Article pubs.acs.org/est
Kinetic Evaluation of Removal of Odorous Contaminants in a ThreeStage Biological Air Filter Dezhao Liu,† Michael Jørgen Hansen,† Lise Bonne Guldberg,‡ and Anders Feilberg*,† †
Department of Engineering, Faculty of Science and Technology, Aarhus University, Blichers Allé 20, DK-8830 Tjele, Denmark Department of Research and Development, SKOV A/S, Hedelund 4, DK-7870 Glyngøre, Denmark
‡
S Supporting Information *
ABSTRACT: Biofiltration is a cost-effective technology for removing air contaminants from animal facilities. Kinetic analysis can be helpful in understanding and designing the process but has not been performed on full-scale filters treating complex mixtures. In this study, kinetics was investigated in a full-scale biological filter treating air pollutants from a pig facility. Due to the high air flow rates used in the filter, both a plug flow model and a model based on complete mixing were tested with respect to kinetic order and Michaelis−Menten kinetics. Application of these models only gave poor to moderate agreement with air filter removal data. Two alternative kinetic models (Stover-Kincannon model and Grau second-order model) adopted from wastewater biofiltration process analysis were introduced to analyze contaminant removal in the biological air filter. Data analysis demonstrated the applicability of these two models with a high degree of precision on contaminant removal in the biological air filter. Whereas the Stover-Kincannon model demonstrated that pollutant removal rates were related to the mass loading rates, the Grau second-order kinetic model indicated that the removal efficiencies were dependent on air loading rates. Therefore, the kinetic data can be used for comparing biofilter performances and for design purposes.
■
INTRODUCTION
phenomena involved and an uneven distribution of biofilm with time and locations inside the filter. Commonly used kinetic models include microkinetic and macrokinetic models.14 Microkinetic models attempt to cover both bioconversion and mass transfer processes using complex sets of equations which need a multitude of parameters to be determined experimentally. The microkinetic model from Ottengraf 15 is a common foundation and a reference of many other microkinetic models. Macrokinetic models, on the other hand, try to avoid detailed individual processes and regard the biofilter as a ‘black box’ and are thus referred to as empirical models. In general, the parameters for kinetic models have been determined in bench- or pilot-scale experiments. However, to the best of our knowledge, only a few studies have included kinetics analyses of full-scale biological air filters, and these were focused on single contaminant removal.16,17 Due to the complexity and hard-to-control environments, applications of these models in the description and simulation of full-scale biological air filters dealing with complex air pollutants have not been investigated so far. In the present study, seven different macrokinetic models were tested to simulate the performance of a full-scale
Odor emission from intensive pig production is a major cause of nuisance for neighbors and may have negative health and environmental effects.1,2 Technologies implemented for contaminant reduction from intensive pig production include slurry oxidation, chemical treatment, and biological air filters or scrubbers.3−5 Biological air filtration processes are increasingly used as a cost-effective technology for odor nuisance abatement from intensive animal production.4−9 Biofiltration employs microorganisms immobilized within a biofilm fixed to a packing material to break down contaminants present in the air stream. The process is relatively complex involving interactions of absorption, adsorption, and biological degradation.10 In order to reduce cost by reducing pressure drop (by reduced filter volume) across the air filter, empty bed residence time (EBRT) is typically kept less than 10 s in air filters for pig production.4,11 With such short residence time, mass transfer resistance between air and water is likely to limit the removal of compounds with high air-to-water partitioning ratios.12,13 Process kinetics is a useful tool for providing a rational basis for process analysis, control, and design. Although numerous applications have demonstrated odor removal in biological air filters, only a limited number of studies have been performed on the process kinetics analysis.14 This may be due to the complexity of the physical, chemical, and microbiological © 2012 American Chemical Society
Received: Revised: Accepted: Published: 8261
April 1, 2012 July 2, 2012 July 9, 2012 July 9, 2012 dx.doi.org/10.1021/es301295m | Environ. Sci. Technol. 2012, 46, 8261−8269
Environmental Science & Technology
Article
biological air filter for livestock production. Apart from other models, the Stover-Kincannon model and the Grau secondorder model, which often have been used for wastewater biofiltration processes, were introduced for the first time to simulate the performance of a biological air filter. The objective of the present study was to evaluate the applicability of different kinetic models for describing the removal of a large number of contaminants in a full-scale biological air filter and to obtain additional knowledge about biofilter function from the kinetic analysis. In a previous study,18 the applicability of Proton-TransferReaction Mass Spectrometry (PTR-MS) for odor measurements in the biological air filter was demonstrated.
a manometer) and damper position and was calibrated by comparison with a Fancom Measurement wing (AT(M) unit 80, Fancom, Panningen, The Netherlands).20 The minimum and the maximum ventilation rates for the biofilter were ca. 6,000 m3 h−1 and 35,000 m3 h−1, respectively. The averaged EBRT is ca. 1.5 s in total. Typically, the pressure drop across the filter is in the range of 30−50 Pa. PTR-MS Measurement of Contaminants Concentrations. A High-Sensitivity PTR-MS (Ionicon Analytik, Innsbruck, Austria) was utilized for measuring contaminants in the biological air filter. PTR-MS has previously been demonstrated to be a useful tool for analysis of pollutants in air emitted from intensive pig production.21 Standard drift tube conditions at 600 V and 2.1−2.2 mbar were applied for the PTR-MS during the measurement period. Contaminants from every stage were measured in turn in a continuous mode with 20 min of measurement for each stage. Measurements were carried out in the middle of the void area between stages to achieve homogeneous air flow and avoid any edge effects on measurements. A more detailed description of the contaminants measurement can be found in a previous study.18 Evaluation Methods. The removal efficiency (RE) for a component in a biofilter system can be expressed as10
■
MATERIALS AND METHODS Three-Stage Biological Air Filter. A full-scale three-stage biological air filter (Figure 1) from SKOV A/S was installed
RE = (Ci − Co)/Ci*100%
(1)
where Ci is the concentration at the inlet (g m−3), and Co is the concentration at the outlet (g m−3). Elimination capacity (EC), on the other hand, is expressed by10 EC = (Ci − Co)*Q /Vf
(2)
where Q is the air flow rate (m3 h−1), and Vf is the bulk volume of filter material (m3). Kinetics Analysis. Several models are commonly used to describe the overall kinetics of biological air filters. First-order kinetics, zero-order kinetics with diffusion rate limitation, or Michaelis−Menten type model were mostly applied for simulation of biodegradation in biofilters. Although the plug flow model has commonly been used for most biological filters,14,22−25 the completely mixed-model was also considered in biological filters for treatment of wastewater26 or waste air.27 Due to the special design of the filter in the present study, the air flow is mixing in the vertical direction (perpendicular to the air flow direction) while passing through the filter but not mixing in the horizontal direction, which is also perpendicular to the air flow direction. In other biological filters this is not the case. Moreover, the air will mix while passing through the void space between stage 1 and stage 2 and the void space between stage 2 and stage 3. Thus, the assumption of a completely mixed system is interesting to test, especially when the residence time is short due to the high air flow rate. The kinetic parameters (rate constants, etc.) that result from the application of different kinetic models are defined for the gas phase concentration of the contaminants. This means that the expressions characterize the overall contaminant removal and include all steps and processes: air−water mass transfer, water-biofilm mass transfer, and microbial degradation. Plug Flow with First-Order Kinetics. First-order kinetics with plug flow has widely been considered for modeling biological filters. Assuming plug flow without dispersion at steady state, the following is obtained from a local mass balance, i.e. QCl + (AΔl)Rx = QCl+Δl + (AΔl)(∂Cl/∂t)
Figure 1. Scheme of the three-stage cross-current biological air filter: 1 - air distribution plate; 2 - stage 1; 3 - air flow; 4 - water flow; 5 - stage 2; 6 - stage 3; 7 - water irrigation; 8 - water supply; 9 - overflow; 10 drainage.
next to a pig production facility in Mors, Denmark (56° 48′ 00″ N, 8° 52′ 00″ E). 350 growing-finishing pigs were fed inside the pig facility during the measurement period (19/8−12/9, 2010). The packing material used in all stages was made of cellulose pads (CELdek 7060-15, Munters AB, Kista, Sweden) with specific surface area of 383 m2 m−3.19 The material consists of specially corrugated cellulose paper sheets with different flute angles, one steep (45°) and one flat (15°) that have been bonded together (Figure 1). Stages 1 and 2 have a depth of 15 cm (H × W × D = 2 × 5.0 × 0.15 m), while the third stage of the filter has a depth of 60 cm (H × W × D = 2 × 5.0 × 0.6 m). Stages 1 and 2 were irrigated with recirculated water with estimated irrigation speeds of 3.3 m3 m−3 h−1 and 2.2 m3 m−3 h−1, respectively. Stage 3 was only supplied by humidified air from stage 2. The electric conductivity in reservoir 2 (under stage 2; 2.94 ± 0.35 mS cm−1; see Table S1 (Supporting Information) for process control parameters) was controlled to be stable by discharging wastewater from reservoir 1 (under Stage 1; 9.46 ± 2.58 mS cm−1; Table S1) and adding fresh water to reservoir 2. The air flow rate was measured by the Dynamic Air system (SKOV A/S, Glyngeore, Denmark) placed in the exhaust outlets from the air filter and logged via Farm Online (SKOV A/S, Glyngoere, Denmark). The Dynamic Air system calculated air flow rate based on pressure difference (via 8262
dx.doi.org/10.1021/es301295m | Environ. Sci. Technol. 2012, 46, 8261−8269
Environmental Science & Technology Q ∂C ∂C =− + Rx ∂t A ∂l
Article
If the reaction follows the zero-order kinetics (Rx= k0, where k0 is zero-order rate constant), then eq 9 could be reduced to the following
(3)
where C is the concentration in the biological air filter (g m−3); ∂C/∂t is the rate of change of concentration (g m−3 h−1) and will equal to zero under a steady-state assumption; A is the filter cross section area (m2); and Rx is the overall reaction rate (g m−3 h−1). Under the first-order kinetics assumption, Rx= k·C, where k is the first-order rate constant (h−1). By integration of eq 3 along the pathway from inlet to outlet, the following formula is established23 Q /Vf = k /ln(Ci /Co)
Ci − Co = k 0·(Vf /Q )
Michaelis−Menten Kinetics in the Complete Mixed System. Instead of assuming that the overall degradation follows first-order kinetics, Michaelis−Menten kinetics is used in this model under a steady-state assumption. Although no reference application is available (to our knowledge) for this model, it is still interesting to test the applicability of the model in this study. Equation 9 can then be reduced to
(4)
Vf Co
And the EC can be expressed as a function of the concentration as follows EC = k(Ci − Co)/ln(Ci /Co) = kCln mean
Q (Ci − Co)
where Cln mean (g m ) is the logarithmic (base e) mean of the inlet and outlet concentration. Rewriting eq 3 leads to a relationship between the RE and the volumetric air loading rate (VALR) of the biofilter as follows
Ci − Co
=
1 − (Co/Ci)1/2 = kd·EBRT = kd·(Vf /Q )
Umax(QCi /Vf ) dC = dt KB + (QCi /Vf ) (8)
(15)
where dC/dt is the rate of contaminant utilization (g m−3 h−1); Umax is the maximum utilization rate constant (g m−3 h−1); and KB is the saturation value constant (g m−3 h−1). In this model, the contaminant utilization rate is expressed as a function of the contaminant loading rate by monomolecular kinetic for biofilm reactors such as biological filters. The Stover-Kincannon model was originally proposed for rotating biological contactor (RBC) systems.31 The model was based on the surface mass loading rate (QCi/A) instead of the volumetric mass loading rate (QCi/ Vf) in the equation above in order to represent relationships to the total attached-growth active biomass in the RBC system. Due to the difficulty in measuring the active surface area that supports the biofilm growth, several authors have revised the model and used the volume of the biological reactors for different applications.26,32 In the present study, volume was used due to the simplicity and its clear relation to the surface area in the biological air filter. Linearization of eq 15 will give the relationship as below26
Plotting [(Vf/Q)/(Ci−Co)] against (1/CLn) results in a straight line with an intercept of 1/Rm and a slope of Km/Rm. First-Order Removal Kinetics in the Completely Mixed System. Under the assumption of a completely mixed system, the following equations are obtained27 (9)
Under first-order kinetics, the equation above can be reduced to Ci − Co = k·Co Vf / Q
(14)
Thus, the model becomes a macrokinetics model and the plot of 1 − (Co/Ci)1/2 against EBRT should give a linear relationship if the assumptions of this model are valid. Stover-Kincannon Model. In this paper, the StoverKincannon model with the assumption of steady state is used as below31,26
(7)
∂C ∂t
(13)
where a is the interfacial surface area per unit volume (m m−3), De is the effective diffusion coefficient (cm2 s−1), b is the distribution coefficient of the component (−), and δ is the biofilm thickness (cm). With the definition of kd = (ak0De/2bCiδ)1/2,23 it follows that kd is a function of the operating conditions of the biofilter system and kd is constant under steady-state conditions. Equation 13 can be reduced to the following
K m ln(Ci /Co) K 1 1 1 · + = m· + R m Ci − Co Rm R m CLn Rm
QCi + Vf ·R x = QCo + Vf
(12)
2
where Rm is the maximum bioreaction rate per unit biofilter volume (g m−3 h−1), and Km is the saturation (Michaelis− Menten) constant (g m−3) (gas concentration). Integration of eq 3 will then give the following result23 Vf / Q
K 1 Co + m Rm Rm
Co = Ci[1 − EBRT (ak 0De /2bCiδ)1/2 ]2
(6)
where G (=Q/Vf) is the VALR (m3 m−3 h−1). This relationship describes that the RE is dependent on the VALR regardless of the actual inlet concentration if the degradation is following first-order kinetics. Plug Flow Model with Michaelis−Menten Kinetics. Michaelis−Menten kinetics is commonly used to characterize rates of biofiltration processes.22,28−30 By application of Michaelis−Menten kinetics, the overall reaction rate Rx is defined as below R mC Rx = Km + C
=
By plotting Vf Co/[Q·(Ci−Co)] against Co, Rm and Km can be obtained if the assumptions of this model are valid. Zero-Order Kinetics with Diffusion Rate Limitation. A zero-order kinetic model with diffusion rate limitation under steady-state conditions was proposed by Ottengraf15 and often used in biofiltration applications24,27
(5)
−3
RE = (Ci − Co)/Ci = 1 − e−k(Vf / Q ) = 1 − e−k / G
(11)
(10)
The first-order rate constant k can then be obtained by plotting [(Ci−Co)/(Vf/Q)] against Co if the assumptions of this model are valid. 8263
dx.doi.org/10.1021/es301295m | Environ. Sci. Technol. 2012, 46, 8261−8269
a
8264
17 34 44 48 58 59 60 62 72 74 86 88 94 102 108 117 122 126 131
ammonia hydrogen sulfide acetaldehyde methanethiol acetone trimethylamine acetic acid dimethyl sulfide 2-butanone propanoic acid 2,3-butanedione butanoic acid phenol+DMDSd C5 carboxylic acids 4-methylphenol indole 4-ethylphenol dimethyl trisulfide 3-methylindole
277 ± 72 27 ± 9 0.83 ± 0.25 1.32 ± 0.33 1.05 ± 0.24 1.62 ± 0.59 48 ± 15 0.51 ± 0.16 0.54 ± 0.15 13 ± 4 0.25 ± 0.07 9.6 ± 3.2 0.47 ± 0.10 2.96 ± 0.90 2.35 ± 0.66 0.18 ± 0.05 0.33 ± 0.09 0.03 ± 0.01 0.11 ± 0.05
mass load (g m day )
−1
79 75 93 13 88 92 99 15 83 99 83 99 88 99 98 95 93 39 89
± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± 7 11 4 13 7 5 1 16 8 1 7 1 3 1 1 1 2 17 8
RE (%) 3749 3395 6733 326 5649 6405 14793 433 4471 13236 4552 14352 4959 12236 10257 9675 7694 2704 9448
k (h )
−1
0.20 0.55 0.00 0.01 0.40 0.40 0.27 0.05 0.36 0.51 0.41 0.61 0.00 0.57 0.20 0.09 0.08 0.42 0.51
R
2
1873 4075 12342 −61 7635 5510 18644 66 5567 16420 6883 17248 8187 13717 11636 9140 8939 5559 8665
k3rd
a
2
0.22 0.43 0.66 0.00 0.60 0.46 0.70 0.00 0.56 0.73 0.67 0.76 0.51 0.63 0.52 0.41 0.65 0.43 0.58
R
first-order kinetics (plug flow) 0.0086 0.14 0.83 2.8 0.53 0.17 0.0002 541 6.9 0.0019 10.5 0.0008 3.0 0.014 0.052 0.21 2.8 259 1.2
intercept 1.209 1.201 1.101 5.146 1.117 1.074 1.002 16.56 1.052 1.007 1.114 1.005 1.194 1.011 1.011 1.022 1.013 1.278 1.029
slope 0.7571 0.8197 0.9795 0.0018 0.8086 0.9845 1.0000 0.0012 0.8438 0.9999 0.8422 0.9999 0.9816 0.9996 0.9992 0.9995 0.9931 0.6678 0.9994
R
2
116 7.2 1.2 1.9 5.9 5000 0.15 526 0.096 1250 0.33 72 19 4.8 0.36 0.004 0.85
Umaxb
Stover-Kincannon model 141 8.6 1.3 2.1 6.4 5009 0.15 530 0.11 1256 0.40 73 19 4.9 0.36 0.0049 0.87
KBc 0.0001 0.0002 2 × 10−5 0.0012 8 × 10−5 3 × 10−5 5 × 10−7 0.0027 0.0001 2 × 10−6 1 × 10−4 2 × 10−6 6 × 10−6 3 × 10−6 5 × 10−6 3 × 10−6 2 × 10−5 0.0003 3 × 10−6
intercept 1.040 0.861 1.019 4.204 0.933 0.991 1.001 7.623 0.932 1.001 0.972 1.000 1.151 1.002 1.003 1.013 1.001 0.884 1.014
slope 0.9297 0.8115 0.9942 0.1312 0.9618 0.9910 1.0000 0.0006 0.9151 0.9999 0.9490 1.0000 0.9898 0.9998 0.9998 0.9996 0.9958 0.5395 0.9853
R2
Grau second-order kinetics 0.033 0.0018 0.0005 1 × 10−5 0.0002 0.0008 1.33 7 × 10−5 0.090 3 × 10−5 0.072 0.001 0.013 0.0064 0.0009 0.0002 1 × 10−6 0.0006
k2 (d−1)
First-order kinetic constant (k) for the third stage of the biological air filter. bMaximum biodegradation rate (g m−3 h−1). cSaturation constant (g m−3 h−1). dPhenol and dimethyl disulfide (DMDS).
MW
compounds
−3
Table 1. Averaged Mass Load (± Standard Deviation) and Averaged Removal Efficiency (RE, ± Standard Deviation) of Contaminants Measured in the Three-Stage Biological Air Filter Together with the Applications of First-Order Kinetics (Plug Flow), the Stover-Kincannon Model, and the Grau Second-Order Kinetics to the Biological Air Filter
Environmental Science & Technology Article
dx.doi.org/10.1021/es301295m | Environ. Sci. Technol. 2012, 46, 8261−8269
Environmental Science & Technology Vf Q (Ci − Co)
=
KB Vf 1 · + Umax QCo Umax
Article
hydrogen sulfide (H2S) had relatively higher RE than other sulfur compounds (e.g., methanethiol), although it has lower solubility in the biofilter liquids.39 Since H2S is usually considered as an easily biodegradable compound,10 solubility and mass transfer are considered as limiting factors for removal of this compound. On the other hand, methanethiol has been suggested to be moderately biodegradable,10 and therefore the removal could be considered to be limited by both biodegradation and mass transfer (solubility). Although methanethiol has a lower Henry’s law constant (atm M−1), the solubilities (i.e., air−water partitioning) of H2S and methanethiol are both pH dependent, and a higher pH will increase the overall solubility40 of both compounds (taking dissociation products into account). However, the biofilter liquids pH (e.g., average 6.95 in reservoir 1; Table S1) is much closer to the pKa value (6.98) of H2S41 than the pKa value (ca. 10) of methanethiol.40 This means that only dissociation of H2S is significant and will cause the overall solubility of H2S to exceed that of methanethiol. Due to the control of the filter, conductivity and pH were relatively constant through the measurement period. Thus, it can be assumed that solubility was relatively constant. In addition, it was demonstrated in a recent study that air−liquid partitioning of sulfur compounds in biotrickling filter liquids in general is close to the partitioning with respect to pure water.39 In addition, the individual EC for butanoic acid and 4methylphenol almost equaled the individual mass loading rate (Figure S1; Supporting Information), and this indicated that the critical elimination capacities 10 of these two compounds were not reached for the biofilter in the present study. On the other hand, elimination capacities of both H 2 S and methanethiol were lower than the mass loading rates of the two contaminants, respectively (Figure S1). In particular, methanethiol had very low EC, and this is ascribed to the limiting factors for the removal mentioned above. Plug Flow First-Order Removal Model. According to eq 4, the EC will depend linearly on the logarithmic mean of the inlet and outlet concentrations, and the linearity will give firstorder kinetic constant if the model assumptions are valid. Examples of such linear relations are shown in Figure S2 and Figure S3 for the three-stage filter and for the third stage of the filter, respectively. Obtained first-order kinetic constants are given in Table 1. As shown in Table 1, only a few compounds gave reasonable correlation coefficient (R2) when regressions were based on the three-stage biofilter. On the other hand, the regressions that were only based on the third stage of the filter obtained higher correlation coefficients except for the sulfur compounds (Table 1). Since the depth of the third stage of the filter is much bigger than the other two stages, the mix and flow condition will be different from that in the other two stages due to the special design of the filter and this will influence the kinetics modeling performance. Despite the complexity of the conditions in the full-scale filter, the linearity obtained in the third stage of the filter is considered to be reasonable, and therefore the assumption of first-order (with plug flow) could basically be applied for the third stage of the filter. By assuming first-order kinetics for the third stage of the biological air filter, the first-order kinetic constants for the third stage of the filter were found to have a negative correlation to natural logarithm of Henry’s law constants as shown in Figure S4. The correlation between averaged removal efficiencies (the third stage) and natural logarithm of Henry’s law constants showed a similar trend. Higher removal efficiencies and higher
(16)
By plotting Vf/[Q(Ci−Co)] as a function of Vf/(QCo), a straight line with an intercept of 1/Umax and a slope of KB/Umax results. Grau Second-Order Kinetics. The Grau second-order model is based on the following kinetic expression33 −
dC = k 2·X ·(C /Ci)2 dt
(17)
where k2 is the second-order removal constant (d−1), and X is the biomass concentration (g m−3). The second-order kinetics was originally based on a broader Monod equation for multicomponent substrate removal by activated sludge33 and was later used for kinetics analysis of several different types of biological reactors34,35 including biological filters36,37 but not air filters. The model could maintain the ‘order of reaction’ analogy to chemical reaction kinetics, which simulates random and gradual diminuition of components with time. The model expresses in this way the decrease of removal rate caused by a reduced number of components and thus the decrease in total substrate concentration with time.33 A previous study based on a two-stage cellulose filter showed a significant decrease of aerobic activity through the filter.38 This result indicated the decrease in the abundance of active micro-organisms along the filter direction. If the degradation rate in eq 17 is related to (C/ Ci) as a consequence of diminished concentration as air passes through the filter, a second-order kinetic expression (eq 17) could give a reasonable simulation of the biodegradation process in the three-stage cellulose filter used in this study. By integration and subsequent linearization of eq 17, the following equation is obtained EBRT ·Ci Ci = EBRT + Ci − Co k 2· X
(18)
If Ci/(k2·X) is considered as a constant, and (Ci−Co)/Ci is replaced by RE, the following equation will then be obtained EBRT = n·EBRT + m RE
(19)
where m and n are constants, and n is close to 1. By plotting EBRT/RE as a function of EBRT, a linear relationship should be obtained under the assumptions of this model.
■
RESULTS AND DISCUSSION Performance of the Three-Stage Biological Air Filter. The averaged volumetric mass loadings of contaminants together with the averaged removal efficiencies by the threestage biological air filter were calculated and are presented in Table 1. Regardless of the very short residence time through the filter, the biofilter generally had a good reduction for most of the contaminants which had relatively low mass loading rates (compared to e.g. air from municipal WWTPs) except for ammonia. On the other hand, volatile sulfur compounds had relatively low removal efficiencies compared to the other contaminants. Especially, methanethiol and dimethyl sulfide (DMS) only exhibited around 15% removal (with high variabilities) in the three-stage biological air filter. Due to the very short residence time and the low solubility of those sulfur compounds in the biofilter liquids,39 the mass transfer process is generally considered as a limiting factor in order to achieve higher removal efficiencies for those compounds.13 However, 8265
dx.doi.org/10.1021/es301295m | Environ. Sci. Technol. 2012, 46, 8261−8269
Environmental Science & Technology
Article
first-order kinetic constants were observed for more watersoluble compounds. If we assume that all the compounds have similar mass transfer coefficients (KLa) from gas to liquid under the same operating conditions (air flow rates, water flow rates, etc.), a lower Henry’s law constant will cause higher concentration potential (CG/H−CL) between the gas phase and liquid phase.10 As a result, the overall mass transfer rate (dCL/dt = KLa(CG/H−CL)) will be higher for compounds with lower Henry’s law constants. Assuming that the biodegradation rates for those compounds are not limiting, those compounds will have higher removal efficiencies in the biofilter. However, due to the very low REs and the high variability (of REs) for methanethiol and DMS (Table 1 and Figure S5), it was difficult to find any correlations for these two compounds (for all models in this study). For H2S, on the other hand, a higher correlation coefficient was observed for the three-stage filter (R2 = 0.55) than the third stage (R2 = 0.43) of the filter although the kinetic constants did not deviate far from each other (Table 1). The measured RE of H2S in the three-stage biofilter is shown as a function of the VALR in Figure S6. The line shows the calculated RE as a result of first-order kinetics (k = 3395 h−1) which seems to give a reasonable agreement with the measured RE. Plug Flow Model with Michaelis−Menten Kinetics. By plotting eq 8, the Michaelis−Menten kinetic coefficients Rm and Km could be obtained (Figure S7 and Table S2). Similar to firstorder kinetics, the linear correlation coefficients for the third stage of the filter were in most cases higher than for the complete three-stage filter. Exceptions include H2S, for which a higher degree of linearity was observed when applying the model to the complete three-stage filter. The Michaelis− Menten kinetics with plug flow assumption seems to be suitable for simulation of some compounds (e.g., carboxylic acids) when applied to the third stage of the filter but not for other compounds (e.g., ammonia and trimethylamine) (Table S2). On the other hand, when applied to the three-stage filter, this model is only suitable for H2S. Rm and Km values for the threestage biofilter are calculated, but they might not even be close to the true values due to the low correlation of the model (Table S2). First-Order Removal Kinetics in the Complete Mixed System. By plotting [(Ci−Co)/(Vf/Q)] against Co, a linear relationship should be obtained if the assumption of this model is correct. However, no linearity was observed for any of the compounds (R2 ≈ 0 for most of the compounds) for the threestage biofilter or the third stage of the biofilter (data not shown). The only exception was ammonia for which a fairly linear relationship for the three-stage filter (R2 = 0.57) was demonstrated. On the other hand, higher linear correlation coefficients were observed for all compounds when applying eq 11 (zero-order kinetic) to the three-stage biofilter (Table S3). The linearity was poor (R2 < 0.1) for all compounds for the third stage of the filter when applying eq 11 (data not shown). Michaelis−Menten Kinetics in the Completely Mixed System. When Michaelis−Menten kinetics is used under the assumption of complete mixing, the plot of Vf Co/[Q·(Ci−Co)] against Co will give Rm and Km from the slope and intercept (Figure S8 and Table S4). In general, reasonable correlation coefficients were obtained for most compounds when applying eq 12 to the three-stage biofilter. On the other hand, poor correlations were observed when the model was applied to the third stage of the filter, as expected from physical considerations (Table S4).
Zero-Order Kinetics with Diffusion Rate Limitation. Poor correlation performance was obtained for most compounds when applying eq 14 to the three-stage biofilter except for a few compounds such as H2S and acetone (Table S3). Stover-Kincannon Model. By plotting Vf/[Q(Ci−Co)] against Vf/(QCo) using the Stover-Kincannon model (Figure 2), Umax and KB could be estimated by the intercept and the
Figure 2. Substrate removal model plot for hydrogen sulfide (a), butanoic acid (b), 4-methylphenol (c), and acetone (d) using the Stover-Kincannon model.
slope (Table 1). For example, Umax and KB were estimated to be 7.2 g m−3 h−1 and 8.6 g m−3 h−1, respectively for H2S. High correlation coefficients (R2 = 0.82) for the regression lines, which prove the applicability of the Stover-Kincannon model, were obtained for most of the compounds except for methanethiol and DMS (Table 1). By rearranging the equation of the Stover-Kincannon model, the following two equations could be obtained24 Vf =
QCi [UmaxCi /(Ci − Co)] − KB
(20)
UmaxCi KB + (QCi /Vf )
(21)
Co = Ci −
Thus, eq 21 could then be employed to determine the outlet concentration for a given volume of filter and inlet concentration. If not, the required volume of the filter could be determined for a given inlet and outlet concentration from eq 20. For example, introduction of the values of Umax (7.2 g m−3 h−1) and KB (8.6 g m−3 h−1) for H2S to the relationship above will give Co = Ci −
7.2Ci 8.6 + (QCi /Vf )
(22)
Thus, the outlet concentration of H2S could be calculated according to eq 22 at a given VALR and inlet concentration. Grau Second-Order Kinetics. A plot applying eq 19 will give a linear regression line between EBRT/RE and EBRT for individual compounds under the Grau second-order kinetics assumption (e.g., Figure 3). The obtained intercept and slope shown in Table 1 are equal to m and n in eq 19, respectively. The high correlation coefficients shown in Table 1 confirmed the applicability of the Grau second-order kinetics for most 8266
dx.doi.org/10.1021/es301295m | Environ. Sci. Technol. 2012, 46, 8261−8269
Environmental Science & Technology
Article
In addition, neither assumptions of plug flow nor a completely mixed system could sufficiently simulate the filter conditions. This may be due to the design of the filter, uneven distribution of biofilm, high air flow rates, and/or large void spaces between the stages. Whereas the plug flow model may be appropriate for the third stage of the filter, the completely mixed-model may be more applicable to the first two stages due to the large void space between the stages. On the other hand, the entire filter may be modeled by a combination of a plug flow model and complete mixed model (e.g., plug flow model for filter body and complete mixed model for void spaces). Furthermore, neither zero-order kinetics (with reaction limitation, completely mixed system) nor zero-order kinetics with diffusion rate limitation gave good regression performance and indicated that the pollutant removal in this study was not following the zero-order kinetics. Due to the low concentrations of contaminants in this study, it was expected that the removal of these compounds follow closely first-order kinetics rather than zero-order kinetics (with reaction limitation).10 On the other hand, zero-order kinetics with diffusion rate limitation to some extent had better regression performance (which was expected) for contaminants with lower solubility in water (such as H2S and 2-butanone) except for methanethiol, DMS, and acetaldehyde. For methanethiol and DMS, no model was able to simulate the data well due to the very low removal efficiencies and high variabilities. However, it is unknown why acetaldehyde in this model behaved differently from other compounds with similar water solubility. One possible explanation is that acetaldehyde could be produced during the biodegradation process. Application of the Stover-Kincannon model and the Grau second-order kinetics model were much more promising than the other models evaluated in the present study. The StoverKincannon model showed that the contaminant degradation rate is dependent on the mass loading rate which is related to inlet concentration and VALR. Without considering kinetic order, this model could be used at any loading condition. At the same time, this model does not depend on simulation of diffusion and fluid dynamics, which are difficult to measure and simulate.35 The Grau second-order model, on the other hand, demonstrated the relationship between RE and EBRT of pollutants in the filter with the assumption of a constant ratio between compound concentration and biomass concentration through the filter. Since EBRT is the reverse of VALR, the Grau second-order kinetic model indicated the connection between RE and VALR in the filter. High regression coefficients (R2) for most of the contaminants obtained from both models confirmed that the Stover-Kincannon model and the Grau second-order model are both suitable tools to represent most pollutants removal in the biological air filter in this study. Furthermore, a plot (Figure S9) of simulated EC against experimental EC demonstrated the better performance of both models compared to other models such as first-order kinetics with plug flow for H2S. Similar results were seen for compounds other than H2S. Both the Stover-Kincannon model and the Grau secondorder kinetics can therefore be employed for biofilter design purposes (e.g., prediction of the performance of a filter or calculation of the required size for a specified efficiency). According to the two models, the design of the filter should focus on increasing the mass loading rate (=QCi/Vf) and decreasing the VALR (=Q/Vf) at the same time. In practice, dealing with a higher concentration (Ci) within the capacity
Figure 3. Substrate removal model plot for hydrogen sulfide (a), butanoic acid (b), 4-methylphenol (c), and acetone (d) using Grau second-order kinetics.
compounds but not for methanethiol and DMS. The model results clearly show that the RE for a specific compound is only dependent on EBRT which is reverse to VALR. The secondorder rate constant (k2) could be estimated according to the equation n = Ci/(k2X). Due to the variation of biomass with time and location in the biological air filter, the biomass concentration (X) within the measurement period was estimated with an average thickness of biofilm of 200 μm for the first and second stage and 20 μm for the third stage of the filter. Based on the averaged inlet concentration, the secondorder rate constants were calculated and are shown in Table 1. By rearranging eq 19, the following relations are obtained
RE =
EBRT n·EBRT + m
(23)
m ·RE (24) 1 − n·RE Thus eq 23 could then be employed to determine RE for a given EBRT. If not, the required EBRT (or VALR) could be determined for a given RE from eq 24. For example, assuming that expected RE for H2S is 90%, then the required EBRT is calculated to be 2.9 s (equivalent to a maximum VALR of 1241 m3 m−3 h−1). Evaluation of Kinetic Models. Seven possible contaminant removal models were applied to the full-scale biological air filter for modeling the performance. First-order kinetics with plug flow assumption only gave low to moderate regression coefficients for the third stage of the filter for most compounds and even lower regression coefficients for the entire filter. Similarly, Michaelis−Menten kinetics with plug flow assumption could only be applied to the third stage of the filter with low to moderate degree of precision. On the contrary, only Michaelis−Menten kinetics could be applied to some extent to the entire filter (not the third stage of the filter) under the complete mixing-assumption. The difference between the regression coefficients obtained for the third stage of the filter and for the entire filter by the same model may partly be due to the different operating conditions for the different stages of the filter. The first two stages use recycling water with a constant irrigation rate and could thus be considered as biotrickling filter stages. The third stage of the filter, on the other hand, is humidified only with water from other stages of the biological air filter and could therefore be considered as a biofilter stage. EBRT =
8267
dx.doi.org/10.1021/es301295m | Environ. Sci. Technol. 2012, 46, 8261−8269
Environmental Science & Technology
Article
(8) Sheridan, B.; Curran, T.; Dodd, V.; Colligan., J. Biofiltration of odour and ammonia from a pig unit- a pilot-scale study. Biosystems Eng. 2002, 82 (4), 441−453. (9) Kristiansen, A.; Pedersen, K. H.; Nielsen, P. H.; Nielsen, L. P.; Nielsen, J. L.; Schramm, A. Bacterial community structure of a fullscale biofilter treating pig house exhaust air. Syst. Appl. Microbiol. 2011, 34 (5), 344−352. (10) Devinny, J. S.; Deshusses, M. A.; Webster, T. S. Biofiltration for air pollution control; Lewis Publishers: Boca Raton, 1999; pp 5−22. (11) Chen, L.; Hoff, S. J.; Koziel, J. A.; Cal, L. S.; Zelle, B.; Sun, G. Performance evaluation of a wood-chip based biofilter using solidphase microextraction and gas chromatography-mass spectroscopyolfactometry. Bioresour. Technol. 2008, 99 (16), 7767−7780. (12) Devinny, J. S.; Ramesh., J. A phenomenological review of biofilter models. Chem. Eng. J. 2005, 113 (2−3), 187−196. (13) Nielsen, A. M.; Nielsen, L. P.; Feilberg, A.; Christensen., K. V. A method for estimating mass-transfer coefficients in a biofilter from membrane inlet mass spectrometer data. J. Air Waste Manage. Assoc. 2009, 59 (2), 155−162. (14) Streese, J.; Schlegelmilch, M.; Heining, K.; Stegmann, R. A macrokinetic model for dimensioning of biofilters for VOC and odour treatment. Waste Manage. 2005, 25 (9), 965−974. (15) Ottengraf, S. P. P.; van den Oever, A. H. C. Kinetics of organic compound removal from waste gases with a biological filter. Biotechnol. Bioeng. 1983, 25 (12), 3089−3102. (16) Spigno, G.; De Faveri, D. M. Modeling of a vapor-phase fungi bioreactor for the abatement of hexane: Fluid dynamics and kinetic aspects. Biotechnol. Bioeng. 2005, 89 (3), 319−328. (17) Yasuda, T.; Kuroda, K.; Hanajima, D.; Fukumoto, Y.; Waki, M.; Suzuki, K. Characteristics of the microbial community associated with ammonia oxidation in a full-scale rockwool biofilter treating malodors from livestock manure composting. Microbes Environ. 2010, 25 (2), 111−119. (18) Hansen, M. J.; Liu, D.; Guldberg, L. B.; Feilberg, A. Application of Proton-Transfer-Reaction Mass Spectrometry to the assessment of odorant removal in a biological air cleaner for pig production. J. Agric. Food Chem. 2012, 60 (10), 2599−2606. (19) Ottosen, L. D. M.; Juhler, S.; Guldberg, L. B.; Feilberg, A.; Revsbech, N. P.; Nielsen, L. P. Regulation of ammonia oxidation in biotrickling airfilters with high ammonium load. Chem. Eng. J. 2011, 167 (1), 198−205. (20) Morsing, S. Method for determination of air flow, exhaust unit and its application. WIPO Patent Application WO/2011/009458, January 27, 2011. (21) Feilberg, A.; Liu, D.; Adamsen, A. P. S.; Hansen, M. J.; Jonassen, K. Odorant emissions from intensive pig production measured by online proton-transfer-reaction mass spectrometry. Environ. Sci. Technol. 2010, 44 (15), 5894−5900. (22) Hirai, M.; Ohtake, M.; Shoda, M. Removal kinetics of hydrogen sulfide, methanethiol and dimethyl sulfide by peat biofilters. J. Ferment. Bioeng. 1990, 70 (5), 334−339. (23) Melse, R. W.; Van Der Werf, A. W. Biofiltration for mitigation of methane emission from animal husbandry. Environ. Sci. Technol. 2005, 39 (14), 5460−5468. (24) Chan, W.; Chang, L. Kinetic behaviors between acetone and composite bead in biofilter. Appl. Microbiol. Biot. 2006, 72 (1), 190− 196. (25) Raghuvanshi, S.; Babu, B. V. Experimental studies and kinetic modeling for removal of methyl ethyl ketone using biofiltration. Bioresour. Technol. 2009, 100 (17), 3855−3861. (26) Yu, H.; Wilson, F.; Tay, J. Kinetics analysis of an anerobic filter treating soybean wastewater. Water Res. 1998, 32 (11), 3341−3352. (27) Kwon, S.; Cho, D. A comparative, kinetic study on cork and activated carbon biofilters for VOC degradation. J. Ind. Eng. Chem. 2009, 15 (1), 129−135. (28) Chung, Y.-C.; Huang, C.; Tseng, C.-P. Rushing Pan, J.Biotreatment of H2S- and NH3-containing waste gases by coimmobilized cells biofilter. Chemosphere 2000, 41 (3), 329−336.
together with a reduced ventilation rate (Q) and/or larger biofilter volume (Vf) is predicted to gain better performance in biofilters. It is noteworthy, that neither the Stover-Kincannon model nor the Grau second-order kinetics could simulate removal of methanethiol and DMS in the filter. Due to the low and relatively varying removal efficiencies (Figure S5), it was impossible to obtain good correlations for both models since RE was used as denominator (after linearization). In addition, both models seem to have better simulation performance on more water-soluble compounds (e.g., acids) than less watersoluble compounds (e.g., sulfides) (Figure S10). This may be due to the fact that less water-soluble pollutants are expected to have more internal transport limitations, which means that the application of the gas phase concentration as modeling parameter is less valid. Thus, a microkinetic model might be necessary for low water-soluble pollutants in order to simulate all the transport limitations together with the biodegradation in the biofilm.
■
ASSOCIATED CONTENT
* Supporting Information S
Tables S1−S4 and Figures S1−S10. This material is available free of charge via the Internet at http://pubs.acs.org.
■
AUTHOR INFORMATION
Corresponding Author
*Phone: +45 8715 7647. Fax: +45 8715 6000. E-mail: Anders.
[email protected]. Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS The Danish Strategic Research Council (No. 2104-08-0017) is acknowledged for financial support.
■
REFERENCES
(1) Schiffman, S. S.; Studwell, C. E.; Landerman, L. R.; Berman, K.; Sundy, J. S. Symptomatic effects of exposure to diluted air sampled from a pig confinement atmosphere on healthy human subjects. Environ. Health Perspect. 2005, 113 (5), 567−576. (2) Hartung, J.; Phillips, V. R. Control of gaseous emissions from livestock buildings and manure stores. J. Agric. Eng. Res. 1994, 57 (3), 173−189. (3) Liu, D.; Feilberg, A.; Adamsen, A. P. S.; Jonassen, K. E. N. The effect of slurry treatment including ozonation on odorant reduction measured by in-situ PTR-MS. Atmos. Environ. 2011, 45 (23), 3786− 3793. (4) Feilberg, A.; Adamsen, A. P. S.; Lindholst, S.; Lyngbye, M.; Schäfer, A. Evaluation of Biological Air Filters for Livestock Ventilation Air by Membrane Inlet Mass Spectrometry. J. Environ. Qual. 2010, 39 (3), 1085−1096. (5) Melse, R. W.; Ogink, N. W. M. Air scrubbing techniques for ammonia and odor reduction at livestock operations: Review of onfarm research in the Netherlands. Trans. ASABE 2005, 48 (6), 2303− 2313. (6) Chen, L.; Hoff, S.; Cai, L.; Koziel, J.; Zelle, B. Evaluation of wood chip-based biofilters to reduce odor, hydrogen sulfide, and ammonia from swine barn ventilation air. J. Air Waste Manage.Assoc. 2009, 59 (5), 520−530. (7) Nicolai, R. E.; Janni, K. A. Biofilter media mixture ratio of wood chips and compost treating swine odors. Water Sci. Technol. 2001, 44 (9), 261−267. 8268
dx.doi.org/10.1021/es301295m | Environ. Sci. Technol. 2012, 46, 8261−8269
Environmental Science & Technology
Article
(29) Krailas, S.; Pham, Q. T. Macrokinetic determination and water movement in a downward flow biofilter for methanol removal. Biochem. Eng. J. 2002, 10 (2), 103−113. (30) Ramírez, M.; Gómez, J. M.; Aroca, G.; Cantero, D. Removal of hydrogen sulfide by immobilized Thiobacillus thioparus in a biotrickling filter packed with polyurethane foam. Bioresour. Technol. 2009, 100 (21), 4989−4995. (31) Kincannon, D. F.; Stover, E. L. Design methodology for fixed film reaction-RBCs and biological towers. Civil Eng. Pract. Des. Eng. 1983, 2 (1), 107−124. (32) Borghei, S. M.; Sharbatmaleki, M.; Pourrezaie, P.; Borghei, G. Kinetics of organic removal in fixed-bed aerobic biological reactor. Bioresour. Technol. 2008, 99 (5), 1118−1124. (33) Grau, P.; Dohanyos, M.; Chudoba, J. Kinetics of multicomponent substrate removal by activated sludge. Water Res. 1975, 9 (7), 637−642. (34) Jin, R.; Zheng, P. Kinetics of nitrogen removal in high rate anammox upflow filter. J. Hazard. Mater. 2009, 170 (2−3), 652−656. (35) Ni, S.-Q.; Lee, P.-H.; Sung, S The kinetics of nitrogen removal and biogas production in an anammox non-woven membrane reactor. Bioresour. Technol. 2010, 101 (15), 5767−5773. (36) Travieso, L.; Sánchez, E.; Borja, R.; Benítez, F.; Raposo, F.; Rincón, B.; Jiménez, A. M. Evaluation of a laboratory-scale stabilization pond for tertiary treatment of distillery waste previously treated by a combined anaerobic filter-aerobic trickling system. Ecol. Eng. 2006, 27 (2), 100−108. (37) Borghei, S. M.; Sharbatmaleki, M.; Pourrezaie, P.; Borghei, G. Kinetics of organic removal in fixed-bed aerobic biological reactor. Bioresour. Technol. 2008, 99 (5), 1118−1124. (38) Juhler, S.; Revsbech, N. P.; Schramm, A.; Herrmann, M.; Ottosen, L. D. M.; Nielsen, L. P. Distribution and rate of microbial processes in an ammonia-loaded air filter biofilm. Appl. Environ. Microb. 2009, 75 (11), 3705−3713. (39) Liu, D.; Feilberg, A.; Nielsen, A. M.; Adamsen, A. P. S. PTR-MS measurement of partition coefficient of reduced volatile sulfur compounds in liquids from biotricking filters. Submitted to Chemesphere 2012, assigned manuscript number: CHEM25707. (40) De Bruyn, W. J.; Swartz, E.; Hu, J. H.; Shorter, J. A.; Davidobits, ́ P. Henry’s law solubility and Setchenow coefficients for biogenic reduced sulfur species obtained from gas-liquid uptake measurements. J. Geophys. Res. 1995, 100 (D4), 7245−7251. (41) CRC Handbook of Chemistry and Physics, 86th, ed.; Lide, D. R., Ed.; CRC Press: Boca Raton, 2005; Section 8−40.
8269
dx.doi.org/10.1021/es301295m | Environ. Sci. Technol. 2012, 46, 8261−8269
