DIAUXIC LAG OF DENITRIFYING BACTERIA IN OXIC/ANOXIC CYCLING UNDER CONTINUOUS FLOW CONDITIONS

Transcription

1 DIAUXIC LAG OF DENITRIFYING BACTERIA IN OXIC/ANOXIC CYCLING UNDER CONTINUOUS FLOW CONDITIONS By DONG-UK LEE A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2005

2 Copyright 2005 by Dong-Uk Lee

3 To God, MJ and my family.

4 ACKNOWLEDGMENTS I would like to truly thank my co-advisors, Dr. Ben Koopman and Dr. Spyros Svoronos, for their guidance and advice throughout my graduate study. Their passion and sincerity toward academic research and teaching will be strong guiding lights for the rest of my life. I also have to acknowledge the other members of my committee, Dr. Angela Lindner, Dr. Atul Narang, and Dr. Samuel Farrah for, their advice and help on my research since I asked them to be on my committee. I would like to appreciate the Alumni Fellowship from the University of Florida for my entire doctoral study. I thank Mr. Chuck Fender and the fellows in the Physical Plant Division of University of Florida at the Water Reclamation Facility for their help and friendliness. I have to thank the fellows in our research group, Anna I. Casasus-Zambrana, Ryan K. Hamilton, Dr. Sung-Hoon Woo, Kiran Durvasula, and Adrian Vega, for their help, support and friendship. Also, I thank Jao Jue, Gautam Kini, Vijay Krishna and other fellows in the Academic Interface Lab for being good friends of mine. Finally, I would like to thank my family and MJ for their endless love, prayer, and, most of all, for being my family. I truly thank my God for preparing everything and leading me here. iv

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS... iv LIST OF TABLES... ix LIST OF FIGURES... xi ABSTRACT... xvi CHAPTER 1 INTRODUCTION REVIEW OF LITERATURE...4 Dynamics of Heterotrophic, Denitrifying Bacteria Switching between Electron Acceptors...4 Effects of Alternating Electron Acceptors...5 Diauxic Lag of Bacteria Switching between Electron Acceptors...8 Factors Affecting the Diauxic Lag of Heterotrophic, Denitrifying Bacteria...12 Bacterial species...12 Length of aerobic phase...12 Dissolved oxygen concentration in aerobic phase...13 Nitrate exposure history of preceding culture...14 Nitrate concentration in anoxic phase...14 Modeling of Denitrification in Activated Sludge...15 Activated Sludge Model No. 1 (ASM1)...15 Modeling of Denitrification with a Cybernetic Approach for Denitrifying Enzyme Kinetics...17 Modeling of Denitrification with Mechanistic Approach for Denitrifying Enzyme Kinetics EXTENSION OF ACTIVATED SLUDGE MODEL NO. 1 TO INCORPORATE DENITRIFYING ENZYME KINETICS...26 Extension with Cybernetic Approach (easm1c)...26 Extension with Mechanistic Approach (easm1m)...29 Comparison of Extended Versions of ASM1 to the Original Version of ASM v

6 Simulation of Dynamics of Bacteria Switching between Oxygen and Nitrate by ASM Simulation of Dynamics of Bacteria Switching between Oxygen and Nitrate by easm1c...36 Simulation of Dynamics of Bacteria Switching between Oxygen and Nitrate by easm1m...48 Re-examination of Results from a Previous Study SIGNIFICANCE OF DENITRIFYING ENZYME DYNAMICS IN BIOLOGICAL NITROGEN REMOVAL PROCESSES: A SIMULATION STUDY...57 Experimental Methods...58 Process Configurations and Modeling...58 Wastewater Composition and Model Parameters...61 Diurnally Varying Flow and Component Concentrations in Influent Wastewater...62 Results and Discussion...64 Simulations of Fed-Batch Process...64 Simulations of BDP Process...65 Optimum Cycle Length as a Function of UVF...67 Conclusions OBJECTIVES GENERAL MATERIALS AND METHODS...70 Bacterial Cultivation...70 Reviving Freeze-Dried Bacteria and Deep-Freezing of Bacterial Cultures...70 Reviving of Frozen Bacteria...71 Preculture Procedure...71 Reactors...71 Overall Layout...72 Fermentor Assembly...72 Feed Reservoir Assembly...75 Autoclaving Procedure and Aseptic Connection of Feed Reservoir Assembly to Fermentor Assembly...75 Inoculation of Fermentor and Initiation of Startup Phase...76 Initiation of Continuous Flow Phase...77 Sampling from Fermentor...77 Monitoring of Contamination of Pure Culture...78 Analytical Measurements...78 Biomass Absorbance...78 Chemical Oxygen Demand...79 Nitrate and Nitrite...79 Nitrate Reductase Activity...79 vi

7 7 METHOD FOR ACHIEVING REPRODUCIBLE INITIAL CULTURE STATES IN STUDY OF BACTERIAL DENITRIFICATION KINETICS...81 Introduction...81 Materials and Methods...82 Results...85 Paracoccus pantotrophus...85 Pseudomonas denitrificans...95 Discussion DIAUXIC LAG OF DENITRIFYING BACTERIA IN A CONTINUOUS FLOW REACTOR I. SINGLE SWITCH FROM OXIC TO ANOXIC CONDITIONS Introduction Materials and Methods Experimental Procedures Modeling Results Determination of Diauxic Lag under Continuous Flow Conditions Experimental Results Modeling Results Discussion DIAUXIC LAG OF DENITRIFYING BACTERIA IN A CONTINUOUS FLOW REACTOR II. ALTERNATING OXIC/ANOXIC CONDITIONS Introduction Materials and Methods Alternating / Cycling under Continuous Flow Conditions Modeling Results Preliminary Simulations Experiments Short cycle length (12 hours) Long cycle length (24 hours) Simulations to Predict Experimental Results using easm1m Discussion Effect of Alternating Cycling on Growth of P. denitrificans Predicted by easm1m Growth Dynamics of P. denitrificans in Alternating / Cycling under Continuous Flow Condition-Experiment Suggestion of a Preliminary Modeling Concept to Show the Effect of Growth Patterns during Phase on Diauxic Lag of P. denitrificans Fast Decrease of Biomass Absorbance during Lag Period SUMMARY AND CONCLUSIONS vii

8 11 FUTURE WORK APPENDIX A B SIMULATIONS OF RESULTS FROM A PREVIOUS STUDY MEASUREMENTS USING HACH TEST TUBES LIST OF REFERENCES BIOGRAPHICAL SKETCH viii

9 LIST OF TABLES Table page 2-1. Factors affecting the diauxic lag of denitrifiers Process rates and stoichiometric coefficients of easm1c Process rates and stoichiometric coefficients of easm1m Characteristics of growth medium in batch simulations Parameters of ASM1 for the batch simulation Design of UF BDP water reclamation facility (train 1 of two parallel trains) Sequence of phases in the fed-batch and BDP processes hour flow-weighted average wastewater composition Stoichiometric and kinetic parameters in the ASM1 and easm1 models Composition of nutrient solution for P. pantotrophus Composition of nutrient solution for P. denitrificans Amount of carbon substrate in feed solutions Nutrients in two feed solutions Summary of experimental results of anoxic batch phases Summary of experimental results of oxic continuous flow phases Comparison of experimental data Experimental conditions Composition of nutrient solution for P. denitrificans Amount of carbon substrate, ammonia in nutrient solution of each stage Nutrients in two feed solutions ix

10 8-5. Calculation procedures of virtual batch curve method Parameters of easm1m for simulation Initial conditions for easm1m simulations Parameters of easm1m after calibration Summary of experimental results Parameters of easm1m for simulations x

11 LIST OF FIGURES Figure page 2-1. Schematics of Mechanistic Denitrification Model Simulation of experimentally observed diauxic lag of Pseudomonas denitrificans, predicted by easm1c Simulation of experimentally observed diauxic lag of Pseudomonas denitrificans, predicted by easm1m Growth of heterotrophic biomass during cyclic simulations with 8 mg/l of DO during oxic phase, predicted by ASM Mass specific and volumetric denitrification rate during cyclic simulation, predicted by ASM Growth of heterotrophic biomass during cyclic simulations with 4 mg/l of DO during oxic phase, predicted by ASM Growth of heterotrophic biomass during cyclic simulations with 2 mg/l of DO during oxic phase, predicted by ASM Growth of heterotrophic biomass during cyclic simulations with 1 mg/l of DO during oxic phase, predicted by ASM Growth of heterotrophic biomass under oxic/anoxic switch Specific nitrate reductase level and activity of heterotrophic biomass under oxic/anoxic switch, predicted by easm1c Growth of heterotrophic biomass during cyclic simulations with 8 mg/l of DO during oxic phase, predicted by easm1c Mass specific and volumetric denitrification rate during cyclic simulation, predicted by easm1c Growth of heterotrophic biomass during cyclic simulations with 4 mg/l of DO during oxic phase, predicted by easm1c Growth of heterotrophic biomass during cyclic simulations with 2 mg/l of DO during oxic phase, predicted by easm1c...46 xi

12 3-14. Growth of heterotrophic biomass during cyclic simulations with 1 mg/l of DO during oxic phase, predicted by easm1c Growth of heterotrophic biomass during cyclic simulations with 8 mg/l of DO during oxic phase, predicted by easm1c Specific nitrate reductase level and specific intracellular nitrate level of heterotrophic biomass under oxic/anoxic switch, predicted by easm1c Growth of heterotrophic biomass during cyclic simulations with 1 mg/l of DO during oxic phase, predicted by easm1m Growth of heterotrophic biomass during cyclic simulations with 0.5 mg/l of DO during oxic phase, predicted by easm1m Growth of heterotrophic biomass during cyclic simulations with 0.1 mg/l of DO during oxic phase, predicted by easm1m Simulation of experimental results from a previous study Process schematics of fed-batch process (top) and BDP process (bottom) showing the fraction of the cycle length or hydraulic residence time occupied by each phase or part of the processes Sequence of phases in the BDP oxidation ditches Effects of anoxic volume fraction and cycle length on performance of fed-batch process predicted by ASM1 and easm1c Effect of unaerated volume fraction (UVF) and cycle length on performance of BDP process Optimum cycle lengths of fed-batch and BDP processes as a function of unaerated volume fraction (UVF) Overall layout of experimental configuration Side view of New Brunswick Bioflo 2000 Fermentor Fermentor assembly Feed reservoir assembly Biomass absorbance profile of Experimental Biomass absorbance during anoxic batch phase (Trial 1, Experimental 1) Biomass absorbance during anoxic batch phase (Trial 2, Experimental 1)...89 xii

13 7-4. Biomass absorbance profile of Experimental Biomass absorbance during anoxic batch phase (Trial 1, Experimental 2): Biomass absorbance during anoxic batch phase (Trial 2, Experimental 2): Biomass absorbance profile of Experimental Biomass absorbance during anoxic batch phase (Trial 1, Experimental 3) Biomass absorbance during anoxic batch phase (Trial 2, Experimental 3) Biomass absorbance profile of Experimental Biomass absorbance during anoxic batch phase (Trial 1, Experimental 4) Biomass absorbance during anoxic batch phase (Trial 2, Experimental 4) Feed inlet configurations Flow and components around CSTR in simulation Determination of diauxic lag under continuous flow condition using virtual batch curve method Biomass absorbance profile (Trial 1) Biomass absorbance profile during anoxic continuous flow phase (Trial 1) Determination of diauxic lag (Trial 1) Biomass absorbance profile (Trial 2) Biomass absorbance profile during anoxic continuous flow phase (Trial 2) Biomass absorbance profile (Trial 3) Biomass absorbance profile during anoxic continuous flow phase (Trial 3) Determination of diauxic lag (Trial 3) Biomass absorbance profile (Trial 4) Biomass absorbance profile during anoxic continuous flow phase (Trial 4) Determination of diauxic lag (Trial 4) Biomass absorbance profile (Trial 5) Biomass absorbance profile during anoxic continuous flow phase (Trial 5) xiii

14 8-17. Determination of diauxic lag (Trial 5) Biomass absorbance profile (Trial 6) Biomass absorbance profile during anoxic continuous flow phase (Trial 6) Determination of diauxic lag (Trial 6) Simulation of experimental result (Trial 6) Simulation of experimental result with calibrated parameters (Trial 6) Simulation of experimental result with calibrated parameters (Trial 5) Simulation of experimental result (short oxic continuous flow phase) Change of biomass absorbance and carbon substrate concentration during diauxic lag and recovery of growth, predicted by easm1m Schematic view of gas supply system Biomass absorbance profile from a typical simulation Biomass absorbance profile during ultimate state in alternating oxic/anoxic cycling (6-hour cycle length) Biomass absorbance profile during ultimate state in alternating oxic/anoxic cycling (12-hour cycle length) Biomass absorbance profile during ultimate state in alternating oxic/anoxic cycling (24-hour cycle length) Biomass absorbance profile during ultimate state in alternating oxic/anoxic cycling (48-hour cycle length) Biomass absorbance profile in alternating oxic/anoxic cycling (180-hour cycle length) Overall biomass absorbance profile of short cycle length experiment (12 hourcycle length) Biomass absorbance profile in alternating cycling (12 hour-cycle length) Overall biomass absorbance profile of long cycle length experiment (24-hour cycle length) Biomass absorbance profile in alternating cycling (24-hour cycle length) Component concentrations during ultimate state (24-hour cycle length) xiv

15 9-13. Component concentrations during the final anoxic phase Simulation of experimental results (12 hour cycle length) Simulation of experimental results (24 hour cycle length) Simulation of experimental results (final anoxic phase, 24 hour cycle length) xv

16 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy DIAUXIC LAG OF DENITRIFYING BACTERIA IN OXIC/ANOXIC CYCLING UNDER CONTINUOUS FLOW CONDITIONS By Dong-Uk Lee August 2005 Chair: Ben L. Koopman Cochair: Spyros A. Svoronos Major Department: Environmental Engineering Sciences The present study was conducted to investigate diauxic lag of denitrifying bacteria under an ultimate state of oxic/anoxic cycling under continuous flow conditions. As preliminary steps, the industry standard Activated Sludge Model No. 1 was extended with denitrification models and a simulation study was conducted to compare predictions of the conventional and an extended version. An experimental system was developed to implement bacterial pure culture growth under continuous flow conditions. The performance of the system was verified by determining the reproducibility of experimental results. Using the experimental system, diauxic lag of denitrifying bacteria was then studied under oxic/anoxic cycling conditions. The experimental system developed in the present study was capable of achieving pure culture of denitrifying bacteria without contamination up to the desirable length of time for experiments. The reproducibility of the length of diauxic lag and the highest xvi

17 anoxic observed specific growth rates were significantly improved by achieving steady state growth of bacteria as a preliminary stage. Diauxic lag of Pseudomonas denitrificans under continuous flow conditions could be characterized by the virtual batch curve method developed in the present study. The easm1m was able to predict the observed diauxic lag under the continuous flow conditions with slight modification of parameters. The experimental results were significantly influenced by the magnitude of biomass accumulation at the feed inlet. Growth patterns in preceding oxic phase were likely to have an effect on length of diauxic lag during consecutive anoxic phase, which could not be predicted by easm1m. Predictions of easm1m on growth of Pseudomonas denitrificans in alternating oxic/anoxic cycling under continuous flow conditions were consistent with results from a previous study. It has been found that Pseudomonas denitrificans could not establish significant anoxic growth during alternating oxic/anoxic cycling under continuous flow conditions, with up to 24 hour cycle length. The easm1m could not fit the growth behaviors with the previous parameters. Furthermore, the diauxic lag after the cycling was significantly longer than the initial lag, which was additional evidence explaining that growth patterns of bacteria in the preceding oxic phases may influence diauxic lag of bacteria in the following anoxic phases. xvii

18 CHAPTER 1 INTRODUCTION Nitrogen removal from wastewater has become more and more important because of a number of reasons, including pollution, eutrophication of receiving water bodies and increasing needs for reuse of reclaimed wastewater. Biological nitrogen removal using activated sludge is a popular method to remove nitrogen from wastewater. Ammonia nitrogen is first oxidized to nitrite and nitrate nitrogen by nitrifying bacteria (nitrification) in activated sludge under oxic conditions in a typical wastewater treatment process utilizing biological nitrogen removal. Nitrite and nitrate nitrogen are then reduced to dinitrogen or other gaseous nitrogen compounds by denitrifying bacteria (denitrification) under anoxic condition. Since the two major reactions take place under different growth conditions in a single sludge biological nitrogen removal process, it is inevitable that bacteria in activated sludge are exposed to cycling oxic/anoxic conditions. Growth dynamics of bacteria can occur in such conditions if bacteria cannot adjust their growth capabilities to repetitive change of growth conditions. Therefore, it is very important to study growth dynamics of bacteria that have important roles in biological nitrogen removal from wastewater. The phenomenon of diauxic lag for bacteria switching between electron donors was discovered at least 60 years ago (Monod, 1942). Subsequently, Kodama et al. (1969) observed a similar lag for bacteria switching between electron acceptors. Experiments with activated sludge and pure culture of Pseudomonas denitrificans have established that the diauxic lag of bacteria switching between oxygen and nitrate as electron 1

19 2 acceptors can last for several hours and depend on the preculture environment, length of aeration period, and dissolved oxygen concentration during the aeration period that precedes anoxic conditions (Liu et al., 1998a, b; Gouw et al., 2001; Lisbon et al., 2002). Effects of aeration period length and dissolved oxygen concentration on diauxic lag of bacteria switching between oxygen and nitrate were successfully modeled (Liu et al., 1998a, b; Casasús-Zambrana, 2001). However, the popular Activated Sludge Models No 1, 2, 2d, and 3 (Henze et al., 2000) cannot portray the diauxic lag phenomenon. This deficiency could result in sub-optimal operational strategies or designs and lead to needless environmental impact on receiving waters or waste of economic resources. Recently, Lee et al. (2004) compared the predictions of an extended version of ASM1 (easm1c) with enzyme kinetics to the predictions of ASM1 for periodically operated nitrogen removing processes (fed-batch and BioDenipho). ASM1 and easm1c gave similar predictions of optimal unaerated volume fraction (UVF) that were consistent with operation of the BioDenipho process at the University of Florida. However, the easm1c predicted substantially longer optimal cycle lengths, which are more consistent with the BioDenipho process operation at the University of Florida than those predicted by ASM1. Furthermore, easm1c predicted a critical cycle length, below which denitrification would cease. The growth dynamics of denitrifying bacteria under alternating oxic/anoxic cycling found by Lee et al. (2004) has never been investigated by experimental work. In early studies of diauxic lag of denitrifying bacteria, growth responses of those bacteria were investigated within few switches between oxic and anoxic conditions (Liu et al., 1998a, b; Gouw et al., 2001; Lisbon et al., 2002). Moreover, the experiments were performed

20 3 under batch conditions where carbon substrates were provided with non-limiting amounts. In such conditions, growth dynamics of denitrifiers could be different from those taking place in real wastewater treatment plants, where concentrations of organic substrates are relatively low. With these needs, the present study was conducted to investigate growth dynamics of denitrifying bacteria with better understandings. Efforts were made to develop a proper experimental setup to implement a bacterial pure culture system under continuous flow conditions. Diauxic lag of Pseudomonas denitrificans in single switch and alternating cycling between oxic and anoxic conditions was studied under continuous flow conditions. The capabilities of an extended version of ASM1 were evaluated with respect to prediction of experimental results.

21 CHAPTER 2 REVIEW OF LITERATURE The phenomenon of diauxic bacterial growth was extensively studied by Monod (1942, 1949). He mentioned that diauxie is characterized by a dual growth cycle which consists of two exponential growth phases separated by a period during which the growth rate reaches a minimum, or becomes negative. He reported that a diauxie could occur when bacteria grew on media where the organic substrate is limiting and consists of mixtures of two carbohydrates. He reported that this phenomenon indicated that each of two exponential growth cycles corresponded to the exclusive utilization of one of the substrates, due to an inhibitory effect of one of the substrates on formation of the enzyme for the other substrate. It has also been found that bacteria may experience lag when they switch between electron acceptors (Kodama et al., 1964, Liu et al., 1998a, b). In the following sections, the diauxic lag of bacteria switching between electron acceptors will be reviewed. The discussion will be focused on diauxic lag of denitrifying bacteria switching their electron acceptors from nitrate to oxygen. Furthermore, several mathematical models to predict the denitrification and related enzyme dynamics will be reviewed. Dynamics of Heterotrophic, Denitrifying Bacteria Switching between Electron Acceptors A number of studies have been performed to investigate the dynamics of denitrification under conditions in which the electron acceptors switch. A general discussion of the effect of alternating electron acceptor on denitrification and growth of 4

22 5 bacteria will be given and the diauxic lag of denitrifying bacteria switching between electron acceptors between oxygen and nitrate will be discussed. Effects of Alternating Electron Acceptors Several investigators have examined the effect of alternating electron acceptors on the dynamics of bacteria. Simpkin and Boyle (1985) investigated variations of nitrate and nitrite reductase activities of activated sludge exposed to alternating aerobic/anoxic conditions. In laboratory sequencing batch reactors (SBRs), anoxic phases were provided during part of the reaction phase or during settling. The highest nitrate reductase activity was found when the feed had a high level of nitrate (> 30 mg/l) and the anoxic period included 4 hours out of a total 6-hour reaction phase, plus one-half hour of anoxic conditions during settling. An intermediate nitrate reductase level was found in a reactor with high nitrate (16 mg/l) but only one-half hour of anoxic conditions. The lowest nitrate reductase activities were found in SBRs with low nitrate and only one-half hour of anoxic conditions (during settling). O Neil and Horan (1995) investigated the effect of oxic/anoxic cycling on nitrification and denitrification in a chemostat that was inoculated with nitrifying activated sludge. They first cycled the growth conditions between 4 hours of aerobic phase and 20 hours of anaerobic phase for 15 days. They performed two experiments involving the same length of oxic and anoxic periods and different feeding patterns. There was no indication of a growth lag in denitrification after the switch from oxic to anoxic conditions. In the third experiment they provided 14 hours of oxic growth conditions. After aeration stopped and dissolved oxygen dropped to zero, denitrification activity remained very low for the remainder of the time monitored, which was 20 hours.

23 6 Baumann et al. (1996) studied the response of Paracoccus denitrificans to changes between aerobic and anaerobic growth conditions in a continuous culture. They first ran the reactor until an aerobic steady state was reached. They then stopped aeration and ran the reactor until an anaerobic steady state was reached. Finally, they restarted aeration and ran the reactor to another aerobic steady state. When the growth conditions were changed from aerobic to anaerobic, the culture did not immediately establish complete denitrification. Nitrite started accumulating immediately after the switch and nitric oxide production began somewhat later. Dinitrogen became the major denitrification product after the intermediates disappeared and the culture established a new steady state. The mrna levels for nitrate reductase and nitrous oxide reductase started increasing immediately after the switch whereas the mrna level for nitrite reductase started increasing somewhat later. Biosynthesis of nitrite reductase was started about 30 minutes after the increase of the mrna level of the enzyme and gradually built up over a period of 30 hours. Baumann et al. (1997a) investigated the effect of repeated alternating aerobicanaerobic conditions on denitrification in continuously-fed cultures of Paracoccus denitrificans and activated sludge. In the case of a Paracoccus denitrificans growth reactor with alternating aerobic (24 h) and anoxic (24 h) phases, the authors grew the bacteria for three cycles of aerobic-anoxic phases. The authors showed measurements of nitrate versus time over a span of 120 hours. Vertical lines in the graph indicated switches between external electron acceptors. In the first cycle, it is apparent that nitrate levels began to decrease immediately after oxygen supply was stopped. In the second cycle, decrease of nitrate levels lagged exhaustion of dissolved oxygen by one hour. In

24 7 the third cycle, the lag lasted two hours. According to the authors, nitrate consumption began immediately after dissolved oxygen was depleted. However, it seems doubtful that one or two hours would be required for oxygen depletion in cycles 2 and 3. These data thus suggest the occurrence of a diauxic lag with the long phase lengths. Denitrification intermediates (nitrite, nitric oxide, nitrous oxide) accumulated during the anoxic phase of the first cycle but not during the anoxic phases of the second or third cycles. The authors performed another experiment with shorter phase lengths, 1.5 h anaerobic and 2.5 h aerobic, through a total of four cycles. With the shorter phase lengths, nitrite accumulated during the anoxic phases and there was negligible nitrous oxide production. The nitrite reductase level increased throughout the experiment, during aerobic as well as anoxic phases. The authors attributed nitrite accumulation to insufficient time for the bacteria to adjust their enzyme synthesis system. The authors also measured mrna for nitrate reductase during the second experiment. The mrna for nitrate reductase decreased during the aerobic phases and increased in the anoxic phases after approximately 0.5 hour of lag period. The characteristics of behaviors of activated sludge under longer alternating aerobic-anaerobic conditions (24 h aerobic and 24 h anaerobic) were similar to those of Paracoccus denitrificans (i.e., denitrification intermediates (e.g., nitrite and nitrous oxide) were accumulated during the first anoxic phase but later disappeared). Upon change to anoxic conditions, nitrate consumption lagged for about 4 hours. Baumann et al. (1997b) studied the effect of change from aerobic to anaerobic growth conditions on the denitrification of a continuous culture of Paracoccus denitrificans at a suboptimal ph. The biomass concentration started decreasing

25 8 immediately after the switch and continued decreasing for 50 hours. This trend approximately followed a dilution and decay curve with a near-zero specific growth rate (simulation not shown). Nitrite and nitric oxide started accumulating almost immediately after the switch. Accumulation of nitrous oxide and dinitrogen started somewhat later. Increases in levels of the mrnas for nitrate reductase and nitric oxide reductase were observed immediately after the switch, whereas levels of the mrna for nitrite reductase began increasing one hour later. However, even though mrna levels for nitrite reductase increased, the amount of nitrite reductase synthesized was low. The authors suggested that biosynthesis of nitrite reductase was inhibited by higher free nitrous acid concentration due to lower ph. Cultures grown under cycling aerobic/anaerobic conditions or strictly anaerobic conditions were less affected by the low ph, indicating that they may have accumulated nitrite reductase over time. Oh and Silverstein (1999) studied the effect of feeding pattern on the mass specific denitrification rate of activated sludge in sequencing batch reactors. They found that the mass specific denitrification rate during anoxic phases decreased and the oxygen uptake rate of the sludge increased as the length of time that the substrate was present during aerobic phases was increased. The lengths of aeration period in the absence of substrate did not influence the mass specific denitrification rate. They concluded that feeding during the aerobic phase led to growth of aerobic (non-denitrifying) bacteria. Diauxic Lag of Bacteria Switching between Electron Acceptors Diauxic lag of bacteria switching between electron acceptors was first reported by Kodama et al. (1969). The authors examined the growth of Pseudomonas stutzeri in the presence of various concentrations of nitrate. The authors reported that the initial growth continued until all nitrate in the culture was consumed. Nitrite was accumulated while

26 9 nitrate was consumed. The lag of growth began after exhaustion of the nitrate. After a period of time, bacterial growth resumed, along with consumption of nitrite. The length of the lag period depended on the initial nitrate concentration (i.e., the higher the concentration, the longer the lag). The authors gave two possible mechanisms to explain this effect: (1) repression of development of the nitrite reducing machinery by nitrate, and (2) competition between nitrate and nitrite for electrons. Since the original description of a lag experienced by bacteria switching between electron acceptors, a number of other investigators have studied this phenomenon. Waki et al. (1980) investigated the effect of aerobic-anaerobic condition change on the growth, carbon source and nitrate consumption, and nitrate and nitrite reductase activity of Paracoccus denitrificans. They reported that the carbon source consumption and the growth of the bacteria stopped for a few hours when the condition was changed from aerobic to anoxic (oxygen absent, nitrate present). During this lag period, incomplete denitrification occurred (i.e., a rapid nitrate consumption was observed, but with a high level of nitrite accumulation). A careful glance at the profile of the bacterial growth in the reference reveals a second lag period that begins after the bacteria stop accumulating nitrite. However, there was still nitrate available at the beginning of this lag period. The specific nitrate reductase activity began to increase after the transition from aerobic to anoxic conditions. About 6 hours were required for the bacteria to reach the maximum nitrate reductase concentration. In comparison, the nitrite reductase activity remained constant for 2.5 h after the transition from aerobic to anoxic conditions and then started to increase as the bacteria started reducing nitrite.

27 10 Robertson and Kuenen (1984) tested denitrification of Thiosphaera pantotropha and Thiobacillus A2 switching their electron acceptor from oxygen to nitrate. Aerobically grown bacteria were exposed to anaerobic conditions with nitrate present as the electron acceptor. Acetate and thiosulphate or a mixture of both was provided as electron donor. Thiosphaera pantotropha produced a gaseous product immediately after the switch, regardless of electron donor. Thiobacillus A2 began to produce gas 3 hours after the switch when the electron donor was acetate, 4 hours after the switch when the electron donor was thiosulphate, and 2 hours in the presence of mixed electron donor. Bonin et al. (1989) examined growth and nitrate and nitrite reductase activity of bacteria exposed to alternating aerobic and anoxic environments. They found that nitrate reductase activity declined under aerobic conditions but was regained under anoxic conditions, once bacteria ended the lag phase and began to grow again. Liu et al. (1998a) exposed samples of activated sludge from a wastewater treatment plant to aerobic and anoxic conditions. They were the first investigators to observe diauxic lag of activated sludge and pointed out that this phenomenon could have significant engineering and economic implications for nitrogen-removing, single-sludge activated sludge processes. The authors reported that both activated sludge and nitrate enrichment denitrifying culture did not grow or grew very slowly for a while during anoxic conditions that followed oxic conditions. The authors modeled the phenomenon using a cybernetic approach. They noted that conventional models of single-sludge wastewater treatment process (e.g., Activated Sludge Model 1; Henze et al., 2000) could not depict the phenomenon of diauxic lag when bacteria switched between electron acceptors. Liu et al. (1998a) suggested that the reason for the onset of diauxic lag during

28 11 denitrification was the lack of enzyme that was required for reduction of nitrate. They hypothesized that the lack of enzyme was due to decay and dilution of the enzyme when the bacteria grew exponentially under aerobic conditions without the synthesis of the enzyme. They suggested that more than one hour of average length of the diauxic lags under conditions of their experiment was quite surprising, because the length of the lags was similar to the length of anoxic phase in a typical BioDenipho process, which is the nutrient removal process utilized at the facility where the samples were obtained. Liu et al. (1998b) studied the growth characteristics of a facultative denitrifying bacterium, Pseudomonas denitrificans. The authors observed that length of aerobic period and presence of nitrate during aerobic periods could affect the length of diauxic lag under subsequent anoxic conditions and successfully modeled these effects using a modified cybernetic approach. In the study of Baumann et al. (1997b), the authors observed that the biosynthesis of nitrite reductase was less inhibited by the low ph when the cultures were grown under cycling aerobic/anaerobic conditions or strictly anaerobic conditions, indicating that they may have accumulated nitrite reductase over time. Hence, it would be interesting to see whether alternating oxic/anoxic conditions results in development of a stable denitrifying continuous culture due to building up of denitrifying abilities over time or failure of denitrification in continuous culture due to diauxic lag. In this point of view, the cycle length of alternating oxic/anoxic conditions will be very important to the continuous denitrifying culture because insufficient length of anoxic condition would result in difficulties in developing denitrifying abilities, such as reductase enzymes.

29 12 Factors Affecting the Diauxic Lag of Heterotrophic, Denitrifying Bacteria Investigations of the dynamics of bacteria switching between oxygen and nitrate have identified several factors that can affect the length of the diauxic lag. These include bacterial species, length of the aerobic phase, dissolved oxygen concentration in the aerobic phase, and nitrate exposure history of the preceding culture. These factors are summarized in Table 2-1. Bacterial species There is some evidence in the literature that diauxic lag of denitrifiers under cyclic oxic/anoxic conditions differs according to bacterial species. For example, Pseudomonas denitrificans have relatively long diauxic lag when they experience an oxic/anoxic switch (Liu et al., 1998b; Lisbon et al., 2001; Casasús-Zambrana, 2002) whereas Paracoccus denitrificans exhibit little or no lag following oxic/anoxic switches (Baumann et al., 1996, 1997a, b). Length of aerobic phase Bonin et al. (1989) examined the effect of alternating changes from aerobic to anoxic conditions on an enzyme level of denitrifying bacteria, Pseudomonas nautica617. The authors reported that, in case of a short aerobic phase, both nitrate and nitrite reduction activities, which were depleted under aerobic conditions, recovered quickly in the following anoxic phase. However, after a long aerobic phase, the start of nitrate reduction activity was delayed for four hours, and the nitrite reduction rate reached only 20% of the original rate before aerobic conditions. Liu et al. (1998b) reported that a pure culture of Pseudomonas denitrificans aerated for a longer period experienced a longer lag than the same culture aerated for a shorter time. This result can be explained using dilution and decay of a denitrifying enzyme, as

30 13 Table 2-1. Factors affecting the diauxic lag of denitrifiers. Factors Effects on length of lag Length of aerobic phase Positive effect on length of lag Dissolved oxygen concentration in aerobic phase Nitrate exposure history of preceding conditions Positive effect on length of lag Presence of nitrate in preceding aerobic phase has negative effect on length of lag suggested by Liu et al. (1998a). A longer period of the aerobic phase provides a higher amount of dilution and decay of the denitrifying enzyme due to the suppression effect of dissolved oxygen on synthesis of denitrifying enzyme during aeration. Dissolved oxygen concentration in aerobic phase Lisbon et al. (2001) investigated the effect of dissolved oxygen concentration during the aerobic phase on the length of diauxic lag during the following anoxic phase. The authors reported that the average length of diauxic lags in the case of the high dissolved oxygen runs was longer than that in the case of the low dissolved oxygen runs. The average specific growth rates in the anoxic phases following low dissolved oxygen aerobic phases were significantly higher than those in the anoxic phases following high dissolved oxygen aerobic phases. The authors computed the ratio of biomass concentration at the end of an aerobic phase to the biomass concentration at the beginning of an aerobic phase. Higher values of the aerobic biomass ratio indicate higher levels of new biomass formed under aerobic conditions. The specific growth rate during the anoxic phase was inversely correlated with the biomass ratio for the preceding aerobic phase, whereas the diauxic lag of bacteria switching between oxygen and nitrate was directly correlated to the aerobic biomass ratio. This is consistent with a mechanism of

31 14 nitrate reductase dilution by growth under aerobic conditions and indicates that the effect of dissolved oxygen was to influence the rate of aerobic growth and, hence, enzyme dilution. Nitrate exposure history of preceding culture Gouw et al. (2001) examined the effect of nitrate exposure history on the oxygen/nitrate diauxic growth of Pseudomonas denitrificans. Their culturing sequence consisted of a pre-culture (bacterial growth in nutrient media that were inoculated from agar plates), an aerobic phase, and an anoxic phase. Three different pre-culture conditions were investigated: (1) anoxic with nitrate present, (2) aerobic with nitrate present, and (3) aerobic with nitrate absent. The effect of presence or absence of nitrate during the aerobic phase was also examined. In the case of aerobic pre-culture, the diauxic lag was long ( h) if nitrate was absent in pre-culture, whereas the presence of nitrate in pre-culture resulted in shorter lags ( h). The presence of nitrate in pre-culture partially compensated for absence of nitrate in subsequent long aerobic phases. (The combination of aerobic pre-culture and aerobic phase, both without nitrate, gave the longest lags.) In the case of anoxic preculture (with nitrate present), presence of nitrate during the following aerobic phase resulted in relatively short diauxic lags or no lags whereas there were always diauxic lags if nitrate was absent during the aerobic phases. The authors hypothesized that key denitrification enzymes might be synthesized under aerobic conditions if nitrate is present. Nitrate concentration in anoxic phase Kodama et al. (1969) investigated the effect of nitrate concentration on the diauxic lag of Pseudomonas stutzeri switching their electron between acceptors although they did not focus on diauxic lag switching between oxygen and nitrate. As the bacteria reduced

32 15 nitrate, nitrite accumulated until all nitrate was removed. The growth was then lagged until the bacteria started reducing nitrite. The authors found that the length of the lag period depended on the initial nitrate concentration (i.e., the higher the nitrate concentration, the longer the lag.) The authors gave two possible mechanisms for this effect: (1) repressed development of the nitrite reducing system by nitrate, and (2) competition between nitrate and nitrite for electrons. Modeling of Denitrification in Activated Sludge Several mathematical models have been developed to predict denitrification of activated sludge. Prediction of denitrification of the activated sludge models will be discussed and two new models capable of depicting diauxic lag of denitrification will be introduced. Activated Sludge Model No. 1 (ASM1) Activated Sludge Models No. 1, 2, 2d and 3 (Henze et al., 2000) were created by the task group on mathematical modeling for design and operation of biological wastewater treatment of the International Water Association. They have become well accepted for modeling of single-sludge biological wastewater treatment processes. The four models have similar expressions for growth of heterotrophic biomass on oxygen and nitrate and control the respective rates using the same switching functionality. ASM1 will be discussed in the present literature review because it is the oldest of the four models and thus has the longest experience base. The complete matrix representation of ASM1 is given in Table 2-2. In ASM1, the process rate for growth of heterotrophic biomass on oxygen is expressed by S S S O ρ 1 = ˆ μ H X B, H K S + S S K O, H + S (2-1) O

33 16 where ρ 1 is the process rate for growth of heterotrophic biomass on oxygen, μˆ H is the maximum specific growth rate of heterotrophic biomass, S S is the concentration of readily biodegradable substrate, S O is the dissolved oxygen concentration, K S and K O,H are the half saturation coefficients for a readily degradable substrate and dissolved oxygen, respectively, and X B,H is the concentration of heterotrophic biomass. by The process rate for growth of heterotrophic biomass on nitrate ( ρ 2 ) is expressed S K S O, H S NO ρ ˆ 2 μ H η g X B, H K S S S K O, H SO K NO S = (2-2) NO where S NO is nitrate plus nitrite nitrogen concentration, K NO is the half saturation coefficient of nitrate nitrogen, and η g is the correction factor for anoxic growth of heterotrophic biomass. In equation (2-1), the effect of oxygen on the rate of growth of heterotrophic biomass on oxygen is portrayed by the following switching function: K OH, SO + S O (2-3) The term approaches 1.0 when dissolved oxygen concentration is high and approaches zero as the dissolved oxygen concentration approaches zero. The effect of dissolved oxygen on the rate of growth of heterotrophic biomass on nitrate is depicted by the following switching function: S O K OH, + K O, H (2-4) The term approaches zero when dissolved oxygen concentration is high and approaches 1.0 as the dissolved oxygen concentration approaches zero. Thus, it has the effect of

34 17 slowing the growth of heterotrophic biomass on nitrate when dissolved oxygen is present in the medium. Modeling of Denitrification with a Cybernetic Approach for Denitrifying Enzyme Kinetics Liu et al. (1998a) proposed a model of denitrification that relied on a cybernetic approach analogous to that of Kompala et al. (1986) to predict the extent of utilization of two alternative electron acceptors (oxygen, nitrate). The proposed model includes the concentrations of two enzymes, E O and E NO, which stand for concentration of oxygenase and nitrate reductase, respectively. Both the specific levels and activities of these enzymes regulate the growth rate of heterotrophic biomass. The process rate expressions for growth of heterotrophic biomass on oxygen ( ρ 1 ) and on nitrate ( ρ 2 ) from ASM1 were modified and process rate expressions for enzyme synthesis and decay ( ρ9 - ρ 12 ) were developed. Since the model was subject to be incorporated into ASM1, the order numbers were assigned to the process rates in a manner consistent with that of ASM1. The #1 and #2 were assigned to the process rate for growth of heterotrophic biomass on oxygen and nitrate, respectively, as in ASM1 and #9 through #12 were assigned to the four additional process rates. The effects of oxygenase level and activity on the process rate for growth of heterotrophic biomass on oxygen ( ρ 1 ) are expressed in the model of Liu et al. (1998a) by multiplying the ASM1 expression by the term e ν / e O O O,max, as follows: ρ e ν O O O ˆ 1 = μh X BH, e O,max KO, H + S O S (2-5)

35 18 where e O represents the specific level of oxygenase (i.e., eo = EO / XB, H), ν O is activity of oxygenase (ranging from 0 to 1), and e O,max is the maximum specific level of oxygenase. The effect of nitrate reductase level on the process rate for growth of heterotrophic biomass on nitrate ( ρ 2 ) is expressed by ρ μ e ν S η NO NO NO ˆ 2 = H gx B, H eno,max KNO + SNO (2-6) where the enzyme variables eno and ν NO and parameters e O,max are analogous to those in equation (2-6). The process rate for synthesis of oxygenase ( ρ 9 ) can be expressed by the following: ρ = α u S X O 9 O O B, H KOH, + S O (2-7) where α O represents a synthesis rate coefficient for oxygenase and u O is cybernetic variable ranging from 0 to 1, which governs the specific oxygenase synthesis rate. The process rate for synthesis of nitrate reductase ( ρ 10 ) can be expressed by the following: S ρ = α u X NO 10 NO NO B, H KNO + SNO (2-8) where parameter α is analogous to that in Equation (2-8) and u = 1 u. The NO process rate for decay of oxygenase ( ρ 11) was assumed to be first order with respect to oxygenase concentration with the same manner that expresses biomass decay ( ρ 4 ), as follows: ρ = β E (2-9) 11 O O NO O

36 19 where β O is oxygenase the decay coefficient. The process rate for decay of nitrate reductase ( ρ 12 ) is described as follows: ρ = β E (2-10) 12 NO NO where β is the nitrate reductase decay coefficient. NO The variables u O, u NO, ν O, and ν NO in the above formulation represent the control actions of the cellular regulatory process of repression-induction and inhibition-action. The cybernetic modeling approach postulates that the bacteria adjust the values of these variables, as well as the values of e O,max and e NO,max, to maximize their instantaneous growth rate. Kompala et al. (1986) showed the solution of the optimization problem to be u O ρ1 / ν O = ρ / ν + ρ / ν 1 O 2 NO (2-11) ν O ρ / ν max / = 1 O (2-12) ( ρ ν ) 1 O ν NO ρ / ν max / = 2 NO (2-13) ( ρ ν ) 2 NO e O,max αo = ˆ μ + β H O (2-14) e NO,max α NO = ˆ μ η + β H g NO (2-15) Liu et al. (1998a) used the above model to successfully simulate the diauxic lags observed in their experiments. In a second paper, Liu et al. (1998b) pointed out the fact that the new model still could not depict longer lags or the effect of length of aerobic phase on the length of

37 20 diauxic lag. They modified the process rate expressions for enzyme synthesis and enzyme activity. Although the cybernetic variable u is retained in the new model, the overall skim for regulating denitrification is no longer analogous to that of Kompala et al. (1986), hence the model was referred to as a modified cybernetic approach. The process rate for synthesis of oxygenase ( ρ 9 ) was modified by adding a second synthesis rate coefficient, α O,2 as follows: ρ = α + α e S X O O 9 O,1 O,2 B, H O e O,max KO, H + S O u (2-16) The process rate for synthesis of nitrate reductase ( ρ 12 ) was modified as follows: ρ = α + α e S X NO NO 10 NO,1 NO,2 B, H NO e NO,max KNO + SNO u (2-17) The expression for oxygenase activity ( ν O ) was changed to provide a sharper transition from inactive to active enzyme. This was accomplished utilizing a logistic function of the ratio e / e as follows: O O,max 1 ν O = e O 4s rco, e O,max 1+ e (2-18) where r CO, is the value of eo / e O,max at which the oxygenase activity is 0.5 and s is the sharpness parameter, which is the slope of the curve at eo / eo,max = rc, O. Similarly, the expression of nitrate reductase activity ( ν NO ) was modified as follows: 1 ν NO = e NO 4s rcno, e NO,max 1+ e (2-19) where parameters s and r CNO, are analogous to those in Equation (2-18).