COD input. The sampling procedure and frequency adopted for the monitoring of the system might have underestimated biomass loss through the effluent or biomass growth and accumulation on the DM due to the small volume of the system. Also, biofilm formation inside the reactor and the piping might have contributed to the missing fraction of the COD mass balance. Considering the low yield value of anaerobic biomass it can, however, be presumed that these underestimated COD fractions do not represent the entire missing COD. It is believed that the missing fraction COD is largely due to an underestimated dissolved methane concentration in the liquid phase. Dissolved methane was calculated assuming thermodynamic equilibrium (eq. 2) and accounted for 8% of the influent COD for all HRTs while CH4 collected in the gas phase accounted for 13-24 % of the influent COD (Fig. 7). The methane yield calculated comprising both the gaseous and dissolved phase, was recorded at between 85 and 133 LCH4 kgCOD-1. These yields are much lower than the theoretical value of 350
kgCOD-1 [25], although some authors also reported lower
methane yields (210-270 LCH4 kgCOD-1) than the theoretical values during operation of AMBRs [13]. The low values obtained in this study support the hypothesis that the underestimated methane production is due to an actual dissolved methane concentration in the liquid phase that is considerably higher than the amount calculated under thermodynamic equilibrium. The oversaturation conditions produced a loss of dissolved
21
methane through the reactor effluent that was not detected, explaining the unaccounted COD fraction. Methane oversaturation in the liquid phase is likely to occur in anaerobic systems working at ambient temperature as reported by several authors [3, 4, 30]. Pauss et al. [30] stated that the liquid-to-gas mass transfer coefficient significantly changes according to reactor configuration and operating conditions and can lead to methane concentrations in the liquid phase that are up to 12 times higher than the equilibrium values. Indeed, if dissolved methane concentration at equilibrium was increased to values indicated by Pauss et al. [30], the missing COD fraction in the mass balances was reduced to less than 5% (data not shown) highlighting the need for proper recovery strategies of methane losses in effluents from low temperature AMBRs [1].
4. Conclusions The results demonstrated that an anaerobic DM can be achieved on a mesh with large pore size (200 µm) treating municipal wastewater at ambient temperature. The large pore size allows for high membrane fluxes (15-20 L m-2 h-1) applying low TMP (lower than 50-100 mbar). The treatment system achieved an average COD removal of higher than 80 % for HRTs higher than 0.5 d. On the contrary at HRTs lower than 0.5 d the average COD removal decreased to 50 - 60%. The results also demonstrated that the biofilm composing the cake layer of the DM significantly concurs in obtaining the high treatment efficiency. COD mass balance suggested that the low HRTs applied to the ADMBR produced dissolved methane concentrations higher than those that are estimated by Henry’s law, thus causing important methane loss through the reactor effluent.
22
5. References [1] A.L. Smith, L.B. Stadler, N.G. Love, S.J. Skerlos, L. Raskin, Perspectives on anaerobic membrane bioreactor treatment of domestic wastewater: a critical review, Bioresource Technol. 122 (2012) 149-159.
[2] A.L. Smith, S.J. Skerlos, L. Raskin, Psychrophilic anaerobic membrane bioreactor treatment of domestic wastewater, Water Res. 47 (2013) 1655-1665.
[3] C. Shin, P.L. McCarty, J. Kim, J. Bae, Pilot-scale temperate-climate treatment of domestic wastewater with a staged anaerobic fuidized membrane bioreactor (SAF-MBR), Bioresource Technol. 159 (2014) 95-103.
[4] R. Yoo, J. Kim, P.L. McCarty, J. Bae, Anaerobic treatment of municipal wastewater with a staged anaerobic fluidized membrane bioreactor (SAF-MBR) system, Bioresource Technol. 120 (2012) 133-139.
[5] W. Verstraete, P. Van de Caveye, V. Diamantis, Maximum use of resources present in domestic “used water", Bioresource Technol. 100 (2009) 5537-5545.
[6] G. Lettinga, S. Rebac, G. Zeeman, Challenge of psychrophilic anaerobic wastewater treatment, Trends Biotechnol. 19 (2001) 363-370. [7] V. O’Flaherty, G. Collins, T. Mahony, The microbiology and biochemistry of anaerobic bioreactors with relevance to domestic sewage treatment, Rev. Environ. Sci. Biotechnol. 5 (2006) 39–55.
[8] J.B. Giménez, N. Martí, J. Ferrer, A. Seco, Methane recovery efficiency in a submerged anaerobic membrane bioreactor (SAnMBR) treating sulphate-rich urban wastewater: evaluation of methane losses with the effluent, Bioresource Technol. 118 (2012) 67-72.
[9] S. Casu, N.A. Crispino, R. Farina, D. Mattioli, M. Ferraris, A. Spagni, Wastewater treatment in a submerged anaerobic membrane bioreactor. J. Environ. Sci. Heal. A. 47
23
(2012) 204-209.
[10] A. Robles, M.V. Ruano, J. Ribes, J. Ferrer, Factors that affect the permeability of commercial hollow-fibre membranes in a submerged anaerobic MBR (HF-SAnBR) system, Water Res. 47 (2013) 1277-1288.
[11] G. Skouteris, D. Hermosilla, P. López, C. Negro, Á. Blanco, Anaerobic membrane bioreactors for wastewater treatment: a review, Chem. Eng. J. 198 (2012) 138-148.
[12] E.R. Hall, P.R. Berube, Membrane bioreactors for anaerobic treatment of wastewaters: phase II, WERF report (Alexandria) and IWA Publishing (London), 2006.
[13] D. Martinez-Sosa, B. Helmreich, T. Netter, S. Paris, F. Bischof, H. Horn, Anaerobic submerged membrane bioreactor (AnSMBR) for municipal wastewater treatment under mesophilic and psychrophilic temperature conditions, Bioresource Technol. 102 (2011) 10377-10385.
[14] A. Spagni, S. Casu, N.A. Crispino, R. Farina, D. Mattioli, Filterability in a submerged anaerobic membrane bioreactor, Desalination, 250 (2010) 787-792.
[15] L. Alibardi, R. Cossu, M. Saleem, A. Spagni, Development and permeability of a dynamic membrane for anaerobic wastewater treatment, Bioresource Technol. 161 (2014) 236-244.
[16] H. Chu, Y. Zhang, X. Zhou, Y. Zhao, B. Dong, H. Zhang, Dynamic membrane bioreactor for wastewater treatment: Operation, critical flux, and dynamic membrane structure, J. Membrane Sci. 450 (2014) 265-271.
[17] M.E. Ersahin, H. Ozgun, R.K. Dereli, I. Ozturk, K. Roest, J.B. van Lier, A review on dynamic membrane filtration: Materials, applications and future perspectives, Bioresource Technol. 122 (2012) 196-206.
[18] H. Liu, C. Yang, W. Pu, J. Zhang, Formation mechanism and structure of dynamic membrane in the dynamic membrane bioreactor, Chem. Eng. J. 148 (2009) 290-295. 24
[19] C. Wang, W-N. Chen, Q-Y. Hu, M. Ji, X. Gao, Dynamic fouling behavior and cake layer structure changes in nonwoven membrane bioreactor for bath wastewater treatment, Chem. Eng. J. 264 (2015) 462-469.
[20] X. Zhang, Z. Wang, Z. Wu, F. Lu, J. Tong, L. Zang, Formation of dynamic membrane in an anaerobic membrane bioreactor for municipal wastewater treatment. Chem. Eng. J. 165 (2010) 175-183. [21] C. Loderer, B. Gahleitner, K. Steinbacher, C. Stelzer, W. Fuchs, Dynamic filtration – A novel approach for critical flux determination using different textiles, Sep. Purif. Technol. 120 (2013) 410-414.
[22] J. Ma, Z. Wang, X. Zou, J. Feng, Z. Wu, Microbial communities in an anaerobic dynamic membrane bioreactor (AnDMBR) for municipal wastewater treatment: Comparison of bulk sludge and cake layer, Process Biochem. 48 (2013) 510-516.
[23] J. Ma, Z. Wang, Y. Xu, Q. Wang, Z. Wu, A. Grasmick, Organic matter recovery from municipal wastewater by using dynamic membrane separation process, Chem. Eng. J. 219 (2013) 190-199.
[24] Y. Kiso, Y.J. Jung, T. Ichinari, M. Park, T. Kitao, K. Nishimura, K.S. Min, Wastewater treatment performance of a filtration bio-reactor equipped with a mesh as a filter material, Water Res. 34 (2000) 4143-4150.
[25] D. Batstone, J. Keller, R.I. Angelidaki, S.V. Kalyuzhnyi, S.G. Pavlostathis, A. Rozzi, W.T.M. Sanders, H. Siegrist, V.A. Vavilin, Anaerobic Digestion Model No.1 (ADM1), IWA Publishing, London, 2002.
[26] APHA, AWWA, WEF, Standard Methods for the examination of water and wastewater, 21st ed., American Public Health Association/ American Water Works Association/ Water Environmental Federation. Washington, DC., 2005.
25
[27] J. Ho, S. Sung, Methanogenic activities in anaerobic membrane bioreactors (AnMBR) treating synthetic municipal wastewater, Bioresource Technol. 101 (2010) 2191-2196.
[28] M.E. Ersahin, H. Ozgun, Y. Tao, J.B. van Lier, Applicability of dynamic membrane technology in anaerobic membrane bioreactors, Water Res. 48 (2014) 420-429.
[29] H. Yeo, H.S. Lee, The effect of solids retention time on dissolved methane concentration in anaerobic membrane bioreactors, Environ. Technol. 34 (2013) 2105-2112.
[30] A. Pauss, G. Andre, M. Perrier, S. Guiot, Liquid-to-gas mass transfer in anaerobic processes: inevitable transfer limitations of methane and hydrogen in the biomethanation process, Appl. Environ. Microb. 56 (1990) 1636−1644.
26
Figure captions
Figure 1. Schematic diagram of the bench-scale ADMBR.
Figure 2. a) TMP variations, applied HRT and b) membrane fluxes for the ADMBR.
Figure 3. a) Trends of the effluent total COD (Eff. CODt), effluent filterable COD (Eff. CODf), total COD removal (CODt Rem.) and filterable COD removal (CODf Rem.) in the ADMBR; b) variations of mixed liquor filtered COD (CODf ML), effluent filtered COD (CODf Effl.), and filtered COD removal due to the DM (CODf Rem.) in the ADMBR.
Figure 4. a) Effluent TSS and VSS variations in the ADMBR. b) MLTSS and MLVSS variations in the ADMBR. Arrows indicate the further seeds addition during the experiments.
Figure 5. Biogas production (a), biogas production fractions emitted from the reactor, membrane module and effluent vessel (b) and CH4 content of the biogas emitted from the reactor and from the membrane module (c).
Figure 6. Variations of the total (CODt) and filtered COD (CODf) concentration and removal (Rem.) in the short-term experiment.
Figure 7. Mass balance COD distributions (%) for a) HRT = 2 d; b) HRT= 1 d; c) HRT=0.5 d.
27
Figure 1. Schematic diagram of the bench-scale ADMBR.
28
6
a)
200 HRT TMP
5
HRT (d)
4 3
100
2
TMP (mbar)
150
50 1 0
Flux (L m-2 h-1)
b)
0
30 25 20 15 10 5 0 0
20
40
60
80
100
120
140
Time (d)
Figure 2. a) TMP variations, applied HRT and b) membrane fluxes for the ADMBR.
29
900
100
800
90
700
80 70
600 Eff. CODt Eff. CODf CODt Rem. CODf Rem.
500 400
60 50 40
300
30
200
20
100
10
0
COD removal (%)
Effluent COD (mg L-1)
a)
0 0
20
40
60 80 Time (d)
100
120
400
140
25 ML CODf Eff. CODf CODf Rem.
b)
20
15 200 10
CODf removal (%)
CODf (mg L-1)
300
100 5
0
0 0
20
40
60
80
100
120
140
Time (d)
Figure 3. a) Trends of the effluent total COD (Eff. CODt), effluent filterable COD (Eff. CODf), total COD removal (CODt Rem.) and filterable COD removal (CODf Rem.) in the ADMBR; b) variations of mixed liquor filtered COD (CODf ML), effluent filtered COD (CODf Effl.), and filtered COD removal due to the DM (CODf Rem.) in the ADMBR.
30
b)
Effluent TSS and VSS (mg L -1)
1000
MLTSS and MLVSS (g L -1)
a)
800
TSS VSS
600 400 200 0 12
MLTSS MLVSS
10 8 6 4 2 0 0
20
40
60
80
100
120
140
Time (d)
Figure 4. a) Effluent TSS and VSS variations in the ADMBR. b) MLTSS and MLVSS variations in the ADMBR. Arrows indicate the further seeds addition during the experiments.
31
a)
150
100
50
0
100 80
b)
60
Reactor Membrane Effluent vessel
40 20
100
CH4 Content (%)
c)
0 Reactor Membrane
80
Biogas production fraction (%)
Biogas production (mL d-1)
200
60 40 20 0 0
20
40
60
80
100
120
140
Time (d)
Figure 5. Biogas production (a), biogas production fractions emitted from the reactor, membrane module and effluent vessel (b) and CH4 content of the biogas emitted from the reactor and from the membrane module (c).
32
CODt 400
CODf
CODf Rem.
CODt Rem.
HRT = 1.5 d
HRT = 0.6 d
100
80 200
70 60
100
COD removal (%)
-1 COD (mg L )
90 300
50 0
40 0
20
40
60
80
100
Time (h)
Figure 6. Variations of the total (CODt) and filtered COD (CODf) concentration and removal (Rem.) in the short-term experiment.
33
Waste 3% a)
Effluent 11%
CH4 (g) 19% Unknown 59%
CH4 (d) 8% b)
Waste 2%
Effluent 13%
Unknown 53%
CH4 (g) 24% CH4 (d) 8%
Waste 1% c)
Effluent 25% Unknown 53% CH4 (g) 13% CH4(d) 8%
Figure 7. Mass balance COD distributions (%) for a) HRT = 2 d; b) HRT= 1 d; c) HRT=0.5 d.
34
Highlights
Anaerobic dynamic membranes can be developed at ambient temperature Dynamic membrane over large pore size mesh can obtain efficient solid rejections Biofilm forming the dynamic membrane significantly contributed to organics removal Methane oversaturation occured due to ambient temperature and low HRTs
35