Mass transfer enhancement and improved nitrification in MABR through specific membrane configuration

1 Mass transfer enhancement and improved nitrification in MABR through specific membrane configuration Authors: M. Castrilloa,*, R. Díez-Monteroa,b, A.L. Esteban-Garcíaa, I. Tejeroa Affiliation: a Environmental Engineering Group, Department of Science and Techniques of Water and the Environment. University of Cantabria. Santander, Spain. b Group of Environmental Engineering and Microbiology, Department of Civil and Environmental Engineering. Universitat Politècnica de Catalunya, Barcelona, Spain *Corresponding author: M. Castrillo Tel: (+34)942 206 711 Email address: [email protected] Abstract: One of the main energy consumptions in wastewater treatment plants (WWTPs) is due to the oxygenation of aerobic biological processes. In order to approach to an energy selfsufficient scenario in WWTPs, Membrane Aerated Biofilm Reactors (MABRs) provide a good opportunity to reduce the impact of aeration on the global energy balance. However, mass transfer limitations derived from poor flow distribution must be tackled to take advantage of this technology. In this work, in order to improve mass transfer between biofilm and bulk water, a specific configuration was developed and studied at laboratory scale, aimed at compactness, energy efficiency and high nitrification rates. Nitrification rates were higher in the innovative configuration than in the conventional one, achieving a Volumetric Nitrification Rate (VNR) as high as 575.84 g NH4-N m-3 d-1, which is comparable with confirmed technologies. Regarding energy consumption due to aeration, a reduction of 83.7% was reached in comparison with aeration through diffusers with the same Oxygen Transfer Efficiency (OTE). These results highlight the importance of hydrodynamic conditions and the membranes configuration on treatment performance. Keywords: wastewater; nutrient removal; mass transfer; MABR; nitrification; energy self-sufficiency Abbreviations: Cs Saturation oxygen concentration mg L-1 HRT Hydraulic Retention Time min HRTm Mean Hydraulic Retention Time min ID Inner diameter μm O2inf Dissolved oxygen in influent mg L-1 O2ef Dissolved oxygen in effluent mg L-1 OD Outer diameter μm OOR Oxygen Outgoing Rate g O2 m-2 d-1 OSR Oxygen Supply Rate g O2 m-2 d-1 OTE Oxygen Transfer Efficiency % OTRbl Oxygen Transfer Rate calculated through measures in bulk water g O2 m-2 d-1 OTRgp Oxygen Transfer Rate calculated through measures in gas phase g O2 m-2 d-1 OUE Oxygen Utilization Efficiency % OUEt Oxygen Utilization Efficiency from the oxygen Transferred % S Membrane surface m2 SOTE Standard Oxygen Transfer Efficiency % SOTR Standard Oxygen Transfer Rate g O2 m-2 d-1 SNLR Specific Nitrogen (as NH4-N) Loading Rate g NH4- N m-2 d 1 SNR Specific Nitrification Rate g NH4- N m-2 d-1 TSS Total Suspended Solids mg L-1 KLa20 Volumetric Mass Transfer Coefficient at 20ºC d-1 α Significance Level of Statistical Tests 1. INTRODUCTION Membrane Aerated Biofilm Reactors (MABRs) for wastewater treatment show a huge potential to reduce the environmental and economic impacts of wastewater treatment plants (WWTPs). In MABRs, a biofilm grows on the outer side of gas permeable membranes. Oxygen is supplied to the inner part of the membranes and it diffuses from the lumen to the biofilm, while substrates diffuse from the wastewater to the outer side of the biofilm. This means that, in contrast to biofilm processes with conventional aeration, in MABRs oxygen and substrate transfer into the biofilm in opposite directions. It is this diffusion mechanism that allows for lower air flowrates and pressure, thus resulting in lower energy consumption. Thanks to this technology it is possible to save about 70% of aeration energy in comparison with fine bubble diffusers (Syron, 2014) and even more in comparison with coarse bubble diffusers (Soreanu et al., 2010). In addition, it provides high adaptability to variable conditions due to the easiness to control the factors that determine gas transfer, such as intramembrane pressure (Downing and Nerenberg, 2008) and periodic venting (Perez-Calleja et al., 2017). In spite of its recognised advantages that have made its incipient commercialization possible, some drawbacks still need to be tackled, especially mass transfer limitations that are among the key factors to achieve high efficiency in MABR. In any biofilm process, in order to achieve high substrate removal rates, high available surface for biofilm is needed. This may be reached by designing high specific surface reactors. As an example, Moving Bed Biofilm Reactors (MBBR) are commonly designed with specific surfaces ranging from 500 to 1200 m2 m-3 depending on carrier characteristics (Barwal and Chaudhary, 2014). This is not the case of MABR, in which such high specific surfaces have led to poor performance. Packing density (%v/v) and specific surface (m2 m-3) in experimental set-ups usually vary from 0.16 (Hwang, Cicek and Oleszkiewicz, 2009a) to 1.3 (Brindle, Stephenson and Semmens, 1998) and from 8.9 (da Silva et al., 2018) to 304 (Hwang, Cicek and Oleszkiewicz, 2009b) respectively. The effect of packing density and membranes configuration on mass transfer has been intensively studied in hollow fibre contactors for oxygen stripping with porous membranes through mathematical modelling as well as experimental work. Among others, it has been concluded that the mass transfer coefficients in randomly packed modules can double those of uniformly-packed ones, which is attributed to the occurrence of transverse flow besides the theoretically predominant axial flow (Costello et al., 1993). It has been also stated that channelling and dead zones are the major phenomena that can take place on high packing modules (Wu and Chen, 2000). However, membranes in MABR show a relevant difference with hollow fibre contactors since a biofilm grows over them; therefore the flow pattern in the bulk liquid does not directly affect the gaseous phase mass transfer between the membrane and the biofilm. When the biofilm adheres to the membrane, there not practically exist a stagnant liquid layer between the membrane and the biofilm so the oxygen transfer is even greater than in absence of biofilm (Osa et al., 1997). Nevertheless, hydrodynamics continues playing a crucial role since solutes must diffuse from the bulk liquid to the biofilm going across the boundary layer formed over the biofilm, and the contribution of the resistance of the liquid boundary layer to the overall resistance is considerably higher than the resistance of the membrane when using thin silicone rubber membranes (Pellicer-Nàcher et al., 2013). Concerning similar conclusions, it has been stated that further efforts to reduce the mass transfer resistance should be focused on improving the hydraulics of the membrane module (Côté, Bersillon and Huyard, 1989). Proper flow pattern is essential to achieve high mass transfer rates and, therefore, high removal rates. Membranes layout not only affects the uniformity of flow velocity field, but also the local flow regime. The importance of fluid flow regime in MABR has been recently highlighted (Nerenberg, 2016). The occurrence of transverse flow giving place to convective transfer can improve reactor performance compared with only diffusion transfer occurrence (Ahmadi, Voller and Semmens, 2006). Wei et al. (2012) presented a reactor design to enhance flow velocity distribution with flow direction almost perpendicular to the membranes. In other work, residence time distribution has been experimentally studied at different recirculation flow rates and theoretical hydraulic retention times. It was stated that the higher the recirculation flow rate is, the more similar to full mixed flow pattern is achieved, but no data about membrane specific surface or packing density are given (Wang et al., 2012). With the aim of improving mixing in MABR, recent advances involving low energy consuming devices have been reported (Syron and Byrne, 2015). In general, nowadays MABR are designed with the membranes in bundles, and bundles in rows. Depending on the specific arrangement, different packing densities and distributions may be found. Therefore different velocity fields appear in the reactor: while water velocity is high around the bundles resulting in channelling, low velocity and even stagnant volumes may take place inside the bundle as computational fluid dynamic (CFD) simulations have revealed (Plascencia-Jatomea et al., 2015; Kavousi et al., 2016). To date, MABR CFD simulations have been applied to homogeneously distributed straight membranes but, in spite of providing highly valuable and detailed information, its usefulness to design full scale reactors is still limited. In this work a specific configuration was developed and studied at laboratory scale aimed at high compactness, energy efficiency and high nitrification rates. This configuration consists on a high packing reactor in which the opposite heads of the bundles are set closer than the length of the membranes, resulting in random membrane curvature and occupation of the whole cross section of the reactor. It was hypothesised that this configuration would reduce the thickness of the liquid boundary layer and, as a result, mass transfer would be enhanced. Furthermore, channelling would be avoided, or at least minimized due to the whole occupation of the cross section of the reactor. In addition, a compact bed could retain solid particles, which might avoid the need for secondary clarifier. The general objective of this study was to elucidate the effect of the membranes configuration on the flow pattern and mass transfer properties at different mixing intensities to optimize mixing energy, which is one of the key factors to achieve energy efficient scalable configurations based on MABR. 2. MATERIALS AND METHODS The feasibility of the specific, innovative configuration of the MABR previously described, was studied in comparison with a conventional one with more loosely and parallel fibre distribution. Comparison was made in terms of physical properties (flow pattern and mass transfer) as well as biological performance in a nitrification process. The effect of water velocity throughout the cross section of the reactor was also studied for both configurations testing with four different values: 1, 5, 10 and 15 m h-1. They were achieved by different internal recirculation flow rates provided by a peristaltic recirculation pump (Cole Parmer Masterflex L/S). The lowest velocity is the equivalent to the velocity that would be provided by the raw wastewater feed flow rate plus a nitrate recirculation rate of 300% of the feed flow rate, without any other mixing system. Regarding physical properties, flow pattern and mass transfer were assessed by means of tracer-response experiments with both bare and colonized membranes and dynamic oxygen transfer experiments in clean water respectively. After inoculation with activated sludge and treating synthetic wastewater, nitrification performance was studied in both configurations at the four water velocities. Oxygen transfer was monitored through the gas phase and the liquid phase in order to assess the influence of configuration and water velocity on Oxygen Transfer Rates (OTR) and Oxygen Transfer Efficiency (OTE) in bare membranes and colonized membranes. 2.1. Experimental set-up This study was carried out in two laboratory-scale reactors in which the traditional configuration (R1) and the innovative one (R2) were implemented (Figure 1). 500 µm (OD) x 300 µm (ID) dense polydimethylsiloxane (PDMS) membranes were provided by Oxymem Ltd. 2000 membranes 0.5 m long were distributed in 4 equal bundles providing a total membrane surface of 1.57 m2. The reactor consisted on a rectangular vessel with section of 0.10 x 0.07 m and 0.65 m high. The membranes were vertically placed and parallel to the axial axis of the reactor in R1 (Figure 1B). The same membranes were used in R2, but the inlet heads were set closer to the outlet heads, achieving a compact configuration where the membranes present deformations and curvatures and occupy the whole cross section of the reactor (Figure 1C). Specifically, the height of the module in R2 was 80% of the height of the module in R1. The water level was placed just above the upper heads, giving a specific surface of 453 m2 m-3 in R1 and 554 m2 m-3 in R2. Packing density calculated as the relation between the volume occupied by the membranes and the volume of the bed zone was 10.25 and 11.82 % for R1 and R2 respectively. Figure 1. Reactor set-up diagram (A), picture of R1 (B) and picture of R2 (C). In the reactor, water was continuously fed through a port located in the upper part of one of the narrower faces, just above the membrane bundles, while the outlet was placed in the opposite face below the membrane bundles. The reactor was operated in co-current mode, using ambient compressed air as gaseous phase. A valve was placed in the outlet of the air stream to regulate intramembrane pressure. The reactor was maintained at room temperature, which means that neither cooling nor heating were used. The average registered temperature is indicated for each specific experiment. A piezometer was placed on the bottom of the reactor; nonetheless no head loss was detected during the whole experimentation. 2.2. Tracer-response experiments 2.2.1. Bare membranes and clean water A concentrated solution of sodium chloride (NaCl) was introduced through pulse injection into the feed stream of the reactor. The amount of NaCl was calculated to have an initial concentration of 500 mg L-1 in the reactor supposing ideal complete mixing. The electrical conductivity of the outlet stream was monitored with an electrical conductivity probe (Hach IntelliCAL™ CDC401), which was correlated to NaCl concentration. The theoretical hydraulic retention time (HRT) was 129 min and the experiments lasted at least three HRT. The residence time distribution (RTD) functions were calculated as in Equation (1), where E(t) is the fraction of fluid that spends a time t in the reactor (min-1), C(t) is the tracer concentration in the outlet stream at a time t (mg L-1), and t is the running time (min). 𝐸𝐸(𝑡𝑡)=𝐶𝐶(𝑡𝑡) ∫𝐶𝐶(𝑡𝑡)𝑑𝑑𝑡𝑡 ∞ 0 (1) The mean hydraulic retention time (HRTm) was calculated as: 𝐻𝐻𝐻𝐻𝐻𝐻𝑚𝑚=∫𝑡𝑡𝐸𝐸(𝑡𝑡) ∞ 0 dt (2) 2.2.2. Colonized membranes and synthetic wastewater Additional tracer tests were performed after several weeks of operation treating synthetic 142.12, 147.33, and 141.60 minutes from 1, 5, 10 and 15 m h-1 respectively). These results highlight the importance of hydrodynamics in this kind of reactors. On the one hand, low velocities do not produce enough hydraulic resistance to force the flow to go through narrow and tortuous conducts. On the other hand, a higher energy investment to reach high velocities not only may be worthless, what is more, it can be detrimental. According to these results, the traditional configuration is more prone to produce shortcircuits, giving place to peaks of concentration higher than the maximum achievable concentration considering fully mixing, while the innovative configuration shows a profile that matches a fully mixed tank model from 5 to 15 m h-1. 3.1.2. Colonized membranes with synthetic wastewater When the flow pattern was studied in presence of biofilm, the concentration profiles were similar to those obtained with bare membranes (Figure 3), but showing more pronounced peaks. The only situations in which a sharp peak was not observed at the beginning of the test were at velocities of 5 and 10 m h-1 in R2. Furthermore, in these situations the initial tracer concentration was very close to the theoretical one (5 mg Li+ L-1). Velocities of 1 and 15 m h-1 presented the peaks suggesting the occurrence of short-circuits or canalization with too low or high velocities also in R2. It could be said that the situation found with bare membranes is accentuated when the biofilm has grown over it. These results suggest that the presence of biomass produces additional hydraulic resistance due to the filling of the conducts. It is coherent with the hypothesis that in the traditional configuration, increasing the hydraulic resistance leads to a preferential circulation through wider conducts, which is detrimental for the performance. However, in the innovative one, flow pattern is similar than in absence of biomass within a certain range of velocities due to the homogeneity of the width in the conducts. Figure 3. Tracer concentration (mg Li+ L-1) versus time (min) in R1 (♦) and in R2 (♦) for the tested water velocities, in experiments with synthetic wastewater and colonized membranes. 3.2. Oxygen transfer characterization 3.2.1. Oxygen transfer with bare membranes and clean water According to flow pattern results in clean water, the tests at the lowest velocity in both reactors and at the highest velocity in R1 show a deviation from a fully mixed tank pattern. Therefore, spatial variations in these situations are expected to affect the KLa determination. As shown in Figure 4a and 4b, in R1 raising the water velocity always resulted in an increase of the KLa20 and the specific OTRbl. The increasing in KLa20 and OTRbl with water velocity could be attributed to the liquid boundary layer reduction due to the increasing in mixing intensity. An optimum was not reached within this range of velocities, which suggests that probably an OTRbl equal or higher than the maximum obtained for R2 could be achieved, but at the expense of higher energy consumption. In R2, from 10 m h-1, higher velocity did not result in better results. KLa20 in R2 was always higher than in R1, in fact, even at 5 m h-1 it was higher than at all the velocities tested in R1. According to Figure 4b, OTRbl seems to present lower differences between both configurations than KLa20, being almost equal at the highest and at the lowest velocities. Nonetheless, the highest OTRbl was achieved with the innovative configuration, and this OTRbl was not reached with the traditional one even at higher velocity. Taking into account the results of the tracer-response experiments, the enhancement at intermediate velocities in the innovative configuration may be attributed to the overall effect of a better flow pattern as well as to the reduction of the liquid boundary layer. This better flow pattern can have different effects: on the one hand there is more membrane surface exposed to the bulk liquid due to the reduction of channelling, while on the other hand solutes are better distributed in the bulk liquid. The latter is more evident when the oxygen is measured in the bulk liquid. When the flow pattern is not proper, solutes may find difficulties to diffuse homogeneously and to arrive to the probe. Therefore, this effect is not so evident when the oxygen is measured in the gas phase. Figure 4. A. KLa20 (d-1) and B. specific OTRbl (g O2 m-2 d-1) at the tested water velocities in R1 (□) and in R2 (■). OTRgp was always higher than OTRbl, independently of the configuration. As an example the representation of OTR as a function of dissolved oxygen for experiments at 10 m h-1 is shown in Figure 5. Oxygen bubbles were formed on the surface of the membranes, especially at low mixing intensities, which may mean that the liquid boundary layer gets oxygenated faster than the bulk liquid. This suggests that there is a lag in bulk water oxygenation with respect to membrane transfer. This lag is also intuited in the first part of the graph in Figure 5, at DO < 2 mg L-1. Although OTR is expectable to take its maximum value at low DO concentrations, this cannot be observed when measures for oxygen transfer characterization are taken in bulk water. The deviation from linearity with respect to measures taken in the gas stream may be explained as the delay of bulk water oxygenation due to mixing properties. Figure 5. OTR (mg O2 min -1) in experiments at 10 m h-1 obtained from gas stream (circles) and from water (triangles) in the traditional and the innovative configuration, as a function of dissolved oxygen (mg O2 L-1). 3.2.2. Oxygen transfer and utilization efficiency with bare and colonized membranes In order to compare OTR in clean water and under biological operation, during nitrification the gas line was monitored in the same way than in the oxygenation tests. As it can be observed in Figure 5, OTR determined through the gas stream as a function of dissolved oxygen in water follows a linear trend. Therefore, the linear function for each situation was obtained and the maximum OTRgp was determined as its value when dissolved oxygen in water was 0 mg L-1. The values of maximum OTRgp and OTR in biological experimentation are presented in Figure 6. As a general rule, in oxygenation tests, OTR always increased with water velocity. However, in biological operation OTR followed this trend only between 1 and 10 m h-1, but at 15 m h-1, OTR decreased to 74 and 93% of OTR at 10 m h-1 for the traditional and the innovative configuration respectively. This could be explained by the fact that, with bare membranes, high water velocity contributes to a reduction of the boundary layer between the membrane and the bulk water and OTR is enhanced, but in presence of biofilm, the reduction of the liquid boundary layer does not directly affect the oxygen transfer, but the interchange of solutes (i.e. NH4-N). The tracer experiments revealed that the highest velocity was detrimental for the flow pattern in the reactor. Therefore, although higher velocity could be producing higher oxygen mass-transfer between the membranes and the boundary layer in absence of biofilm, as the results of OTRgp reveal, the distribution of the solutes in the bulk liquid is not appropriate to allow the enhancement of mass transfer between the membrane and the biofilm. Figure 6. OTR (g O2 m-2 d-1) in R1 (white background) and in R2 (black background) at the tested water velocities, measured under biological operation (plain) and calculated from oxygen transfer characterization tests for DO=0 mg L-1 (stripes) (♦) Error bars represent 95% confidence intervals of the mean. In biological operation OTR was between 1.29 and 2.64 times higher than OTRgp in oxygenation tests. The highest absolute augmentation occurred at 10 m h-1, followed by 5 m h-1 and finally the lowest augmentation occurred at 1 and 15 m h-1 in both configurations. In R1, under biological operation no differences in OTR were found between 1 and 5 m h- 1 and between 5 and 15 m h-1, but highly significant differences appeared between all the other possibilities. In R2, the differences between the assayed velocities were less remarkable. Between 1 and 5 m h-1 and between 10 and 15 m h-1 no differences were found, but they appeared between all the other couples, although with less significance than in R1. Significant differences between both configurations only were found at 15 m h-1, meaning that, taking into account only oxygen mass balances in the gas phase it cannot be concluded that the innovative configuration significantly enhances the mass transfer in MABR. The significance values of the Student’s T-test and Tukey's HSD test belonging to this analysis are shown in Appendix section A.1. Regarding OTE, in oxygen transfer characterization with clean water and bare membranes it ranged from 4.36 to 14.30 %, being higher in R2 except for the lowest water velocity, where it was almost equal than in R1 (Figure 7). Since OSR was always the same, OTE in biological experimentation followed the same trend with respect to water velocity than OTR. The maximum OTE reached was 27.89%, which was achieved without implementing optimization strategies in the gas line, like venting (Perez-Calleja et al., 2017), since its optimization was not the aim of this work. It took place at 10 m h-1 in the compact configuration. In the MABR technology the OTE with biofilm is higher than the one measured with clean water, due to the driving force produced by the biofilm. This is a particular difference with conventional activated sludge systems, where the OTE in operation is lower than the OTE measured with clean water. Figure 7. Oxygen Transfer Efficiency (%) in R1 (white background) and in R2 (black background) at the tested water velocities, measured under biological operation (plain) and in oxygenation tests with clean water and bare membranes through measures in the gas phase (striped). Error bars represent 95% confidence intervals of the mean. The oxygen transfer efficiency per unit energy in the experimental conditions of this work, takes the value of 19.6 kg O2 kWh-1. The OTE of 27.89% is in the upper range of standard OTE (SOTE) reported for different diffuser systems at 4.5 m column water submergence (Metcalf and Eddy, 2004); however diffuser systems require a pressure to overcome the water column and friction losses. Furthermore, OTE of porous diffusers is known to significantly decrease under process conditions, on the contrary than in MABR. There are several factors that reduce OTE with respect to SOTE (i.e. surface tension, temperature, clogging, etc.). Taking typical values of the corrector factors (Metcalf and Eddy, 2004), a reduction of at least 30% is expectable. With the resulting OTE, considering fine bubble diffusers, oxygen transfer efficiency per unit energy would be 3.19 kg O2 kWh-1. In other words, a reduction of 83.7% in energy consumption for aeration is possible by making use of MABR technology with the appropriate physical configuration. Not all the oxygen supplied was utilized in nitrification; therefore OUE and OUEt were calculated according to Equations (6) and (7) and presented in Figure 8. OUEt did not present significant differences between both configurations at the tested velocities, but the highest significance occurred at 10 m h-1 (α=0.068). In the case of the traditional configuration, the increase of water velocity resulted in the rise of OTR, but not in a higher oxygen utilization, therefore OUEt decreased between 1 and 10 m h-1. Significant differences occurred only between 1 and 10 m h-1. Regarding the compact configuration, OUEt was maintained almost constant for all the water velocities tested, in fact, no significant differences appeared between different velocities. Consequently, OUE is not improved when raising the water velocity in the traditional configuration, on the contrary than in the innovative one, where OUEt is practically constant and therefore OUE is proportional to OTR. Figure 8. Oxygen Utilization Efficiency (%) from supplied oxygen (OUE) (striped) and from transferred oxygen (OUEt) (plain), in R1 (white background) and in R2 (black background) at the tested water velocities. Error bars represent 95% confidence intervals of the mean. 3.3. Nitrification performance Nitrification performance in this work is referred to membrane surface (SNR in g NH4-N m-2 d-1) and to reactor volume (Volumetric Nitrification Rate (VNR) in g NH4-N m-3 d-1) and their values are graphically presented in Figure 9. Mean values and standard deviation of SNR and VNR are shown in Appendix section A.2. The significance values of the Student’s T-tests and Tukey's HSD tests belonging to this analysis are shown in Appendix section A.3. Figure 9. A. SNR (g NH4-N m-2 d-1) and B. VNR (g NH4-N m-3 d-1) for the traditional (white) and the innovative (black) configurations at the tested velocities. Error bars represent 95% confidence intervals of the mean. For all the cases tested, SNR is higher in the compact system that in the traditional system, although at low water velocity (1 m h-1) differences are not significant. At 5 m h-1 significant differences appear between both configurations and at 10 and 15 m h-1 the differences are highly significant. The most significant differences take place at 10 m h-1. SNR and VNR were compared in pairs between different velocities within each configuration. Regarding R1, no differences are found between the two lowest velocities and between the two highest ones, while highly significant differences appear between the lowest and the two highest velocities and between the two intermediate ones. Significant differences occur between 5 and 15 m h-1. In R2 there are no statistically significant differences between 5 and 15 m h-1, but there exist highly significant differences between all the other couples, except between 1 and 15 m h-1, that are only significant. According to these results, it can be said that 10 m h-1 is the optimum water velocity for both configurations. Accordingly to OTR, SNR and VNR also decreased at the highest water velocity tested, although this trend was statistically noticeable only in the case of the innovative 25. Wei, X., Li, B., Zhao, S., Qiang, C., Zhang, H., Wang, S. 2012. COD and nitrogen removal in facilitated transfer membrane-aerated biofilm reactor (FTMABR). Journal of Membrane Science 389, 257-264, 10.1016/j.memsci.2011.10.038. 26. Wu, J. and Chen, V. 2000. Shell-side mass transfer performance of randomly packed hollow fiber modules. Journal of Membrane Science 172 (1-2), 59-74, 10.1016/S0376-7388(00)00318-5. APPENDIX A.1. Oxygen transfer and utilization efficiency with bare and colonized membranes The tables below show the significance level (α) resulting from the statistical analysis for means comparison between two groups of measures. Two asterisks on a α value means that there exists highly significance difference between both groups, one asterisk means that there exists significance difference between both groups, while no asterisks mean that there is not statistically significance difference. A.1.1. Comparison of OTR between R1 and R2 at different velocities. The table shows the significance values (α) of Student’s T-test. Velocity (m s-1) α 1 0.188 5 0.053 10 0.222 15 0.002** A.1.2. Comparison of OTR between pairs of velocities in R1. The table shows the significance values of post-hoc Tukey's HSD test. Velocity (m s -1 ) 1 5 10 15 1 0.058 0.000** 0.002** 5 0.000** 0.083 10 0.001** 15 A.1.3. Comparison of OTR between pairs of velocities in R2. The table shows the significance values of post-hoc Tukey's HSD test. Velocity (m s -1 ) 1 5 10 15 1 0.107 0.001** 0.002** 5 0.014* 0.045* 10 0.822 15 A.1.4. Comparison of OUEt between R1 and R2 at different velocities. The table shows the significance values (α) of Student’s T-test. Velocity (m s-1) α 1 0.269 5 0.569 10 0.068 15 0.165 A.1.5. Comparison of OUEt between pairs of velocities in R1. The table shows the significance values of post-hoc Tukey's HSD test. Velocity (m s -1 ) 1 5 10 15 1 0.226 0.037* 0.740 5 0.594 0.704 10 0.158 15 A.1.6. Comparison of OUEt between pairs of velocities in R2. The table shows the significance values of post-hoc Tukey's HSD test. Velocity (m s -1 ) 1 5 10 15 1 0.926 1 0.202 5 0.940 0.434 10 0.192 15 A.2. Mean values ± standard deviation of SNR (g m-2 d-1) and VNR (g m-3 d-1) for the traditional and the compact configurations at the tested velocities. SNR (g NH 4 -N m-2 d-1) VNR (g NH 4 -N m-3 d-1) Water velocity (m h-1) Traditional configuration Compact configuration Traditional configuration Compact configuration 1 0.670±0.049 0.688±0.059 238.54±17.44 280.84±23.88 5 0.693±0.080 0.942±0.168 246.77±28.46 383.86±68.44 10 0.956±0.153 1.411±0.165 340.34±54.34 575.84±67.17 15 0.904±0.118 1.118±0.110 321.95±42.17 455.94±44.82 A.3. SNR (g NH4-N m-2 d-1) and VNR (g NH4-N m-3 d-1) results. The tables below show the significance level (α) resulting from the statistical analysis for means comparison between two groups of measures. Two asterisks on a α value means that there exists highly significance difference between both groups, one asterisk means that there exists significance difference between both groups, while no asterisks mean that there is not statistically significance difference. A.3.1. Comparison of SNR between R1 and R2 at different velocities. The table shows the significance values (α) of Student’s T-test. Velocity (m s-1) α 1 0.563 5 0.013* 10 0.001** 15 0.009** A.3.2. Comparison of SNR between pairs of velocities in R1. The table shows the significance values of post-hoc Tukey's HSD test. Velocity (m s -1 ) 1 5 10 15 1 0.982 0.001** 0.006** 5 0.002** 0.014* 10 0.838 15 A.3.3. Comparison of SNR between pairs of velocities in R2. The table shows the significance values of post-hoc Tukey's HSD test. Velocity (m s -1 ) 1 5 10 15 1 0.018* 0.000** 0.000** 5 0.000** 0.131 10 0.005** 15 A.3.4. Comparison of VNR between R1 and R2 at different velocities. The table shows the significance values (α) of Student’s T-test. Velocity (m s-1) α 1 0.006** 5 0.003** 10 0.000** 15 0.000** A.3.5. Comparison of VNR between pairs of velocities in R1. The table shows the significance values of post-hoc Tukey's HSD test. Velocity (m s -1 ) 1 5 10 15 1 0.982 0.001** 0.006** 5 0.002** 0.014* 10 0.838 15 A.3.6. Comparison of VNR between pairs of velocities in R2. The table shows the significance values of post-hoc Tukey's HSD test. Velocity (m s -1 ) 1 5 10 15 1 0.018* 0.000** 0.000** 5 0.000** 0.131 10 0.005** 15