@prefix vivo: . @prefix edm: . @prefix ns0: . @prefix dcterms: . @prefix dc: . @prefix skos: . vivo:departmentOrSchool "Applied Science, Faculty of"@en, "Civil Engineering, Department of"@en ; edm:dataProvider "DSpace"@en ; ns0:degreeCampus "UBCV"@en ; dcterms:creator "Barañao, Pablo Andres"@en ; dcterms:issued "2009-11-02T20:38:11Z"@en, "2003"@en ; vivo:relatedDegree "Master of Applied Science - MASc"@en ; ns0:degreeGrantor "University of British Columbia"@en ; dcterms:description "Activated Sludge Model N°3 (ASM3) was chosen as a basis to model an activated sludge system treating effluents from a mechanical pulp and paper mill. The wastewater treatment plant (WWTP) selected, located in Port Alberni, consists of a pH adjuster, primary clarifier (not modeled), five complete mixed bioreactors in series, and a secondary clarifier. The plant treats an average of 74,000 m³ d⁻¹, with the following average influent characteristics (primary effluent): COD of 594 g m⁻³, BOD of 250 g m⁻³, TSS of 27 g m⁻³, pH of 7.2, and temperature of 32°C. A simplified version of the ASM3 model (excluding nitrification and anoxic conditions) was successfully calibrated for this plant. A single set of parameters, with minor variations, was able to fit a variety of batch test respirographs, and performed relatively well when a full-scale simulation was performed using the calibrated parameters values over a time period of nine days. Three measurement campaigns were undertaken in order to calibrate the model properly. The most sensitive parameters (Y[sub STO], Y[sub H], b[sub H], k[sub STO], μ[sub H], K[sub STO], and k[sub H]), as well as the wastewater influent COD fractions (S[sub I], S[sub S], X[sub I], and X[sub S]), were evaluated using three complementary tools: batch respirometric tests, analytical measurements, and mass balance equations. Even though the authors of ASM3 advises not to apply that model for industrial wastewaters and outside the temperature range of 8 - 23°C, the model was found suitable for modelling the COD removal in a WWTP treating pulp and paper effluents at temperatures around 32°C. The importance of this finding is that the application of widely used models, developed originally for municipal wastewater, in pulp and paper applications would simplify considerably the task of modelling and designing WWTP for this industry. Some assumptions of the model, however, proved not to be applicable and some adjustment would be necessary in building a dynamic model for the treatment of these effluents. The least sensitive parameters were not calibrated, so they were either assumed from the literature (f[sub SI], f[sub XI], K[sub X], and K[sub S]) or assumed to be equal to other estimated parameters (b[sub STO] and θ[sub T,bsto], assumed to be equal to b[sub H] and θ[sub T,bH] respectively). Sensitivity and structural identifiability analysis were also performed of the ASM3 simplified model. Only the calibration of the heterotrophic, growth rate (μ[sub H]) presented some identifiability problems, which were solved by estimating an auxiliary parameter (μ[sub OBS]). A few novel calibration procedures were developed for estimating some of the parameters (b[sub H], θ[sub T,bH]), which could be used for any type of effluents. In addition, different methods were used for the estimation of many of the model parameters and wastewater fractions, and some recommendations were done in order to select the best method for estimating different model components."@en ; edm:aggregatedCHO "https://circle.library.ubc.ca/rest/handle/2429/14560?expand=metadata"@en ; dcterms:extent "10487081 bytes"@en ; dc:format "application/pdf"@en ; skos:note "MODELLING CARBON OXIDATION IN PULP MILL ACTIVATED SLUDGE SYSTEMS: CALIBRATION OF ASM3 by PABLO ANDRES BARANAO Lie, P. UNIVERSIDAD CATOLICA DE CHILE, 1995 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF APPLIED SCIENCE in THE FACULTY OF GRADUATE STUDIES (Department of Civil Engineering) We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA August 2003 © Pablo A. Baranao, 2003 in presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of O l l / I L b ^ ' The University of British Columbia Vancouver, Canada DE-6 (2/88) A B S T R A C T Activated Sludge Model N°3 (ASM3) was chosen as a basis to model an activated sludge system treating effluents from a mechanical pulp and paper mill. The wastewater treatment plant (WWTP) selected, located in Port Alberni, consists of a pH adjuster, primary clarifier (not modeled), five complete mixed bioreactors in series, and a secondary clarifier. The plant treats an average of 74,000 m 3 d\"1, with the following average influent characteristics (primary effluent): COD of 594 g m\"3, BOD of 250 g m\"3, f SS of 27 g m\"3, pH of 7.2, and temperature of 32°C. A simplified version of the ASM3 model (excluding nitrification and anoxic conditions) was successfully calibrated for this plant. A single set of parameters, with minor variations, was able to fit a variety of batch test respirographs, and performed relatively well when a full-scale simulation was performed using the calibrated parameters values over a time period of nine days. Three measurement campaigns were undertaken in order to calibrate the model properly. The most sensitive parameters (Y S TO, Y h , bH, ksro, UH, K S T O , and kH), as well as the wastewater influent COD fractions (Si, Ss, Xi, and Xs), were evaluated using three complementary tools: batch respirometric tests, analytical measurements, and mass balance equations. Even though the authors of ASM3 advises not to apply that model for industrial wastewaters and outside the temperature range of 8 - 23°C, the model was found suitable for modelling the COD removal in a WWTP treating pulp and paper effluents at temperatures around 32°C. The importance of this finding is that the application of widely used models, developed originally for municipal wastewater, in pulp and paper applications would simplify considerably the task of modelling and designing WWTP for this industry. Some assumptions of the model, however, proved not to be applicable and some adjustment would be necessary in building a dynamic model for the treatment of these effluents. ii The least sensitive parameters were not calibrated, so they were either assumed from the literature (fsi, fxi, Kx, and Ks) or assumed to be equal to other estimated parameters (bsTo and 0T,bsto, assumed to be equal to bn and 0T,bH respectively). Sensitivity and structural identifiability analysis were also performed of the ASM3 simplified model. Only the calibration of the heterotrophic, growth rate ([in) presented some identifiability problems, which were solved by estimating an auxiliary parameter (UOBS). A few novel calibration procedures were developed for estimating some of the parameters (bH, Qr.bn), which could be used for any type of effluents. In addition, different methods were used for the estimation of many of the model parameters and wastewater fractions, and some recommendations were done in order to select the best method for estimating different model components. iii T A B L E OF C O N T E N T S A B S T R A C T II T A B L E OF C O N T E N T S IV LIST OF T A B L E S X I LIST OF F I G U R E S X I V A C K N O W L E D G E M E N T S X V I D E D I C A T I O N X V I I 1. I N T R O D U C T I O N 1 2. T H E P U L P I N G A N D P A P E R M A K I N G PROCESS 4 2.1 INTRODUCTION TO THE MECHANICAL PULPING PROCESSES 5 2.2 T H E BLEACHING PROCESS ....7 2.3 BASICS OF PAPERMAKING 8 2.4 CHARACTERISTICS OF MECHANICAL PULP AND PAPER EFFLUENTS 9 2.5 MECHANICAL PULP AND PAPER MILL WASTEWATER TREATMENT 11 3. A C T I V A T E D S L U D G E M O D E L L I N G 13 3.1 INTRODUCTION TO MATHEMATICAL MODELLING 13 3.1.1 Usefulness and classification of mathematical models 14 3.1.2 Complexity of the models 15 3.2 HISTORY OF MODELLING FOR MUNICIPAL AND PULP AND PAPER EFFLUENTS 16 3.2.1 Development of models for municipal wastewaters , 17 3.2.2 Development of models for pulp and paper effluents 18 3.3 MODEL VERIFICATION 20 3.3.1 Model Sensitivity 21 3.3.2 Model Identifiability 21 3.3.2.1 Structural identifiability 22 3.3.2.2 Practical identifiability : 23 3.3.3 Overcoming sensitivity and identifiably problems 24 3.4 MODELLING WASTEWATER CHARACTERISTICS 25 iv 3.4.1 Biodegradable organic matter 29 3.4.1.1 Readily biodegradable COD ; 30 3.4.1.2 Slowly biodegradable COD 30 3.4.2 Inert organic matter 32 3.4.2.1 Inert soluble COD 32 3.4.2.2 Inert particulate COD 33 3.4.3 Volatile organic compounds 34 3.4.4 Temperature 34 3.4.5 pH. 37 3.4.6 Toxicity 38 3.4.7 Nutrients • 40 3.5 MODELLING THE MICROORGANISM POPULATION 42 3.5.1 Modelling heterotrophic biomass 43 3.6 MODELLING THE MAJOR PROCESSES 45 3.6.1 Hydrolysis 46 3.6.2 Organic matter removal 49 3.6.2.1 Biooxidation by microorganisms 49 3.6.2.2 Storage of soluble substrate 51 3.6.2.3 Volatilization 51 3.6.2.4 Sorption ,.' 52 3.6.3 Endogenous respiration of microorganisms 53 3.7 PROCEDURES FOR ESTIMATION OF MODEL PARAMETERS 55 3.7.1 Analytical and respirometric techniques commonly used 56 3.7.1.1 Analytical techniques 56 3.7.1.2 Respirometric techniques 56 3.7.2 Estimation of wastewater characteristics 58 3.7.2.1 Readily biodegradable COD (Ss) 58 3.7.2.2 Slowly biodegradable COD (Xs) 59 3.7.2.3 Soluble inert COD (Si) 60 3.7.2.4 Particulate inert COD (XT) 61 3.7.3 Estimation of stoichiometric parameters 62 v 3.7.3.1 Yield coefficients 62 3.7.3.2 Estimation of fxi 64 3.7.3.3 Estimation of fsi 65 3.7.4 Estimation of kinetic parameters 67 3.7.4.1 Endogenous respiration rate, bH and bsTO 67 3.7.4.2 Estimation of kSTo,Ks, PH , and KSTO 68 3.7.4.3 Estimation of 9T for different kinetic parameters 70 3.7.4.4 Hydrolysis parameters k H and K x 71 4. RESEARCH OBJECTIVES AND METHODOLOGY.. 73 4.1 RESEARCH OBJECTIVES 73 4.2 RESEARCH METHODOLOGY 73 4.2.1 Selection and description of the mill 73 4.2.2 Selection of the model 75 4.3 DESCRIPTION OF THE ASM3-BASED MODEL 76 4.3.1 Model components 77 4.3.2 Processes in the model 79 4.3.3 Stoichiometry of the model 79 4.3.4 Kinetic expressions of the model 80 4.3.5 Restrictions and constraints 81 4.4 EXPERIMENTAL PROGRAM : 81 4.4.1 Sensitivity analysis 82 4.4.2 Identifiability analysis 83 4.4.3 Sampling program '. 83 4.4.4 Analytical methods used • 84 4.4.4.1 Filtration 84 4.4.4.2 Chemical oxygen demand (COD) 84 4.4.4.3 Ultimate biochemical oxygen demand (UBOD) 85 4.4.4.4 Total and volatile suspended solids (TSS and VSS) 86 4.4.4.5 Clarification by coagulation and flocculation 86 4.4.4.6 Ammonia (NH3-N) and ortho-phospahe (P04\") 86 4.4.4.7 Alkalinity 87 vi 4.4.5 Quality assurance program 87 4.4.5.1 Field blanks 87 4.4.5.2 Duplicates 87 4.4.5.3 Standard curves 88 4.4.6 Respirometric methods used 56' 4.4.7 Data from historical database 90 4.5 MODEL PARAMETER ESTIMATION METHODS 92 4.5.1 Parameters estimated by analytical measurements 92 4.5.1.1 Observed heterotrophic yield (YOBS) 92 4.5.1.2 Soluble inert influent COD fraction (fs,)in) 92 4.5.1.3 Readily biodegradable COD fraction (fss.in)..., 92 4.5.1.4 Slowly biodegradable COD fraction (f x s >in) 93 4.5.1.5 Heterotrophic biomass (XH) 93 4.5.1.6 Endogenous respiration rate coefficient (bH) 94 4.5.2 Parameters estimated by respirometry 95 4.5.2.1 Storage yield (Y S T o) 9 5 4.5.2.2 Observed heterotrophic yield ( Y H ) 96 4.5.2.3 Readily biodegradable COD fraction (f s s) 97 4.5.2.4 Endogenous respiration rate coefficient QOH) 97 4.5.2.5 Temperature coefficient of bH (OT.OH) 98 4.5.2.6 Oxygen saturation constant (K02) 99 4.5.2.7 Curve fitting for estimation of k S T o , M«, K S TO , and k H 99 4.5.2.8 Observed maximum growth rate (UOBS) 99 4.5.3 Parameters estimated by mass balances 100 4.5.3.1 Particulate inert COD fraction (fXi,in) 100 4.5.3.2 Slowly biodegradable COD fraction (f x s) 101 4.5.4 Methods used for model assumption verification 101 4.5.4.1 Absence of volatile compounds in the influent 101 4.5.4.2 TSS to VSS relation '. 101 4.5.4.3 pH close to neutrality 102 4.5.4.4 Generation of Si during endogenous respiration 102 vii 4.5.4.5 Nutrients are not limiting for heterotrophic bacteria 103 4.5.4.6 Assumptions checked with the results 103 4.5.5 Statistical methods 103 5. R E S U L T S A N D DISCUSSION: C A L I B R A T I O N OF A S M 3 F O R P O R T A L B E R N I A C T I V A T E D S L U D G E S Y S T E M 104 5.1 SENSITIVITY ANALYSIS 104 5.2 IDENTIFIABILITY ANALYSIS 107 5.2.1 Without hydrolysis 107 5.2.2 Incorporating hydrolysis 110 5.2.3 Restrictions to identifiability due to sensitivity of parameters Ill 5.2.4 Selection of identifiable parameters to be estimated by curve fitting Ill 5.2.5 Temperature coefficients 112 5.3 VERIFICATION OF SOME MODEL ASSUMPTIONS 112 5.3.1 Absence of volatile COD in the influent 113 5.3.2 TSSto VSS relation 114 5.3.3 pH close to neutrality 115 5.3.4 No generation of Sj during endogenous respiration 116 5.3.5 Nutrients are not limiting for heterotrophic bacteria 117 5.4 MODEL CALIBRATION 119 5.4.1 Parameters not calibrated 119 5.4.2 Estimation of model parameters 121 5.4.2.1 Yield coefficients estimation 121 5.4.2.2 Wastewater characterization 123 5.4.2.3 Kinetic parameters estimation 131 5.5 SUMMARY OF THE CALIBRATED MODEL 143 6. E N G I N E E R I N G S I G N I F I C A N C E OF THIS R E S E A R C H 145 6.1 POSSIBLE APPLICATIONS OF THIS RESEARCH 145 6.2 FULL-SCALE SIMULATION USING THE CALIBRATED MODEL 146 6.3 MAJOR LIMITATIONS OF THE PRESENTED RESULTS 148 7. C O N C L U S I O N S 149 viii 7.1 SENSITIVITY AND IDENTIFIABILITY ANALYSES 150 7.2 WASTEWATER CHARACTERIZATION 151 7.3 PARAMETER ESTIMATION 152 8. R E C O M M E N D A T I O N S F O R F U T U R E D E V E L O P M E N T 154 9. L I S T O F T E R M S , S Y M B O L S , A N D A C R O N Y M S 155 10. G L O S S A R Y 158 11. R E F E R E N C E S 160 A P P E N D I X 1: R E L A T I O N S H I P B E T W E E N M O B s A N D M H 176 A P P E N D I X 2: S E N S I T I V I T Y A N A L Y S I S 178 APPENDIX 2.1 SENSITIVITY FUNCTION FOR M A T L A B ® USED IN ANALYSIS 178 APPENDIX 2.2 SENSITIVITY RESULTS 184 A P P E N D I X 3: S A M P L I N G C A M P A I G N S D E S C R I P T I O N S 186 APPENDIX 3.1: SAMPLING CAMPAIGN N°l 186 APPENDIX 3.2: SAMPLING CAMPAIGN N°2 187 APPENDIX 3.3: SAMPLING CAMPAIGN N°3 188 A P P E N D I X 4: S U M M A R Y O F R E S P I R O M E T R I C T E S T S 189 APPENDIX 4.1 DATA FROM BATCH TESTS PERFORMED MIXING WASTEWATER WITH SLUDGE 189 APPENDIX 4.2 SUMMARY OF BATCH TESTS INFORMATION 199 APPENDIX 4.3 DATA FROM LONG-TERM BATCH TESTS .„ 201 . APPENDIX 4.4 RESPIROMETRIC MEASUREMENTS FOR ESTIMATING M 0 B S 202 APPENDIX 4.5 RESPIROMETRIC INFORMATION FOR ESTIMATING K 0 2 204 APPENDIX 4.6 RESPIROMETRIC INFORMATION FOR ESTIMATING ©T,BH 206 A P P E N D I X 5: S U M M A R Y O F S A M P L I N G C A M P A I G N N°l R E S U L T S 207 A P P E N D I X 6: D A T A U S E D F R O M H I S T O R I C A L D A T A B A S E 212 A P P E N D I X 7: M A T L A B ® F I L E U S E D F O R F U L L - S C A L E S I M U L A T I O N 213 A P P E N D I X 8: P H O T O S 219 ix A P P E N D I X 9: A N A L Y T I C A L R E S U L T S O F E X P E R I M E N T S F O R M O D E L P A R A M E T E R S E S T I M A T I O N 222 APPENDIX 9.1 COD RESULTS FROM LONG-TERM BATCH TEST FOR ESTIMATING B h 222 APPENDIX 9.2 COD RESULTS FOR ESTIMATING THE PRODUCTION OF SI DURING ENDOGENOUS RESPIRATION 222 APPENDIX 9.3 COD RESULTS FOR ESTIMATING Y 0 B S 223 APPENDIX 9.4 LONG-TERM BOD TESTS 223 x LIST OF TABLES Table 2.1: Yield of different pulping process 5 Table 2.2: Typical characteristics of waters from mechanical pulping 10 Table 4.1: Definition of components and symbols of the model 77 Table 4.2: Stoichiometric matrix of ASM3 aerobic sub-model 80 Table 4.3 Kinetic expressions of ASM3 aerobic sub-model 80 Table 4.4: Coefficient of variation for different analytical methods 88 Table 4.5: Parameters obtained from mill historical database 91 Table 5.1: Sensitivity of model parameters based on assumed values 105 Table 5.2: Sensitivity of model parameters based on calibrated values 106 Table 5.3: Evolution of COD and COD s during abundant aeration of wastewater 113 Table 5.4: TSS/VSS ratio of samples from different sampling points 114 Table 5.5: Statistical information of pH series between January 1st until December 17th.... 115 Table 5.6: Alkalinity concentrations in Port Albemi WWTP 116 Table 5.7: Ammonia and ortho-phosphate concentrations at Port Alberni WWTP 117 Table 5.8: Confidence intervals for ammonia and ortho-phosphate at Port Albemi WWTP 118 Table 5.9: Parameters not calibrated and their assumed values 120 Table 5.10: Estimation of Y S T o 1 2 1 Table 5.11: Estimation of Y H 122 Table 5.12: Analytical estimation of Ss,in 124 Table 5.13: Results of physical-chemical estimation of S s 125 Table 5.14: Wastewater characterization of soluble COD fractions 126 Table 5.15: Biodegradable and inert COD fractions 127 Table 5.16: Summary of COD influent fractions 131 Table 5.17: Best fit values for parameter K S T O 138 Table 5.18: Best fit values for parameters kSTo, UH , and kH 140 Table 5.19: Summary of calibrated model parameters (kinetics parameters at 30°C) 144 Table A2.1: Sensitivity of model parameters based on assumed values 184 Table A2.2: Sensitivity of model parameters based on calibrated values 185 Table A3.1: First sampling campaign chronology 186 Table A3.2: Daily measurement matrix 187 Table A4.1: Respirometric information obtained from Test 2-1 189 xi Table A4.2: Respirometric information obtained from Test 4-1 190 Table A4.3: Respirometric information obtained from Test 4-3 191 Table A4.4: Respirometric information obtained from Test 5-1 192 Table A4.5: Respirometric information obtained from Test 5-2 193 Table A4.6: Respirometric information obtained from Test 7-1 194 Table A4.7: Respirometric information obtained from Test 7-3 195 Table A4.8: Respirometric information obtained from Test 8-1 196 Table A4.9: Respirometric information obtained from Test 8-3 197 Table A4.10: Respirometric information obtained from Test 9-1 198 Table A4.11-A: Summary of batch tests information 199 Table A4.11-B: Summary of batch tests information 200 Table A4.12: Endogenous respiration Test N° 1 (T =24.4 °C) 201 Table A4.13: Endogenous respiration Test N° 2 (T = 30.2°C) 201 Table A4.14: Respirometric data from batch test 9-2 202 Table A4.15 Decrease of OUR due to a decrease on DO in the presence of substrate 204 Table A4.16 Decrease of OUR due to a decrease on DO in the absence of substrate, Test 1 204 Table A4.17 Decrease of OUR due to a decrease on DO in the absence of substrate, Test 2 205 Table A4.18 Decrease of OUR due to a decrease on DO in the absence of substrate, Test 3 205 Table A4.19 OUR during a rapid increase in temperature to endogenous sludge 206 Table A5.1: COD results for sequences 1 to 7 at different plant locations 207 Table A5.2: Soluble COD results for sequences 1 to 7 at different plant locations 207 Table A5.3: TSS results for sequences 1 to 7 at different plant locations 208 Table A5.4: VSS results for sequences 1 to 7 at different plant locations 208 Table A5.5: NO2\" + N0 3\" results for sequences 1 to 7 at different plant locations 208 Table A5.6: N0 2\" results for sequences 2, 4, and 5 at different plant locations 209 Table A5.7: Ammonia results for sequences 1 to 7 at different plant locations 209 Table A5.8: TKN results for sequences 1 to 6 at different plant locations 209 Table A5.9: Soluble TKN results for sequences 1 to 5 at different plant locations 210 Table A5.10: PO4\" results for sequences 1 to 7 at different plant locations 210 Table A5.11: Total P results for sequences 1 to 6 at different plant locations 210 xu Table A5.12: Soluble total P results for sequences 1 to 5 at different plant locations 211 Table A9.1: Particulate COD results and XH estimation from a long-term batch test 222 Table A9.2: COD and COD s results from a long-term batch test without the addition of substrate 222 Table A9.3: COD and particulate COD results from a high F/M ratio batch test (9-2) 223 Table A9.4: Long-term BOD Test N° 1 with filtered and unfiltered influent and effluent.. 223 Table A9.5: Long-term BOD Test N° 2 with filtered and unfiltered influent and effluent.. 224 xiii LIST OF FIGURES Figure 3.1: Effect of temperature on biological oxidation rate (Eckenfelder, 2000) 35 Figure 3.2: Schematic diagram of a respirometer with a computer connection 57 Figure 3.3: Breakpoint 'B ' and area that was assumed to represent the oxygen used during storage process. Data from Test 9-1, performed with F/M of 0.9 and temperature of 21.6°C 63 Figure 3.4: Influence of the parameter fsi in the estimation of S^n 66 Figure 3.5: Different regions of a respirographs and their implications on parameter estimation 69 Figure 4.1: Diagram of Port Alberni WWTP 74 Figure 4.2: COD fractions in ASM3 simplified model 78 Figure 5.1: Influent pH during 2002 115 Figure 5.2: Soluble COD in a batch test with sludge in endogenous respiration state 117 Figure 5.3: Long term total and soluble BOD effluent 123 Figure 5.4: Long-term influent BOD total and soluble 128 Figure 5.5: Wastewater characterization summary 131 Figure 5.6: Endogenous respiration coefficient based on particulate COD measurements.. 132 Figure 5.7: Endogenous respiration decay of OUR in two batch reactors, at 24°C and 30°C respectively 133 Figure 5.8: Specific OUR versus temperature of different batch tests 134 Figure 5.9: Increase of OUR of sludge in endogenous state during a rapid increase of temperature 135 Figure 5.10: Plot of oxygen switching function versus oxygen concentration during growth/storage processes 136 Figure 5.11: Plot of oxygen switching function versus oxygen concentration during endogenous respiration 137 Figure 5.12: Two respirographs at very low F/M ratio 139 Figure 5.13: Two respirographs at low F/M ratio 139 Figure 5.14: Variation of kSxo and uH with temperature 141 Figure 5.15: Variation of kH with temperature 142 Figure 5.16: Estimation of UOBS by respirometry 143 Figure 6.1: Influent flow (Qin) and influent COD (CODin) for validation period 147 xiv Figure 6.2: Full-scale simulation performed with the calibrated model to historical data from the Port Albemi WWTP 147 Figure A8.1: Cells 1A and IB from Port Albemi WWTP 219 Figure A8.2: Secondary clarifier from Port Albemi WWTP 219 Figure A8.3: Respirometer with temperature-controlled water recirculation system 220 Figure A8.4: Three long-term batch respirometric tests 220 Figure A8.5: Computer set-up for recording the dissolved oxygen concentration 221 Figure A8.6: The author with the sampling device 221 xv ACKNOWLEDGEMENTS I would like to thank Professor Eric R. Hall for his guidance and support throughout this research. Many thanks also go to Susan Harper and Paula Parkinson for their assistance with the analyses performed at the UBC Civil Engineering laboratory. I would like also to thank Larry Cross and Port Alberni Specialties for letting us to work at their facilities, and for all the help provided. Thanks also to the crew for their help and food (John, Bob, Peter, and Paul). I would like to the institutions that made this research possible through their financial support: the Natural Sciences and Engineering Research Council of Canada (NSERC); The National Commission for Scientific and Technological Research of Chile (CONICYT); and the Chilean Safety Association (ACHS). My thanks also to the people who encouraged me to come to Canada and doing this research, including Fernando Baranao and Pilar Diaz (my parents), Eduardo Undurraga, Eugenio Cantuarias, Angel Mansilla, Martin Fruns, Alejandro Tapia, Jean-Luc Antoine van Den Ende, Juan Francisco Urmeneta, Gonzalo Pulido, and Gloria Herrera. Many other relatives and friends should be also included in this list. I would like to thank also to Ing-Wei Lo (Wayne), for his company and help; the UBC Pulp and Paper Centre, for its complete library and kindness; and to McGraw-Hill Companies, for kindly let me use a figure from the book Industrial Water Pollution Control. Special thanks to Robin B. Clark for his very useful (and needed) editing help with the English. My biggest thanks to my family, to whom I dedicate all this work. xvi To Paulina, my wife, and to Alberto and Trinidad, our children, who went from Chile to Canada to accompany me during this research. \"If you shut your door to all errors, truth will be shut out.\" - Rabindranath Tagore xviii 1. Introduction Historically, the pulp and paper industry has been considered to be a major consumer of natural resources and energy, and a significant contributor of pollutant discharges to the environment (Thomson et al, 2001). Among all the impacts generated by the pulp and paper industry, pollution to water bodies is the impact of major concern (Ali and Sreekrishnan, 2001). In fact, the Canadian situation shows that pulp and paper mills use more water and produce more water pollution than any other industry in the country (Christie and McEacher, 2000). Increasing public concern on environmental care and new regulations about effluent discharges are demanding more stringent control over the mill effluents. Since 1992, with the implementation of the federal Pulp and Paper Effluent Regulations, Canada's pulp and paper mills have been required to achieve high contaminant removal efficiency in their treatment systems. Energy conservation, water reduction and environmental management, along with pollution reduction, have played an important role in coping with these environmental challenges (Greenbaum, 2002; Kantardjieff et al, 1997; Schroderus et al, 2002). To meet the challenge of more stringent effluent regulations, the Canadian pulp and paper industry has adopted biological systems to treat its effluents, reducing considerably its environmental impacts. However, there is little knowledge of the particular characteristics of these effluents, their biodegradability, and the biological processes involved in their treatment. Therefore, pulp and paper wastewater treatment plants have been designed and operated mainly based on the knowledge of municipal wastewater, regardless of the differences between them. For a better understanding of the biodegradability of pulp and paper effluents and the processes involved in their treatment, developing a mathematical model is a very useful approach. A mechanistic model, in particular, is necessary for understanding the system, because it is a mathematical representation of the biological processes that are responsible of the degradation of the organic matter. A 'black box' model may be useful for control or designing, but those models are not helpful for understanding the processes involved. 1 Modelling is, in fact, a very important part of the design and operation of wastewater treatment plants. A good model of the pulp and paper treatment dynamics would not only help in designing the most appropriate treatment plants, but also in improving the control of plants that are already operating. Better models allow better design and better control of the systems, which result finally in lower costs in building and operating treatment facilities and an improved environment. Modelling is a very active research area, and many models are developed each year to describe the processes involved in treating the organic material in different types of wastewaters and to predict treated effluent characteristics. New models have replaced early models that were originally labeled as \"general models\" by their authors. In comparison to the many models that have been developed for application to municipal wastewater, the simulation of pulp and paper wastewater treatment is at an early stage, and there still is no satisfactory model for pulp and paper wastewater treatment. The particular characteristics of pulp and paper wastewaters, such as high organic content, low nutrient content, high proportion of readily biodegradable substrate, high temperature, and toxicity, may require the use of a different activated sludge model than the models used for municipal wastewater, or should require at least a specific calibration of an existing model. Previous works carried out at the University of British Columbia have made some advances in this task, calibrating Activated Sludge Model N°l (ASM1) to bleached kraft pulp mill effluents (BKME) (Stanyer, 1997), and building a new model based on Activated Sludge Model N°2 (ASM2) for mechanical pulp effluents (Sreckovic, 2001). However, the parameter values obtained by both researchers were significantly different from each other, and contrasting as well from typical municipal wastewater values, especially those found by Sreckovic (2001). As a result, further work was needed in order to check the similarity or dissimilarity of the pulp and paper model parameters in relation to the widely available information about domestic sewage. The objective of this research was to contribute to the development of a mechanistic model for predicting the organic removal in biological wastewater treatment plants treating pulp and paper wastewater. Since there are a number of models and plenty of information available for 2 municipal wastewaters, the easiest and most useful way of building a model for pulp and paper effluents would be testing and calibrating a model developed for municipal applications, and propose modifications only if it is required. The wastewater treatment plant selected for developing this research was an activated sludge system treating mechanical pulp wastewaters, located in Port Alberni, British Columbia. Therefore, the literature review is focused mainly on the mechanical pulp and paper processes and effluents, and on the activated sludge treatment system modelling. 3 2. The Pulping and Papermaking Process Pulping refers to any process by which a fibrous raw material is reduced to a fibrous mass (Smook, 2002). Before wood became the major fibre source (around 1840s), many other materials had been used as fibre source for the pulping process, such as: bamboo, mulberry, fishnets, rags, hemp waste, linen and cotton. Increasing shortages of raw materials during the 18th century made it evident that a process for utilizing a more abundant fibrous material was needed. Different wood pulping processes were gradually developed that relieved the paper industry of dependency upon cotton and linen rags and made modern large-scale production possible (India Paper, 2003). Wood is a natural composite material that consists of fibres separated by lignin. Its complexity is high due the fact that the wood fibre is, in itself, a composite made of fibrils, lignin and hemicellulose; and fibrils are also a composite consisting of micro fibrils (Sundholm, 1999), made mainly of cellulose. Pulping of wood is the way by which the bonds are systematically ruptured within the wood structure, so the fibres can be separated from each other without being too damaged. For this purpose, the lignin and the hemicellulose are either dissolved or more or less softened. The purpose of all pulping processes is to separate the fibres from the wood and make them suitable for papermaking (Sundholm, 1999). Rupturing can be accomplished mechanically, thermally, chemically, or by combinations of these treatments. Existing pulping processes for wood pulp are broadly classified as mechanical, chemical, or semi chemical (Smook, 2002). After the major pulping process is finished, there are many processes before the pulp is ready for the papermaking process, including: defibreing, de-knotting, brown stock washing, screening, centrifugal cleaning, thickening, pumping, storage, blending, drying, and bleaching. In relation to the present research, only mechanical pulping is described in more detail. In addition, for their relative significance and environmental importance, bleaching and papermaking processes are also explained briefly. 4 2.1 Introduction to the mechanical pulping processes In the mechanical pulping process, the wood and the fibres are fatigued by vibrational forces that separate the fibres until the structure loosens up. In practice, there are two major ways to produce mechanical pulp on an industrial scale: by pressing wood blocks against a revolving stone (grinding); and by disintegrating wood chips in a disc refiner (refining). The grinding process was the first industrial pulping process developed (in the 1840s). In the 1880s and '90s the hot grinding process was developed, which obtained much longer fibres due the high temperature and press loads. In the 1930s, the thermomechanical pulping (TMP) process using wood chips was developed in Sweden as an evolution of the refiners of the time. Finally, in 1938, at Blandin Paper (United States) a refiner was used to defibre chips soaked in chemicals. Several other mills followed suit, and the semi-mechanical pulping processes developed in the direction of the high-yield sulfite process and in the direction of the chemi-thermomechanical pulping (CTMP) processes of today (Sundholm, 1999). Mechanical pulping processes have the advantage of converting up to 95% of the dry weight of the wood into pulp, but require prodigious amounts of energy to accomplish this objective (Smook, 2002). These high pulping yields have a direct impact in reducing the organic matter that is discharged by mills with this technology, because less wood ends up going into the wastewater. Table 2.1 compares the yields of different mechanical pulping process with chemical pulping. Table 2.1: Yield of different pulping process Pulping Process Yield (%) Stone groundwood (SGW) 90-98 Refiner mechanical pulp (RMP) 90 - 95 Thermomechanical pulp (TMP) 90 - 97 Chemi-thermomechanical pulp (CTMP) > 90 Chemical pulps 40-55 Sources: (Smook, 2002; Sundholm, 1999) 5 Mechanical pulping is the method that produces most of Canada's pulp, representing the 47% of the total pulp produced in Canada, ahead of bleached kraft pulping (BKP) with 44% of the Canadian total. Furthermore, Canada is the country that produces the most mechanical pulp, accounting for one third of the world's total. On a world scale, however, mechanical pulping only represents the 20% of the total production, whereas chemical pulping accounts for the 69% of the total (Pulp & Paper International, 2001). There are several mechanical pulping processes, but the four most important are: stone groundwood (SGW), refiner mechanical pulp (RMP), thermomechanical pulp (TMP), and chemi-thermomechanical pulp (CTMP) (from Smook, 2002). In the stone groundwood (SGW) process, pulp is produced by pressing blocks of wood against an abrasive rotating stone surface. The wood blocks are oriented parallel to the axis of the stone so that the grinding action removes somewhat intact fibres. Refiner mechanical pulp (RMP) is produced by the mechanical reduction of wood chips in a disc refiner. The process usually employs two high consistency refiner stages operating in series. RMP produces a longer-fibred pulp, which is stronger, freer and bulkier than conventional SGW. Thermomechanical pulping (TMP) is a modification of the RMP process. It involves steaming the raw material under pressure for a short period of time prior to and during refining. The steaming serves to soften the chips, with the result that the pulp produces a greater percentage of long fibres and fewer shives than RMP. Chemi-thermomechanical pulping (CTMP) is a modification of the TMP process. A modest chemical impregnation is incorporated during the steaming stage, which serves to improve the bonding properties of the pulp. In conjunction with established peroxide bleaching technology, the CTMP process produces high-brightness pulp at a reasonable cost. 6 2.2 The bleaching process The objective of bleaching pulp is to increase brightness, which is accomplished by different means for mechanical and chemical pulps. While chemical pulp bleaching is based mainly on lignin removal, mechanical pulp bleaching is based on elimination of colored chemical groups in lignin. For that reason, bleaching of mechanical pulp is often referred to as lignin-retaining bleaching or brightening (Sundholm, 1999). Modern mechanical pulp bleaching is achieved through a continuous sequence of process stages utilizing different chemicals and conditions in each stage, usually with washing between stages (Smook, 2002). There are several possible bleaching processes, four of which are discussed here. Peroxide bleaching is normally performed with hydrogen peroxide (H2O2) (Sundholm, 1999), which can be used at different temperatures depending on the bleaching objective. When applied at moderate temperature up to 60°C, peroxide is an effective lignin-preserving bleaching agent. At slightly higher temperatures (i.e., 70 - 80°C), peroxide is used to provide increments of brightness and to improve brightness stability. Peroxide bleaching is strongly affected by pH, which must be adjusted and buffered at about 10.5 for best results (Smook, 2002). Sodium hydrosulphite (Na2S2C>4) is used for decoloring the lignin by reducing the colored quinoid structures to colorless phenolic compounds, retaining the advantages of high yields because the lignin is not solubilized. This method is not capable of providing a high brightness level or a stable brightening effect, but is sufficient for improving the eye appeal of newsprint and other products. Oxygen (O2) alone is not a good bleaching agent because of its lack of selectivity. However, the addition of magnesium \"protector\" into the process can successfully control cellulose degradation up to a certain degree of delignification. Typically, it is possible to reduce the lignin content by up to 50% in a single oxygen prebleach stage (and even further with a two-stage system) without excessive cellulose degradation. Environmental concerns have increased 7 the interest in this process, which allows reducing the amounts of chlorine ( C I 2 ) or chlorine dioxide ( C I O 2 ) used. Caustic soda is used in the so-called extraction stage, which solubilizes lignin made susceptible to alkali in the previous stages and reactivates the pulp for further oxidation. The final pH of this stage should be above 10.8, otherwise the solubilization would be incomplete. 2.3 Basics of papermaking The art of true papermaking had its origin in China around 100 A.D. It was discovered that a mat of fibres (obtained by macerating the inner bark portion of bamboo) could be formed by filtering a pulp suspension through a fine screen. The mat, after pressing and drying, was found suitable for writing and drawing (Smook, 2002). The first commercially successfully paper machine was built by the Fourdrinier brothers in 1799. Since then, that machine has undergone a continual evolution of change. The basic components of today's paper machine are: the stock inlet, which distributes the papermaking fibres uniformly across the machine from back to front; - the headbox, which distributes the paper stock onto the moving forming wire; the Fourdrinier wire, which is a moving wire that form the fibres into a sheet and enables the furnish to drain by gravity and dewater by suction; - the press section, where the sheet is conveyed through a series of presses where additional water is removed and the web is consolidated; the dryer section, where most of the remaining water is evaporated and fibre bonding develops as the paper contacts a series of steam-heated cylinders; the calender section, where the sheet is pressed between metal rolls to reduce thickness and smooth the surface; and - the reel, where the dried, calendered paper is wound onto a reel (based on Smook, 2002). In the Fourdrinier wire (forming section), approximately 95% of the water is removed from the paper stock by gravity or suction, an additional 4% is removed by the presses, 0.8% is 8 evaporated, and approximately 0.4% of the original water remains in the finished paper (based on Smook, 2002). Most of the water and the fibres collected are recycled during this process; however, the forming and the pressing section are the major contributors of effluents during the papermaking process. 2.4 Characteristics of mechanical pulp and paper effluents There are many factors affecting the composition and quality of the effluent from a pulp and paper mill, including the pulping processes used, bleaching sequence, papermaking process, wood used (softwood or hardwood), tree species, technology used (depending basically on the year of construction), purification steps, retention aids, etc. For this reason, even though the effluents are usually classified only based on the pulping processes used, the wastewater characterization is highly site-dependent. The dissolved and particulate substances in the process waters of mechanical pulp and paper systems originate from several sources (from Sundholm, 1999): soluble and particulate substances mainly from the debarking; soluble and particulate components of wood during the high-temperature conditions of the defibreization process (hemicellulose, pectins, lignin, extractives, and inorganic salts); residuals from papermaking chemicals with the re-circulated water from papermaking to pulping; and - leakages from sealing, lubrication, and dissolution of equipment materials into the process. Some parameters used to measure the pollution level or quality of the mechanical pulp and paper mill effluents are the following. - B i o c h e m i c a l oxygen d e m a n d (BOD) is the amount of oxygen required by bacteria while stabilizing biodegradable organic matter under aerobic conditions. It is a widely used test to determine the pollutional strength of wastewaters in term of the oxygen they will consume if discharged into natural water systems. BOD is measured as a function 9 of time. For example, BOD5 will measure the oxygen consumed after 5 days of incubation; - Chemical oxygen demand (COD) is another widely used test for estimating wastewater organic strength. This test measures the total quantity of oxygen required for oxidizing chemically the organic matter in the effluent; - Total suspended solids (TSS) generally refers to the matter that, retained in a filter of 2.0 pm or smaller, remains as residue upon evaporation and drying at 103°C to 105°C; - Acute toxicity, measured as lethal concentration (LC), is the concentration of the effluent estimated to produce death in a specified number of test organisms in a certain time period. For example, 96h LC50 is the concentration that produces 50% mortality over a period of 96 hours; - Resin acids, as a substance group, have been identified as a major source of toxicity in mechanical pulp effluents, especially in CTMP wastewaters (Ali and Sreekrishnan, 2001), where they can contribute between 60 to 90% of the toxicity. Resin acids are present in softwoods in a much higher quantity than in hardwoods (Leach and Thakore, 1976, as cited in Ali and Sreekrishnan, 2001); and Fatty acids, together with resin acids, are the main contributors to the toxicity of pulp and paper effluents, especially for CTMP (Cornacchio and Hall, 1988). Table 2.2 presents a summary with typical ranges of these parameters for different mechanical pulp and paper mill processes. Table 2.2: Typical characteristics of waters from mechanical pulping Pulping process Unit SGW TMP CTMP BOD7 gm\"3 10,000 - 12,000 1 1,600 4-25,000 1 2,000 3 - 35,000 1 COD gm\"3 30,000 - 40,000 1 50,000 - 80,000 1 5,000 3 - 100,000 1 TSS gm\"3 209 4 180 3 -220 4 Acute toxicity 96h L C 5 0 3 - 12 2 1.3-35 2 0.2 - 5 5 Resin acids gm\"3 1 - 8 3 2 2-21 2 1 3 - 560 2 Fatty acids gm\"3 0.3 3 -500 2 Sources: 1 Sundholm (1999);2 Cornacchio and Hall (1988);3 Wilson and Frenette (1988); 4 McAllen (1988);5 Almemark and Frostell (1988). 10 2.5 Mechanical pulp and paper mill wastewater treatment As each pulp and paper mill is a large, complex and highly interactive operation, the treatment of wastewaters from these mills tends to become mill-specific. In general, a wastewater treatment plant (WWTP) treating pulp and paper effluents consists of two stages: primary and secondary treatment. Primary treatment typically consists of a primary sedimentation process, which reduces the TSS load of the raw effluent. The solids, which consist principally of residual fibres from the pulp and papermaking process, are collected as underflow sludge. This sludge requires further treatment or disposal, and it is typically thickened and incinerated due its high calorific value. The major objective of the secondary treatment process is reducing the dissolved organic matter of the effluent, since the majority of the particulate fraction is removed in the primary treatment. There are different alternatives for this secondary treatment (biological v/s physical-chemical; aerobic v/s anaerobic; mesophilic v/s thermophilic). Among all the options, the mesophilic aerobic biological treatment system has proven to be the best way of treating most wastewater effluents generated by the pulp and paper industry (Ali and Sreekrishnan, 2001; Springer, 2000). Mesophilic refers to the range of temperatures used, typically in the range of 25 - 40°C, in contrast to the thermophilic process that requires temperatures in the range of 55 -65°C. For treating mechanical pulp effluents, anaerobic treatment has the problem that tannins, resin acids, and long chain fatty acids are toxic to anaerobic and methanogenic bacteria (Ali and Sreekrishnan, 2001). There are different types of aerobic treatments, the most common are the aerated stabilization basins (ASB) and the activated sludge process. ASB are large lagoons with mechanical surface aerators, which considerably increase the rate of biodegradation compared with a natural biodegradation in the environment. However, the most popular aerobic biological treatment systems are activated sludge processes, which consist of one or more aeration basins and secondary clarifiers. Activated sludge systems usually have the highest efficiency of treatment, in the range of 90 - 95% BOD removal (McAllen, 1988), and therefore require less 11 land area than other less efficient biological processes. They are also very stable and reliable although not entirely devoid of troubles (Springer, 2000). An additional advantage of aerobic treatment is that natural wood extractives (i.e., resin and fatty acids) are readily biodegradable aerobically (Turk, 1988). Activated sludge treatment has proven to be able to remove most of the wood extractives from pulp and paper effluents, reducing their concentrations to levels below those that have been suggested to cause acute toxicity or disturb the hormonal balance of aquatic organisms (Kostamo et al, 2003). Zero effluent technology is an alternative to biological treatment of pulp and paper effluents. It means that no process effluent is discharged. Al l effluents from the mill are collected in storage ponds, the suspended solids are removed by mechanical clarifiers, and the removed sludge is dewatered and burned in a waste fuel boiler. The clarified effluent is concentrated by evaporation and incinerated in a recovery boiler. Zero effluent is an elegant way to eliminate effluent pollution, but it is expensive (about 25% to 30% extra capital cost of a CTMP pulp mill) (Sundholm, 1999), so currently the state of the art technology is a partial close loop of the waste water system plus secondary treatment (Stratton et al, 2003). 12 3. Act ivated s l u d g e mode l l ing Activated sludge systems are highly complex, and involve biological, physical-chemical, and hydraulic processes combined. Therefore, the use of models is indispensable for designing and controlling activated sludge plants, which may vary depending on the objective of the plant and the characteristics of the wastewaters. Activated sludge modelling is almost as old as activated sludge technology itself. When investigators started aerating wastewater trying to oxidize the organic matter in the early 1910s, they found that the growth of some microorganisms might greatly increase the degree of purification obtained. In trying to explain this fact the first models were developed, a task that has continued until today. 3.1 Introduction to mathematical modelling Models are representations of reality and often, because of the complexity of the reality, they are also simplifications. Models can be physical (i.e. pilot plants) or conceptual (i.e. equations) representations of the system being modeled. A mathematical model is a particular form of representation, which attempts to capture, in the form of equations, certain characteristics of a system for a specific use (or purpose) of the model (Hangos and Cameron, 2001). Ultimately, a model is a 'machine' that transforms inputs to outputs by defined relations (Chui and Chen, 1989). The outputs of a model are those variables the model user is interested in, and the inputs consist of disturbances and manipulated variables that affect the outputs. The most important feature of a state-space model is the introduction of state variables, which act as mediators between the inputs and the outputs. Then, the state of the system is defined as the values of the state variables at any instant of time (Dochain and Vanrolleghem, 2001). The equations (model structure) have three components: variables, constants and parameters. Variables are used to represent inputs, outputs and states; constants are values that never change throughout all possible applications of the model; and parameters are components whose value may vary depending the circumstances of the application, so their value has to be determined for each particular application (calibration). 13 3.1.1 Usefulness and classification of mathematical models The usefulness of mathematical modelling is closely related to the purpose we have to create and use these models. The two major objectives in using models in biological wastewater treatment are the process design and the determination of optimal operating conditions. Other common wastewater model applications are to increase the understanding of the bioprocess mechanisms, trouble shooting, estimation of the 'true' capacity of a WWTP, and the training of operators and process engineers. Mathematical models can be classified according to many different criteria, such as the following. Depending on the purpose of the model, it can be a model for understanding or a model for prediction/forecasting. Models applied for understanding aim to increase knowledge of system behaviour. The objective is to develop a simple, though universal model of the system under consideration that gives an adequate description of reality as it was observed (Reichert, 1994). This kind of model never can be validated; it can only be disproved or confirmed based on the deviations between model predictions and measurements. Models applied for prediction aim to provide an accurate and fast image of a real system's behaviour under different conditions. They can be used either to forecast futures states of the system or to predict system behaviour under hypothetical scenarios (Dochain and Vanrolleghem, 2001). Depending on the method used to model the internal processes of the system, the models can be mechanistic or the so-called 'black box' models. A mechanistic model tries to describe with equations the internal processes that are actually occurring in the system. A 'black box' model is an empirical model, which has a structure that is not necessarily compatible with the underlying physical, chemical or biological reality (Tulleken, 1992), and can be calibrated only looking for the relationships between the inputs and the outputs, regardless of the way they are connected. 'Black box' models are generally useful for the application in complex or incompletely understood processes, or as a part of more general models. The so-called 'grey box' models are in between those two approaches, using some mechanisms of the system and empirical functions where the mechanisms are complex or unknown (Olsson and Newell, 1999). 'Grey box' models may also be called stochastic models because they sometimes 14 incorporate stochastic terms to describe the unknown residual variation of the data (Carstensen etal, 1995). Depending on the variation over time of the system, the models can be steady state or dynamic. Steady state models assume there is no variation of the system over time, which is never true in biological wastewater treatment systems, because of frequent changes in feed flow rate and composition and the permanent biomass recycling within the treatment plant (Carucci et al, 2001). However, it could be possible to consider 'steady state' conditions in terms of average values rather than constant values (Olsson and Newell, 1999) so they may be sufficient for design purposes (Sheffer et al, 1984; Argaman, 1995). Dynamic models are required if dealing with influent disturbances and for wastewater treatment operation and control (Sheffer et al, 1984). Typically influent characteristics, flowrates and temperature vary by factors of two to ten (Olsson and Newell, 1999). And finally, because almost no phenomenon is totally deterministic, it is not possible to write deterministic models that allow exact calculation of the future behaviour of phenomena. Nevertheless, it is possible to derive a model that can be used to calculate the probability of a future value lying between specified limits. Such a model is called a stochastic model (Box et al, 1994). In principle, it means that some random variable with a certain probability distribution is added to each process variable at each time instant (Olsson and Newell, 1999). Combining the use of more than one type of model generates a hybrid model. For example, a 'black box' model can be coupled to a mechanistic model, in order to improve the results, by working with the residuals of the first model. 3.1.2 Complexity of the models Microbial systems are very complex and the models describing them can be very complicated. The degree of complexity, although somewhat subjective, can generate an additional way of classifying the models between simple or complex models. A simple model is characterized by few equations and parameters while a complex model has many equations and parameters (Dochain and Vanrolleghem, 2001). It is expected that a more complex model will describe or 15 predict the system better than a model with less equations or components, so the modelling exercise can be seen as a trade-off between accuracy in fitting the data versus simplicity. Fortunately, relatively simple models have proven to be satisfactory for describing the performance of many biochemical operations (Grady et al, 1999), and, simulation results have proven not to change significantly if the number of states and parameters of complex models are reduced (Jeppsson and Olsson, 1993; Sheffer et al, 1984). In addition, a larger number of parameters increases the error associated with the parameter estimation (Dochain and Vanrolleghem, 2001). Therefore, it is obvious that, with equal accuracy, the simplest model should be chosen. Sometimes, however, additional complexity in a model improves the results to a certain degree. In that case, the decision about complexity will depend on the accuracy required and if the improvements are 'worth' the extra components required. On the contrary, it is often seen that some models are simplified when the objectives of the simulations do not require high accuracy (i.e. Steffens et al, 1997). Some mathematical tools for testing the model parsimony could help with this analysis and avoid the overparameterization problem, which happens when the number of identifiable parameters is lower than the original number of parameters (more details in Dochain and Vanrolleghem, 2001; Olsson and Newell, 1999). 3.2 History of modelling for municipal and pulp and paper effluents The history of activated sludge modelling is basically the history of modelling for municipal effluents. The activated sludge process was developed around 1913 by Clark and Gage in Massachusetts, USA, and by Ardern and Lockett in 1914, in Manchester, England (Metcalf & Eddy, 2003) for removing the organic load of domestic wastewaters. Since then, many models have been developed in order to explain the processes involved in the activated sludge method, and most of the knowledge is related to municipal applications. Models involving industrial effluents are a relatively new concern, and there is relatively little information about them. The pulp and paper industry is not an exception to this reality. Therefore, models developed for municipal effluents are described in order to understand the 16 development of models applied to the pulp and paper industry, as well as for comparison purposes. 3.2.1 Development of models for municipal wastewaters Wastewater treatment plants (WWTPs) are complex systems;. hence, modelling a complete plant requires many different models working together in order to describe the different processes (i.e. hydraulics, growth of different types of microorganisms, substrate removal, aeration, settling of particulate compounds, etc.). Since the major objective of the treatment is the removal of organic matter, which is performed mainly by microorganisms, the 'core' of an activated sludge model will be the so-called biokinetic model. This model describes the depletion of organic matter and/or nutrients, and their transformation into microorganisms (biomass) and/or other by-products. However, for the implementation of the biokinetic model, depending on the information available, other models may also be required (hydraulic, clarification and/or aeration). Before 1987, when Activated Sludge Model No.l (ASM1) was published (Henze et al, 1987), the various models developed at that time had only little use. ASM1 received contributions from other models (i.e. Dold and Marais, 1986), so the final result consolidated the existing knowledge at that time. ASM1 is a dynamic model that incorporates the processes related to C O D degradation and nitrification under aerobic and anoxic conditions. ASM1 has been widely used and validated under different conditions, and it has served as the 'core' for numerous models and as a basis for further model development, with a number of supplementary details (Henze et al, 2000). Activated Sludge Model N°2 (ASM2) (Gujer et al, 1995) and ASM2d (Henze et al, 1999) used ASM1 as a basis, incorporating new knowledge related to excess phosphorus removal. Additional processes related to phosphorus removal, introduced by Wentzel et al. (1989), and some processes under anoxic and anaerobic conditions complemented ASM1. These models have also been used and validated under different conditions, and have proven to be useful in designing biological nutrient removal systems. 17. After more than ten years of experience with the application of ASM1, the International Association on Water Quality (IAWQ, now International Water Association, IWA) Task Group proposed the Activated Sludge Model No.3 ( A S M 3 ) , which should correct most of the defects detected in the ASM1 and which could become a new standard for future modelling (Henze et al, 2000). ASM3 has not been used and validated as much as its predecessors, but it has demonstrated the ability to describe process dynamic behaviour satisfactorily, similar to ASM1. In addition, research has shown that ASM3 model generates better simulation results than ASM1 in situations where the storage of readily degradable substrate is dominant (i.e. wastewater treatment plants treating industrial wastewater with a high amount of COD) (Koch et al, 2000). At the present time, modelling of biological wastewater treatment processes is certainly a very active research area (Dochain and Vanrolleghem, 2001). In addition to the ASM models, many other models have also been developed for municipal wastewater activated sludge systems with variable success. Some examples are: dynamic mechanistic models (Barker and Dold, 1997; Billing and Dold, 1988; Makinia and Wells, 2000; Mao and Smith, 1995; Winkler et al, 2001), 'steady state' mechanistic models (Argaman, 1995; Ubisi et al, 1997; Wentzel and Ekama, 1997), and a dynamic 'black box' model (Chunsheng and Poch, 1998). 3.2.2 Development of models for pulp and paper effluents In comparison to the many models that have been developed for application to municipal wastewater, the simulation of pulp and paper wastewater treatment is at an early stage. The particular characteristics of pulp and paper wastewaters, such as high organic content, low nutrient content, high proportion of readily biodegradable substrate, and high temperature and toxicity, may require the use of an activated sludge model that is different from the models used for municipal wastewater, or would require at least a specific calibration of an existing model. The National Council for Air and Stream Improvement ( N C A S I ) of the United States developed a dynamic model for modelling the treatment of pulp and paper effluents (NCASI, 1986) in activated sludge systems as well as in aerated stabilization basins. The objective of the model, however, was not modelling COD removal, but as a screening tool for the identification 18 of the extent of intermedia transport of specific organic compounds in WWTPs. This model considers four different removal pathways: forced stripping, natural volatilization, biosorption and biodegradation. The model was tested for predicting the removal of chloroform, methanol, resin acids and some chlorophenolics, proving relatively good prediction for particular compounds (NCASI, 1990). However, when tested for predicting the organic removal and the solids production of a pulp and paper WWTP, the results were not satisfactory and were comparable to the predictions of a simpler steady state model (Eis et al, 1990). Attempts at calibrating models originally developed for municipal effluents have also been made. Research conducted by the Forest Research Institute of New Zealand (Slade et al, 1991) tried to apply ASM1 to the treatment of bleached kraft mill effluents (BKME) in an aerated lagoon. Although the methodology adopted for the calibration and wastewater characterization was developed for municipal wastewaters, it was found to be applicable to BKME (Slade et al., 1991). However, when applied to the full scale simulation, the model response for the effluent COD and suspended solids (SS) concentration was poor, which was assumed to be due to the inadequate determination of input parameters or due to a failure of the model to include a parameter or process that is vital to the adequate modelling of the biological treatment of BKME (Slade et al. (1994), as cited by Sreckovic, 2001) Previous works carried out at the University of British Columbia (UBC) have also made some advances in the task of calibrating Activated Sludge Model N°l (ASM1) to BKME (Stanyer, 1997), and building a new model based on Activated Sludge Model N°2 (ASM2) for chemi-thermomechanical pulping (CTMP) effluents (Sreckovic, 2001). However, the parameter values obtained by both researchers were significantly different from each other, and were also different from the municipal wastewater typical values, especially those found by Sreckovic (2001). In addition, Helle and Duff (2003), working with activated sludge treating BKME, identified up to five soluble readily biodegradable fractions with different biodegradation rates. They concluded that the interpretation of the respirometric data assuming only one wastewater fraction resulted in over prediction of substrate removal rates. An experience with the application of a hybrid biokinetic model to pulp and paper activated sludge systems showed that the hybrid model did not generate any mechanistic model improvement (Sreckovic, 2001). The method used was the selection of the input nodes for a 19 neural network based on the results of the cross-correlation analysis of the input data and the mechanistic model residuals. 'Black box' models have been also developed for pulp and paper wastewaters. For example, Mujunen et al. (1998) developed a model based on partial least squares for effluent quality and sludge bulking. It was found that the major trends of purification efficiency could be predicted reliably. However, the results of modelling effluent parameters, total nutrients, and COD indicated a lack of relevant information. CH2M H I L L developed a steady state model that has been successfully used to size and evaluate activated sludge systems for a variety of industrial wastewaters. However, that model did a poor job of predicting the performance of a pulp and paper WWTP when tested with a different set of data, other than the data used for calibrating it (Eis et al, 1990). 3.3 Model Verification Once a model has been developed, the verification stage consists of confirming that the model is applicable in practice. This verification process may be divided into two steps: verifying that the behaviour of the model is reasonable for the complete range of conditions the model was designed for; checking that the model parameters values can, in fact, be obtained. The first step in model verification usually involves setting up the model with a reasonable set of parameters and then systematically examining its behaviour to changes in input conditions, system states, and parameters; and its behaviour in limiting and extreme situations. The goal is to delineate the operating region in which the model behaves sensibly, and to determine if this region includes the region for which the model was developed for (Olsson and Newell, 1999). The task of checking the calibration feasibility and determining which parameters to adjust is performed with two tools: a sensitivity analysis and an identifiability analysis. As an alternative, the selection of the parameters to adjust may be done on the basis of process and model knowledge rather then on the sensitivity analysis, (i.e. as in Van Veldhuizen et al, 1999) 20 3.3.1 Model Sensitivity The sensitivity analysis of a mathematical model will give valuable information about the dependence of different state variables on the model parameters. Equation 3.1 shows the general equation used for dynamic sensitivity calculations, by examining the change in the function f( ) for changes in the model parameters p. dx/dt = f(x,p) [3.1] Where x is the vector of state variables. Overparameterized models have some parameters that are very insensitive to the model components, making their estimation unfeasible. Assuming reasonable values for insensitive parameters can alleviate this problem (Olsson and Newell, 1999). 3.3.2 Model Identifiability The identifiability problem of a mathematical model is related to the possibility of giving a unique value to each of its parameters. The problem can be divided into structural identifiability and practical identifiability. Structural identifiability is related to the possibility of determining uniquely the value of the parameters, based on the model structure, under 'ideal' data conditions. Practical identifiability, on the other hand, is related to the quality of the data and their 'information' content. Are the available data informative enough for identifying the model parameters and for giving accurate values (Dochain and Vanrolleghem, 2001)? The central question of the identifiability analysis can be formulated as follows: \"assuming that a certain number of state variables are available for measurement (structural identifiability) or on the type and quality of available data (practical identifiability), can we expect to give via parameter estimation a unique value to the model parameters?\" (Dochain and Vanrolleghem, 2001). 21 In particular, two important features characterize the dynamic models describing activated sludge processes (Vanrolleghem et al, 1995): models are often highly complex, and they are usually non-linear systems incorporating a large number of state variables and parameters; and there is, generally speaking, a lack of cheap and reliable sensors for on-line measurements of the key state variables, in particular those involved in the model. 3.3.2.1 Structural identifiabilitv Godfrey and Distefano (1985, as cited in Carstensen et al, 1995), showed that the four parameters in a typical microbial model (U.H, Ks, YOBS and bn), consisting of two equations with Monod-kinetics, are theoretically identifiable. However, the additional complexity incorporated in models such as ASM1 made them overparameterized, so the order of the deterministic model has to be reduced in order to be able to estimate the parameters (Jeppsson and Olsson, 1993; Larrea etal, 1992). Dochain et al. (1995) demonstrated analytically that, based only on respirometry, a simplified version of ASM1 (not including nitrification), as well as three other models tested, were not structurally identifiable and only a smaller set of the combinations of the original parameters were structurally identifiable. Therefore, separate experiments or a priori information about some parameters are required to provide the necessary information to estimate all of the individual parameters (i.e. yield heterotrophic coefficient). The mathematical determination of the identifiable parameters or combinations of them is very complicated without the aid of symbolic algebra software packages and a large computational capacity. In addition, there are at least six different structural identifiability methods available in the literature, and it is very difficult to a priori select the best method to test structural identifiability of a dynamic model (Dochain and Vanrolleghem, 2001). Some of the identifiability methods used for nonlinear systems, which correspond to the typical activated sludge models, are: Taylor series expansion, generating series, local state isomorphism, and transformation into linear model in the parameters. 22 An alternative approach was suggested by Weijers and Vanrolleghem (1997), who performed a sensitivity analysis in which only a subset of parameters composed of the most sensitive was tested for identification. The Fisher Information Matrix was computed for all possible subsets of parameters selected from the sensitivity analysis. This method was demonstrated to be successful in selecting an identifiable subset of parameters and in avoiding computational problems. The selection of a reduced set of parameters was crucial since the number of calculations required for the analysis increases dramatically with the number of parameters to be evaluated. 3.3.2.2 Practical identifiability Even though some parameters can be structurally identifiable, it does not mean that they are practically identifiable. Practical identifiability plays an important role during the calibration process of model parameters. When a limited set of data is used for parameter estimation, the problem of highly correlated parameters can disturb the uniqueness of parameter estimation (Vanrolleghem et al., 1995). In that case, a change in one parameter can be compensated almost completely by a proportional shift in another parameter, still generating a satisfying fit between the experimental data and the model predictions. In addition, the numerical algorithms for non-linear parameter estimation show poor convergence when faced to correlated parameters or noisy data, and the estimates become very sensitive to the initial parameter values given to the, algorithm (Dochain and Vanrolleghem, 2001). As a result, the parameters estimated in this way may have little physical meaning. Two useful tools for evaluating the precision of parameter estimates (practical identifiability) are the eigenvalue decomposition of the covariance matrix (as used in Brouwer et al., 1998; Vanrolleghem et al., 1995) and the Fisher Information Matrix (as in Weijers et al., 1996). These two methods are indeed related, so only the Fisher Information Matrix is explained in more detail below. However, since a detailed description of the method is complicated and highly theoretical, the reader who is interested in more detail may refer to Dochain and Vanrolleghem (2001). 23 The objective in finding optimal parameters is to minimize Equation 3.2. This goal is achieved by maximizing the Fisher Information Matrix ('F' in Equation 3.3), which expresses the information content of the experimental data. J(9) = Zi=,N(yi(e') - Y i ) T Qi (yi(6') - Y i) [3.2] F = Z i = 1 N (dy/de (ti))T Qi(5y/50 (ti)) [3.3] Where yi and yj(6') are vectors of N measured values and model predictions at times tj (i = 1 to N) respectively, Qi is a square matrix with used-supplied weighting coefficients, and dy/dQ are the output sensitivity function. Qi is typically chosen as the inverse of the measurement error covariance matrix. The evaluation of the sensitivity functions is a central task in the practical identifiability study, because it quantifies the dependence of the model predictions on the parameter values. Under this method, approximate standard errors for the parameters can be calculated by using Equation 3.4, where V is the covariance matrix, which is equal to the inverse of the Fisher Information Matrix (Equation 3.5). 3.3.3 Overcoming sensitivity and identifiably problems Identifiability problems are not new, but they have increased in importance since models have become more complex and since respirometry became the key operational tool for model calibration. Identifiability is an issue when many parameters are estimated from a single test (typically a respirograph). Different approaches are commonly used for overcoming the problems that respirometry cannot resolve completely. The easiest way of calibrating difficult parameters is by not estimating them, but rather by assuming their value from the literature (i.e. in Weijers and Vanrolleghem, 1997). The problem o(0i) = VVij [3.4} V = F - i [3.5] 24 here is choosing the parameter value that the modeller believes is the most suitable for that specific model application. An alternative is to try to optimize the experimental design in order to obtain the maximum useful information from a single test. Choosing an appropriate food to microorganisms (F/M) ratio, an appropriate reactor configuration, adding a pulse of substrate during the experiment, and a proper measurement strategy may improve the accuracy and reduce some identifiability problems (i.e. Dochain and Vanrolleghem, 2001; Larrea et al, 1992; Lukasse et al, 1997) Performing additional analytical tests (i.e. COD, VSS, etc.) is always an option for determining some parameters or wastewater characteristics that are difficult to estimate by respirometric tests (i.e. Dochain et al, 1995). Even different respirometric tests, based on alternative approaches, can give extra information (i.e. long term tests for determining bn). Some authors also performed batch tests using synthetic wastewater in order to estimate some kinetic parameters (i.e. Brouwer et al, 1998). The main advantage of using synthetic wastewater is that the composition is known and, by eliminating the wastewater characterization from the parameter estimation, most of (if not all) the identifiability problems are eliminated. Finally, if a physical meaning of the parameters is not searched, the problem may be avoided arguing that the identifiability problem is a theoretical problem with no practical implication (i.e. Sreckovic, 2001). This approach may transform a mechanistic model into a 'black box' model. 3.4 Modelling wastewater characteristics Models for describing biological wastewater treatment require a model of the influent wastewater according to particular objectives. While some models can focus on specific chemical compounds, other models may require a measure of the total organic content of the wastewater, or others may need to incorporate further organic fractions, nutrient concentrations, or even organic functional groups such as carbohydrate, protein, and lipid content (as suggested by Dircks efa/., 2001a). 25 Modelling of wastewater characteristics is receiving increasing attention with the application of models, and many researchers have stressed the extreme importance of an adequate knowledge of wastewater characteristics (Barker and Dold, 1997; Dold and Marais, 1986). Once the desired model components or states have been determined, the next step is to measure them in the influent wastewater (inputs). The appropriate estimation of the wastewater characteristics has been shown to be one of the primary tools for calibrating a model, and that nearly all other parameters can be assumed as their default values (Dupont and Sinkjaer, 1994). The modelling of the wastewater is always linked to a particular biokinetic model, and sometimes wastewater components determined to be identical by two different models may represent different physical properties of the wastewater. In addition, the analytical methods used and the assumptions made for calibrating a particular component may vary its connotation. Thus, the task of reviewing the different wastewater models is not easy. However, there are some wastewater characteristics that have become standard among different models, so they are used as a basis in the following review. Still, the wastewater fractions described have been narrowed according to the scope of the present research. The main wastewater characteristics considered by most of the models are: organic content, oxygen concentration, nutrients, and solids. Other characteristics that may be relevant are temperature, pH, and alkalinity. In pulp and paper systems, the volatile compounds, the toxicity, and some particular compounds (i.e. AOX, resin acids, fatty acids, etc.) could be also of interest. Although some concentration of microorganisms can be present in the wastewater, the common assumption to neglect influent biomass is sustained, not only in the actual absence of bacteria in the influent, but mostly by the theory that the bacterial diversity in activated sludge is a product of selection by the environment rather than inoculation by the wastewater (Curtis et al, 1998). 26 For estimating the concentration of organic matter in wastewater, three measures are commonly used: biochemical oxygen demand (BOD), total organic carbon (TOC) and chemical oxygen demand (COD). These three measures are used for modelling purposes, however, COD has been most widely used as the standard measure of organic concentration since it is \"undoubtedly the superior measure because it alone provides a link between electron equivalents in the organic substrate, the biomass and the oxygen utilized\" (Gaudy and Gaudy, 1971). In addition, the organic matter (COD will be used from now on) may be subdivided into a number of categories. The first important subdivision of wastewater COD is based on biodegradability, and two broad groups can be identified: the biodegradable COD and the inert (or non-biodegradable) COD. Based on the biodegradation rate, biodegradable COD may be sub-divided into readily biodegradable COD (Ss) and slowly biodegradable COD (Xs). Conversely, inert COD may be sub-divided according to its physical properties into soluble inert COD (Si) and particulate inert COD (Xi). Nutrients can be divided into two classes: macronutrients, which are required in large amounts, and micronutrients, which are required in only small amounts. The major macronutrients are nitrogen (N) and phosphorus (P), which are those of main interest for activated sludge systems. Other macronutrients are sulfur (S), potassium (K), magnesium (Mg), sodium (Na), calcium (Ca) and iron (Fe). The major micronutrients for microorganisms are chromium (Cr), cobalt (Co), copper (Cu), manganese (Mn), molybdenum (Mo), nickel (Ni), selenium (Se), tungsten (W), vanadium (V) and zinc (Zn) (Madigan et al, 1997). For municipal WWTPs, sufficient nutrients are generally present to satisfy the microorganisms requirements (Metcalf & Eddy, 2003). Paper mill effluents are often deficient in nitrogen and phosphorus (Gostick, 1990), but they usually contain sufficient levels of the other macro and micronutrients (Hynninen, 1998), so problems in treating pulp and paper effluents attributable to insufficient micronutrients appear to be rare (NCASI, 2000). Consequently, only the nitrogen and phosphorus are discussed in more depth. Oxygen is a key element in activated sludge systems, especially under aerobic conditions, because it acts as an electron acceptor during the respiration of the microorganisms responsible for COD degradation. The growth rate of the microorganisms is influenced by the dissolved 27 oxygen (DO) concentration in the mixed liquor, so it will be an essential component of any model of an aerobic system. Since the model of oxygen transfer from the gaseous to the liquid phase is relatively complicated, and the DO is commonly available from on-line measurements, this model is not discussed in this thesis. Further, it has been demonstrated that DO levels have no effect on COD removal in activated sludge systems treating pulp and paper effluents at levels above 1 g m\"3 (Mobius, 1989), which are typically achieved by the WWTP under study. Solids are a fundamental element of most activated sludge models because their measurement is assumed to correspond to the sum of the particulate wastewater fractions (X), in contrast with those assumed to be soluble (S). These particulate wastewater fractions typically include the heterotrophic biomass (XH), the slowly biodegradable COD (Xs, assumed to be particulate), and the inert particulate COD (XT). Other wastewater fractions that are assumed to be particulate are the autotrophic biomass (X A , microorganisms responsible of the nitrification) and the substrate that has been stored by the microorganisms (i.e. XSTO)-Temperature, p H and alkalinity are three environmental characteristics that can play an important role in activated sludge systems, by inhibiting or enhancing some biological processes. They are not only important as influent wastewater characteristics, but also along the whole WWTP. Alkalinity is generally incorporated into the models in order to obtain an early indication of possible low pH conditions (Gujer et al, 1999), so this model component can be omitted if pH is consistently in the optimum range (between 6.5 and 7.5). In fact, most of the applications of ASM models do not incorporate alkalinity. Even very complete calibration studies using ASM3 have not included it (i.e. Koch et al, 2000). Volatile compounds (generally organic) are only important when air stripping is a relevant removal mechanism (as it is in the NCASI model), which is not usual in the most commonly used models. Sometimes chloroform is present in BKME, in which case the removal of this component by air stripping could be modeled independently of the general biokinetic model. Toxicity is another wastewater characteristic especially relevant for industrial effluents, and toxicity tests are used for a variety of reasons. The test methods can be classified into those for estimating the toxicity effects in the environment, and those for estimating the effects on the wastewater treatment system. Since studying the toxicity effectvin the environment is beyond 28 the scope of this research, only the toxic effect of the primary effluent (secondary influent) on the activated sludge system is discussed. 3.4.1 Biodegradable organic matter The biodegradable organic matter can be considered as the COD fraction that can be degraded (or oxidized) by microorganisms. Part of that organic matter is transformed into new cells, and the rest is used for energy requirements. The classification of organic matter into biodegradable and non-biodegradable could seem simple, but in reality the division between biodegradable and non-biodegradable (inert) is not very clear. There could be some COD fraction that, being . potentially biodegradable, would require so much time to be oxidized by the microorganisms (compared with the time scale of a particular WWTP), that it could be considered as inert in a model for that plant. The biodegradable COD fraction can be subdivided regarding the degradation rate of the different biodegradable organic compounds present. However, since there are hundreds of different compounds with different degradation rates, classifying them into two, three or more biodegradable fractions is quite arbitrary. However, for modelling purposes, the biodegradable COD is usually subdivided into only two fractions: readily biodegradable COD (Ss) and slowly biodegradable COD (Xs). In fact, practical experience has showed that the full-scale simulation models are not that sensitive even with the division between slowly and readily biodegradable COD (Hulsbeek et al, 2002). In order to simplify the analytical identification of the different biodegradable COD fractions (i.e. Ss and Xs), it is useful to assume that the biodegradation rates are related to the physical properties. Consequently, Ss is assumed to be soluble while Xs is assumed to be particulate and it needs to be hydrolysed before being utilized. This assumption has been found to fit relatively well with experience, but there is a number of authors that have proposed alternates to this classification. The main advantage of this classification is its simplicity, which allows modelling of the biodegradable fraction based on the same simple test used to determine the soluble and particulate matter. However, in complex wastewaters with high fractions of soluble slowly biodegradable substrate, or more than one type of particulate slowly biodegradable substrate, this assumption has to be revised. 29 3.4.1.1 Readily biodegradable COD Readily biodegradable COD (Ss) is the COD fraction that is directly available for biodegradation by the microorganisms. This fraction is assumed to be soluble, and it has a high biodegradation rate. Ss is also assumed to come in the influent wastewater, as well as being produced during the hydrolysis process. Some authors have found that the readily biodegradable COD (Ss) could be subdivided into two fractions (i.e. Ss,i and Ss,2) based on their different degradation rates (Brouwer et al, 1998; Spanjers and Vanrolleghem, 1995). A more detailed model of the Ss fraction is also required for modelling biological phosphorus (Bio-P) removal. Models like ASM2 (Gujer et al, 1995), ASM2d (Henze et al, 1999), or the Barker and Dold (1997) model divide Ss into complex (or fermentable) substrate, and in the form of short chain fatty acids (SCFA, or fermentation products). This classification is important to model Bio-P removal, where the poly-P microorganisms use only the SCFA fraction of Ss for phosphorus sequestration, but it does not have a major effect on heterotrophic growth. However, newer Bio-P models neglect the fermentation of Ss and use only one Ss fraction (Rieger et al, 2001; Siegrist et al, 2002). 3.4.1.2 Slowly biodegradable COD The slowly biodegradable COD (Xs) is composed of high molecular weight, colloidal and particulate organic substrates, which must undergo exocellular hydrolysis before they are available for degradation. Regarding Xs, the major differences among the models are not related to the biodegradation rates but to the physical features of Xs (particulate, soluble, or colloidal). Different researchers have reached different conclusions working with different wastewaters. The importance of classifying Xs as particulate, colloidal or soluble is not only related to the analytical measurements used to characterize the wastewater or the sludge. Physical properties 30 (soluble and particulate) are very useful to model the clarification process, and the assumption that the particulate matter is separated from the soluble matter in two different flow streams (underflow and overflow respectively) works very well in the practice. Then, the amount of X s degraded in the WWTP will be related with the X s retention time, which will be directly related to its physical composition. Most models assume Xs is made up of particulate and colloidal COD. However, when those same models postulate that filtration is an appropriate way of determining Xs, they are assuming that it has neither a soluble nor a colloidal component, which would pass through a 0.45 pm pore size filter (i.e. Barker and Dold, 1997; Gujer et al, 1999). Therefore, there exists a gap between the theoretical models and their practical application. In relation to the colloidal fraction of X s , sometimes called rapidly hydrolysable COD, X R , (as in Henze, 1992; Sollfrank and Gujer, 1991), Torrijos et al. (1994), working with domestic wastewater, found that the COD colloidal fraction made up the slowly biodegradable fraction of the wastewater with very low content of readily biodegradable compounds. Ginestet et al. (2002) found that, for seven municipal WWTPs, the major part of the colloidal fraction was readily hydrolysable COD, followed by inerts or very slowly biodegradable COD, and biomass. It has been reported many times that there exists analytically soluble COD that is actually slowly biodegradable (i.e. Carucci et al, 2001; Ginestet et al, 2002), which contrasts with the assumption that all slowly biodegradable COD is particulate. That fact was incorporated into the model used by the simulation software EFOR, based on ASM1, which includes a soluble slowly biodegradable COD fraction (Dupont and Sinkjaer, 1994; Pedersen, 1992; Pedersen and Sinkjaer, 1992). On the extreme, and working with BKME, Slade (2003) found that all the slowly biodegradable COD was soluble for the effluents tested. On the other hand, Torrijos et al. (1994) found that the soluble COD fraction did not contain compounds with a low degradation rate for the municipal wastewater tested. Whether or not Xs is produced in the WWTP, or whether all of it is coming with the influent, is not totally clear. ASM1, which adopted the death regeneration theory from Dold et al. (1980), suggested that a fraction of the cells, after decay, becomes available for new growth in the form of Xs. Nevertheless, later evidence related to the accumulation of stored products inside the cell 31 (adopted by ASM3) suggested that the only source of Xs is the influent wastewater, and there is no such death, rather only an endogenous decay. 3.4.2 Inert organic matter The inert COD is the fraction of the COD that cannot be further degraded in the treatment plant. Therefore, the quality of inert is related to the system of interest, and is not an intrinsic characteristic of the wastewater. The significance of the inert fractions in biological treatment is such that its correct determination can be considered more important than the kinetics of the biodegradable part of the influent COD (Artan and Orhon, 1989) The inert COD fraction can be sub-divided according to its physical properties into soluble inert COD (Si) and particulate inert COD (Xi). While a fraction of the inert COD originates with the influent (soluble and/or particulate), some inerts may be generated inside the activated sludge system, during the hydrolysis and the endogenous respiration processes. The inerts generated in the system are usually called residual microbial products, and may be soluble or particulate. 3.4.2.1 Inert soluble COD The inert soluble COD (Si) is the soluble COD fraction that cannot be further degraded in the activated sludge system modeled. Si may have three possible origins: the influent wastewater, residual microbial products generated during the hydrolysis of Xs, and residual microbial products generated during the endogenous respiration process. However, there is still no agreement about these different sources of Si. For example, some evidence has demonstrated that there is no production of Si during biomass decay (Sollfrank et al, 1992). The different relative importance of each source and the simplification of model structures have excluded the assumption of soluble microbial products (SMP) generation during endogenous respiration in most models. Further, even though the production of residual microbial products during the hydrolysis of Xs may be included in some models, the typical proposed coefficient of zero renders this source irrelevant in practice. 32 Due to the difficulty in distinguishing among the different sources of Si, most models group all soluble inert COD under only one state variable (called here Si, disregarding the original denomination in each discussed model) (Barker and Dold, 1997; Dold et al, 1980; Gujer et al, 1995; Gujer et al, 1999; Henze et al, 1987; Henze et al, 1999; Makinia and Wells, 2000). Some authors, however, make the distinction and identify soluble residual microbial products as SP (Babuna et al, 1998; Orhon et al, 1999b). Working with effluents from a papermill using mechanical pulp, Franta el al. (1994a) found that only about 10% of the residual organics in the effluents had a microbial origin, and 90% of the organics were compounds originally present in the influent. 3.4.2.2 Inert particulate COD Inert particulate COD (Xi) is the particulate COD which is not degraded in the activated sludge system modeled. These solids are non-biodegradable (inert) in a process sense, because they are not metabolized under the conditions present in activated sludge systems (Van Loosdrecht and Henze, 1999). Xi may have two origins: the influent COD and the endogenous respiration process. Some models differentiate between these two different sources by having two different variables for these particulate inert fractions. For example, in ASM1 (Henze et al, 1987) the solids resulting from decay are called 'particulate products arising from biomass decay' (Xp), and in Barker and Dold (1997) they are named 'endogenous mass' (ZE). For simplicity, ASM2 and ASM3 (Gujer et al, 1995; Gujer et al, 1999) group both inert particulate fractions in Xi , because it is impossible to differentiate between them in reality. In terms of the volatile suspended solids (VSS), most pulp fibres in pulp and paper mill wastewater are essentially non-degradable, and hence almost all influent VSS are non-degradable (Eckenfelder, 2000). 33 3.4.3 Volatile organic compounds Organic compounds that have a boiling point at less than or equal to 100°C and/or a vapour pressure greater than 1 mm Hg at 25°C are generally considered to be volatile organic compounds (VOCs) (Metcalf & Eddy, 2003). These compounds may have a public health significance, but from the modelling perspective, their evaporation during the activated sludge process will produce a COD 'loss' in the system. Then, the presence of VOC may affect the COD mass balance if it is not considered properly. In activated sludge systems incorporating anaerobic zones, this loss of COD may be around 20% of the influent COD, but less in anoxic-only systems. This loss of COD is apparently associated with the fermentation process, whether due to the generation of gas or through the production of volatile compounds that are released from the system (Barker and Dold, 1997). In aerobic systems there should not be generation of volatile compounds, so the only source of VOC should be the influent wastewater. VOCs can be an important wastewater constituent in some pulp and paper effluents, such as bleached kraft pulp mill effluents (BKME), due to the common presence of chloroform in these effluents. In fact, the NCASI model for pulp and paper wastewater treatment incorporates air stripping as one of the removal mechanisms of organic compounds (NCASI, 1986), and it was validated for the volatile compounds chloroform and methanol (NCASI, 1990). The results of these validation tests concluded that air stripping accounted for around 80% of the chloroform removal, but for less than 1% of the methanol removal. 3.4.4 Temperature Temperature is an important characteristic of the wastewater that affects many processes occurring in an activated sludge system, such as: biological reaction rates (growth, hydrolysis and decay), aeration efficiency, the nitrification process, sedimentation rates of particulates, hydraulics (affecting the physical properties of water), etc. Therefore, temperature should be an essential component of any model describing the biological treatment of pulp and paper effluents, and a proper calibration for this parameter is also critical. 34 When used to treat municipal wastewaters, the activated sludge process is operated at temperatures as low as 2 to 5°C, and it is typically operated in the temperature range of 8 -20°C. In contrast, effluents from the wood-processing industry are usually warm, and the activated sludge process still performs well at up to 35 - 37°C. Above 40°C, however, the effluent first has to be cooled. It has been reported that, for pulp and paper mill wastewater biological treatment systems, decomposition is often 3 to 5 times faster at 35°C than at 20°C (Hynninen, 1998). The need for cooling effluents with temperatures higher than 40°C responds to the fact that temperature speeds the biodegradative reaction up to a certain point, after which, its effects are detrimental for the wastewater treatment system. Figure 3.1 shows the effect of temperature on biological oxidation rate, which increases according to the Arrhenius equation until approximately 31°C. Between 31°C and 39°C the reaction rate is relatively constant, and beyond 39°C, the reaction rate decreases due to floe dispersion. 1 Floe dispersion i • i 4 32 39 Mixed liquor temperature (°C) Figure 3.1: Effect of temperature on biological oxidation rate (Eckenfelder, 2000) Even though the evidence of the lower efficiency achieved by activated sludge systems operated at temperature above 40°C is extensive, there are also some experiences that demonstrate the contrary. For example, Barr et al. (1996), working with KPME, observed good COD removal without decreased sludge settleability in the range between 41 - 50°C. 35 Temperature is often incorporated in models by modifying the kinetic rate constants for the biological processes, whereas the stoichiometric parameters, as well as the wastewater characteristics, are assumed to be independent of temperature. However, it has been shown that temperature may also affect stoichiometry as well as apparent wastewater composition (Sollfrank a/., 1992). The effects of temperature on stoichiometry may include changes in the activated sludge composition (the lower the temperature, the higher the COD/VSS ratio) (Sollfrank et al, 1992), and changes in the net yield of biological cells (the lower the temperature of an activated sludge system treating BKME, the higher the observed yield of biological cells) (Gill and Ross, 1990). The effects of temperature on wastewater composition may include changes in the soluble effluent COD (the lower the temperature, the higher the fraction of undegraded COD) (Sollfrank et al, 1992), More complicated effects, usually not included in activated sludge models, may include changes in the population structure of microorganisms. Brdjanovic et al. (1997), measuring the effect of temperature on a biological phosphorus removal system, determined that the composition of the bacterial population shifted with temperature. Gill and Ross (1990), working with BKME, observed that effluent TSS increased when aeration basin temperature exceeded 36°C to 38°C, due to a decrease the number of free-swimming higher life forms. There is also evidence of a negative effect of temperature on biological reactions, which is not common for normal heterotrophic processes. Krishna and Van Loosdrecht (1999) observed a decrease in the storage rate with an increase in temperature. The most common way to model the effect of temperature on reaction rates is the simplified version of the Arrhenius Equation shown in Equation 3.6. Even though this equation may be valid in a temperature range of 4 to 31°C (Eckenfelder, 2000), each model may restrict the temperature range further (i.e. ASM2 recommends a range 10 - 25°C and ASM3 is limited to a range 8 - 23°C). 36 k ( T ) = k ( T R E F ) - 0T\"' [3.6] Where k is any reaction rate coefficient (i.e. maximum growth rate, endogenous respiration coefficient, storage rate, hydrolysis etc.), T is the temperature in °C, T R E F is a reference temperature (typically 20°C for domestic wastewaters), and 6 is the temperature coefficient for that reaction rate coefficient. The values of 0, using Equation 3.6, have been found to vary from 1.03 to 1.12 for different processes and a variety of wastewaters. ASM3 incorporated a variation of the latter temperature equation in an exponential form, substituting 0 for e 9 r (see Equation 3.7). This new 0j coefficient will take values in the range of 0.03 to 0.11. k(T) = k(T R E F )- e 9 T ' ( T - T r e ° [3.7] For pulp and paper systems, the recommended value of 0 is 1.072 (Springer, 2000), equivalent to a 0T of 0.07. This value is the same as the proposed 0T for municipal wastewater systems. Other models of the effect of temperature on activated sludge systems have also been developed for specific purposes. For example, Berube and Hall (2000) developed a model for modeling the effect of temperature on methanol removal kinetics for synthetic kraft pulp mill condensate using a membrane bioreactor. The model corrected the simplified Arrhenius equation in the temperature range between 55 and 70°C. 3.4.5 pH The pH of the activated sludge system is also a key factor in the growth of microorganisms. Most bacteria cannot tolerate pH levels above 9.5 or below 4.0. Generally, the optimum pH for bacterial growth lies between 6.5*and 7.5 (Metcalf & Eddy, 2003), which is also the range for which most models have been developed (i.e ASM3 6.5 - 7.5; ASM2 6.3 - 7.8). Few quantitative equations have been developed for modelling the influence of pH over the kinetics of heterotrophic growth, so the usually adopted approach is to avoid pH far from 37 neutrality. Alkalinity is incorporated in the models with this objective, since its buffering capacity can be used as a warning for pH values outside the expected range. In the case of pulp and paper effluents, the best pH for biological treatment seems to be in the range 7 - 7.5, while the range 6.8 - 8 is workable. A low pH favours the growth of yeasts present in the biomass (Hynninen, 1998). Working with BKME, Helle (1999) reported that pH did not affect the growth yield coefficient and decay coefficient over the normal range of pH employed by a WWTP, and only with pH above 9.5 or less than 5.5 was microbial decay accelerated. 3.4.6 Toxicity Toxicity is the potential or capacity of a test material to cause adverse effect on living organisms (American Public Health Association et al, 1998). The effect of a toxic compound over an activated sludge system may be lethal or sublethal. Lethal effects are observed, for example, after the occurrence of peak concentrations of certain compounds that kill all bacteria of a certain type. However, the most typical response is a sublethal effect in the form of inhibition of certain processes (i.e. microorganisms growth). It is possible to consider that the mechanisms are probably the same for inhibition and toxicity, with reactions rates becoming progressively slower as more damage occurs (Grady et al, 1999). In wastewater treatment and the activated sludge process, the microorganisms involved include prokaryotic and eukaryotic types, and not all microorganisms respond to all toxic substances released to sewer in the same way (Larson (1991), as cited in Dalzell et al, 2002). For example, nitrifying organisms are sensitive to a wide range of organic and inorganic compounds and at concentrations well below those concentrations that would affect aerobic heterotrophic organisms (Metcalf & Eddy, 2003; Nowak et al, 1995). The inhibition of the processes due to toxicity depends not only on the concentration of toxic substances, but also on factors such as temperature, nutrient level, types of microorganisms present, and the ability of the microorganisms to adapt. 38 Toxic compounds are normally found in some industrial effluents, but not frequently in municipal wastewaters. In particular, pulp and paper effluents may contain some toxic compounds, whether they are added in the pulping process or are originally present in the wood. Components added during the process, which may be toxic to the treatment system, are mainly those related with the bleaching process; and toxic wood components are typically fatty acids and resin acids. However, process discharges from the wood-processing industry do not normally give rise to toxicity problems (Hynninen, 1998). Even the nitrification process has been found to be insensitive to paper mill effluent (Dalzell et al, 2002). The acclimation capacity of the microorganisms is very important in dealing with toxicity. It has been proven that acclimatized bacteria can growth normally in the presence of toxic compounds at concentrations 50 to 80 times the toxic threshold that would affect non-acclimatized bacteria (Pollard and Greenfield, 1997). This may explain observations of identical concentrations of the same toxic compound having different effect on the bacteria present in activated sludge samples obtained from different WWTPs (as in Gernaey et al., 1999). Acclimation of the biomass to a specific wastewater composition is assumed to be complete when the specific oxygen uptake rate (SOUR) of the residual organic concentration reaches a steady state condition (Eckenfelder and Grau, 1992). Acclimatization also allows bacteria to better resist the effect of a toxic spill. Working with BKME, Larisch and Duff (1996) concluded that activated sludge that has been acclimatized to a bleaching agent was more resistant to the shock loading of a sudden bleaching agent spill. However, even with an acclimatized bacterial population, WWTPs have to be protected from peak loads (Nowak et al, 1995). Commonly, the evaluation of wastewater toxicity is done by comparison of the response of the activated sludge system before and after the addition of the potentially toxic wastewater sample (Dalzell et al, 2002; Vanrolleghem et al, 1996). The effect measured in this manner is called inhibition, because generally it causes a decrease of the degradation process more than the death of the cells. This methodology underlies the assumption that under the normal working state, the wastewater has no toxic effect on the activated sludge microorganisms. For example, Volskay and Grady (1990) developed a technique called Respiration Inhibition Kinetics Analysis (RIKA) to quantify the toxic, or inhibitory, effect of xenobiotic compounds on 39 biogenic-carbon removal in biological wastewater treatment systems. This process has been further developed in following research (Kong et al, 1994; Kong et al, 1996; Nowak and Svardal, 1993), making it more automatic and suitable for including nitrification inhibition. For modelling inhibition, concentrations of toxic substrates and parameters of inhibitors have to be inserted into the kinetic equations of the model. Depending on the manner in which the toxic substance acts, there are different ways to model its effect on the microbial kinetics. For example, it is possible to change the coefficients for the maximum growth rate and half saturation coefficient (pn may decrease; and Ks may increase or decrease, depending upon whether the inhibitor is competitive, non-competitive, uncompetitive or mixed) (Volskay and Grady, 1988). In the presence of more than one toxic substance, inhibitor concentrations and constants cannot be determined for each different compound. Consequently, virtual (or multicomponent) inhibitor concentrations and coefficients used should be estimated from the variations in the measurable kinetic parameter half-saturation coefficients and maximum growth rate respectively (Nowak et al, 1995). 3.4.7 Nutrients Nitrogen and phosphorus are the most important nutrients in activated sludge systems. The control of these compounds is becoming more important not only because these are key elements for the efficiency of the biological treatment, but also because new and more stringent regulations are restricting the discharge of these nutrients to the environment. Domestic wastewater typically contains nitrogen and phosphorus far in excess of the amount needed to stabilize the limited quantity of organic matter present (Sawyer et al, 1994). Then, the goal in municipal wastewater nutrient management is, typically, to remove as much nutrient as possible in order to minimize the effects of its discharge in the environment. On the other hand, it is well recognized that effluents from the pulp and paper industry tend to be deficient in nitrogen and/or phosphorus. As a consequence, these two nutrients are commonly added to mill wastewaters to improve biological treatment (NCASI, 2000). The 40 incorrect dosing of nutrients is a common cause of problems in the activated sludge process treating pulp and paper effluents (Hynninen, 1998), which, in turn, may cause a decrease of the microorganisms growth rate in the range of 75 - 80% for different types of pulp and paper mills (Eckenfelder, 2000). The nutrient load to a WWTP should satisfy the requirements for growth of new biomass formed from the degradation of organic matter. The biomass contains approximately 12.3% nitrogen and 2.6% phosphorus. As sludge age increases and the biomass decomposes, its nitrogen content falls to around 7% and its phosphorus content to 1% (Hynninen, 1998). Nitrogen is typically added in the form of ammonia (NH3) or in its ionized form ammonium (NH/), because it is the form most heterotrophic bacteria are capable of using. However, nitrogen gas can also be a nitrogen source for certain bacteria (Madigan et al, 1997), which has been observed in the aerobic treatment of forest industry wastewaters. Hynninen and Viljakainen (1995) observed that if the dissolved nitrogen in the effluent does not exceed 1.5 -2 g m\"3, microorganisms will obtain nitrogen from the air; and Gapes et al. (1999) developed a technology based on fixing atmospheric nitrogen that can effectively treat N-limited BKME without the addition of supplemental nitrogen. Like nitrogen, the importance of phosphorus as a nutrient in biological methods of wastewater treatment is such that its determination is essential with many industrial wastes and in the operation of WWTPs (Sawyer et al, 1994). The usual form of the phosphorus as a nutrient is orthophosphate (PO43\") (Madigan et al, 1997), which is the component included in all models that include phosphorus removal processes. The dynamics of phosphorus removal are complex, and involve different processes under aerobic, anoxic and anaerobic conditions (storage, growth, lysis and hydrolysis), in addition to the reactions related to the nitrification-denitrification processes. However, since phosphorus removal was not an objective of the WWTP under study, a more detailed explanation of the phosphorus dynamics is beyond the scope of this research. 41 3.5 Modelling the microorganism population Living single-cell microorganisms that can only be seen with a microscope are responsible for the activity in biological wastewater treatment. The basic functional and structural unit of all living matter is the cell. Living organisms are divided into either prokaryote or eukaryote cells as a function of their genetic information and cell complexity. The prokaryotes have the simplest cell structure, and include bacteria, blue-green algae (cyanobacter), and archaea. In contrast, the eukaryotes are much more complex and contain plants and animals and single-celled organisms of importance in wastewater treatment, including protozoa, fungi, and green algae (Metcalf & Eddy, 2003). Activated sludge systems present a mix of different bacterial species and other microorganisms acting as bacteria predators (protozoa, rotifers, etc.). Since modelling all the relations and processes of all these different type of microorganisms is too complex, models have introduced only those microorganisms responsible for the major biodegradation processes as model components. The removal and stabilization of dissolved and particulate organic matter found in wastewater is accomplished biologically by a variety of microorganisms, principally bacteria. Heterotrophic bacteria (heterotrophs) are used to oxidize the dissolved and particulate carbonaceous organic matter into simple end products and additional biomass. Microorganisms are also used to remove nitrogen and phosphorus in wastewater treatment processes. Specific bacteria (nitrifiers) are capable of oxidizing ammonia to nitrite and nitrate, while heterotrophs can reduce the oxidized nitrogen to gaseous nitrogen. For phosphorus removal, biological processes are configured to encourage the growth of a certain group of heterotrophs (phosphorus accumulating organisms, PAOs) with the ability to take up and store large amounts of inorganic phosphorus (based on Metcalf & Eddy, 2003). For modelling purposes, in addition to these three groups of bacteria (heterotrophs1, nitrifiers and PAO), evidence that some bacteria have carbon storage capability has added additional complexity to the models, leading to the concept of structured biomass models. Under a 1 Despite the fact that P A O bacteria are a group of heterotrophic bacteria, in this thesis, heterotrophic bacteria are understood as non-PAO heterotrophs, unless the contrary is specified. 42 structured biomass model, a certain bacterial population is divided into two or more model components, one representing the actual bacterial active biomass, and one or more model components corresponding to the different compounds stored inside the cell. In addition, new evidence has shown that other microorganisms (such as fungi and protozoa) might play a relevant role in COD removal, bacterial decay, and in the production of particulate inert material (Mathys, 1991; Van Loosdrecht and Henze, 1999). For the purposes of this research, only the heterotrophic bacteria were of interest, so the modelling of the microorganism population was restricted to this type of bacteria, neglecting all other microorganisms and the processes that involve them. 3.5.1 Modelling heterotrophic biomass Heterotrophs are responsible for hydrolysis of particulate substrates (Xs) and the metabolism of all degradable organic substrates. Heterotrophs can grow aerobically and many of them also anoxically (denitrification process). They are modeled as a particulate component (XH), which can be separated from soluble components in a clarifier. Most models represent heterotrophic bacteria as a \"heterotrophic biomass concentration\", regardless of the number of bacteria present. Then, the growth of bacteria is modeled as an increase in the heterotrophic biomass concentration instead of as the multiplication of individual bacteria. This assumption is based on observations that determined that only negligible cell multiplication takes place at low substrate concentrations, because cell replication is a demanding process energetically so that the increase in cell mass reflects only the increase in molecular polymer content in the biomass. Consequently, weight changes do not necessarily reflect similar changes in cell number (Chudobae/a/., 1992). Since many different bacteria compose the heterotrophic biomass, many different groups can be identified, such as: those which can and which cannot grow under anoxic conditions; those which have and which do not have storage capacity; those that are slow and fast growers, etc. The way of modelling these different groups may vary depending on the groups that are to be modeled. 43 An approach that has been widely used for modelling the anoxic/non-anoxic capacity is the incorporation of kinetic \"reduction factors\" to the kinetic process rates. These factors simulate the work done by a smaller group of microorganisms as a smaller rate affecting the whole heterotrophic biomass (i.e., as in Barker and Dold, 1997; Gujer et al, 1995; Henze et al, 1999). This assumption has been shown to be able to predict reasonably well in practice and reduces significantly the complexity of the models. For modelling the storage capacity of microorganisms, structured biomass models were developed. Structured biomass was first included in the models of the biological phosphorus removal process. ASM2 (Gujer et al, 1995) modeled PAO bacteria with two cellular internal stored components: poly-phosphates (Xp P ) from ortho-phosphate uptake; and ploy-hydroxy-alkanoates (PHA) and glycogen (grouped in one component called XPHA) from the uptake of organic fermentation products. The growth of the PAO bacteria was assumed to occur only from the use of stored substrate X P H A - Further evidence showed that heterotrophic bacteria also have the ability to store substrate, and ASM3 (Gujer et al, 1999) used similar assumptions for modelling the growth of heterotrophic bacteria in the presence of readily biodegradable COD, modelling the heterotrophic bacteria using two components: the microorganisms themselves; and an internal storage product, which includes PHA and glycogen, as well as other storage products. Evidence has shown that heterotrophic growth can occur directly from readily biodegradable substrate (Ss) as well as from storage substrate (XSTO)- This fact has induced some proposals that model heterotrophs as two groups, with and without storage capacity (Hanada et al, 2002), as well as proposals that assign to one heterotrophic biomass the capacity of storage and growth from storage substrate as well as directly from Ss (Beccari et al, 2002). Even though stored substrates have not been measured analytically in activated sludge systems treating pulp and paper effluents, there is some evidence of the occurrence of metabolic reactions inside the cell that appear to continue for a prolonged period of time even when the respiration rate had reached he level of endogenous respiration (Franta et al, 1994b), which suggests the occurrence of stored substrate for these effluents. 44 The differentiation of heterotrophs depending on their growth rate (fast and slow), as proposed by Chudoba et al. (1992), may have little impact on the dynamic modelling of activated sludge systems, but it could be crucial when designing tests for determining the kinetic parameters of the heterotrophs. If the food to microorganism (F/M) ratio is too high (above 2 - 4) in a certain test, the kinetic expressions derived from that test might be only representative of the fast-growing group of microorganisms. More complicated models have also been developed for microbial growth. For example, Daigger and Grady (1982) proposed a comprehensive conceptual model in which the heterotrophic biomass was partitioned into four components: storage products, synthetic component, structural component, and the enzymes responsible for the synthesis of the precursors of the components. This model also incorporated nine equations and many parameters, making it very difficult to calibrate. 3.6 Modelling the major processes As was explained before, activated sludge systems are very complex and involve many different processes acting together. A model including all processes is beyond the scope of this research, so only the processes related to the organic matter removal were modeled. This sub-model is commonly called the biokinetic model, because it describes the kinetics of the bioprocesses. Four major processes make up the biokinetic model: hydrolysis, growth, storage, and endogenous respiration (decay). Heterotrophic bacteria are responsible, or are assumed to be responsible, for these four processes, so the presence of other microorganisms is neglected. Two additional processes that may be related to organic matter removal are volatilization and sorption. These processes are not included in the biokinetic model because they are considered physical-chemical processes. However, volatilization and sorption are discussed because there is,some evidence that supports their importance in pulp and paper applications. 45 3.6.1 Hydrolysis Hydrolysis of slowly biodegradable COD (Xs) is a very important step of activated sludge modelling, especially when Xs accounts for the bulk of the biodegradable COD content of the wastewater, as is typically the case for both domestic sewage and industrial wastewaters (Insel et al, 2002). In municipal wastewaters, the particulate and colloidal fractions account typically for 40 - 50% of the total organic content after primary clarification (Levine, 1991, as cited by Eliosov and Argaman, 1995), and in industrial wastewaters it depends highly on the type of industry. Currently, all activated sludge models for organic carbon and nutrient removal involve a hydrolysis step to quantify the fate of slowly biodegradable substrate as a model component (Orhon et al, 1999a). The general definition of hydrolysis is the breakdown of a polymer into smaller units, usually monomers, by the chemical addition of water (Madigan et al, 1997). In wastewater treatment applications the processes of hydrolysis summarizes all mechanisms that make slowly biodegradable substrates (Xs) available for use by the microorganisms for growth and/or storage. It is generally assumed that the products of hydrolysis are Ss and Si. As for other biological processes, hydrolysis rate depends on temperature (Lishman and Murphy, 1994). In reality, since Xs is made up of a vast array of organic compounds, the hydrolysis process is a way of modelling the sequence of complex reactions ranging from physical entrapment and adsorption to enzymatic breakdown prior to biochemical oxidation and synthesis. As separate and individual identification and description of these reactions is practically impossible in activated sludge systems (Orhon et al, 1999a), a one or two step hydrolysis process is conveniently adopted by most models to reflect the cumulative effect of all these complex interactions. Particulate organics are rapidly taken up into the activated sludge floes, but their degradation is relatively slow. In order to use particulates as substrate, bacteria release extracellular enzymes that break down and solubilize particulate matter. It is generally accepted that the rate of hydrolysis is much slower than that of the removal of soluble substrate produced by hydrolysis. Thus, it is the hydrolysis rate that determines the overall rate of particulate organics degradation (Eliosov and Argaman, 1995). 46 Equations The hydrolysis process is commonly defined by means of a surface-limited reaction kinetics (Barker and Dold, 1997; Dold et al, 1980; Gujer et al, 1995; Gujer et al, 1999; Henze et al, 1987; Orhon et al, 1999a), as presented in Equation 3.8. This equation has proven to be the appropriate model for the hydrolysis of Xs that adequately describes the oxygen uptake rate (OUR) profiles obtained at different F/M ratios with a single set of parameters. Alternative equations may be used as approximations of Equation 3.8. These approximations have proven to be accurate under relatively narrow conditions ranges. dXs/dt = - k H - ( X S / X H ) / ( K X + X S / X H ) - X H [3.8] The most common simplification of Equation 3.8 is a first-order rate expression with respect to Xs (Equation 3.9), assuming that Kx » XS/XH and that the hydrolysis is not affected by XH (i.e. Kappeler and Gujer, 1992; Sollfrank and Gujer, 1991; Spanjers and Vanrolleghem, 1995; Wanner et al, 1992). dXs/dt = - k H - X s [3.9] Kinetic expressions similar to Monod (Equation 3.10) have also been found suitable to describe the hydrolysis process using pure cultures (Mino et al, 1995) or in activated sludge systems with a single organic compound (Goel et al, 1998a). dXs/dt = - k H - X s/(KX+Xs)- X H [3.10] Additional Xs fractions More complex approaches for modelling hydrolysis incorporate a further subdivision of Xs into two fractions: rapidly hydrolysable COD (XR or SH, depending whether it is assumed to be soluble or particulate) and slowly hydrolysable COD (Xs). This assumption might be reasonable for wastewaters with a high fraction of Xs, where it could be misleading to characterize this fraction by a single hydrolysis rate. In attempting to correlate different hydrolysable fractions with their physical properties, different authors have reached non-conclusive results. Sollfrank and Gujer (1991), working 47 with domestic wastewater, reported success by assuming the rapidly hydrolysable COD (XR) to be particulate, and assuming both rapid and slow hydrolysis to follow a first order reaction. Orhon et al. (1999a) found that the rapidly hydrolysable COD may conveniently be identified as the soluble fraction of the slowly biodegradable COD (SH), and this dual hydrolysis model provides a much better interpretation of hydrolysis kinetics with markedly different rates associated with these fractions. Ginestet et al. (2002), working with municipal wastewater from seven different WWTPs, found that rapidly hydrolysable COD was present in all physical fractions: soluble, colloidal, and particulate. It was found also that the hydrolysis kinetics of the rapidly hydrolysable COD could be modeled by a first order reaction, while the hydrolysis kinetics of Xs could be modeled by a limited surface reaction. When working with multiple Xs fractions, two approaches can be used to model the stoichiometry of the hydrolysis process: parallel hydrolysis (Janning et al, 1998; Orhon et al, 1998; Sollfrank and Gujer, 1991) or sequential hydrolysis (i.e. Confer and Logan, 1997). Parallel hydrolysis allows modelling of the utilization of different fractions independent of each other, and as a result allows greater flexibility in model calibration. Sequential hydrolysis has to be considered when the accumulation of intermediate hydrolysis products should be described by the model (Morgenroth et al, 2002). Some models incorporate two bacterial populations that grow separately (degrading either predominantly slowly or readily biodegradable substrate), which have been supported by recent results (Morgenroth et al, 2002). Effect of electron acceptor The effect of the electron acceptor on hydrolysis is not totally clear. Some experimental evidence indicates a dependence of hydrolysis on the available electron acceptor, which is incorporated in some models as an empirical reduction factor affecting the hydrolysis rate under anoxic and anaerobic conditions (i.e. in Barker and Dold, 1997; Gujer et al, 1995; Henze et al, 1987). On the other hand, more recent experimental evidence suggests that hydrolysis is independent of the electron acceptor (Goel et al, 1998a, 1999; Mino et al, 1995). The latter assumption is incorporated in ASM3 with a single hydrolysis equation that applies regardless of the redox conditions of the system. 48 3.6.2 Organic matter removal The wastewater organic matter may be removed in activated sludge systems in a number of ways, the most important of which are: bio-oxidation by microorganisms (cell growth), storage of substrate, volatilization (or air stripping), and sorption. It is important to distinguish among these different contributions to the overall COD removal, in order to better understand the dynamics of the process and to build up a useful mechanistic model for activated sludge systems. 3.6.2.1 Biooxidation by microorganisms The oxidation of organic matter by microorganisms is by far the most important mechanism of COD removal in activated sludge systems treating both municipal (Metcalf & Eddy, 2003) and industrial wastewaters (Eckenfelder, 2000). Approximately one third of the COD is removed during the growth process for use as a source of energy for this process. The remaining two thirds of the COD are transformed into new biomass, such that they are not actually removed in this stage. The amount of new biomass generated per unit of COD removed is characterized by an observed \"yield\" coefficient (YOBS) , which is around 0.67 g CODxh (gCODss)\"1. For simplicity, this yield determination is generally based on a 'black box' approach, without modelling all different metabolic pathways of the COD (Dircks et al, 2001a). In addition to the COD removed during the growth process, the remaining COD is removed in the form of microorganism biomass. A clarification process can separate the microorganisms, which present a density higher than water, from the liquid. Typically, greater than 99 percent of the suspended solids can be removed in the clarification step (Metcalf & Eddy, 2003). Part of the microorganism mass is recycled to the biotreatment process in order to maintain a certain biomass concentration, and the rest is taken out from the system through a wastage flow. Whether the heterotrophs use external substrate, internal substrate, or a combination of both for growth is not totally clear. Before ASM3, most models included a one-step growth process from Ss to X H . ASM3 included two processes in series: storage of Ss to X S T O , followed by a 49 growth from X S T O to X H . New evidence supports a concept of parallel growth using both routes at the same time (Beccari et al, 2002; Krishna and Van Loosdrecht, 1999), which may complicate the model quite a bit. In fact, ASM3 assumes growth only from stored substrate, not based on evidence, but because that assumption significantly simplifies the model. The kinetics of growth on soluble substrate has been widely modeled using the well-known Monod kinetics (Equation 3.11), in part because evidence supports this model but also because of mathematical convenience (Gujer et al, 1999). u=u H - S/(Ks + S)- X H [3.11] Where u is the heterotrophic growth rate, U H is the maximum heterotrophic growth rate, S is the substrate concentration, and Ks is the substrate saturation coefficient. Generally it is assumed that the substrate is Ss exclusively. Based on existing experimental information, a U.H value of around 5 - 6 d - 1 is generally adopted for domestic sewage (at 20°C). Results in the same range have also been reported for different industrial effluents, provided that compounds with inhibitory action were removed prior to the test (Sozen et al, 1998). In contrast, the kinetics of growth on stored substrate have been modeled using surface limited kinetic dynamics (Equation 3.12, ASM3) as well as first-order kinetics (Equation 3.13 as in Van Aalast-Van Leeuwen et al, 1997). Growth on X S T O appears to be independent of the type of electron acceptor present in the system and independent of the solids retention time (SRT) of the system (Beun et al, 2002). U = U - H - ( X S T O / X H ) / ( K S T O + X S T O / X H ) - X h [3.12] u = k- X S T O [3.13] The amount of X H produced from X S T O is characterized by a heterotrophic yield ( Y H ) coefficient, which represents the amount of X H produced by each unit of X S T O consumed. Since the storage is an intermediate step in the growth process, only a fraction of the energy required is used in this stage. Therefore, Y H has to be higher than Y O B S -50 3.6.2.2 Storage of soluble substrate Storage processes play a dominant role in the activated sludge process. It seems that bacteria use storage capacity to provide substrate for growth when the extracellular substrate is depleted. In this manner, these bacteria can balance their growth rate in dynamic processes (Van Loosdrecht et al, 1997). Since there is almost no natural environment with a constant supply of substrate, it is therefore logical that bacteria have a storage pool of substrate in their cells (Van Loosdrecht and Henze, 1999). Under this process, the COD removed is partially oxidized and partially accumulated and stored inside the cells (Cech and Chudoba, 1983). Another yield coefficient (YSTO) is used for stoichiometry in this process, which represents the amount of COD stored per unit of COD degraded. Based on an energy balance, Y S T O can be obtained from Y O B S and Y H using Equation 3.14. • Y S T O = Y O B S / Y H [3.14] As mentioned, ASM3 incorporated the storage process based on the large amount of evidence of substrate accumulation, as a way to predict the substrate flux into storage, assimilation and dissimilation respectively (Gujer et al, 1999). It was assumed that all Ss is first accumulated as X S T O , which is definitely not observed in reality, but which considerably simplifies the model. The first kinetic equation for modelling the storage process by heterotrophs was developed by van Aalast-van Leeuwen et al. (1997), measuring the acetate uptake during a feast period. In the absence of any other model, they chose the simplest one: zero-order kinetics. Later, ASM3 modeled the storage kinetics with a Monod equation, based on the soluble nature of the substrate (Equation 3.15). The Monod equation is also more flexible, being able to simulate the effect of first order kinetics by choosing the parameters properly. k = k S T O - S / (K S + S)- X H [3.15] 3.6.2.3 Volatilization Volatilization is a mass transfer process in which a constituent in the liquid is transferred to the gas phase. This process groups forced air stripping and the natural volatilization processes. This 51 mechanism may contribute to the loss of COD from an activated sludge bioreactor during the transfer of oxygen to the system (Grady et al, 1999). The more volatile organic compounds are the most susceptible to volatilization, although they may also undergo biodegradation. Volatilization is not typically incorporated in models for COD removal (i.e. ASMs), but it is included in the NCASI model (NCASI, 1986) for the treatment of pulp and paper effluents. It was observed that, using subsurface aeration, this process was only relevant to chloroform removal, being responsible for 80% of the total observed chloroform depletion. In contrast, it was observed that volatilization did not affect the removal of phenol, tetrachloroguaiacol and hexachlorobenzene. In the case of mechanical pulp and paper mill effluents, Sreckovic (2001), from two sets of experiments conducted at the Port Alberni mill, could not find significant differences in the initial COD and the final COD of samples aerated for two hours. He concluded that an insignificant amount of volatile COD was present in the mill process effluent, or that all volatile compounds had been volatilized before the wastewater reached the primary clarifier outlet (Sreckovic, 2001). Also working with four different CTMP effluents, Dubeski (1993) concluded that, during biological treatment, little of the reduction in volatile organics could be attributed to volatilization. Similarly, Mathys (1991), working with TMP effluents, found that the BOD5 losses by evaporation were negligible under the conditions tested. 3.6.2.4 Sorption Some organic matter can be sorbed onto the biomass and other solids in the bioreactor, such that they will be removed from the system with the waste sludge. The sorption process is not frequently incorporated in models for COD removal because it is assumed the sorption of organic compounds onto biological solids is rapid, so it is included as one non-limiting step in the hydrolysis process. The NCASI model incorporated this mechanism even though the sorption of pulp and paper compounds onto the biological floe has also been demonstrated to occur rapidly. When testing the model, it was observed that sorption was significant for two of the four compounds tested (tetrachologuaiacol and hexachlorobenzene), accounting for between 74 and 99% of the overall 52 removal. It was also observed that 75% of the sorbed substrate was removed with the waste sludge and 25% remained in the treated effluent. With the incorporation of the storage process, new evidence has considered more seriously the effect of biosorption as a relevant mechanism in the activated sludge system. Beccari et al. (2002) proposed a modification of ASM3 incorporating a biosorption and/or accumulation step, which demonstrated a better fit to the whole range of experimental results. 3.6.3 Endogenous respiration of microorganisms The concept of endogenous respiration (consumption of oxygen in the absence of external substrate) was first observed by Porges et al. in 1953. In 1958, Herbert modeled similar findings as a mathematical equation for the endogenous respiration (known as Herbert-Pirt relation). In 1962, based on experimental evidence, Dawes and Ribbon concluded that endogenous respiration in the absence of external substrate is due the conversion of intracellular reserve material for maintenance purposes (Van Loosdrecht and Henze, 1999). It is generally accepted that a fraction of the biomass 'consumed' in this process is transformed into inerts through the generation of particulate residual products (usually assumed to be Xi). A factor of 0.2 is widely used for estimating the amount of Xi produced per unit of biomass lost, although that factor may vary depending on the organic carbon source (Orhon et al, 1999b). The assumption of inerts generation in the models is based on the observation that the model fits the data better and reduces complexity (Dold and Marais, 1986). On the other hand, there is evidence that no accumulation of inerts needs to occur during the endogenous respiration process (Van Loosdrecht and Henze, 1999), and other mechanisms instead of endogenous respiration are the responsible of the generation of inerts (i.e. predation by protozoa). Mathematically, the endogenous respiration process is modeled as a first order process with respect to the active biomass concentration (Equation 3.16). dXH/dt = - b H - X H [3.16] 53 Where t>H is the endogenous respiration coefficient for heterotrophs (d\"1). An alternative concept is the maintenance process, which mathematically, is equivalent to endogenous respiration, but which assumes the use of external substrate for energy requirement. These concepts cannot be easily distinguished from each other experimentally, provided external and internal substrate is always available. In addition, if a biological system is modeled with structured biomass, endogenous respiration and maintenance are essentially the same (Van Loosdrecht and Henze, 1999). A different approach, the death-regeneration theory, was incorporated in ASM1 by postulating that the death of a portion of living organisms causes the lysis of their cell membrane, allowing the growth of new biomass on the biodegradable fraction of the cell protoplasm and an increase of the concentration of non-biodegradable compounds. Even though this cyclic process can adequately describe the activated sludge processes, that fact does not prove that the decay and lysis really occur. The endogenous respiration (or decay) parameter for heterotrophic biomass (bn) has a different connotation in models adopting the death-regeneration theory (i.e. ASM1) than in models using the endogenous respiration concept (i.e. ASM3). However, it is easy to convert one to another using Equation 3.17. b H ) A S M i = b H / ( l - Y 0 B S - (1-fxi)) [3.17] Some later models incorporated the death-regeneration theory (i.e. Barker and Dold, 1997; Gujer et al, 1995) while others switched back to the 'traditional' endogenous respiration concept, even though using ASM1 as a model basis (i.e. Dochain et al, 1995; Gujer et al, 1999; Kappeler and Gujer, 1992; Sollfrank and Gujer, 1991; Spanjers and Vanrolleghem, 1995). Structural problems with the decay/hydrolysis process as in ASM1 have also been discussed in the literature (Keesman and Spanjers, 2000), which leads to systematic modelling errors. New evidence, regarding the death mechanisms of bacteria, suggests that most bacteria do not die easily, instead they become dormant (Kaprelyants et al, 1993; Kaprelyants and Kell, 1996). 54 Dormancy can be defined as a reversible state of metabolic shutdown (Kaprelyants et al, 1993), under which bacterial cells are characteristically more resistant to environmental insults than cells in any other recognized physiological state. It has also been shown that being nonviable or nonculturable is not the same as being dead (Kell et al, 1998). Even after death, cell lysis and the use of the protoplasm for growth has not been demonstrated. Predation by larger forms of microorganisms (i.e. protozoa, worms, etc.) is an alternative concept to model the fate of bacterial cells. In fact, it is believed that a large part of the observed endogenous respiration is due to predation (Krishna and Van Loosdrecht, 1999). In spite of the uncertainties related to the decay process, the first-order kinetic model has remained as the most accepted way of modelling it. Since it has been proven to predict reasonable well the dynamics of all these iterations combined, no further complexity has been added. 3.7 Procedures for estimation of model parameters There is great diversity in experimental methods for estimating model parameters in activated sludge models, depending on the approach used and on the selected model. A brief review of some of the methods commonly used for calibrating these models is presented. Special attention is paid to batch and respirometric tests, which have proven to be easy and reliable ways of estimating model parameters (Chudoba et al, 1992; Henze, 1992; Kappeler and Gujer, 1992; Vanrolleghem and Vandaele, 1994). In addition to the procedures used for estimating single model parameters, full calibration of complex models such as ASM2 or ASM3 usually requires the use of an iterative calibration approach (Brun et al, 2002; Rieger et al, 2001; Vanrolleghem et al, 1999a). Highly correlated parameters and the impossibility of identifying single parameters from single experiments make it necessary to perform more than one calibration process (tuning procedure). 55 3.7.1 Analytical and respirometric techniques commonly used 3.7.1.1 Analytical techniques The most widely used analytical measures of wastewater composition are chemical oxygen demand (COD), biochemical oxygen demand (BOD), and total suspended solids (TSS), as described in Section 2.4; and volatile suspended solids (VSS). Volatile suspended solids (VSS) corresponds to the fraction of the TSS that can be volatilized and burned off when ignited at 500 + 50°C. In general, VSS are presumed to be organic matter, although some organic matter will not burn and some inorganic solids break down at high temperatures (Metcalf & Eddy, 2003). Filtration is also widely used in order to differentiate the soluble and particulate fractions, a key element of wastewater modelling. Typically, a pore size of 0.45 pm is used for this purpose, however, some authors recommend the use of a smaller pore size such as 0.1 pm in order to obtain the truly soluble fraction. Flocculation followed by clarification is an alternative method for separating the truly soluble components. 3.7.1.2 Respirometric techniques Respirometry is a commonly used technique for the characterization of wastewater and activated sludge and constitutes a well-established method to assess the state of microbial activity and for the calibration of microbial kinetic models (Marsili-Libelli and Tabani, 2002). Respirometry is based on the principle that aerobic microorganisms consume oxygen, so a decrease in the oxygen concentration of an un-aerated reactor should be equal to the oxygen used by the microorganisms. The reactor or chamber used is called a respirometer, and it is composed of: an air-hermetic chamber, a dissolved oxygen (DO) probe, an air diffuser stone, a stirrer (magnetic or mechanic), and a thermometer (when not measured by the DO probe). Additional desirable 56 features are a water bath for temperature control, a computer interface, and an automatic oxygen controller. Figure 3.2 illustrates a typical respirometer. Magnetic stir plate Figure 3.2: Schematic diagram of a respirometer with a computer connection The procedure for measuring the OUR is based on aeration and non-aeration cycles. The oxygen concentration of a sample is raised to a certain level (i.e. 6 g m\"3) during the aeration part of the cycle, at which point the aeration is stopped. The DO concentration is then monitored continuously until the DO concentration reaches another pre-determined level (i.e. 4 g m\"). The slope of the straight line of the DO concentration versus time is equal to the OUR of the sample at that moment. There are a number of different tests that can be performed with this technique, however, batch tests have received special attention since they are relatively easy to perform and the results have been shown to adequately represent the characteristics of the system (Chudoba et al, 1992; Henze, 1992; Hulsbeek et al, 2002; Kappeler and Gujer, 1992; Spanjers et al, 1994; Spanjers and Vanrolleghem, 1995; Vanrolleghem and Vandaele, 1994). The use of a continuous-feed reactor is difficult because a long period (several days) is generally required to reach steady state, which is difficult to maintain for a reactor fed with real wastewater (Spanjers and Keesman, 1994). Batch tests can be classified into three groups, depending on the sample being analyzed. 57 - Pure wastewater: this test can be used for determining the heterotrophic biomass composition of the influent (as in Wentzel et al, 1999). Since the amount of bacteria is low in not-seeded wastewater, relative small OUR levels are typically measured and many hours are necessary in order to allow bacteria for substantial multiplication. Pure mixed liquor (sludge): this test is useful for estimating the endogenous state of the microorganisms (typically used to determine bH), which requires the absence of external substrate. A mixture of sludge and wastewater: this test is by far the most useful, because it shows the biodegradation kinetics of the microorganisms in the presence of substrate. A wide range of different tests can be further distinguished, depending basically on the F/M ratio and on the use of filtered or raw wastewater. On one extreme, it can be used by inoculating wastewater with a small amount of sludge (F/M > 5 on COD basis); and on the other extreme can be used by adding a small amount of wastewater to a reactor with sludge in the endogenous state (F/M ~ 0.01). Then, F/M may vary over a range of 4 orders of magnitude, allowing for a variety of possibilities. 3.7.2 Estimation of wastewater characteristics There are a number of tests used for estimating the different wastewater characteristics. The selection of the proper method is not easy and depends on many factors. The task is made more difficult by the fact that there are no 'standard methods' for this characterization; so only some 'proposed methods' are available in the literature. 3.7.2.1 Readily biodegradable COD (Ss) Three methods for determining the readily biodegradable COD (Ss) fraction were selected. 1) The first method, proposed by ASM3 (Gujer et al, 1999), estimates Ss with the aid of a respirometric bioassay. Y O B S is required in order to obtain Ss, which can be estimated using Equation 3.18. S s (batch) = JOUR • d t / ( l - Y 0 B s ) \" [3-1 8] 58 Where the term (JOUR dt) refers to the oxygen consumed by substrate utilization (endogenous respiration consumption has to be subtracted) during the test and't' is the time. This method was originally proposed by Kappeler and Gujer (1992). 2) The second method, proposed by Mamais et al. (1993), is a rapid method for estimating Ss based on the physical properties of this fraction. Assuming that the influent truly soluble COD consists of Ss and Si, and that the influent Si is equal to the truly soluble effluent COD from an activated sludge plant with an SRT > 3 days, they estimated Ss as the difference between the truly soluble influent COD minus the truly soluble effluent COD. The truly soluble COD is measured after filtering (0.45 pm) a flocculated and clarified sample. This method eliminates the colloidal component of the soluble COD fraction. 3) Spanjers and Vanrolleghem (1995) developed a method for rapid wastewater characterization using very low F/M ratio (0.005 - 0.05 on a COD basis) batch tests and parameter identification with the aid of computational optimization tools. With this method, many model parameters can be determined simultaneously, including Ss. However, this method is susceptible to identifiability problems. 3.7.2.2 Slowly biodegradable COD (X 5 days, the concentration of Ss and Xs in the reactor is negligible compared with the amount in the feed. This method is also,proposed by other authors (i.e. Grady et al, 1999). 2) Roeleveld and Van Loosdrecht (2002) suggest estimating Xs from COD tests together with long term BOD tests using raw and filtered samples of influent and effluent. The 59 biodegradable C O D ( B C O D ) is obtained by multiplying the ultimate B O D ( B O D U L T ) estimated by a correction factor as shown in Equation 3.19. The estimation of the ultimate B O D can be done by measuring the 8001,2,4,6,8 and using the model to describe the B O D as function of incubation time. The parameter fBoD can be calculated using Equation 3.20, which usually takes values between 0.1 - 0.2. B C O D = (l/(l-f B O D))- B O D U L T [3.19] f B 0 D = fxi- YOBS [3.20] Xs is estimated by the difference between B C O D and Ss. 3) Kappeler and Gujer (1992) proposed estimating X s by curve fitting based on respirometric information from a batch test with unfiltered wastewater. This method has been adopted by many other researchers with the development of better optimization tools for the curve fitting procedure (i.e. Spanjers and Vanrolleghem, 1995). 3.7.2.3 Soluble inert COD (Si) The analytical determination of the inert COD is the most difficult task in wastewater characterization and, because only biodegradable components can be estimated by respirometry (Vanrolleghem et al, 1999b), other analytical tests have to be used for determining the inert COD fractions (Germirli et al, 1991; Orhon et al, 1999b; Petersen et al, 2001). Also, it has been shown that the fraction of inert COD in the influent depends on the sludge age of the WWTP (Brdjanovic et al, 2000; Hulsbeek et al, 2002; Van Veldhuizen et al, 1999) and on some external factors, such as temperature (Sollfrank et al, 1992). Three different approaches for determining the soluble inert COD (Si) fraction were reviewed. 1) When Si is being determined for an existing low loaded WWTP, the most common approach is to estimate Si as the effluent soluble COD (Gujer et al, 1999; Metcalf & Eddy, 2003; Sollfrank et al, 1992). Under this assumption, all Si is assumed to originate with the influent, and the generation of soluble residual microbial products is negligible. For a high loaded WWTP, Henze (1992) proposed correcting that estimation 60 by subtracting 1.5 times the effluent BOD 5 from the effluent COD s . An alternative empirical correction is to estimate Si as 90% of CODs,eff (Roeleveld and Van Loosdrecht, 2002; Siegrist and Tschui, 1992), which is based on practical experience with wastewaters from the Netherlands and Switzerland, respectively. 2) ASM1 (Henze et al, 1987), as well as ASM3 (Gujer et al, 1999), assume that the concentration of Si in a batch reactor does not change. Therefore, ASM1 recommends aerating activated sludge until the concentration of soluble COD remains constant. This amount should be equal to Sj. 3) Finally, complex tests have been developed for the assessment of the initial Si when it is desired to distinguish whether a soluble inert fraction was originally present in the influent or was generated within the plant. The methods consist of running two (Germirli et al, 1991) or three (Orhon et al, 1999b) parallel batch reactors, which can identify the fraction that corresponds to the residual microbial products generated during the hydrolysis and the endogenous respiration processes. The main problem with these tests is that they may identify as biodegradable some fractions of very slowly biodegradable COD that are actually inert for the time scale of the WWTP being modeled. 3.7.2.4 Particulate inert COD (X,) The estimation of Xi is frequently based on suspended solids mass balances. Two approaches can be followed: estimate Xi directly from a sludge production mass balance, as proposed by ASM1 (Henze et al, 1987), or to estimate Xs from the mass balances and Xi by difference after the other COD fractions have been estimated (i.e. Roeleveld and Van Loosdrecht, 2002). These methods based on mass balances, as long as they incorporate prediction of the solids production, can act to tune the model to the particular wastewater under study and compensate for any error made in other parameter estimates (i.e. YOBS or bH). As an alternative, some authors propose using Xi in the influent as a main calibration factor in the modelling process (Hulsbeek et al, 2002) or using a suspended solids mass balance (Meijer et al, 2001). 61 3.7.3 Estimation of stoichiometric parameters 3.7.3.1 Yield coefficients The determination of yields coefficients is very important not only because these are essential parameters in the models but also because the determination of many other parameters is based on the estimation of yield coefficients. There are two general approaches for determining yield coefficients: direct measurements of biomass generated (as COD or VSS) versus COD consumed; and curve fitting based on respirometric measurements. Estimating of YOBS The most common approach for estimating YOBS is that proposed in ASM1, in which a batch reactor with filtered wastewater is inoculated with an aliquot of acclimatized biomass. Then, COD and CODs are measured periodically, and YOBS can be calculated using Equation 3.21. The COD corresponding to cells is calculated as the particulate COD, as presented in Equation 3.22. YOBS = A cell COD / A COD s [3.21] cell COD = COD - COD s [3.22] A variation of the previous method was proposed by Sollfrank and Gujer (1991), using respirometry for estimating the biomass produced. YOBS is estimated, then, using Equation 3.23, which can also be written as shown in Equation 3.24. YOBS = (A COD-1 OUR- d t ) /ACOD s [3.23] YOBS = 1 - (IOUROBS • dt) / S s consumed [3-24] Where the term (JOUROBS dt) refers to the oxygen used for growth and storage, not including the oxygen used for the endogenous respiration process. Curve fitting of the predicted OUR (as presented in Equation 5.2 for ASM3) to the measured OUR is also a method that can be used for estimating YOBS along with other parameters, as in Spanjers and Vanrolleghem (1995). However, due to the importance of the correct estimation 62 of this parameter and the identifiability uncertainty of this method, normally this approach is not used. Estimating of YSTO The storage yield coefficient is a relatively new parameter, and there are few published methods for measuring it. However, the respirometric methods proposed are similar to those used for estimating Y O B S -The first method for the specific determination of Y S T O , proposed by Karahan-Gul et al. (2002a), is a graphic method based on respirometric measurements. Y S T O is calculated using Equation 3.25. Y S T O = 1 - (JOURSTO • dt)/COD degraded [3.25] Where the term (JOURSTO • dt) refers to the oxygen used only during the storage process. It is assumed the oxygen used during the storage process can be determine graphically as the area under the O U R curve and above the diagonal between the initial O U R and the breakpoint (B) of the respirograph, as presented in Figure 3.3. -10 30 70 110 150 Time (min) Figure 3.3: Breakpoint ' B ' and area that was assumed to represent the oxygen used during storage process. Data from Test 9-1, performed with F/M of 0.9 and temperature of 21.6°C. 63 As for YOBS , curve fitting of the predicted OUR (as presented in Equation 5.2 for ASM3) to the measured OUR (as in Koch et al, 2000) is also a method that can be used for estimating YSTO together with other model parameters. Estimating of YH The heterotrophic yield coefficient, as used in ASM3 for growth only from stored substrate, is also a relatively new parameter, which can be calculated based on the YOBS and YSTO using Equation 3.14 rearranged in the form of Equation 3.26 As for YOBS and Y S T O , curve fitting of the OUR is also a method that can be used for estimating Y H , as in (Koch et al, 2000), which can be estimated together with other model parameters. The quantity of substrate oxidized per unit of biomass synthesized will be greater for a growth response than for a storage response (Chudoba et al, 1992), which means that the storage yield (YSTO) should be higher than the observed yield coefficient (Y 0 Bs)- This conclusion is in accordance with the proposed values for those parameters by the ASM3 model. 3.7.3.2 Estimation off™ The inert fraction of the biomass (fXi), which is equal to the fraction of inert particulate products generated during endogenous respiration, is typically assumed to be 20% (Henze et al, 1987; Kappeler and Gujer, 1992). This parameter is not equivalent to ASM1 %', which is affected by the death-regeneration cycle of the biomass. However, fp and fxi can be easily converted from one to the other by using Equations 3.27 and 3.28. In the application of ASM1, it is generally recommended to assume fxi as 0.2 and estimate fp based on the differences in YOBS (Siegrist and Tschui, 1992; Vanrolleghem et al, 1999b). Y H - YQBS / YSTO [3.26] fxi = f P / ( l - Y Q B S (1-fp)) [3.27] fp = fxi- (l-YoBs)/(l-fxr YQBS) [3.28] 64 Working with ASM1, it has been demonstrated that, even though fp is theoretically identifiable by measuring OUR together with VSS, it is not practically identifiable due to its high correlation with other model parameters. (Keesman et al, 1998a; Wanner et al, 1992). One relatively simple approach for determining fXi was originally developed by Germirli et al. (1991) and improved later by Orhon et al. (1999b). The test consists basically of running two parallel batch reactors starting with the same initial COD, one with filtered wastewater and the other with glucose. Assuming that all of the glucose is biodegradable, it is possible to calculate the Xi produced by the difference between the inert fractions remaining in the reactors with filtered wastewater and glucose, respectively. Working with pulp and paper effluents, Slade et al. (1991) assumed for fp the default value proposed by ASM1. On the other hand, Sreckovic (2001) estimated fp as 0.342, which corresponds to a value of fxi of 0.55, which is extremely high compared with the proposed default value of 0.2. 3.7.3.3 Estimation off™ The production of soluble inert COD (Si) in activated sludge systems, specifically during the hydrolysis process, is not questioned but seldom incorporated in the models. It is accepted that soluble microbial products (SMP) are generated during the hydrolysis of Xs, and the inert fraction of those SMP contributes to Si. However, the coefficient fsi, which represents the amount of Sr generated per unit mass of X s hydrolyzed, has not been included in ASM1, the Barker and Dold (1997) model, and most of the ASM 1-based models. In contrast, Sollfrank and Gujer (1991), ASM2, and ASM3 incorporate this coefficient in the hydrolysis process, but its proposed value is always 0.00. This underestimation of the generation of Si may be due to a variety of reasons: the assumption that all Si comes with the influent is easy and it works well; that assumption is consistent with the calibration methods typically used; it is impossible to distinguish between different fractions of Si depending whether they were produced or coming with the influent; and finally because the determination of fsi requires additional tests that are labourious and still sensitive to analytical errors. 65 The assumption that all Si comes with the influent is directly related to the assumed value of fsi as zero. Assuming fs, greater than zero would make it necessary to estimate the initial fxs including that fraction that will be transformed to Si during the process, complicating the estimation of fxs.in- Equivalent results are obtained assuming fsi = 0 and that all Si,eff was initially in the influent (Figure 3.4). Assuming fS i >. 0 Assuming f s i - 0 Influent Effluent Influent Effluent Figure 3.4: Influence of the parameter fsi in the estimation of Si,jn The same experiment described for estimating fXi can be used for estimating fsi (Orhon et al, 1999b). The test consists of running two parallel batch reactors started with the same initial COD, one with filtered wastewater and the other with glucose. Assuming all glucose is biodegradable, it is possible to calculate the Si generated in hydrolysis by difference between the final CODs of the two reactors (Germirli et al, 1991). Incorporating a third reactor with unfiltered wastewater it is possible to estimate Si generated in the endogenous respiration process, which does not correspond to a parameter in most models. The main problem with these tests is that, since they are very long (> 15 days) compared with the time scale of the WWTP, they may identify as biodegradable some fractions of very slowly biodegradable COD that are actually inert for the time scale of the WWTP being modeled. 66 3.7.4 Estimation of kinetic parameters 3.7.4.1 Endogenous respiration rate, bH and bsm The endogenous respiration (or decay) parameter for heterotrophic biomass (bH) has a different connotation in models adopting the death-regeneration theory (i.e. ASM1) than in models using the 'traditional' endogenous respiration concept (i.e. ASM3). Only the traditional endogenous respiration approach is discussed, however, it is easy to convert one to another using Equation 3.17. The endogenous respiration coefficient estimates the reduction in time of the heterotrophic active mass in the absence of external substrate. The two major approaches for estimating bn differ in the way they estimate the heterotrophic mass. The respirometric method for determining bH was proposed by Sollfrank and Gujer (1991), based on the test design first proposed by Marais and Ekama (1976, as cited in Vanrolleghem et al., 1999b). In this approach, bH is obtained as the slope of the straight line resulting from the plot of the logarithm of the respiration rate obtained from a non-fed batch reactor. This method assumes that the decrease in the respiration rate is associated with a decrease in the heterotrophic mass. This method is coherent with that proposed by Keesman et al. (1998b), who described the endogenous respiration rate in a respirogram by a exponential function of time. An alternative estimation can be obtained from the same batch test but estimating the decrease in the heterotrophic biomass from the decrease of the particulate COD. Based on this method, the determination of bn is obtained from the slope of the straight line resulting from the plot of the logarithm of the particulate COD versus time. This method may also be used with total COD if all soluble COD is inert and will not change during the experiment (as in Koch et al., 2000) 67 3.7.4.2 Estimation ofksm.Ks, UH, and Ksm Not considering storage of substrate (as in ASM1) Two methods for determining UOBS and one for determining Ks are described. Note that POBS here is the equivalent to u H in ASM1, and that Ks in this section is considered as in ASM1, which is not comparable to the Ks used in the rest of the thesis. Kappeler and Gujer (1992) proposed estimating POBS from the OUR during a batch test with a high F/M ratio. It is assumed that unlimited heterotrophic growth occurs during the first period of the test, while the concentration of Ss remains sufficiently high. The slope of a plot of the natural logarithms of the OUR gives a straight line with slope (POBS - bH). Therefore, an independent estimation of bn is required. The alternative method for estimating POBS, and the most common method used for estimating Ks, is by curve fitting the predicted OUR (as presented in Equation 5.2 for ASM3) to the measured OUR. The parameters can be estimated by a graphical comparison (or a optimization routine) of the measured OUR with the simulated one. Curve fitting has become a very common method for parameter estimation. For example, Spanjers and Vanrolleghem (1995) used this method for estimating both POBS and Ks (as well as other parameters and wastewater characteristics), and Kappeler and Gujer (1992) proposed this method only for estimating Ks. Considering storage of substrate (as in ASM3) Since the ASM3 model has twice the number of parameters than does ASM1 for modelling bacterial growth, the calibration process is more difficult and more susceptible to identifiability problems when working with curve fitting. Still, ksxo, Ks, PH , and K S T O can be determined analytically by measuring the detectable stored products, but the method is labourious and sensitive to analytical errors. The easiest method for estimating ksTO, Ks, PH , and K S T O , as used by Koch et al. (2000), is by curve fitting the predicted OUR (as presented in Equation 5.2 for ASM3) to the measured 68 OUR. In fact, using OUR measurements together with analytical methods for measuring the stored products with municipal wastewater as substrate, Carucci et al. (2001) and Beccari et al. (2002) observed that parameter estimation based on curve fitting to OUR curves was acceptable, if correspondence with measurable stored substrates is not required. Despite the fact that all parameters are related, and modifying the value of one will affect the determination of the others, all these parameters can be determined from one single respirograph by fitting different regions of the curve. Figure 3.5 shows the different regions and which parameters can be determined based on them. Time Figure 3.5: Different regions of a respirographs and their implications on parameter estimation. In Figure 3.5, Region 1 accounts for the OUR due to the endogenous respiration process, and it is considered as the baseline for the other processes, assuming there is no increase nor decrease of OUR associated with endogenous respiration during the test. Region 2 is assumed to represent mainly the oxygen used for storage, so that is the key region for calibrating ksro- Region 3 shows a decrease of the storage process due to the depletion of Ss, so this is the ideal region for calibrating Ks, since it is assumed to represent the decrease of the storage rate at decreasing Ss concentrations. 69 The breakpoint B is assumed to represent the amount of stored substrate at the end of the storage process, which is related to the growth rate by the X S T O / X h ratio, so this point is useful for calibrating u.H. Finally, Region 4 is assumed to represent the decrease of the growth rate as X S T O is being used, so that region is good for calibrating K S T O -Since the storage and the growth processes are lumped together, there is still no test for estimating specifically u H and k s T O without curve fitting, as there is for U . 0 BS - However, Equation 3.29 shows a valid relationship for comparing u H and U.OBS- The mathematical derivation of this equation is presented in Appendix 1. Then, Equation 3.29 gives a reference value that may help, by estimating U 0 B S analytically, to avoid the risk of gross identifiability errors. U H > U O B S [3.29] 3.7.4.3 Estimation of 6T for different kinetic parameters In spite of the great importance of temperature on biological reactions, estimation of the temperature coefficients for the different kinetic parameters is not often considered as a relevant task in the calibration process, so typically the default values for 0TS are assumed (i.e. Koch et al, 2000; Rieger et al, 2001). Furthermore, methods for the estimation of 0Ts are often excluded from compilations of procedures for parameter estimation (i.e. Kappeler and Gujer, 1992; Vanrolleghem et al, 1999b). The only method found for the estimation of temperature coefficients (9TS) is the error minimization of regressions using the Arrhenius Equation (Equations 3.6 and 3.7) with data tests performed at different temperatures. All reviewed literature uses tests performed at a constant temperature, and no tests were found that performed with a change in the temperature during the test. The Arrhenius Equation, as presented in Section 3.4.4, is the most common way to model the effect of temperature on reaction rates, as well as the method used by ASM3 for modelling the change of all kinetic parameters ( k s T O , U H , bH, and bsTo) with temperature. Therefore, the following analysis is valid for determining 0T for any kinetic parameter that uses the Arrhenius 70 Equation. Rearranging Equation 3.7, and taking logarithms of both sides, Equation 3.30 is obtained. ln(k(T)/k(TREF)) = GT > k • (T-TREF) [3.30] Therefore, 0T,k can be obtained as the slope of the straight line from the plots of ln(k(T)/k(TREF)) versus (T-TREF), or just from the plots of ln(k(T)) versus T, because ln(k(TREF)) and TRE F are constants that do not affect the slope of the curve. At least three independent estimations of each kinetic parameter, performed at different temperatures, are required in order to estimate Oj.k. 3.7.4.4 Hydrolysis parameters k» and Ky Experience has shown that hydrolysis is a process that is difficult to estimate in practice (Eliosov and Argaman, 1995). In fact, the rate of hydrolysis in activated sludge systems cannot be determined directly by analytical measurements (Lishman and Murphy, 1994), and techniques for quantifying it include complex measurements of specific hydrolytic enzymes, specific hydrolytic intermediates or specific end products (Morgenroth et al, 2002). The determination of hydrolysis kinetic parameters is complicated even more, considering that this process has a different connotation in models adopting the death-regeneration theory (i.e. ASM1) than in models using the endogenous respiration concept (i.e. ASM3). Furthermore, the use of different kinetic expressions generates additional difficulties (i.e. the use of first-order reaction kinetics by many authors instead of the saturation kinetics proposed by most models). For ASM1, the hydrolysis parameters are used mainly to fit Region 4 of the curve presented in Figure 3.5. That 'tail' is assumed in ASM1 to be due to the hydrolysis of Xs, so this process has a major role in ASM1. On the other hand, ASM3 assumes that Region 4 is due to growth on X S T O , SO Region 5 would be used for estimating hydrolysis parameters, hydrolysis being a process with a lower relative importance than in ASM1. The most common approach for determining hydrolysis kinetic parameters is based on curve fitting of the OUR equation to OUR measurements (i.e. Gujer et al, 1999; Henze et al, 1987; Kappeler and Gujer, 1992; Koch et al, 2000; Spanjers and Vanrolleghem, 1995), which 71 requires model calibration to the experimental OUR data, provided that necessary information on wastewater characteristics and other kinetic and stoichiometric parameters have been previously obtained and included in the evaluation (Vanrolleghem et al, 1999b). This method assumes that hydrolysis is the rate limiting step that determines respiration rates (Dold et al, 1980). The major concerns in using this experimental approach for estimating hydrolysis parameters are: the effect of inaccurately pre-selected parameters on the evaluation of the OUR data, and identifiability problems that generate not one, but a series of parameter couples for hydrolysis with equal statistical validity, even if all other kinetic and stoichiometric parameters are assessed or selected properly (Grady et al, 1996; Insel et al, 2002; Orhon et al, 1999b). Temperature can also affect hydrolysis parameter determination, because hydrolysis may stop being the rate-limiting process in the degradation of Xs. This effect has been identified for ASM 1-based hydrolysis processes for temperatures higher than 20°C. This can be illustrated by comparing the temperature coefficients of the hydrolysis rate and the growth and/or the storage rate. In ASM1, the proposed 0j for kn is approximately 50% higher than 9T proposed for P O B S -In contrast, in ASM3, the proposed 0T for kH is approximately 50% lower than 0T proposed for ksTo and P H . Therefore, this impact may be not valid for the hydrolysis process as modeled in ASM3. 72 4. R e s e a r c h object ives a n d m e t h o d o l o g y 4.1 Research objectives The objective of this research was to assist with the development of a dynamic mechanistic model for predicting carbon oxidation (COD removal) in activated sludge plants treating pulp and paper wastewater. This research was focused on COD removal only. Based on field data analysis, this is the main concern in pulp and paper effluents. Nutrients are of less concern due to their low concentrations. Additional objectives of this research were: to increase the knowledge on pulp and paper activated sludge systems by estimating the wastewater characterization and the most important model parameters; - to compare the parameter estimations from this research with previous work done at the University of British Columbia, especially with the values reported by Sreckovic (2001), which were found to be greatly different from those for municipal wastewaters; to evaluate the suitability of a model developed originally for domestic wastewaters for use in modelling a pulp and paper WWTP outside some recommended ranges; - to identify the difficulties in the application of the selected model and propose changes in order to make it more suitable for modelling pulp and paper effluents treatment; and - to recommend new calibration procedures or to recommend, from the ones available in the literature, those more appropriate for these effluents. 4.2 Research methodology 4.2.1 Selection and description of the mill The mechanical pulping process was selected for developing this research because mechanical pulping is the major pulping process used in Canada, with 46% of the Canadian total (Pulp & Paper International, 2001). Regarding the type of treatment, activated sludge technology was chosen because it is the most popular treatment alternative among Canadian pulp and paper mills (Archibald and Young, 2003). 73 The Port Alberni division of NorskeCanada was chosen because this mill, located in Port Alberni, British Columbia, operates a mechanical pulping process and an activated sludge system suitable for applying and calibrating a dynamic model. An extensive historical database and previous research completed at the mill were also additional reasons for choosing that particular WWTP. The activated sludge system located in Port Alberni ('the plant') treats the pulp and paper mill effluents generated in the following processes of Alberni Specialties pulp mill: groundwood, paper mill, CTMP, power boiler and woodroom. The wood used in the pulping process is 100% softwood (mainly Douglas fir, western red cedar and hemlock). The bleaching process uses a sequence of hydrogen peroxide, sodium hydrosulfite, oxygen, and caustic soda. Port Alberni WWTP consists of a pH adjuster, primary clarifier (not modeled), five complete mixed bioreactors in series (IA and IB of 2,125 m 3 each, and 2, 3 and 4 of 5,250 m 3 each), and a secondary clarifier. The design of the bioreactors suggests the flow characteristics can be modeled as a cascade of five completely stirred tank reactors (CSTRs). There was no tracer information available at the time of the study. A more detailed description of the plant can be found in Koning et al. (1994). A diagram of the plant is presented in Figure 4.1. Influent Primary sludge Primary effluent .. IA IB 2 3 4 o o o o o o o o 0 o o o O O O 0 o o o o o o o o o o O 0 o o o o Recycle Modeled Effluent Wastage F i g u r e 4.1: Diagram of Port Alberni WWTP The plant treats an average of 74,000 m 3 d\"1, with the following average influent characteristics (primary effluent): COD of 594 g m\"3, BOD of 250 g m\"3, TSS of 27 g m\"3, pH of 7.2, and temperature of 32°C. Nitrogen, in the form of ammonia and urea, is added to the recycle sludge 74 in order to supply the nutrient requirements of the bacteria. Phosphorus is not added in the plant but it is added upstream, so its concentration is not limiting for the microorganisms. 4.2.2 Selection of the model Activated Sludge Model N°3 (ASM3) (Gujer et al, 1999) was selected as the biokinetic model for a variety of reasons: there is significant evidence that supports the basic assumptions of the model, mainly about the accumulation of storage products inside the cell (Carucci et al, 2001; Dircks et al, 2001b; Goel et al, 1998b, 1999; Van Loosdrecht and Heijnen, 2002; Van Loosdrecht et al, 1997) and the endogenous respiration approach for modelling the biomass decay (Van Loosdrecht and Henze, 1999); - ASM3 has been recommended for wastewater treatment plants (WWTPs) treating industrial wastewater with a high amount of chemical oxygen demand (COD), where the storage of readily biodegradable substrate is dominant (Koch et al, 2000), which is the case for the WWTP being modeled; it has been proposed that metabolic reactions inside the microbial cells appear to continue for a prolonged period of time in sludge from a paper mill (Franta et al, 1994b), which suggests the relevance of stored products in the treatment of wastewaters of this type; - it is a very simple model under aerobic conditions (simpler than ASM1); and it is the state-of-the-art model, and it is expected to be a new standard in modelling activated sludge systems. Since the degradation of CTMP wastewater organic material in the activated sludge process is mainly due to microbial activity, the absence of other substrate removal processes in ASM3, such as volatilization or adsorption, should not present a problem. Further, ASM3 was simplified by eliminating the anoxic processes, because the full-scale plant has no anoxic reactors. Finally, since the nitrogen level of the Port Alberni WWTP is very low and most of the nitrogen added is used as a nutrient by the heterotrophic microorganisms, the effect of the 75 nitrifiers was neglected. Therefore, the nitrification module of ASM3 was not included in this simplified model. 4.3 Description of the ASM3-based model The model presented corresponds to a subset of ASM3, which does not include seven out of the twelve equations (anoxic equations and processes related to autotrophic organisms), and the components nitrogen gas (SN2), ammonia (SNH4), nitrate plus nitrite nitrogen (SNOX) and alkalinity (SALK)- This model is only valid under aerobic conditions and at a pH close to neutral. The basic or structural assumptions of ASM3 are: - there is only one readily biodegradable substrate component; there is only one slowly biodegradable substrate component, which is particulate and undergoes hydrolysis before being degraded; there is one heterotrophic group of microorganisms with storage capacity; all microorganisms are treated as if they were the same species, and individuals are ignored; - particulate inert substrate can be coming with the influent or be generated during the endogenous respiration process; soluble inert substrate can be coming with the influent or being generated during the hydrolysis process; oxidation is the only mechanism of COD removal, so the effect of other mechanisms such as volatilization, adsorption or predation are negligible; the heterotrophic growth-limiting nutrients are organic carbon, oxygen, and nitrogen, and all other nutrients are always available in excess (including phosphorus); the value of the oxygen saturation constant is the same for growth, storage and endogenous respiration; any anaerobic metabolic response is negligible; stochastic phenomena are negligible; and neither mixing nor oxygen transfer is limiting. Additional assumptions, for this particular implementation of ASM3, are: 76 COD fractions are constant over time; - nitrogen as a nutrient is not limiting for heterotrophic bacteria; - nitrifying activity is negligible; effect of pH is negligible; acclimatization to toxicity is assumed; - parameters are constant over time; and influent biomass is assumed equal to zero. 4.3.1 Model components In ASM3, concentrations of soluble compounds are characterized by S and particulate compounds by X. Soluble compounds can only be transported by water, whereas particulate compounds can be concentrated by sedimentation and thickening processes in clarifiers. Table 4.1 presents a summary of all model parameters and their units, and Figure 4.2 shows the COD fractions of the model. Table 4.1: Definition of components and symbols of the model Component Symbol Unit Oxygen So2 g 02 m\"3 Inert soluble organic material s, g COD nf3 Readily biodegradable organic substrates Ss g COD m\"3 Inert particulate organic material X , g COD rn 3 Slowly biodegradable substrates X S g COD m\"3 Heterotrophic organisms X H g cell COD m\"3 Cell internal storage products of heterotrophic organisms X S T O g PHA COD m\"3 Definition of soluble compounds 1. Dissolved oxygen (S02): dissolved oxygen can be directly measured in the activated sludge reactors. For this application, S02 is included only for modelling inhibition of biological reactions (switching function). Therefore, gas exchange reactions and stoichiometric computations of S02 were not considered. 77 2. Inert soluble organic material (Si): the prime characteristic of Si is that these organics cannot be further degraded in the treatment plant modeled. This material was assumed to be part of the influent and may be produced in the context of hydrolysis of particulate substrates, Xs. 3. Readily biodegradable organic substrates (Ss): this fraction of the soluble COD is directly available for consumption by heterotrophic organisms. For simplification, it was assumed that all these substrates are first taken up by heterotrophic organisms and stored in the form of X S T O -Tota l C O D Readily Slowly Soluble Particulate biodegradable biodegradable inert inert *s s, Biomass Storage products X STO Figure 4.2: COD fractions in ASM3 simplified model Definition of particulate compounds 1. Inert particulate organic material (Xi): this material is not degraded in the activated sludge system modeled. Xi may be a fraction of the influent and is produced in the context of biomass decay. 2. Slowly biodegradable substrates (X s): X s are high molecular weight, soluble, colloidal and particulate organic substrates which must undergo cell external hydrolysis before they are available for degradation. It was assumed that the products of hydrolysis of Xs are either Ss or Si. All Xs is contained in the influent and none is generated in the decay process. 78 3. Heterotrophic organisms (XH): these organisms were assumed to be the 'allrounder' heterotrophic organisms. These organisms are responsible for hydrolysis of Xs and can metabolize all degradable organic substrates. 4. Cell internal storage products of heterotrophic organisms (XSTO): it includes poly-hydroxy-alkanoates (PHA), glycogen, etc. It occurs only associated with X H ; it is, however, not included in the mass of X H . X S T O is only a functional compound required for modelling but not directly identifiable chemically. 4.3.2 Processes in the model 1. Hydrolysis of slowly biodegradable COD: this process makes available all slowly biodegradable substrate (Xs) contained in the influent to the system. 2. Storage of readily biodegradable COD: this process describes the storage of Ss in the form of cell internal storage products (XSTO)- This process requires energy, which is obtained from aerobic respiration. It was assumed that all substrates first become stored material and later are assimilated to biomass. 3. Growth of heterotrophs: the substrate for the growth of heterotrophic organisms (XH) was assumed to consist entirely of stored organics (XSTO)-4. Endogenous respiration of biomass: this process describes all forms of biomass loss and energy requirements not associated with growth: decay, maintenance, endogenous respiration, lysis, predation, motility, death, and so on. 5. Respiration of storage products: this process is analogous to endogenous respiration. It assures that X S T O decays together with biomass. 4.3.3 Stoichiometry of the model Table 4.2 presents the stoichiometric matrix (or Peterson Matrix) of the model, the structure of which has become well known since the introduction of ASM 1. 79 Table 4.2: Stoichiometric matrix of ASM3 aerobic sub-model Processes S02 Si Ss Xi Xs X H X S T O 1 Hydrolysis fsi 1 -1 2 Storage of S s - ( 1 - Y S T O ) -1 Y S T O 3 Growth of heterotrophs - ( 1 - Y H ) / Y H 1 - 1 / Y H 4 Endogenous respiration X H -(1-fxO fxi -1 5 Respiration of X S T O -1 \"1 4.3.4 Kinetic expressions of the model The kinetic expressions of ASM3 are based on switching functions (hyperbolic or saturation terms, Monod equations) for all soluble compounds consumed. This form of kinetic expression was chosen not because of experimental evidence, but rather, for mathematical convenience, stopping all biological activity when concentrations approach to zero. Table 4.3 summarizes the kinetic expressions of the model. Table 4.3 Kinetic expressions of ASM3 aerobic sub-model Process Process rate equation 1 Hydrolysis k H • ( X S / X H ) / ( K X + X S / X H ) • X H 2 Storage of S s k S To • S02/(K02+S02) • SS/(KS+SS) • X H 3 Growth of heterotrophs pH- So2/(Ko2+So2)' (X S TO/XH ) / (KSTO+XSTO/XH ) • X H 4 Endogenous respiration of X H bH • So2/(Ko2+So2)' X H 5 Respiration of X S T O b S TO • S02/(K02+S02) • X S T O 80 The effect of temperature on the process rates was modeled with a modified Arrhenius equation affecting the kinetic rate parameters (k or u), as presented by Equation 3.7. Due to the relatively high temperatures present in the modeled WWTP, the reference temperature chosen was 30°C, instead of the 20°C typically used for domestic wastewaters. 4 .3.5 Restrictions and constraints Since ASM3 was developed for the simulation of domestic wastewaters, the authors suggested not applying it outside the range of conditions for which it was developed, by proposing the following limitations: 1. it is not advised to apply it to industrial wastewaters; 2. temperature should be in the range of 8 - 23°C; 3. the concentration of nitrite cannot be elevated; 4. activated sludge system should not have anaerobic reactors; 5. pH should be in the range of 6.5 to 7.5; and 6. the activated sludge system should not have a very high load or small SRT (< 1 day). The NorskeCanada Port Alberni WWTP is outside the recommended conditions according to criteria one, two and three, but it does meet criteria four, five and six. This fact should not present any difficulty; in fact, one of the objectives of this research was to test the model for pulp and paper effluents outside the recommended condition ranges. Further, the high concentration of nitrite is of less relative importance because nitrification is not included in the selected sub-model. 4.4 Experimental program A sensitivity analysis of the aerobic ASM3 was performed in order to identify the most sensitive parameters. The least or non-sensitive parameters were not calibrated and, therefore, were assumed from the literature. In addition, an identifiability analysis was performed in order to recognize, from the group of sensitive parameters, those parameters that can be adequately identified (calibrated) from respirometric measurements. Proper tests or alternative methods were adopted in order to 81 calibrate those parameters that were not identifiable in this way. The samples used in the tests were obtained from three measuring campaigns. In addition to the experimental tests and analyses, a one-year data window from the mill historical database was obtained. These data were used for performing flux and solids mass balances. They were also useful to validate the model in a full-scale basis, as presented in Chapter 6. 4.4.1 Sensitivity analysis In the sensitivity analysis, all parameters were checked against all model components. This sensitivity analysis was initially performed based on assumed parameter values, and subsequently was checked with the calibrated parameter values (iterative procedure). The initial values of the model components (state variables) for the sensitivity analysis were estimated based on the data obtained from the measuring campaigns and from the historical mill database. The sensitivity was calculated for the system using the dynamic differential equation as expressed in Equation 3.1. Numerically, the sensitivity of each state variable with respect to each parameter was calculated as the variation of the function f(x,p) (as in Equation 3.1) for a variation of 1% in the value of each parameter value, as presented in Equation 4.1. That sensitivity value was calculated for each state variable in each of the five cells. The total sensitivity value was calculated as the sum of the sensitivity values in the five cells (Equation 4.2). Sensitivity! j = f(x, pi) c e l l j - f(x, p) c e l l j [4.1] Sensitivity = Sj Sensitivity) j [4.2] Where Sensitivity^ is the sensitivity of the state variables with respect to the parameter ' i ' in Cell j , Sensitivity! is the total sensitivity of the state variables with respect to the parameter ' i ' , and pi is the parameter vector that modified its element ' i ' by multiplying it by 1.01. The routine used in the sensitivity analysis was that proposed by Olsson and Newell (1999) and performed with MATLAB®. The computational code of this routine is presented in Appendix 2. 82 4.4.2 Identifiability analysis Only the structural identifiability of the model was analyzed. For a complete identifiability analysis, a practical identifiability analysis should be done as a complement to the structural identifiability (for example, using the Fisher Information Matrix, as proposed by Weijers et al. (1996)). However, a practical identifiability analysis exceeded the scope of this research. The structural identifiability was analysed using the Taylor series expansion method for the oxygen uptake rate (OUR) around time zero, similar to the analysis done by Dochain and Vanrolleghem (2001) for a simple model based on ASM1. 4.4.3 Sampling program Three measurement campaigns were undertaken in order to calibrate the model. The first campaign was completed over a five day period, to perform a screening of many parameters along the complete WWTP together with the measurement of some model parameters. Six sampling points were selected and samples were taken twice a day at every point. The samples were taken considering the HRT of the different compartments of the WWTP in order to group different samples into sequences. Each sequence was characterized based on the time of the first sample, which corresponded to the influent. Therefore, the sample corresponding to Cell 4 was taken approximately 8.5 hours after the influent sample of the same sequence, and so on. Seven sequences were collected over four days, and the samples were analyzed for COD, CODs, TSS, VSS, ammonia, and ortho-phosphate. Some samples were also analyzed for alkalinity. In addition, some of these samples were used for respirometric batch tests within the day they were taken. The sampling points were the influent to the secondary treatment (primary effluent), Cell 1 A, Cell 4, recycle sludge, wastage sludge, and effluent. The second and third campaigns had the objective of obtaining fresh wastewater, sludge, and effluent in order to estimate specific model parameters that could not estimated from the previous measurements. Only samples of the influent, effluent, and from Cell 4 were taken during these campaigns, which were analyzed for COD, CODs, TSS, and VSS. 83 A more detailed description of the sampling campaigns is presented in Appendix 3. Preservation techniques are described together with analytical methods when applicable. 4.4.4 Analytical methods used Different analytical methods were used in order to measure different characteristics of the samples. Most of the methods were based on the recommendations in Standards Methods for the Examination of Water and Wastewater (American Public Health Association et al, 1998), which were complemented with methods proposed in research articles (mainly related to respirometry). 4.4.4.1 Filtration Filtration was used with two different objectives: to measure suspended and volatile suspended solids (TSS and VSS respectively), and to obtain the soluble fraction of a sample. Two different pore size filters were used for the two different objectives. TSS and VSS concentrations were measured using 1.5 urn pore size filters, equivalent to the filters used by the mill personnel. The soluble fraction of the samples was obtained by filtering through 0.45 um pore size filters, which is the most commonly used method to separate the soluble fraction as well as the method recommended by ASM3. The filtration of all samples during the WWTP measuring campaign was performed right after taking the sample at the Port Alberni mill laboratory. Note that the term 'soluble' is used in this thesis to identify the fraction that will pass through 0.45 um pore size filters, which includes the 'truly soluble' and the 'colloidal' fractions. Therefore, this definition is directly related to the way it was obtained more than its actual physical nature. 4.4.4.2 Chemical oxygen demand (COD) The COD was measured using the closed reflux, colorimetric method, with a 2 mL sample size, as proposed by Standards Methods (American Public Health Association et al, 1998). The digestion solution used contained mercuric sulphate (HgS04) in order to avoid interferences 84 from chloride present in the wastewater (from salt water coming with the logs stored or transported in the ocean). Two vial ranges were used: a low range (from 0 to 200 g COD m\"3) and a high range (from 20 to 900 g COD m\"3). The low range was used for effluent samples or when a low COD was expected, and the high range was used in all other situations. Total and soluble COD were measured for all samples. Almost all samples were digested right after being taken, at Port Alberni mill laboratory. However, the absorbance measurement and the calibration procedure were performed approximately one week later at the environmental laboratory of Civil Engineering, University of British Columbia (UBC). A few samples were digested at UBC, mainly for checking purposes. When preservation was required, pH < 2 and refrigeration at 4°C was used. 4.4.4.3 Ultimate biochemical oxygen demand CUBOD) Ultimate BOD was measured following the method proposed by Roeleveld and Van Loosdrecht (2002), measuring the BOD during the first ten days, and estimating the ultimate BOD by fitting data to the BOD model (Equation 4.3). The tests were run in duplicate to improve accuracy. Nitrification inhibitor (Hach, formula 2533) was added to all samples as recommended by the manufacturer. The samples for UBOD analysis were taken at Port Alberni WWTP and transported to UBC, Vancouver, during the day, and prepared immediately after arriving. Incubation temperature was 20°C. Total and soluble UBODs were estimated, using unfiltered and filtered samples respectively. The dilution water was seeded with 100 mL L\"1 of effluent, as recommended by the mill BOD procedure. BOD(t) = UBOD- (l-exp(-t- k B 0 D)) [4.3] The instruments used were a dissolved oxygen (DO) meter YSI Model 52, and a DO probe YSI Model 5905. The accuracies of the instruments used are, for the meter, 0.01 g m\"3 and, for the probe, the greater between 0.1 g m\"3 or 2% of the reading. The equipment was calibrated at the beginning of each series of measurements. 85 4.4.4.4 Total and volatile suspended solids (TSS and VSS) The TSS and VSS were measured as proposed by Standards Methods (American Public Health Association et al., 1998), using a filter pore size of 1.5 um. The temperature used for drying the TSS was 104°C, and 550°C as used for igniting the VSS. The sample size was variable depending on the solids content of the sample, and ranged from 2 mL for wastage sludge samples up to 200 mL for effluent samples. Solids measurements of the WWTP sampling campaign were performed at Port Alberni laboratory, as well as solid measurements corresponding to the respirometric tests undergone at the mill site (see Appendix 4 for details). Solids measurements of the respirometric tests performed at UBC were analyzed at UBC. The balances used were Mettler AC 100 (at Port Alberni laboratory) and Ohaus Adventurer (at UBC), both with a readability of 0.0001 g. These balances were used also for weighing the chemicals used in other analytical methods as well as in the preparation of stock solutions. 4.4.4.5 Clarification by coagulation and ftocculation The clarification procedure proposed by Mamais et al. (1993) was used in order to obtain the truly soluble fraction of the wastewater. The samples were flocculated by adding 1 mL of a 100 g L\"1 zinc sulphate solution to a 100 mL wastewater sample and then mixing vigorously with a magnetic stirrer for approximately 1 minute. The pH of the mixed sample was then adjusted to approximately 10.5 with 6 M sodium hydroxide solution and the sample allowed to settle quiescently for a few minutes. The clear supernatant, filtered trough a 0.45 um pore size, was assumed to correspond to the truly soluble fraction of the sample. All analyses using to this method were performed at UBC. 4.4.4.6 Ammonia (NHi-N) and ortho-phospahe (PO/) The concentration of ammonia and ortho-phosphate were measured with a flow injection analyzer (Quickchem 8000, Lachat instruments) at UBC. All samples were filtered. Ammonia 86 samples were preserved at pH lower than two, and those for ortho-phosphate were preserved with phenolic mercuric acetate. 4.4.4.7 Alkalinity The alkalinity was measured as proposed by Standards Methods (American Public Health Association et al, 1998), using the titration method (2320 B) and samples filtered with pore size of 0.45 pm. The results are given in mole HCO3\" m\"3, which can be converted to g CaCCV m\"3 multiplying by the conversion factor 50 g CaC03\" (mole HCO3\")\" 1. 4.4.5 Quality assurance program 4.4.5.1 Field blanks As described in the analytical methods, some samples were transported from Port Alberni to UBC. Field blanks were used in order to correct the results for the possible effect of handling, transportation and storing of the samples. When a difference existed between the field blank and the analysis blank, the field blank was considered as the true blank for the result purposes. 4.4.5.2 Duplicates The precision of the methods used was estimated by calculating the coefficient of variation of the samples that were duplicated at the field. The coefficient of variation corresponds to the standard deviation of the samples divided by their mean. A total of 146 out of 480 analyses were duplicated in this way, which represents a 30% of the total, well above 5% recommended as a minimum (American Public Health Association et al, 1998). The results of this analysis are presented in Table 4.4. 87 Table 4.4: Coefficient of variation for different analytical methods Analysis Coefficient of Number of Percentage variation duplicates of total (%) (%) TSS 8.0 9 12 VSS 9.3 9 12 COD 5.6 87 41 BOD 3.0 18 48 NH 3 -N 1.8 10 25 P0 4\" 1.1 13 32 4.4.5.3 Standard curves Standard curves were made for calculating the results of COD, NH 3 -N, and P04\". The stock solutions used in building the curves were prepared according to the recommendations of Standard Methods (American Public Health Association et al, 1998). The three standard curves achieved a good fit to the standards, with and R 2 > 0.999 in all cases. For the calculation of solids (TSS and VSS) and BOD, the accuracy and precision of the calculations relied on the accuracy and precision of the instruments used (balance and DO probe/meter respectively), as described in Section 4.4.4. 4.4.6 Respirometric methods used The respirometric batch tests were performed in a typical respirometer (see Figure 3.2), continually stirred. The dissolved oxygen (DO) probe (Model YSI 5739, submergible) was connected through the DO meter (YSI Model 52) to a computer, where the DO concentrations and temperatures were recorded. The accuracies of the DO probe and DO meter were 0.01 g 3 3 m\" and 0.2 g m\" respectively. The equipment was calibrated at the beginning of each day of use. 88 The oxygen uptake rate (OUR) was calculated as the slope of a linear regression (least square method) of the DO concentrations. Nitrogen in the form of ammonia was added at concentration of 10 g m\"3 NH3 -N. This concentration was enough to supply the nitrogen requirements to all respirometric tests, based on a ratio BOD 7 :N of 100:3, as proposed by Franta et al. (1994a) when working with pulp and paper effluents. Even lower nitrogen requirements have been found in the literature for pulp and paper wastewaters (i.e. BOD 7 :N of 100:1 as presented in Vaananen (1988), as cited in Dubeski, 1993). The nutrient source used was the mill nutrient stock solution, which corresponds to a standard solution of Urea Ammonium Nitrate (nutrient code 28-0-0, based on NPK composition). All the tests were performed with the addition of 10 g m\"3 of nitrification inhibitor (Hach, formula 2533). Two types of batch respirometric tests were performed, which can be classified as depending on whether wastewater was, or was not, added to the sludge in an endogenous respiration state. Three tests were performed without the addition of wastewater in order to estimate the endogenous respiration decay of the sludge. The sample of sludge was aerated for several days, while measuring the OUR twice a day. These two tests were run at different temperatures. Twelve tests were performed mixing wastewater with sludge, at different F/M ratios and at different temperatures. The OUR was monitored approximately every 7 minutes over a period of two to six hours. Different relative volumes of wastewater and sludge controlled the F/M ratios of the batch tests. Before starting each test, the sludge sample was aerated for several hours in order to achieve an endogenous state. The wastewater itself was aerated for several minutes, and the OUR, COD and COD s were measured for both wastewater and sludge. During the test, the OUR was monitored by periodically raising the DO to 6 g O2 m\"3, switching the air off and recording the decrease in DO until the DO reached 4 g O2 m\"3, at which point the air was switched on again and the cycle repeated (similar to that proposed by Randall et al, 1991). COD and CODs were measured again at the end of the test. The wastewater samples were collected at the primary clarifier outlet, and the sludge samples used for respirometric measurements were collected from aeration basins ' IA ' and '4' of the plant. These tests may be subsequently classified into three groups, depending on the F/M ratio used. 89 - Five \"very low F/M\" tests were performed at F/M ratios in the range of 0.01 to 0.02 (COD basis). In these tests, volumes in the range of 70 - 75 mL of wastewater were added to 2000 - 2500 mL of sludge in an endogenous respiration state. - Five \"low F/M\" tests were performed at F/M ratios in the range of 0.6 to 0.9 (COD basis). In these tests, volumes in the range of 400 to 500 mL of sludge in endogenous respiration state were added to 1500 - 2000 mL of wastewater. These tests are in the range typically considered low F/M (< 1 - 2) in order to prevent mixed culture microorganisms from substantial multiplication (Chudoba et al, 1992). Two \"high F/M\" tests were performed at F/M ratios around 5 (COD basis). In these tests, volumes around 2 mL of sludge were added to 2000 mL of filtered wastewater. The temperature was controlled in all respirometric tests by circulating water using a thermostat water bath connected to the batch reactor. The temperature was kept constant during all tests within a range of + 2°C. A detailed description of each test and its condition is available in Appendix 4. Pictures of the respirometers used as well as the temperature controlling system are available in Appendix 8. 4.4.7 Data from historical database One-year of data were obtained from the mill historical database, corresponding to the complete 2002 calendar year. A frequency of three data points per day was generated from the database; however, the extracted data corresponded to averaged data of those parameters collected at a higher frequency (i.e. influent flow measured on-line available at frequencies as high as every thirty seconds); and to repeated raw data for those available at a lower frequency (i.e. those measured twice a day). Therefore, the total number of available data per measurement varied from, 1100 for those measured on-line, to 150 for BOD, which was the parameter collected with the lowest frequency. Table 4.5 presents the parameters obtained from the historical database, the original 90 frequency, the obtained frequency, and the type of measurement (grab, 24 hours average, or on-line). Table 4.5: Parameters obtained from mill historical database Parameter Location Original Obtained Type Frequency frequency COD Influent 2 per day 2 per day Grab Influent 1 per day 1 per day 24 hours average Effluent 2 per day 2 per day Grab Flows Influent > 1 min\"1 3 per day On-line Recycle 2 per hr 3 per day On-line Wastage 2 per hr 3 per day On-line Nutrient addition > 1 min\"1 3 per day On-line TSS Influent 2 per day 3 per day Grab Influent 9 per hr 3 per day On-line Effluent 2 per day 2 per day Grab Effluent 1 per day 1 per day 24 hours average Cell 4 2 per day 2 per day Grab Recycle 2 per day 2 per day Grab Wastage 2 per day 2 per day Grab VSS Cell 4 2 per day 2 per day Grab Temperature All five cells 1 per hr 3 per day On-line Dissolved oxygen All five cells > 1 min\"1 3 per day On-line BOD 5 Influent 3 per week 3 per week Grab Effluent 3 per week 3 per week Grab pH Influent 20 per hr 3 per day On-line Effluent 20 per hr 3 per day On-line OUR Cell 4 1 per day 1 per day Grab A data screening was performed in order to discard some data or regions of data not representative of the process (i.e. five days data were deleted due to a mill shut down during August 2002). Data visualization along with outlier tests were performed at this stage. Al l the data used is available in a floppy disk attached to this thesis, as presented in Appendix 6. 91 4.5 Model parameter estimation methods As presented in Section 4.4, three approaches were used in order to estimate different model parameters: off-line analytical techniques, respirometry, and mass balances using historical data. Some parameters were estimated with more than one method in order to check different methodologies and to assess which method was most appropriate for the case under study. The parameter estimation based on fitting respirometric measurements was performed using an iterative approach. Therefore, a first estimation of these parameters was done based on some assumed parameters (i.e. temperature coefficients), which were replaced with estimated values later, and used in a secondary calibration iteration. 4.5.1 Parameters estimated by analytical measurements 4.5.1.1 Observed heterotrophic yield (YQRS) The observed heterotrophic yield (YOBS) was calculated based on the changes on the soluble and particulate COD measurements during a batch test with a high F/M ratio, in which an aliquot of sludge was introduced to a batch reactor with filtered wastewater. The change in the particulate COD was assumed to correspond to cell growth, and the change in soluble COD was assumed to correspond to the consumption of readily biodegradable COD. Y O B S was finally estimated using Equation 4.4. Y O B S = A part COD / A CODs [4.4] 4.5.1.2 Soluble inert influent COD fraction (fyj^) The soluble inert COD (Si) present in the influent was assumed to be equal to the soluble COD effluent (CODs>eff)- Therefore, the influent Si fraction (fsi,in) was calculated by dividing the CODs.eff by the total influent COD of the same sampling sequence (corresponding to the influent sample taken 16 hours before). 4.5.1.3 Readily biodegradable COD fraction Cfe,„) Two analytical techniques were used to estimate the readily biodegradable COD fraction. 92 The first method was based on the difference between the soluble COD at the beginning and at the end of a \"low F/M\" batch test. It is assumed that all the soluble COD consumed during the test corresponds to Ss. Then, fSs,,n is estimated using Equation 4.5. fS S )in = ACODs/COD i n [4.5] The second method used for estimating fSs was that proposed by Mamais et. al. (1993), which assumes that the Ss fraction corresponds to the truly soluble fraction of the influent minus the truly soluble fraction of the effluent. The truly soluble fraction was obtained by coagulation, flocculation and filtration, as described in Section 4.4.4.5. Equation 4.6 was used for estimating fss under this method, being CODS(tr) the truly soluble COD. fS ,in = (CODsCrXin - CODS(,r),eff) / COD i n [4.6] 4.5.1.4 Slowly biodegradable COD fraction (fxsjJ Analytically, the slowly biodegradable COD fraction present in the influent (fxs.in) was estimated from the difference between the biodegradable COD (BCOD) and the readily biodegradable COD and the of the sample. The BCOD was estimated using Equation 3.19, which corresponds to the ultimate BOD (UBOD) times a factor (fBoo) that accounts for the amount of COD that is transformed to inerts during the test. The factor f BoD can be obtained by multiplying the values of fxi and YOBS, as presented in Equation 3.20. Finally, the influent Xs fraction corresponds to the influent biodegradable COD minus the influent readily biodegradable COD, divided by the total influent COD, as presented in Equation 4.7. fxS.in = (BCOD in - SS,in) / COD i n [4.7] 4.5.1.5 Heterotroph ic biomass (XH) The heterotrophic biomass (XH) concentration was estimated to be equal to the VSS concentration of the sample. It is important to point out that it was assumed that the concentration of VSS in the sample was equal to the biomass concentration measured as COD units. This assumption is based on a TSS and VSS balance using the solids-to-COD coefficients proposed in ASM3. The complete derivation is as follows: 93 TSS and VSS may be calculated using Equations 4.8 and 4.9, being irss,xj and ivss,xj the TSS and VSS to COD coefficient for the particulate component Xj. TSS (g TSS m\"J) - iiss.BM • XRM+ iTss,xsto ' XSTO + iTss,xs' Xs + i T s s , x i ' Xi [4.8] -3 VSS (g VSS m\") = ivss.BM • XBM + ivss,xsto \" XSTO + ivss.xs • Xs + ivss,xi' Xi [4.9] Where XBM is the biomass component, assumed to be XH only. Subtracting Equation 4.9 from Equation 4.8, and replacing the proposed values for the parameters ITSS,XJ and ivss,xj, Equation 4.10 is obtained. TSS - VSS (g SS m\"3) = (0.9 - 0.75) • X H (g C O D x h rn3) [4.10] Assuming TSS and VSS are related as presented in Equation 4.11, being 'a' a constant for the activated sludge system, and replacing Equation 4.11 in Equation 4.10, Equation 4.12 is obtained. a TSS = VSS [4.11] X H (g C O D x h m\"3) = (a - 1) / 0.15 • VSS (g SS m\"3) [4.12] In conclusion, X H , measured in COD units, is directly proportional to VSS, measured in solids units. Further, the measuring campaign indicated that the coefficient a had a constant value of around 1.15 for the plant under study, so Equation 4.12 results finally in Equation 4.13. X H (g CODxh m\"3) = VSS (g SS m\"3) [4.13] In other words, it is possible to conclude that, in this WWTP, the amount of Xs and Xi present in the VSS parameter is equivalent to the fixed solids of XH present in the TSS parameter. 4.5.1.6 Endogenous respiration rate coefficient (b») The endogenous respiration rate coefficient (bH) was estimated based on the decrease on the particulate COD in a batch test during several days. It was assumed that the decrease in the particulate COD was due to a decrease of biomass, compensated by the generation of Xi during the decay. The value of bn can be obtained from the slope of the plot of the logarithms of X H 94 over time. Since the relative amounts of X H and Xi change with time in a dynamic way, the variation of X H was estimated from the decrease of the particulate COD using Equation 4.14. Therefore, XH(T.) is obtained from Equation 4.15. [4.14] A XH(t) = A part COD(t) / (1 - fXI), [4.15] XH(t) = X H(t-l) + AX H (t) 4.5.2 Parameters estimated by respirometry 4.5.2.1 Storage yield (Ysrn) The coefficient YSTO was estimated using the method proposed by Karahan-Gul et al. (2002a), using Equation 3.25. The term JOURSTO\" dt refers to the oxygen used during the storage process, which can be graphically represented by the area under the curve shown in Figure 3.3. Analytically, the area between time i (ti) and time j (tj) was calculated using the trapezium equation, as presented in Equation 4.16. Only the storage OUR (OURSTO) was used for this purpose, so the OURs corresponding to the endogenous respiration and to the growth processes were subtracted from the total (measured) O U R . IOURSTO- d t = (ti - tj) • ( O U R S T o , i + OURSTOJ) / 2 [4.16] The OUR corresponding to the endogenous respiration process was assumed equal to the OUR of the sludge before adding the wastewater, corrected by the volume dilution. The OUR corresponding to the growth process was assumed to be equal to the diagonal between the initial time and the point corresponding to the breakpoint B of the respirograph (as illustrated in Figure 3.3). The tests were performed with a low F/M ratio (between 0.5 and 0.9) in order to obtain a sufficient number of OUR measurements during the storage stage to minimize the error in the calculation of YSTO, as was suggested by Karahan-Gul et al. (2002a), but lower than 2 to prevent mixed culture microorganisms from undergoing substantial multiplication (Chudoba et al, 1992). 95 4.5.2.2 Observed heterotrophic yield (YH) The coefficient YH was estimated based on the observed yield coefficient (YOBS), using Equation 3.26. The value of YOBS can be determined also by respirometry using Equation 3.24. The tests to determine Y O B s were performed at very low F/M ratios (0.01 - 0.02) in order to ensure that all XSTO was transformed into XH during the test. The amount of oxygen utilized for growth and storage (observed) is shown graphically in Figure 4.3. Similarly to the calculation for estimating YSTO, the area under the observed O U R (OUROBS, also called exogenous O U R ) curve was calculated using the trapezium equation (Equation 4.16). The OUROBS, that is, the O U R generated due to the storage and growth processes only, was calculated subtracting, from the total O U R , the O U R corresponding to the endogenous respiration process. The amount of Ss consumed during the \"very low F/M\" tests was calculated using Equation 3.18, using the value of YSTO calculated previously. T e s t 5-2 ( T = 4 0 . 7 ° C ) Observed 0 25 50 75 Time (min) Figure 4.3: The shaded area was assumed to represent the oxygen consumed by the \"observed\" growth process during a very low F/M batch test 96 4.5.2.3 Readily biodegradable COD fraction (fc) The readily biodegradable COD (Ss) consumed during a batch test is estimated by respirometry using Equation 3.18. The Ss fraction (fssjn) is estimated by dividing the Ss consumed by the initial COD, and compensating by the dilution factor, as presented in Equation 4.17. Note that this estimation requires the value of the parameter Y O B S -Where Vol si udg e is the volume of the sludge and V o l w w is the volume of wastewater used in the batch test. 4.5.2.4 Endogenous respiration rate coefficient (bn) The endogenous respiration rate coefficient (bn) was estimated in two ways, using different respirometric approaches. The first approach was the traditional way of estimating bH. Assuming the decay rate as a first-order reaction, bH was taken as the slope of the curve from a plot of the logarithm of OUR versus time (Sollfrank and Gujer, 1991) during several days in a batch test. The second method for estimating bH was based on the specific OUR (SOUR, which is the OUR per unit of biomass) of sludge in an endogenous respiration state. This method is based on the fact that, in absence of external substrate, and assuming fxi constant, the OUR of sludge in an endogenous state depends on b H and X H only (Equation 4.18). Then, dividing both sides of Equation 4.18 by X H (Equation 4.19), it can be concluded that SOUR depends only on the value of bH- Therefore, bn can be estimated from the values of SOUR of different measurements of sludge in an endogenous respiration state, corrected by the effect of temperature. (Ss consumed / COD i n) • (Vol siUdge + Vol w w ) / Vol WW [4.17] OUR = (l-fXi)- b H - X H [4.18] SOUR = (l-fXi)- b H [4.19] 97 This second approach has not been found in the literature, so it may be considered a novel method for the estimation of bn. Since this method can use the information generated by measurements performed to estimate other kinetic parameters, the author believes that it is simpler and requires less work than the traditional method using long-term batch tests. Besides, this method also allows the estimation of the temperature coefficient of bn (6T,I>H), as described in the next section. 4.5.2.5 Temperature coefficient ofbft (Orj^) The temperature coefficient of bH (OT^H) was estimated in two different ways, both using respirometry. The first method consisted of a rapid increase in the temperature of sludge in the endogenous state, so the resulting increase in the OUR was proportional to the increase of b^ This method is based on the fact that, assuming XH and fxi are constant during the test, the OUR of sludge in an endogenous state depends only on bn (Equation 4.18). Therefore, an increase in the OUR during an increase of the temperature may allow for the calculation of the parameter 0x,bH. Assuming bn varies with temperature according to the Arrhenius equation (Equation 3.7), 0i,bH can be obtained as the slope of the curve that plots the logarithms of OUR versus temperature during the increase in temperature. The second method for estimating 9T,bh was based on the second approach for estimating bn, using the SOUR of sludge in the endogenous respiration state. Assuming bH varies with temperature according to the Arrhenius equation (Equation 3.7), 9x,bH can be obtained as the slope of the curve that plots the logarithms of the SOUR of different batch tests versus temperature. These two methods have not been found in the literature, so they may be considered novel methods for the estimation of Gj.bh- - The author believes that both, depending on the circumstances, are simpler than the traditional approach of making a series of long-term tests at different temperatures. 98 4.5.2.6 Oxygen saturation constant (Km) The oxygen saturation constant (K02) was estimated using the method proposed by Kappeler and Gujer (1992), measuring the O U R continuously while the oxygen concentration decreases from above 4 g m\"3 to a value close to zero. It is assumed that the decrease in the O U R is due to the lack of oxygen as well as due to the decrease in the substrate (Ss) concentration. The effect of the decrease on Ss concentration is estimated by comparing the O U R at high oxygen concentration at the beginning and just after the test. Therefore, the theoretical maximum O U R ( O U R MAX ) is calculated by interpolation between the initial and final O U R . Then, the measured switch function is calculated by dividing the measured O U R at low DO and the O U R M A X -Finally, K02 is estimated by curve fitting the switch function model (Equation 4.20) to the measured switch function values. Switch Function (02) = S 02/(K 0 2 + S02) [4.20] 4.5.2.7 Curve fitting for estimation ofkvrn, jig, Ksm, and UH The values of ksTo, PH , K S T O , and k H were estimated based on the analysis of different respirographs (respirometric curves) from batch tests, at different temperatures and at different F/M ratios. The curve fitting of the OUR equation to the measured OUR was performed using MATLAB®, with the least squares method, numerically minimized using the secant algorithm (Ralston and Jennrich, 1978). During the curve fitting, all other parameters were maintained constant at their calibrated or assumed values. 4.5.2.8 Observed maximum growth rate (JUQRS) The observed maximum growth rate (POBS) was estimated from the slope of the curve plotting the logarithm of the OUR in a test with a high F/M. Since the slope of that curve is actually an estimation of (POBS - bH), POBS was finally estimated subtracting the value of bH from the value of that slope. It was assumed that the growth was not limited; so all substrate, oxygen and nutrients were in excess. It was also assumed that the maximum growth rate of the microorganisms that grow in this test (sometimes called 'fast growers') was representative of the maximum growth rate of the original microbial culture. This parameter is not part of the 99 ASM3 simplified model, so its estimation was used as an auxiliary tool for the estimation of PH-4.5.3 Parameters estimated by mass balances 4.5.3.1 Particulate inert COD fraction (fyrj^) The X] fraction (fxi.in) was estimated in two different ways: independently using a full-scale sludge production mass balance, and using the other C O D j n fraction estimations to perform an C O D j n balance. The first method for estimating fxi,in was performed by minimizing the error sum of squares when predicted sludge production rates were compared to measured rates as a function of the solids retention time (SRT), as proposed by ASM1. The full-scale data were obtained from the WWTP historical database. The SRT was calculated based on Equation 4.21, the estimated sludge production was calculated using Equation 4.22, and the measured sludge production rate was calculated using Equation 4.23. Equation 4.22 accounts for the sludge produced due to heterotrophic biomass growth, production of particulate inerts during endogenous respiration, and incoming particulate inerts. SRT = V o l - X / ( Q w a s t e - X w a s t e + Qefr X e f f ) [4.21] Px,estimated = Qin- [X I j i n + (l+fxiv bH) • Y 0 B S - (COD i n-S,, i n-Xi, i n)/(l+bH- SRT)] [4.22] Px ,measured Qwaste' Xwaste [4.23] Where Vol is the total bioreactor volume (21,000 m ), X is the solids concentration (g TSS m~) and Px is the sludge production rate (g d\"1). The second method, using an influent C O D balance, requires independent estimations of the other CODin fractions (fss.in, fsi.in, and fxs.in). Equation 4.24 was used for this estimation. X i > i n = C O D I N - S s , i „ - S , , i „ - X s , i „ [4.24] 100 4.5.3.2 Slowly biodegradable COD fraction (fys) The Xs fraction (fxs.in) was estimated by an influent COD mass balance, using Equation 4.24, rearranged in form of Equation 4.25. This method requires,independent estimations of the other CODin fractions (f s s, fsi, and fXi). Xs,in = C O D i n - Ss,in ~ S i , i n - X^jn [4.25] Note that both, X i and Xs may be estimated independently, however, the total CODi„ balance is necessary in order to adjust any estimation error. Otherwise it is possible to end up with more or less COD than that actually coming from the influent. 4.5.4 Methods used for model assumption verification The objective of the assumptions is primarily to simplify the model structure and calibration, and even though their verification is not required (in fact, they were assumed as true), to check their validity is always desired because that will reduce the uncertainties of the model. Since verifying all the assumptions was not possible, only a few assumptions were checked a priori, independently of the experimental results. Other assumptions were verified during the calibration procedure, which are discussed together with the results. 4.5.4.1 Absence of volatile compounds in the influent The absence of volatile compounds in the influent was tested by aerating an unfiltered sample of influent for 4 hours, and measuring the total and soluble COD every hour. Beyond this time, corresponding to 50% of the HRT of the bioreactors, almost all the organic matter would have been biodegraded, so the effect of volatilization would be negligible. The addition of a biocide (phenol mercuric acetate) to the wastewater avoided interference from the possible presence of bacteria in the influent. 4.5.4.2 TSS to VSS relation As presented in Section 4.5.1.5, a constant ratio TSS/VSS equal to 1.15 is an important assumption for the estimation of the heterotrophic biomass (XH), in COD units, as the VSS, 101 measured in g m\"3. For checking this assumption, the TSS/VSS ratio was calculated seven times in four different plant locations (Cell 1 A, Cell 4, return sludge, and wastage sludge). 4.5.4.3 pH close to neutrality The range of pH was checked using visualization and statistical techniques over the data from the historical database. The data were tested for outliers, and the average and a 95% confidence interval were calculated after. In addition, alkalinity was also measured in the WWTP, and the switch function for alkalinity was calculated. The value of the saturation constant for HCO3\" (K ALK) was assumed equal to the default of ASM3. The switch function, as presented in Equation 4.26, is a generalization of the Monod Equation, whose value decreases when the concentration of soluble substrate (S) drops close to zero, and tends to a value of one when S increases. S.F. = Si/(K s i + SO [4.26] Where S.F. is the switch function value, Si is any soluble substrate or nutrient (i.e. alkalinity, ammonia or ortho-phosphate) and Ks; is the saturation constant for Sj. 4.5.4.4 Generation of Sr during endogenous respiration Even though ASM3 does not consider the generation of Si during the endogenous respiration process, the active debate around the generation of SMP makes it desirable to check for this possibility. The generation of soluble inert COD (Si) during endogenous respiration was measured during a long-term batch test, in which sludge was aerated without the addition of substrate. The soluble COD (COD s) was measured periodically, so an increase in COD s would indicate the generation of Si. 102 4.5.4.5 Nutrients are not limiting for heterotrophic bacteria Nitrogen and phosphorus are added in the Port Alberni WWTP to ensure they are not limiting factors for heterotrophic bacteria, therefore, the assumption that both are non-limiting may be considered reasonable. However, in order to verify the actual nutrient concentration in the WWTP, ammonia and ortho-phosphate were measured seven times in Cells 1A and 4. The availability of nutrients was evaluated calculating the switch functions for ammonia and ortho-phosphate for the lower range of conditions found at the WWTP. The values of the heterotrophic saturation constants for ammonia and ortho-phosphate were assumed from values available in the literature (Gujer et al., 1999; Rieger et al, 2001 respectively). 4.5.4.6 Assumptions checked with the results Other assumptions, checked with the model calibration results, were: - the presence of only one readily biodegradable substrate fraction; and the value of the K02 is the same for growth/storage and for endogenous respiration. 4.5.5 Statistical methods The uncertainty of the parameter estimates was calculated at a level of confidence of 95%, based on the standard deviation (a) of a normal distribution (mean + 1.645*o~). The values outside the range [mean + 3*a] were checked for outliers. The outlier verification consisted mainly of visualization techniques with more than one time series at a time. No outliers were deleted based only on statistical analysis. 103 5. Resul t s a n d d i s c u s s i o n : Cal ibrat ion of A S M 3 for Port A l b e r n i act ivated s l u d g e s y s t e m 5.1 Sensitivity analysis The sensitivity of each state variable to each model parameter was calculated by evaluating the variation in the state variable, in every plant compartment, after a 1% increase in the parameter. The total sensitivity of each parameter was computed as the sum of the sensitivities in each of the five plant compartments. Table 5.1 presents the results of the sensitivity analysis performed with the assumed parameter values. Empty spaces in the table represent zero values. For checking purposes, the sensitivity of the model parameters was examined again after the calibration procedure was finished. Table 5.2 presents the results of the sensitivity analysis performed with the calibrated parameters values (from Table 5.19). Most of the parameters maintained their overall sensitivity, however, there are some parameters that become more or less sensitive with their calibrated values. The three parameters associated with the endogenous respiration, bn, fxi, and O-r.bH, are less sensitive with the calibrated values, even though fxi did not change its value. This decrease may be mainly explained because the absolute value of the calibrated bH is much lower than the default value (0.13 instead of 0.46 @ 30°C), which reduces the impact in the generation of Xi during endogenous respiration. On the other hand, the two parameters associated with heterotrophic growth, PH , and 0T,^H, increased their sensitivity. This increase may be explained due to the increase in the value of both parameters (from 4.6 to 13, and from 0.07 to 0.08 respectively). Note that the increase in the sensitivity of pn, cannot be observed in Tables 5.1 and 5.2, because it is high in both, but it can be checked in Appendix 2 if required. 104 o X X X X C/2 C/3 CO l-l OJ -*-» • 8 3 S s (A X Q O U 60 X Q O U 60 X Q O U 60 Q O O 60 Q O U 60 Q O U 60 IO 60 OJ > CJ e o > •a o o •A .2 > o o o o o o o s •3 e 3 •3 OJ a 40 O H 3C T3 e a s s s o > o > > oo o Tt - H ' X Q o u 60 E a •3 (U o cu > s 3 •3 oj O O 60 s 3 •3 •a 43 60 43 60 43 60 S S o > •a 43 60 43 JS JS 60 60 60 O & -js > T3 (N O T3 CU 4= 60 oo JS 60 43 60 CO S 3 -3 E o o OJ > o cj > o > c3 * > o CJ > £ £ £ o o o & b Tt o CJ > o o CJ > o in Q O U 60 T3 O H X Q O x Q O 60 ao X Q O u 60 Q O u 60 X Q O O 60 X Q O U 50 o f- ac C D O H O d> o X X Q O O 60 O > 43 60 60 -a 60 42 60 42 60 42 60 42 60 4= © > , 1 X X Q O U 60 o 42 60 42 60 42 •a •SP 11 •a £ > Q O U 60 o o o X Q O U 60 CO Q O O 60 > > 42 42 42 g g 60 60 60 — -2 42 42 42 49 42 & 42 O > CO Q O U 60 O o > > IO 60 O > o > 42 ^ 42 60 60 42 60 42 60 SP 42 42 60 o § o (2 > > ro O 8 - © ,_; 00 —< o ^ o as r» P o C3 Q o IT) o o 1/1 l-l X Q O U 60 o 60 Q O U 60 X Q O x Q O 60 60 X Q O 00 Q O u o 60 60 X Q O U 60 X P o u 60 o H C D C D C D H C D Finally, the increase in the value of K02 from 0.2 to 0.3 increased the sensitivity of this parameter substantively, because this parameter affects the rates of storage, growth and endogenous respiration in the range of oxygen concentrations that is present in the plant compartments. Some conclusions obtained from this sensitivity analysis are the following. The state variables Si, Xj and Xs are relatively insensitive to all parameters. Therefore, the estimation of the wastewater initial fractions of these components has little effect on the overall calibration procedure. The modelled oxygen concentration is especially sensitive to the parameters K02, ksTO, K S T O , M-H, Y S T O , and Y H - Therefore, these are the parameters that are most suitable for estimation by respirometry. - In contrast, the modelled oxygen concentration is very insensitive to the parameter Ks, so that this model parameter is not suitable for estimation by respirometry. - Al l state variables were relatively insensitive to the model parameters fxi and fsi, so the estimation of these parameters is not an easy task. - X S T O is the state variable that is sensitive to the higher number of parameters, which reflects the importance of this model component in the ASM3 structure. Appendix 2 presents the numerical values obtained from the sensitivity analyses, using assumed values and using calibrated values, as well as the criteria for classifying the values into high, medium, low, and very low. 5.2 Identifiability analysis 5.2.1 Without hydrolysis For evaluating the structural identifiability analysis, using the Taylor series expansion method for OUR around time equal to zero, several derivatives of the OUR were required (Equation 5.1). In order to simplify that task, the biomass (X H , here X for simplicity) was assumed to be constant, and the initial stored substrate was assumed to be zero (XSTO (0) = 0). It is necessary to point out that only the exogenous OUR (Equation 5.2) is included in this analysis, which is referred as OUR only for simplicity. These assumptions are reasonable for the kind of 107 respirometric experiments performed, and where this analysis is used. In addition, for simplicity, the hydrolysis process was not included in the analysis (Xs = 0) in a first stage. OUR (t) = OUR (0) +1 • dOUR/dt (0) +12/2! • d2OUR/dt2 (0) + ... [5.1] OUR = ( l -Y STO) \" ksTo\" S/(Ks+S) • X - p H - ( 1 - Y H ) / Y H - (XSTO/X ) / (KSTO ' X [5.2] +XSTO) In the ASM3 simplified model, the OUR can be written as in Equation 5.2, where the only two parameters that vary are the substrate (Ss, here S for simplicity) and X S T O , SO their derivatives are required (Equations 5.3 and 5.4 respectively). dS/dt = -k S T 0 - S/(KS+S)- X [5.3] dXsTo/dt = Y S T o - ksro- S/(KS+S)- X - p H / Y H - (X S T O /X) / (K S T o • X [5.4] +XSTO) Computing the first three derivatives of OUR (t) at t = 0, Equations 5.5, 5.6, and 5.7 are obtained. OUR (0) = (1-YSTO) ksro X S 0/(KS + S0) [5.5] dOUR/dt (0) = - (1-YSTO) ksro2 X 2 K s S0/(KS+S0)3 + ... t5-6] (I-YH)A^H ' PH X Y S T O ksTo So / (KSTO (Ks+So)) d2OUR/dt2 (0) = (1-YSTO) k S T O 3 X 3 K s S0 (Ks-2- S0) / (K s+S0)5 + ... [ 5 - 7 ] (1-YH)/YH- pHXYsToksToSo/(KSTo(Ks+So)- ... { ksio K s X/(Ks+So)2 + 2YSTO k S To S0/((KS+S0) K S T 0 ) + P H / ( Y h K S T 0 ) } Where So is equal to S (0), the concentration of Ss at t = 0. Grouping the model parameters into six parameters 6; (i = 1 to 6), Equations 5.8 to 5.13 are obtained. 108 ei=ksToX(l-Y S To) t 5- 8] e2 = (l-YsTo)S0 [5-9] e3 = ( l -Y S To)Ks [5-10] e4 = ( l -Y H ) /Y H u„X [5.11] e5 = ( l-Y H )KsToX [5.12] 06 = (I -YH ) Y S T O / (1-YSTO) [5.13] Replacing parameters 0j into Equations 5.8 to 5.13, Equations 5.14 to 5.16 are obtained. OUR (O) = 0 , 0 2 / ( 0 2 + 03) douR/dt(O) = e1202e3/(02+03)3 + 0i02e4e6/((02+03)- e5) d2OUR/dt2 (0) = 0,30203(03-202)/(02+03)3 + . . . 01020406/((92+03)05)- {0ie3/(02+03)3 + 2 0,0206/((02+03)05) + 64/65} Higher order derivatives will give the same combinations of parameters, so it is possible to conclude that only these six 'combinations' of model parameters (0js) are theoretically (or structurally) identifiable through respirometry. However, in order to calculate the six 0(s, at least six equations are needed, which would require accurate calculations for five derivatives of the OUR. It is easy to realize that uncertainty in the analysis and experimental errors does not allow the measurement of the OUR with such degree of detail. Therefore, for practical purposes, it is possible to assume that no more than three or four combinations of parameters can be estimated from a single respirograph, and the remaining parameters have to be assumed or estimated in an alternative way. Some practical implications of these results are the following. - Model parameters that were not identifiable by fitting the exogenous OUR curve, because they were not present in any 0j, were Ko 2 , bH, bsTo, fxi, and fsi-109 [5.14] [5.15] [5.16] - Y H and Y S T O can only be estimated together using respirometry if all six 0jS are estimated and all other parameter values and wastewater characteristics are known, which is impossible in practice. This is due to the fact that these parameters were present in all Equations 5.8 to 5.13. Therefore, it was not possible to estimate both yield together based on respirometry. - Further, the estimation of any yield coefficient ( Y O B S or Y H ) would require the estimation of at least three 0jS and knowledge of the vajues of three additional parameters or wastewater COD fractions, which was not convenient in a parameter estimation procedure. Therefore, it can be concluded that it is preferable to estimate both yield coefficients in an alternative way instead of using curve fitting of the modelled OUR. - The third model component that was not likely to be estimated by curve fitting of the OUR equation was the biomass concentration (X in Equations 5.8, 5.11 and 5.13), precisely because it appeared in three 0,s, requiring three times more information than other model parameters that appeared in only one 0. - Finally, assuming that the parameters Y O B S , Y h , and X are known, all the remaining model parameters (ksTO, So, Ks, |J.H, and K S T O ) can be theoretically identifiable based on curve fitting to the OUR curve. The selection of a subset of these parameters may take into account other factors, such as the results of the sensitivity analysis or the availability of alternative estimation methods. 5.2.2 Incorporating hydrolysis If hydrolysis is incorporated in the analysis, the derivative of S (Equation 5.3) has to be modified as presented in Equation 5.17. If a similar analysis is performed under this new condition, three additional groups of parameters become identifiable ( 0 7 , 0g, and 09, as presented in Equations 5.18, 5.19, and 5.20), which incorporates the three model parameters related to the hydrolysis process: kn, Kx, and the initial concentration of Xs (Xs,o). 110 dS/dt = - ksxo • S/(KS+S)- X + k H X s X/(K X X + X s ) [5.17] e7 = ( l - Y S T o ) k H X [5.18] ©8 - (1-YSTO) Xs,o [5.19] 89 - (1-YSTO) Kx X [5.20] In this case, for the same reasons as before, YSTO and X were the parameter and wastewater characteristic with major identifiability problems. Therefore, assuming that Y S TO and X are known, the parameters k H and Kx become identifiable as well as the wastewater characteristic Xs,o. However, the total number of identifiable GjS from a single respirograph remains the same (three or four). 5.2.3 Restrictions to identifiability due to sensitivity of parameters Using a larger number of respirographs in the estimation procedure, would add degrees of freedom to the system, allowing the estimation of some additional parameters. Nevertheless, the estimation of extra model parameters by respirometry is limited by the sensitivity of the parameters to the oxygen concentration. Then, the sensitivity analysis is important in order to determinate the 'candidate' parameters to be estimated using respirometry, and in discarding those that are not sensitive enough to the oxygen concentration to be estimated by respirometry. In practice, this estimation is done by minimizing the errors while fitting the model to the OUR curve obtained in a batch test (curve-fitting). The method used for this procedure was least squares, numerically minimized using the secant algorithm (Ralston and Jennrich, 1978). 5.2.4 Selection of identifiable parameters to be estimated by curve fitting The sensitivity analysis indicated that the parameters that are most suitable for estimation by respirometry are K02, ksTo, KSTO, PH, YSTO, and Y H . From those, K02 and one yield (YSTO or Y H ) are not structurally identifiable, so only four of those six parameters are theoretically identifiable using curve fitting to respirometric curves. I l l However, only three of these four 'candidates' were finally chosen to be estimated by curve fitting: ICSTO, K S T O , and un, leaving both yield coefficients to be estimated independently. Despite the low sensitivity of the oxygen concentration to the hydrolysis rate (kH), this parameter was also selected to be estimated using curve fitting because there are no simple alternative methods for estimating this parameter. A large number of respirographs used in the calibration procedure (ten), fitted independently, helped to reduce the uncertainty of the parameter estimations. In summary, only four kinetic parameters were chosen to be estimated using curve fitting, mainly based on the sensitivity and structural identifiability analyses: ksTO, UH, K S T O , and kH. Al l other model parameters were either assumed or estimated in an alternative way. 5.2.5 Temperature coefficients Since three out of the four model parameters to be estimated with the aid of curve-fitting, are temperature-dependent (kSTo, U H , and kH), and because the temperature in the WWTP is not constant, it was desirable to estimate the temperature coefficients (0TS) for those parameters. Considering that three estimations of each parameter are required for estimating its 9j, they can be estimated if enough respirograms are used during the parameter estimation procedure, adding sufficient degrees of freedom. Therefore, in order to allow the determination of the four kinetic parameters plus the three temperature coefficients, ten respirographs were used in the calibration procedure, which were the result of tests performed at different temperatures and under different F/M ratios. 5.3 Verification of some model assumptions Before estimating the model parameters, some of the assumptions of the model were tested in order to verify their validity and, therefore, reduce the uncertainty on the calibration procedure. Five assumptions were tested, corresponding to those that can be easily verified using analytical techniques. They were: 112 - the absence of volatile COD in the influent; - the relationship of TSS to VSS (assumed to be 1.15); - pH is close to neutrality; - the absence of generation of Si during endogenous respiration; and - nutrients were not limiting for heterotrophic bacteria. Other assumptions were checked during the calibration of other parameters (i.e. verifying good fit in the curve fitting process). 5.3.1 Absence of volatile COD in the influent One test was performed by aerating unfiltered wastewater for 4 hours, and measuring COD and CODs periodically. The test was performed with the addition of a biocide (phenolic mercuric acetate) in order to prevent the consumption of substrate by microorganisms that could have been present in the wastewater. The results (presented in Table 5.3) indicated that there is negligible loss of COD during the first 3 hours of aeration (< 3%). The differences in COD observed during this period are in the range of the experimental error of COD (+ 4.8% according to Standard Methods (American Public Health Association et al, 1998)). Only during the fourth hour, does the COD lost seem to be more significant. Table 5.3: Evolution of COD and CODs during abundant aeration of wastewater Time (hr) COD CODs COD/CODini g COD m\"3 g COD m\"3 (%) 0 757 652 100% 0.5 - 662 -1 765 667 101% 2 745 645 98% 3 733 654 97% 4 711 620 94% Mean 742 650 98% 113 Since the major COD removal occurs in the fist cell (simulations indicate that more than 95% of the readily biodegradable COD is consumed in Cell 1 A), the effect of volatilization after the third hour have little effect on modelling the plant under study. Therefore, neglecting the effect of volatilization seems to be an appropriate assumption for modelling Port Alberni WWTP. 5.3.2 TSS to VSS relation The total suspended solids (TSS) and volatile suspended solids (VSS) were measured for all samples during the first measuring campaign. Table 5.4 presents the TSS/VSS ratio for samples from six sampling sequences taken from Cell IA, Cell 4, recycle sludge, and wastage sludge. The results show a very consistent value of the ratio TSS/VSS around 1.15 (VSS/TSS = 0.87) for the sludge from Cell IA, Cell 4, and wastage. The recycle sludge presented a ratio TSS/VSS around 1.10, consistently lower than the sludge from the other sampling points. The reason for this difference is unknown. However, it can be concluded that the value 1.15 is a very good assumption for the ratio TSS/VSS, especially considering the samples used for respirometry (where this assumption was used) were taken from Cells 1A and 4 exclusively. Table 5.4: TSS/VSS ratio of samples from different sampling points Sequence Cell IA Cell 4 Recycle Wastage 2 1.16 1.16 . 1.13 1.15 3 1.16 1.16 1.15 1.15 4 1.15 1.15 1.02 1.14 5 1.15 1.15 1.03 1.16 6 1.15 1.14 1.14 1.14 7 1.13 1.15 Mean 1.15 1.15 1.10 1.15 Std. Dev. 0.01 0.01 0.05 0.01 For additional information of TSS and VSS measurements, see Appendix 5 where are presented the results of all analytical measurements of the first sampling campaign. 114 5.3.3 pH close to neutrality Visualizing the data corresponding to influent pH during the year 2002 (Figure 5.1) it is easy to verify that influent pH is consistently between 6 and 8. Only three points were considered outliers (Feb 7 th, Oct 9 th, and Nov 5th), and the series was considered only representative until December 17th, so the points beyond that date were also eliminated. Table 5.5 presents the statistical information of the pH series described. 0 -| , , , , , C 1-Jan 2 - M a r 1-May 3 0 - J u n 2 9 - A u g 2 8 - O c t 2 7 - D e c Figure 5.1: Influent pH during 2002 Effluent pH presented a similar behaviour than influent pH, being consistently between 6 and 8, but with a mean slightly lower. Effluent pH statistical information is presented in Table 5.5. Therefore, it is possible to conclude that, with a probability higher than 95%, the pH along the WWTP is around neutrality (range 6.0 - 8.0), which can be considered reasonable. Table 5.5: Statistical information of pH series between January 1st until December 17th Indicator Influent Effluent Mean 7.07 6.52 Std. Dev. 0.39 0.23 95% confidence interval 6.44 - 7.70 6.14-6.90 Probability of range 6.5 - 7.5 0.79 0.53 In addition, the alkalinity concentration was measured six times at different locations along the WWTP (influent, Cell IA, and effluent). All measurements were close to 2 mole HCO3\" m\"3, ranging 1.94 - 2.54 mole HCO3\" m\"3 and averaging 2.12 mole HCO3\" m\"3 (see Table 5.6). The 115 95% confidence interval for the alkalinity was 1.7 - 2.5 mole H C 0 3 \" m\"3. Using the ASM3 default value for KALK (0.1 mole HCO3\" m\"3), the switch function for the lower end of the 95% confidence interval is equal to 0.94. This result indicates that the alkalinity can be neglected and, together with the pH information, suggests that pH is not only close to neutrality but also is not likely to change suddenly due to the stability that the alkalinity gives (buffer). Table 5.6: Alkalinity concentrations in Port Alberni WWTP Sample location Alkalinity mole HCO3\" m 3 Influent 1.84 Influent 2.14 Influent 2.08 Cell IA 2.14 Cell IA 2.54 Effluent 1.96 Mean 2.12 Standard deviation 0.24 5.3.4 No generation of S, during endogenous respiration The possible generation of soluble inert microbial products, that would be measured as Si, were evaluated in a long-term batch test (43 days), with periodic measurements of soluble COD. The results showed a quick decrease of the COD s during the first nine days, followed by a relatively constant concentration of CODs (Figure 5.2). The constant concentration of CODs at the end of the test may be assumed to correspond to the initial concentration of Si present in the sample rather than an accumulation of soluble inerts during the test. It can be discussed whether a constant concentration of CODs is due to an absence of generation of Si or whether it is due to a Si generation rate that is lower than the consumption rate of a very slowly biodegradable CODs fraction. However, this theoretical problem has little effect in practice, and it could be assumed that if CODs decreases continuously, the rate of generation of Si during endogenous respiration is negligible. 116 Therefore, it can be concluded that the absence of Si generation during the endogenous respiration process was a good assumption when modelling Port Alberni WWTP. 200 T 0 15 30 45 Time (days) Figure 5.2: Soluble COD in a batch test with sludge in endogenous respiration state 5.3.5 Nutrients are not limiting for heterotrophic bacteria In order to verify the nutrient concentration in the WWTP, ammonia and ortho-phosphate were measured seven times in Cells 1A and 4, and the results are presented in Table 5.7. Table 5.7: Ammonia and ortho-phosphate concentrations at Port Alberni WWTP Sequence Ammonia (g NH3 -N m\" ) Ortho-phosphate (g PO4 •-P m\"3) Cell IA Cell 4 Cell IA Cell 4 1 1.30 0.52 4.4 4.5 2 0.41 0.19 4.7 3.8 3 0.47 1.02 3.6 3.6 4 1.01 0.99 4.6 4.1 5 1.63 0.61 7.5 4.1 6 0.74 0.01 5.4 3.9 7 0.13 0.24 5.2 4.7 Mean 0.81 0.51 5.06 4.10 Standard deviation 0.53 0.39 1.22 0.39 117 The ammonia concentrations are relatively high (0.8 and 0.5 g m\"3 in Cells IA and 4 respectively), but with a large standard deviation. Therefore, only confidence intervals of 90% for the ammonia concentration can be calculated, which are presented in Table 5.8. On the other hand, the phosphorus concentrations are constantly high in both Cell 1A and Cell 4 (> 3.5 g m\"3 always), with a relatively low standard deviation. The confidence intervals of 95% for the ortho-phosphate concentration are presented in Table 5.8. In order to estimate the effect of the lowest nutrient concentrations expected (estimated as the lower end of the 90% confidence interval of the nutrient concentrations) on the microorganisms' kinetics, the value of the switch function was calculated for ammonia and for ortho-phosphate (see Equation 4.26). The value of the switch function represents a measure of the inhibition of the biological processes due to the low nutrient concentration. For example, a switch function of 1.00 indicates no inhibition at all, and a switch function of 0.50 indicates a 50% inhibition on the growth process, which is the only process affected by the nutrient concentrations. A more detailed explanation about switch functions can be found in Section 4.5.4.3. The ammonia switch function, applied to the lower end of the 90% confidence interval in Cells IA and 4, were 0.93 and 0.67 respectively. These results indicate that the ammonia concentration is typically not limiting in Cell IA, but may become limiting in Cell 4 (Table 5.8). This is the result of the stringent environmental regulations applied to Port Alberni pulp and paper mill, which limit the discharge of ammonia due to its toxicity to fish. Under those conditions, the addition of ammonia has been minimized to balance the requirements of bacterial growth, but also to limit the residual in the effluent. Table 5.8: Confidence intervals for ammonia and ortho-phosphate at Port Alberni WWTP Ammonia Ortho-phosphate Cell IA Cell 4 Cell IA Cell 4 ^ . ^ ^ ^ ^ \" 9 0 % \" 9 0 % \" 9 5 % 9 5 % Confidence interval (g m\"3) 0.14- 1.47 0.02 - 1.00 3.05 - 7.07 3.49 - 4.70 Saturation constant (g m'3) 0.01a 0.01a 0.01 l b 0.01 l b Minimum Switch function value 0.93 0.67 > 0.99 > 0.99 a ASM3 default value of saturation constant for SNH4 for heterotrophic organisms b Value of saturation constant for SPO4 for heterotrophic organisms (Rieger et al, 2001) 118 In opinion of the author, the low probability of ammonia limitation in Cell 4 is not sufficient reason for discarding the assumption that ammonia is not limiting for the heterotrophic bacteria. Adding ammonia, as an extra model component, would incorporate much more additional complexity without a significant gain in accuracy. Incorporating an ammonia switch-function for the heterotrophic organisms can be justified better if ammonia needs to be incorporated into the model for other purposes. The results also indicate that the ortho-phosphate concentration is not limiting for the heterotrophic organisms in the Port Alberni WWTP. In both Cells IA and 4, the switch function for ortho-phosphate, applied to the lower end of a 95% confidence interval for PO4\" concentration, is higher than 0.99. For additional information of nutrient concentrations, see Appendix 5 where are presented the results of all analytical measurements of the first sampling campaign. 5.4 Model Calibration The objective of the model calibration was to assign to all parameters values that permit the model to best represent the behaviour of the system. This task is not easy using ASM3, because some parameters are low in sensitivity and others are not easily identifiable. Therefore, not all parameters could be calibrated and the assigning of values for some parameters was based on values found in the literature. The ASM3 simplified model has 17 parameters: four stoichiometric and 13 kinetic. Of these, seven parameters were assumed and the remaining ten parameters were calibrated. 5.4.1 Parameters not calibrated Based on the sensitivity analysis, two stoichiometric parameters were assumed from the literature. The values of the parameters related to the production of X i during endogenous respiration and production of Si during hydrolysis (fxi and fsi respectively) are widely used and difficult to determine experimentally, as it was concluded from the sensitivity analysis, so they were not calibrated for this particular WWTP. 119 In addition, the hydrolysis saturation constant (K x) was also assumed, due to identifiability problems. The identifiability analysis showed that Kx cannot be determined together with the hydrolysis rate (kH) because only one of these two parameters could be estimated based on curve fitting to the OUR curve. The saturation constant for substrate Ss (Ks) was also assumed because of its low sensitivity with respect to the oxygen concentration and because of identifiability problems. The value of 10 g CODss m\"3 was adopted because better fittings were achieved using this value instead of 1 g CODss m\"3 proposed by ASM3. The sensitivity analysis showed that the respiration rate for X S T O (bsTo) was insignificant compared to the other respiration rates, so it could not be estimated from the batch respirometric tests. Therefore, bsTo was assumed to be equal to the endogenous respiration rate for X H (b^, which is a reasonable assumption used by other authors (i.e. Koch et al, 2000), and also is in agreement with ASM3 default values. A summary of the assumed parameter values and the their sources is presented in Table 5.9. Table 5.9: Parameters not calibrated and their assumed values Parameter Description Value Unit Source fsi Production of Si in hydrolysis 0.0 g C O D s i (g CODxs)\"1 ASM3 fxi Production of Xi in endogenous respiration 0.2 g C O D x i ( g COD™™)- 1 ASM3 K x Hydrolysis saturation constant 1.0 g CODxs m\"3 ASM3 K s Saturation constant for substrate Ss 10 g CODss m° Koch et al, 2000 bsTO Endogenous respiration rate for X S T O b H . d\"1 This study OT.DSIO Temperature coefficient of storage endogenous respiration coefficient 6T,bH -This study 120 5.4.2 Estimation of model parameters 5.4.2.1 Yield coefficients estimation Storage yield (YSTO) Five respirometric measurements gave an average YSTO = 0.90 + 0.06, with a range between 0.85 and 0.95. The measurements were performed with a low F/M ratio batch tests. The results are presented in Table 5.10. T a b l e 5.10: Estimation of Y S TO Test F/M JOURSTO Ss consumed YSTO g C O D (g VSS)'1 g 0 2 m\"3 g C O D s s m\"3 g C O D x s t o (g C O D s s ) \" 1 7-1 0.54 35.4 328 0.89 7-3 0.85 20.8 ' 285 0.93 8-1 0.75 27.9 263 0.89 8-3 0.74 11.7 226 0.95 9-1 0.90 43.6 297 0.85 This storage yield is in the range found in the literature. For municipal wastewaters, YSTO varies from 0.80 (Koch et al, 2000) to 0.97 (Karahan-Gul et al, 2002b). For synthetic substrates, YSTO has been measured between 0.78 and 0.87 (Karahan-Gul et al, 2002b) Observed yield of heterotrophic biomass (YOBS) Two respirographs were used for estimating the value of YOBS, with identical results (YOBS = 0.68; see Table 5.11). Both measurements were performed with very low F/M ratio batch tests. An additional analytical estimation of YOBS was performed based on COD measurements, by applying Equation 4.5 to the COD changes during a batch test (N° 9-2) with high F/M ratio (240 on COD basis). 121 Table 5.11: Estimation of Y H Test F/M jOURoBS Ss consumed YOBS g C O D (gVSS)\"1 g 0 2 m\"3 g CODss rn 3 gCODxh (g CODssX 1 4-1 0.013 16.1 50.8 0.68 5-2 0.023 25.5 79.8 0.68 YOBS = (162 g C O D x h m\"3 - 74 g C O D x h m\"3)/(578 g C O D S s m\"3 - 449 g C O D S s m\"3) [5.21] YQBS = 0.68 g CODXH (g CODssX 1 [5.22] The result obtained with this method (YOBS = 0.68) was equal to the estimation performed using respirometry, confirming the equivalence of both approaches for the estimation of this parameter. The obtained value is in the range of those found in the literature. For example, 0.67 is the value that is widely used as a default value for YOBS, which is the value proposed by ASM1. For ASM3, YOBS can be calculated as 0.54, and from Koch et al. (2000) Y 0 B s can be calculated as 0.64. Yield of heterotrophic biomass (YH) The yield of heterotrophic biomass was based on the estimations of YOBS and YSTO, using Equation 3.27. Based on the estimated value of YOBS (0.68), the value of YH was estimated as 0.76 + 0.04 respectively, which was calibrated value for YH. Again, this parameter is in the range of those found in the literature. While ASM3 proposes a value for Y H = 0.63, Koch et al. (2000) estimated Y H = 0.80. 122 5.4.2.2 Wastewater characterization Inert soluble COD (Si) Si was estimated as the soluble fraction of CODeff. This estimation is based on the assumption that all Si,eff was originally present in the influent, because there was no generation of Si in the system (fsi = 0). Nine measurements estimated Si to be between 0.11 and 0.17 times CODi n, with an average of 0.14 + 0.03. Table 5.14 presents the results of the wastewater characterization of the soluble COD fractions. In order to verify the inert characteristics of the treated effluent, a long term BOD test was performed with unfiltered and filtered effluent. Equation 3.19 was used for converting BOD units into biodegradable COD (BCOD) units. The parameter feoD was estimated using Equation 3.20 and the values of fxi and YOBS equal to 0.2 and 0.68 respectively (corresponding to the assumed and the calibrated values respectively), resulting in Equation 5.23. The results showed that 94% of the soluble CODeff appeared to be inert COD, and 6% corresponded to compounds with a very slow biodegradation rate (ICBOD = 0.065 d\"1 « k B 0 D = 0.18 d\"1, biodegradation rate estimated for Port Alberni influent). A plot of the long term BOD is presented in Figure 5.3. Best fit for total BOD: B O D U L T = 20, k = 0.065; R 2 = 0.999; and best fit for soluble BOD: B O D U L T = 6; k = 0.065; R 2 = 0.992. f B o D = 1/(1-0.2- 0.68)= 1.16 [5.23] 20 0 0 5 10 15 Time (days) 20 25 30 • Total BOD A Filtered BOD Figure 5.3: Long term total and soluble BOD effluent. 123 Readily biodegradable COD (Ss) The Ss fraction was estimated by three different methods. A first estimation of Ss was required in order to calculate the yield coefficients ( Y S T O and Y H ) , so the reduction in CODs within a batch test was assumed to be equal to the Ss initially present in the sample (Tests 7-1, 7-3, 8-1, 8-3, and 9-1 in Table 5.12). Table 5.12: Analytical estimation of Ss,i, Test C O D i n g COD m\"3 CODs, i n i g COD m\"3 CODs,end g COD m\"3 fss,in 7-1 775 536 208 0.56 7-3 775 582 297 0.44 8-1 664 530 267 0.48 8-3 664 523 297 0.41 9-1 845 672 375 0.42 The average fraction of Ss,in estimated based on these five analytical measurements was 0.46 + 0.10. Once the yields were calculated, a second set of Ss estimations was made by respirometry, using Equation 3.18. Nine respirometric tests were used for this estimation (the same five used for the analytical Ss,in calculation plus Tests 4-1, 4-2, 5-1, and 5-2). The results of these Ss,in estimations are presented in Table 5.14. The average fraction of Ss,in estimated based these nine respirometric measurements was 0.49 + 0.27. A third estimation of the Ss.in fraction, based on the physical-chemical method proposed by Mamais et al. (1993), was also used in order to check its validity with the effluents tested. Three samples of influent and two samples of effluent were used for this test, which results are summarized in Table 5.13. 124 Table 5.13: Results of physical-chemical estimation of Ss Unit Influent Effluent Total COD gm\"3 826 138 Soluble COD gm\"3 705 123 Truly soluble COD gm\"3 636 74 Truly soluble fraction % 77% 54% From these results, an estimation of Ss,in equal to 562 g m\"3 is obtained, which represents a fraction equal to 0.68 of COD;n. A comparison of the three fss.in estimations obtained using different methodologies shows that the methods based on CODs and on respirometry gave similar results (0.46 and 0.49 respectively). These similar results reduce the uncertainty associated with the estimation of Y S T O , based on COD estimation of Ss,in- On the other hand, the physical-chemical method based on the truly soluble COD results in estimation that is much larger (0.68). Finally, the estimation obtained using respirometry (fss,in = 0.49) was selected as the calibrated value for fss.in, since this is the method proposed by ASM3 for estimating this parameter. High concentrations of readily biodegradable COD, in comparison with municipal wastewaters, have been observed in pulp mill effluents before (Babuna et al, 1998; Dalentoft and Thulin, 1997; Franta et al, 1994a; Franta et al, 1994b; Helle and Duff, 1997; Mao and Smith, 1995). Summary of soluble COD fractions The characterization of the soluble fraction of the COD allows drawing the following conclusions: - regarding its physical characteristics, 84% of the total COD influent is soluble, 90% of which is truly soluble, and the remaining 10% of the COD s,i„ (8.4% of CODm) could be assumed to be a colloidal fraction; 14% of the influent COD can be considered to be inert in this WWTP, which is composed by 60% of truly soluble compounds and 40% of colloidal components; - 49% of C O D i n is readily biodegradable by the microorganisms; and 125 |co Q •3 u I co ~1= Q Q O U OO CO oo\" Q O cj o co CJ cS +-» 00 ca Q O U 60 ~o ii +^ oo •s S Q 9 o, o u u 6 0 Q Q o o u u 60 00 H o o vo m CN VO O N VO d d d d o T f r o fe i n e N u M « I-a u u 'E cu s o u 'tx J3 r- V O O s CN I/O T f T f T f d d d d o O N o o . CO m o o CN V O T f T f CO CO CN VO d d d d d d d d d 00 O N 00 IT) CO IT) r-- iri C O C O C O C O O N T f CN C O C O CN o o ii CJ T3 CJ a a fe T 3 -o U i ii fe T f vo ~ vo vo r o <—i r--vo vo C O V O T f o vo T3 ii 0 o VO CO CN CJ >/0 o r -O N O N >n T f T f m CO T f O O r - VO VO T f vo VO t-- r - r - r - vo vo o o u o m - o o o o O N 00 oo 00 oo oo oo oo oo oo CJ ii CJ CJ ii CJ CJ CJ QJ H H H H H H H H H T f VO T f T f CN _ T f 1 - H ^ H ^ H i i • i *—< d d d d d d d d d d r-- VO vo i > CO n o 00 vo T f 00 o o 00 00 o o 00 O N O N 00 (-- vo o o 00 d d d d d d d d d d d d d d e #o « N 'E cu HH» « Q O U T f - H T f I O T f C O I O N UO CO T f CN O N 00 CO CO CN OO 00 00 t T f T f CO r - l CN ~ H CN CN CN co r o ^ > ^ * cS cd a Q Q Q Q CO VO O N T f d o C O vo C O 00 C O V O OH 00 CJ (-1 l £ ' • s e oo C O • cd B -*-» oo CJ O OH OO CJ £ o o CJ •1 •4—• 13 cj +^ e CJ 00 CJ kH OH , 1 / 1 CO O O H—» 13 3 O * CJ CJ -a a £ oo oo cd 00 VO CN by difference, 21% of C O D j n is soluble but neither inert nor readily biodegradable, so it was considered to be slowly biodegradable. Approximately half of this soluble fraction of Xs is truly soluble and the other half is colloidal. Particulate COD fractions (Xs and Xi) According to the data presented in Table 5.15, the particulate material only accounted for 16% of the C O D j n in this system, so its relative importance is low compared with activated sludge systems treating municipal wastewater, for which the particulate fraction is around 44%) of the C O D i n (ASM3 default value). Slowly biodegradable COD (Xs) The slowly biodegradable C O D fraction (fxs.in) was estimated in two different ways: an analytical procedure and by an influent C O D mass balance. The analytical method for estimating fxs.in was based on an ultimate B O D ( U B O D ) test performed with unfiltered and filtered influent, as proposed by Roeleveld and Van Loosdrecht (2002). Applying a similar analysis to that used for determining the Si fraction of the influent, the conversion from B O D to biodegradable C O D was made using Equations 3.19 and 3.20. Figure 5.4 presents the plot of the B O D versus time of this test. The best fit was obtained, for the total B O D (R2 = 0.99), with BODULT =500 g m\"3 and k = 0.18 d\"1; for soluble B O D (R2 = 0.97), with BODULT = 400 g m\"3 and k = 0.13 d\"1. Table 5.15 presents the results of the biodegradable and inert fractions of the influent, differentiating the soluble and the particulate components. Table 5.15: Biodegradable and inert C O D fractions Unit Total Soluble Particulate BODULT g 0 2 m\"3 500 400 100 C O D g C O D rn 3 826 705 121 B C O D g C O D m3 575 460 115 127 Time (days) • Total B O D A Soluble B O D Figure 5.4: Long-term influent BOD total and soluble Based on these results, the particulate fraction of the biodegradable COD accounts for 14% of C O D i n . However, as it was calculated before, the slowly biodegradable fraction has also truly soluble and colloidal fractions corresponding to 9% of the C O D i „ each, totalizing an Xs fraction equal to 0.32 of C O D j n . The colloidal fraction, sometimes called rapidly hydrolysable COD (Henze, 1992; Sollfrank and Gujer, 1991), may not be neglected in this wastewater characterization as its fraction is comparable to the actual particulate fraction of Xs. Using the second approach, based on a COD i n mass balance (Equation 4.25), an independent estimation of fxi,in is required. Using the estimation of fxi,jn based on overall solids mass balance in the plant (fxi.in = 0.07, from next section), fXs,in was estimated to be 0.30. Subsequently, the particulate fraction of Xs was calculated by difference (Xs,part,in = 0.30 - 0.18 = 0.12 of C O D i n ) . Comparing the results from both estimation methods, it is possible to conclude that both give similar results for estimating the Xs fraction (0.32 and 0.30 respectively). If the interest is only the estimation of fxs.in, the first approach is preferred because, even though it is more labourious, it avoids the problem of accumulation of errors of the second method. If the interest is to estimate all influent COD fractions, a decision has to be made in order to choose whether 128 fxs,in or fXi,in will be estimated independently, so the other can be estimated by an influent COD mass balance. Since the estimations of fxs.in were relatively similar for both estimation methods, and because the independent estimation of fxs.in was preferred above the influent COD balance for that parameter (see following section), the method based on the influent COD mass balance was selected for estimating fXs,in, resulting in a calibrated value of 0.30. Particulate inert COD (Xi) The particulate inert influent COD fraction (fxi,in) was estimated in two different ways: independently using a full-scale sludge production mass balance, and using the other CODjn fraction estimations to perform a CODjn balance. Using the first method, fxi,jn was estimated by comparing the predicted sludge production rates with the measured rates as a function of the SRT. This full-scale sludge production mass balance was performed using a five day data period, corresponding to the week of the first measuring campaign. That week was selected because it is'the most representative of the wastewater characterization process. The SRT, the estimated sludge production rate, and the measured sludge production rate were calculated using Equations 4.21, 4.22, and 4.23 respectively. The information used for this estimation was obtained from the mill historical database, which included the following measurements: - influent flow; - wastage flow; - effluent flow (by difference); - temperature (average five cells); - influent COD; - effluent TSS; and - wastage TSS. 129 The first four measurements (flows and temperature) were obtained with a frequency of one hour. The COD and solids were obtained with a frequency of two data per day, which was complemented with the results of the sampling campaign performed during that week, resulting in a data frequency of four data per day. In addition, the estimated values of Y 0 B S , b H and 0T,bH were used (0.68 g CODXh (g CODSs)\"', 0.13 d\"1 @ 30°C, and 0.04 respectively), as well as the estimation of the influent Si fraction (0.14 of C O D i n ) . The result of the first method, which minimized the error sum of the squares between the sludge production estimated and measured, was an estimation of fxi.in equal to 0.07 of C O D i n . The second method for estimating fxi,in, based on the previous estimations of fsi,in, fss,in, and fxs.in (0.14, 0.49, and 0.30, respectively) and Equation 4.24, resulted in an estimation fxi.in equal to 0.07 of C O D j n . Comparing the results from both estimation methods, it is possible to conclude that the both estimation procedures gave equivalent estimations of Xi . Therefore, if a checking test is not required, the estimation of Xi based on an influent COD mass balance should be preferred because it gives an accurate estimation with little work. The independent estimation of Xi based on a plant mass balance could be used for checking not only the Xi , i n fraction, but also the estimation of other parameters used in this procedure (i.e. Y O B S and bH). Wastewater characterization summary Figure 5.5 presents a summary of the wastewater characterization, and Table 5.16 compares this characterization to other relevant characterizations. It is interesting to note that the wastewater characterization results obtained in this research were very similar to the ASM3 proposed values for municipal wastewaters, which supports the idea of using ASM3 as a valid model for Port Alberni WWTP. In contrast, the presented wastewater characterization differed from the obtained by Sreckovic (2001), who worked with the same WWTP, in estimating the biodegradable COD fractions. 130 Xs (colloidal) 12% Figure 5.5: Wastewater characterization summary Table 5.16: Summary of COD influent fractions COD Other Pulp and Paper references Municipal fraction This study Stanyer (1997) Slade Sreckovic wastewater (2 mills) (2003) (2001) (ASM3 default) Ss 0.49 0.24 0.44 0.42 0.67 0.43 Si 0.14 0.36 - 0.32 0.33 0.17 0.13 X s 0.30 0.42 0.23 0.11 0.10 0.33 X, 0.07 0.07 0.03 0.14 0.06 0.11 Mill type CTMP/SGW BKME BKME/ TMP BKME CTMP/ SGW -5.4.2.3 Kinetic parameters estimation Endogenous respiration rate (bH) The endogenous respiration rate (bH) was estimated in three different ways. The first two approaches were similar but differed in the way in which the biomass was estimated (one 131 analytically and the other by respirometry). The third approach was based on the SOUR of many different batch tests. The first method used for estimating bu was based on estimating the decrease in biomass as the decrease of the particulate COD. The coefficient bu was estimated as the slope of the curve from a plot of the logarithm of X H versus time. The test, performed at 23°C, estimated bH = 0.08, which, converted to 30°C, corresponds to bH = 0.11 d\"1 (Figure 5.6). T = 2 3 ° C |y= -0.08x+ 0.03 j R2 = 0.95 • 0 2 4 6 8 Time (days) Figure 5.6: Endogenous respiration coefficient based on particulate COD measurements The conversion of all kinetic measurements at a temperature T to the reference temperature (30°C) was calculated using The Arrhenius Equation (Equation 3.7) rearranged as Equation 5.24. Since the value of the parameter 6T depends on the value of the parameter being estimated, an iterative approach was used, as explained in Section 4.5. k(30°C) = k(T) / e e T ( T - 3 0 ° c > [5.24] Using the second approach, two estimations of bn were performed by monitoring the decrease in X H indirectly as the decrease of OUR. These two long-term batch tests were performed without wastewater addition, at temperatures of 24°C and 30°C (Figure 5.7). Similarly, the coefficient bH was estimated as the slope of the curve from a plot of the logarithm of OUR versus time (Sollfrank and Gujer, 1991). The curve fitting was good in both tests (R2 > 0.90), resulting in an estimation of bn equal to 0.11 d\"1 and 0.13 d\"1 for 24.2 and 30.2°C respectively. Converting those values to 30°C and taking the average, the respirometric estimation of bH was 0.13 d\"1. 132 0 2 4 6 8 10 0 2 4 6 Time(d) Time (d) Figure 5.7: Endogenous respiration decay of OUR in two batch reactors, at 24°C and 30°C respectively. These two methods for estimating bH, based on estimations of the biomass decrease, fit relatively well with the experimental data, and may prove the validity of the first-order reaction for modelling the decay of X H . The third approach (second respirometric method) estimated b H from the SOUR measured with sludge in endogenous respiration state in different batch tests, using Equation 4.17 rearranged as Equation 5.25. A more detailed explanation of this method is presented in Section 4.5.2.4. bH = SOUR/(l-f Xi) [5.25] The measurements of the OUR from ten respirometric batch tests were used in this method, results of which are presented in Figure 5.8. The biomass concentration (g CODXh L\"1) was estimated as the concentration of VSS (g SS L\"1). Correcting the results to 30°C (using Equation 5.24), b H is estimated equal to 0.14 + 0.02 d\"1, which is very similar to the results obtained from the first two approaches, and with the advantage of being calculated without the performance of any other additional test. Finally, the value 0.13 d\"1 was selected as the calibrated value for bn for this activated sludge system, which is lower than the default values of ASM3, but in the range of other values estimated for pulp and paper wastewaters: 0.05 d\"1 and 0.10 d\"1, (Stanyer, 1997); 0.16 d\"1 (Slade 133 et al, 1991) (all reference values are transformed to ASM3 endogenous respiration equivalent and corrected to 30°C using Equation 5.24). Estimation of 6r,bh Two different respirometric approaches were used in order to estimate the temperature coefficient of bn (O-r.bh). The first approach consisted of a rapid increase in the temperature of sludge in an endogenous state, so the resulting increase in the OUR was proportional to the increase of b ^ Figure 5.9 presents the outcome of this test, which resulted in a 0T,bh = 0.04. 0.8 _ 0.6 1 ZdZ-y = 0.04x-1.09 R2 = 0.99 25 30 35 40 45 T e m p e r a t u r e (°C) Figure 5.9: Increase of OUR of sludge in endogenous state during a rapid increase of temperature 134 The second method for estimating 0T,bh was based on the endogenous respiration level of ten different batch tests carried out at eight different temperatures. The method was based on the assumption that the specific OUR (SOUR) only depends on the temperature. Therefore, 0 T,bh can be estimated as the slope of the straight line formed by the logarithms of the SOUR from different batch tests versus their temperatures; which is presented in Figure 5.8. The estimation of 0T,bh using this method was 0.04, which is the same as that obtained from the first method. The results of both approaches suggest these two methods give equivalent results. The major advantages of the second approach are that it uses available information and allows the estimation of bH as well. When bH does not need to be estimated, or is estimated without using the SOUR approach, the first approach is very convenient due to its simplicity and its ability to be performed quickly. Oxygen saturation constant (K02) The oxygen saturation constant was estimated using the method proposed by Kappeler and Gujer (1992), starting with an oxygen concentration of 4.0 g O2 m\"3 and finishing with a concentration of 0.2 g O2 m' . Figure 5.10 presents a plot of the oxygen switching function (as O U R measured over OURMAX) versus oxygen concentration. The results present a good fit with K02 = 0.29 for oxygen concentrations above 1 g O2 m\"3, but a better fit with K02 = 0.60 g O2 m\" for oxygen concentrations lower than 1 g O2 m\" . Since the oxygen concentration is consistently above 1 g O2 m'3 in all the cells of the WWTP, 0.3 g O2 m\"3 was selected as the calibrated value for K02. The value for K02 is typically assumed equal to 0.2, which is the default value used in all ASM models, so it is not easy to find independent estimations of this parameter. As an example of a different value, Kappeler and Gujer (1992) estimated K02 to be equal to 0.25, which is close to the estimated value determined from this research. For checking the assumption that set K02 to the same value during the growth/storage processes as the endogenous respiration, the value of K02 during endogenous respiration was also estimated using the same methodology. 135 The value of K02 for endogenous respiration was measured three times, resulting in an average of 0.24 g O2 m\"3. The results, presented in Figure 5.11, showed that K02 for endogenous respiration varies approximately in the range 0.15 - 0.30 g O2 m\"3, which is in accordance with ASM3 default value (0.2 g O2 m\"3) as well as with the estimated value for the growth/storage process (0.29 g 0 2 m\"3). 1.2 0.0 1.0 2.0 3.0 4.0 5.0 Oxygen Concentration (g m-3) Figure 5.10: Plot of oxygen switching function versus oxygen concentration during growth/storage processes 1.00 O x y g e n c o n c e n t r a t i o n ( g 0 2 / m 3 ) Figure 5.11: Plot of oxygen switching function versus oxygen concentration during endogenous respiration These results suggest that the value of the K02 is the same for growth/storage and for endogenous respiration and is valid for this activated sludge system. In addition, given that the 136 results of both estimation procedures are comparable, the author recommends using sludge in an endogenous respiration state, which makes the estimation simpler and less sensitive to measurement errors. This recommendation contrasts with the one proposed by Kappeler and Gujer (1992), who recommend to perform the test in the presence of a small concentration of readily biodegradable COD. Curve fitting for estimation ofksro, UH, KSTO, and kn The values of ksxo, |^H, KSTO, and kn were estimated based on the analysis often respirographs (respirometric curves) from batch tests at different temperatures (22°C - 42°C) and at different F/M ratios (from 0.01 to 0.90). These tests were not replicates of the same test but all of them were performed under different conditions (by choosing different temperatures and F/M ratios). A detail description of each test can be found in Appendix 4. During this curve fitting calibration procedure, all other estimated parameters were kept constant at their calibrated values, including the wastewater COD fractions. For example, even though the storage yield (YSTO) may have been calculated for a specific batch test, the value for YSTO used during the curve fitting of that test was the estimated average value (0.90) and not the one calculated for it, which actually is expected to fit even better to the data. The same criterion was used with all other parameters and wastewater fractions. In a fist stage of the iterative calibration procedure, the parameter KSTO was estimated. The selection of this parameter to be the first estimated was mainly due to two reasons: - because KSTO does not depend on temperature, so the degree of variability of this parameter was expected to be low; and - because the high correlation between KSTO and \\XH makes it difficult for their joint estimation. A first calibration procedure, performed by curve fitting of the ten respirographs and adjusting the four parameters ksTO, UH, K S TO , and kH, resulted in estimations of KSTO ranging between 0.05 and 0.5 g COD X s t o m\"3, with a mean of 0.3 g COD X sto m\"3 (see Table 5.17). 137 In a second stage, and maintaining K S T o constant at 0.3 g COD X s to m , k Sro, PH, and kH, were estimated by repeating the best-fit procedure with the same ten respirographs. The results of this curve fitting process are presented in Table 5.18. The last row of the table presents the average of the values 'corrected' to 30°C, which are the estimated values for those parameters. Table 5.17: Best fit values for parameter KSTO Test Temperature KSTO °C g CODxsto (g CODxh)-' 2 32 0.25 4-1 25 0.05 4-3 26 0.30 5-1 42 0.20 5-2 41 0.20 7-1 25 0.44 7-3 22 0.50 8-1 30 0.05 8-3 32 0.50 9-1 22 0.50 Mean - 0.30 As examples, some results are plotted in Figures 5.12 and 5.13, where the measured O U R (points) are presented together with the best-fit model simulation (line). Figure 5.12 presents two respirographs performed at very low F/M (~ 0.015) and at different temperatures (25°C and 41°C respectively). Figure 5.13 presents two respirographs performed at low F/M (~ 0.6) and at different temperatures (25 and 32°C respectively). 138 0 50 100 150 200 250 0 40 80 120 160 Time (min) Time (min) Figure 5.13: Two respirographs at low F/M ratio. The results showed that a single set of parameters (Y S TO , Y H , f x i , K 0 2 , K s bH, bs-ro, K x , K S T O , and ksro), with some variations only in two parameters (kn and pn), was able to fit the data from different batch tests. It is clear that the variations on kH and P H could have been lower if other parameters were allowed to change during the fitting process (i.e. Ks and KSTO)- For example, pn adopted a value of 14.9 with the best bit achieved when KSTO adopted a value of 0.05 (instead of p H = 70.6 with K S T O = 0.3). The relatively good fit of the model in the different regions of the respirographs suggests that the kinetic expressions from ASM3 are appropriate to model the processes represented. 139 Table 5.18: Best fit values for parameters ksio, M H , and ku Test Temperature ksTO U H k H °C d\"1 d\"1 d\"1 2 32 15.0 18.4 0.1 4-1 25 14.7 70.6 41.7 4-3 26 7.5 14.6 3.2 5-1 42 19.5 31.1 2.1 5-2 41 17.5 32.0 3.3 7-1 25 13.3 1.8 11.4 7-3 22 13.0 1.8 12.7 8-1 30 18.8 2.9 12.3 8-3 32 24.6 5.9 40.5 9-1 22 14.8 1.5 16.5 Calibrated value 30 15.3 17.5a 8.1 Calibrated vale based only on low F/M tests The only exception was (j.H, for which two different values were estimated, depending on the F/M ratio. The average of the tests performed at low F/M (2, 4-3, 5-1, and 5-2, excluding 4-1) gave a value of u« =17.5 + 1.3 d\"1 (@ 30°C), and the average of the tests performed at very low F/M (7-1, 7-3, 8-1, 8-3, and 9-1) gave a value of [iH = 3.1 ± 2.1 d\"1 (@ 30°C). The parameter U O B S was estimated in order to have a reference value for choosing between these two different estimations of p.H (see further section). It is important to point out that U H has a relative importance similar to the hydrolysis rate in ASM1 (both explain the respirometric 'tail' after the consumption of Ss), so even though its symbol is the same as that of the ASM1 maximum heterotrophic growth rate ( U « , A S M I ) , U H is not comparable to | J . H , A S M I -The shape of the respirographs does not suggest the presence of more than one readily biodegradable fraction. Only Test 7-1 (Figure 5.13) presents one point that may be considered 140 as an extra Ss fraction with a higher biodegradation rate. However, no significant improvement would be achieved in the fit by complicating the model with an extra Ss fraction. The identifiability problem affecting the hydrolysis maximum rate was verified in the curve fitting procedure to the respirograms. It was found difficult to distinguish, in a single respirometric test, the contributions to the OUR from the utilization of Xs and from the growth from XSTO- Therefore, even though more than one physical fraction of Xs was detected, a single biodegradation rate seems to predict the hydrolysis of Xs properly and it would be impossible to calibrate more than one Xs hydrolysis rate. Estimation of 6j,kSTO ond 6T,^M The values of 0 T , k S T o and 0T,HH were obtained from the variation of those parameters estimated from respirographs conducted at different temperatures. Seven and four data values were used for estimating 0 T , k S T O and 0T,HH respectively. In the estimation of 0T,HH, the four data values used were those corresponding to tests performed at very low F/M ratio, because the parameter PH was finally calibrated based on those values. Figure 5.14 presents the plots of the logarithms of ksjo and pn versus the temperature, as well as the best-fit curves with their equations. Therefore, the estimations of 0 r , k S T O and 0T,HH are 0.02 and 0.05 respectively, which show a very low effect of temperature over the storage process in the temperature range of 22 - 42°C. These values are lower than the ASM3 proposed values, both equal to 0.07. Temperature (°C) A In(muh) • In(ksto) Figure 5.14: Variation of k s T o and pn with temperature 141 Estimation of BTMH Similarly, the value of 0T,kH was calculated from the plots of ln(kH) versus temperature (Figure 5.15). Seven values of k H were used for this estimation, including data from batch tests performed with filtered and unfiltered wastewater. The estimation of 0T,kH was calculated as -0.09 in the temperature range of 22 - 42°C. This result contrasts with the typical values found in the literature (~ 0.04) especially because it is negative, which represents a decrease in the hydrolysis process at higher temperatures. 3 - i -2.5 -_ 2 -1.5 -~ 1 -0.5 -0 — 20 Figure 5.15: Variation of kn with temperature Estimation of JU0BS The parameter combination (P-OBS - bH) can be obtained as the slope of the curve plotting the logarithm of the OUR in a test with high F/M, where U O B S is equivalent to ASM1 u«- Two estimations of U O B S were performed in this way, with the results equal to 5.3 and 8.0 d\"1 (at 27.0 and 25.9 °C respectively), with an average of 8.6 d\"1 (converted to 30°C equivalent using Equation 5.24 and GT.HODS = 0.07). The plot obtained from one of these tests is presented in Figure 5.16. y = -0.09x + 4.73 R 2 = 0.91 25 30 35 40 45 Temperature (°C) 142 7 1 -0 -I 1 1 1 1 , 1 , 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 1.6 Time (days) Figure 5.16: Estimation of UOBS by respirometry This estimation of POBS can be considered to represent a minimum value for p H (Equation 3.29). Therefore, the tests performed at very low F/M gave a more reliable estimation of PH (17.5 d\"1 > 8.6 d\"1) than the tests performed at low F/M (3.1 d\"1 < 8.6 d'1). The reason for this difference is not known. Consequently, the most probable value for PH would be the estimation of 17.5 d\"1, based on batch tests performed at a very low F/M ratio. To estimate POBS may also be useful to compare the results of this calibration with others based on the more widely used ASM1 structure. 5.5 Summary of the calibrated model Table 19 presents a summary of all the calibrated model parameters as well as those assumed from the literature. The results are compared with ASM3 default values and, whenever it was possible, with the results presented by Sreckovic (2001), who calibrated a model based on ASM2 to the same pulp and paper mill. 143 Table 5.19: Summary of calibrated model parameters (kinetics parameters at 30°C) Symbol Units ASM3 Default3 Calculated values. This study Sreckovic (2001) Kinetic parameters ' Hydrolysis rate constant k H d\"1 6.0 8.1 7.9 Hydrolysis saturation constant K X g X s g - % 1.0 1.0b 0.5 Aerobic storage rate constant k s T O d\"1 10 15.3 -Saturation constant for oxygen K02 g 0 2 m\"3 0.2 0.3 -Saturation constant for substrate S s K S g COD m\"3 2.0 10 c 158 Saturation constant for storage K S T O g Xsto g' 1 X H 1.0 0.3 -Heterotrophic maximum growth rate UH d-' 4.0 17.5 -Observed maximum growth rate M-OBS d\"1 12 d 8.6 39 Aerobic endogenous respiration rate of X H b H d\"1 0.4 0.13 0.53 Aerobic respiration rate for X S T O b s T O d\"1 0.4 0.13 e -Temperature coefficient for k H 0T,kH - 0.04 - 0.09 -Temperature coefficient for k Sxo 0T,kSTO - 0.07 0.02 -Temperature coefficient for u H 9 T , H H - 0.07 0.05 -Temperature coefficient for b H and bsTo ^T.bH - 0.07 0.04 -Stoichiometric parameters Production of X i in endogenous respiration fx, g X , g 1 X„ 0.20 0.20 b 0.55 Production of Si in hydrolysis fsi g S, g 1 X S 0.00 0.00 b -Yield of stored products per S s Y S T O g X S T O g 1 Ss 0.85 0.90 -Yield of heterotrophic biomass growth on X S T O Y H g X H g 1 X S T O 0.63 0.76 -Observed yield Y Q B S g X H g\"1 S s 0.68 0.68 0.54 • a To make the kinetic parameters comparable, it was assumed that ASM3 9 T values were constant up to 30°C b Assumed equal to ASM3 default value, not calculated c Assumed equal to Koch et. al (2000) estimated value, not calculated d ASM1 proposed value, converted to 30°C equivalent e Assumed equal to bn estimated value, not estimated independently 144 6. E n g i n e e r i n g s igni f icance of this research The relevance of this research from the engineering perspective is directly related to the usefulness of models. As discussed in Section 3.1.1, the major objectives using models in biological wastewater treatment are the process design and the determination of optimal operation conditions. Thus, a proper calibrated model may help in designing the best technical and economical WWTPs, and assist in operating existing plants to achieve the best performances and maximum efficiency. Models can also be used to identify bottlenecks in the operation of a certain WWTP, to forecast the impacts of changes in the operating conditions, to determine the suitability of a certain technology for a given problem, or to assist in the selection of the best solution if more than one technology is available. 6.1 Possible applications of this research A model used to design proper facilities may help to save money in the construction of new plants (i.e. smaller bioreactors, smaller pumps, better land use, etc.), where effluent quality predictions are needed to meet the regulation standards. The modelling applied to the design process may also be used in existing facilities, so simple modifications in the plant configuration (i.e. divide a bioreactor into two parts, or change the feeding pattern) can produce significant improvement in the effectiveness of the WWTP. When applied to the operation of a WWTP, a model can help to optimize some operational parameters (i.e. recycle flows, aeration flows, nutrient addition), which can also have a significant effect on the operational costs. Further, improving the efficiency of the plant can allow the treatment of a larger effluent load in the same facility (i.e. due to a mill upgrade) without compromising the quality of the effluent. Finally, this calibrated model increased the knowledge related to activated sludge systems applied to the pulp and paper industry, so additional information is available now to better evaluate the suitability of this technology for treating pulp and paper effluents. If similar models are available for other treatment options (i.e. trickling filters, aeration stabilization basins, etc.); a decision about the best technical treatment alternative can be made. 145 In practice, all these possible applications of models are put into operation with the aid of simulators. There are many different simulators (software designed to perform simulations) in the market, which can simulate the effects of different changes in the layout or operation conditions of a WWTP. Simulators allow running many different alternatives at a minimum cost and effort, without the difficult and risky necessity of changing the operation of the actual plant. 6.2 Full-scale simulation using the calibrated model It seems evident at this stage, that the objective of models is to solve 'real' problems, which require 'full-scale' application. The development and calibration of models attempts to be the most accurate representation of what is happening in the existing WWTP (or expected to be in a projected facility), so a full-scale validation of the model is essential. The full-scale validation is not an easy task. In order to be statistically representative of the system under study, it would require a complex statistical analysis that is beyond the scope of this research. However, a full-scale simulation was performed with the calibrated model, applied to a nine day period of data from the Port Alberni WWTP historical database. The period used for the full-scale simulation was chosen due to its highly dynamic influent COD, so some dynamic response in the effluent and in the sludge characteristics was expected. The influent COD and the influent flow are presented in Figure 6.1. All other data from this series can be found in Appendix 6, starting July 28 th and finishing July 6 th. The simulation was performed using MATLAB® and the routine presented in Appendix 7. 146 3500 - 1000 ^ 2500 4 -Qin CODin Figure 6.1: Influent flow (Qin) and influent COD (COD in) for validation period The results, presented in Figure 6.2, correspond to the measured and predicted concentrations of VSS in Cell 4 and effluent COD. These results suggest the model is able to predict relatively well the major trends of the system, but more frequent data would be necessary to validate the model properly due to the highly dynamic nature of the activated sludge system. 0 1 2 3 4 5 6 7 8 9 Time (days) VSS model • VSS measured • COD eff measured COD eff model Figure 6.2: Full-scale simulation performed with the calibrated model to historical data from the Port Alberni WWTP 147 6.3 Major limitations of the presented results The most important limitations of the model include: - the assumptions impose some limitations, specially about the type of the available information (i.e. only total COD is available from measurements so, even though it is not true, COD fractions have to be estimated to be constant); - conditions outside the normal range of operation of the WWTP may create situations that the model is not able to predict (i.e. pH far from neutrality, low concentrations of nutrients, toxic spills, too-low oxygen concentration , etc.); the model can only predict the states of variables included in the model so, without incorporating ammonia or other types of microorganisms (i.e. autotrophic or filamentous organisms), the model cannot predict the dynamics or the effects that those parameters can produce (i.e. nitrification or bulking events); and since the model was calibrated for a single WWTP, and the wastewater characteristics of the pulp and paper industry are very site-dependent, the values of the parameters presented should not be used in a different plant without verification. 148 7. Conclusions The ASM3 model was successfully calibrated for an activated sludge plant treating mechanical pulp and paper wastewaters. A single set of parameters, with minor variations, was able to fit a variety of batch test respirographs, with values similar to those found in the literature. Some assumptions of the model, however, proved not to be applicable and some adjustment would be necessary in building a dynamic model for the treatment of these effluents. Four influent COD wastewater fractions were estimated (Si, Ss, Xi, and Xs) as well as the most important model parameters ( Y S T O , Y H , bH, ks-ro, P H , K S T O , and kH). These estimations increase not only the knowledge of the activated sludge systems applied to pulp and paper effluents but also on the scarce application of ASM3 to industrial wastewaters. The values of all model parameters were found to be similar to those found in the literature, including those calibrated for municipal WWTPs. The only exception was to those presented by Sreckovic (2001). In the same way, the wastewater characterization was found to be similar to that proposed by ASM3 for domestic wastewaters. Even though the authors of ASM3 advises not to apply that model for industrial wastewaters and outside the temperature range of 8 - 23°C, ASM3 was found to be suitable for modelling the COD removal in a WWTP treating pulp and paper effluents at temperatures around 32°C. It was found that the high content of readily biodegradable COD and the low content of slowly biodegradable COD were important factors in making ASM3 suitable for modelling the effluents of this WWTP. The use of widely used models, developed originally for municipal wastewater, in pulp and paper applications would simplify considerably the task of modelling and designing WWTP for this industry. Sensitivity and structural identifiability analyses were performed for the ASM3 simplified model. Only the calibration of the heterotrophic growth rate (pn) presented some identifiability problems, which were addressed by estimating an auxiliary parameter (POBS)-149 A few novel calibration procedures were developed for estimating some of the parameters (bH, 0T,DH), which gave similar results to the traditional methods. The author believes that these methods can be used for any type of effluents, with less work than the traditional methods. In addition, different methods were used for the estimation of many of the model parameters and wastewater fractions, and some recommendations were done in order to select the best method for estimating different model components. Finally, a full-scale simulation performed with the calibrated parameters values over a time period of nine days suggested the validity of the calibrated model for the application on Port Albemi WWTP. More detailed conclusions, related to the sensitivity and identifiability analyses, wastewater characterization, and parameter estimation, are presented in Sections 7.1, 7.2, and 7.3 respectively. 7.1 Sensitivity and Identifiability analyses The sensitivity analysis indicated that the model components Si, Xi , Xs, fxi and fsi are relatively insensitive to all parameters, so curve fitting is not recommended for their estimation. Further, the estimation of these model components has little effect on the overall calibration procedure. On the other hand, X S T O is the model component that was most sensitive to parameters values, which reflects the importance of this model component in the ASM3 structure. The sensitivity analysis indicated also that the parameters that are most suitable for estimation by respirometry were K02, ksTO, K S T O , PH , Y S T O , and Y H . However, the identifiability analysis concluded that, of those, K02, Y S T O and YH were not structurally identifiable, so only ksro, K S T O , and PH are theoretically identifiable using curve fitting to respirometric curves. An additional constraint imposed by the identifiability analysis showed that, using curve fitting to the exogenous OUR, the parameters bH, bsTO, fxi, and fsi were not structurally identifiable by respirometry. 150 7.2 Wastewater characterization The COD characterization identified a large fraction of influent COD (CODjn) present as readily biodegradable COD (52% of COD i n), which can be assumed to be truly soluble. It was found also that the estimation of this fraction was temperature-dependent. The biodegradation of this fraction was found to be well explained assuming only one fraction with a single degradation rate. The slowly biodegradable influent COD (X s,in) was found to have particulate, colloidal, and truly soluble fractions (12%, 9% and 9% of COD i n respectively). This X s , i n fraction was lower than that reported for domestic wastewaters, making the hydrolysis process of lower relative importance when dealing with pulp and paper effluents. Therefore, physical-chemical methods were found not be accurate for determining Xs. Despite the different physical fractions, a single hydrolysis process for Xs fit the experimental data well. The soluble inert influent COD fraction (fsi.m) was found to account for 14% of the CODjn. The majority of this fraction was observed to be truly inert (94%), while a small portion was observed to be made up by very slowly biodegradable compounds (6%). The assumption that all effluent Si was originally present in the influent, together with assuming there is no Si generated during hydrolysis, was found to be a good assumption that simplified the calibration procedure. The particulate inert COD fraction was estimated as 7% of the CODjn. The estimations based on the sludge production rate and based on a CODjn mass balance gave equivalent results, so the latter method may be preferred because of its simplicity. Al l these wastewater fractions are in the range of those found in the literature for pulp and paper effluents. In addition, some assumptions regarding the wastewater characteristics were checked: the absence of volatile COD in the influent, the absence of bacteria in the influent, the pH close to neutrality; and that nutrients are not limiting for heterotrophic bacteria. All those assumptions 151 were found to be reasonable. Only the presence of bacteria in the influent was suggested by some results, which may require further verification and quantification. 7.3 Parameter estimation The existence of two heterotrophic yield coefficients complicates the estimation of these parameters, because the traditional method for estimating the heterotrophic yield only allows for the estimation of a single yield. Different respirometric techniques were used for these estimations. Finally both yield estimations were in the range of the values found in the literature. Batch tests were useful for estimating model parameters. High, low F/M and very low F/M ratio tests were both required in order to estimate different parameters. Very low F/M ratio were found better for calibrating the kinetic model parameters (especially uu). Curve fitting was an essential tool in parameter estimation, because the complexity of the equations in ASM3 makes it impossible to design experiments for direct measurement of some parameters, especially those related to the storage and . growth processes. However, identifiability problems of ASM3 required the use of additional methods, other than respirometry, for a complete calibration. The results indicated that, under the conditions observed in the WWTP, where Kx » X S / X H , a first order reaction could model the hydrolysis process properly, reducing the number of model parameters and, therefore, the complexity of the model. The estimation of the observed growth rate (UOBS) , equivalent to the ASM1 maximum growth rate ( U H . A S M I ) , was found useful as a reference value for the estimation of U H and for comparison purposes. Al l parameter estimation values were in the range of those found in the literature. This finding contrasts with the results of Sreckovic (2001) who, using ASM1 for modelling the treatment of the same pulp mill effluents, estimated some parameters differently. 152 Two novel methods for estimating 9T,bH and one for estimating b H were presented in this thesis, which can be comparative much simpler than the traditional methods for estimating these model parameters. 153 8. R e c o m m e n d a t i o n s for future deve lopment A validation of the capability of this calibrated model for predicting the carbon oxidation in this activated sludge system would require a more complete full or pilot scale test. Since many variables in WWTPs have very short time-scales (< 1 hour) (Steffens et al, 1997), high frequency data are required in order to have good quality dynamic information to properly validate the model. The development of a new model may require more detail on the soluble COD fractions. The inclusion of an extra soluble slowly biodegradable COD fraction, or assuming all slowly biodegradable COD is soluble, may describe these effluents better than a single particulate Xs fraction. Whether an extra soluble fraction includes the colloidal material or not should depend on the measurement techniques used for filtering. The presence of heterotrophic bacteria in the influent should be checked in order to identify the possible effect of this biomass source in the system. The nitrifying module of ASM3 could be incorporated in future model calibrations. 154 9. List of Terms, symbols, and acronyms AOX Adsorbable organic halogens -ASB Aeration stabilization basin -ASM Activated sludge model BCOD Biodegradable COD g COD m\"3 b H Endogenous respiration rate for X H ('traditional', as used in ASM3) . d\"1 b H , A S M l Endogenous respiration rate for X H under death-regeneration theory (used in ASM1) d\"1 BKME Bleached kraft mill effluent BKP Bleached kraft pulping -BOD Biochemical oxygen demand g 0 2 m\"3 bsTO Aerobic respiration rate for X S T O d\"1 COD Total Chemical Oxygen Demand g 0 2 m\"3 CODs COD soluble g 0 2 m\"3 CSTR Completely stirred tank reactor -CTMP Chemi-thermomechanical pulp -DO Dissolved oxygen -F Mass flow rate gd\"1 F/M Food to microorganisms ratio g CODss (gCODxhV1 fp Fraction of biomass leading to particulate products (as used in ASM1) gCODxi (g CODxbm)-' fsi Production of Si in hydrolysis gCODsi (gCODxsV1 fxi Production of Xi in endogenous respiration gCOD X i (g CODxbm)\"' fsi.in Soluble inert COD fraction in the influent -fsS,in Readily biodegradable COD fraction in the influent -fxi.in Particulate inert COD fraction in the influent -fxS.in Slowly biodegradable COD fraction in the influent -HRT Hydraulic retention time d 155 lSS,BM SS to C O D ratio for biomass X H gSS ( g C O D x b m ) - 1 iss,xi SS to C O D ratio for X R g S S ( g C O D x i ) - 1 iss,xs SS to C O D ratio for X S gSS (g C O D x s ) \" 1 isS,XSTO SS to C O D ratio for X S T o g SS (g C O D x s t o X 1 K A L K Saturation constant for alkalinity for X H mole HC0 3\" m\"3 k H Hydrolysis rate constant g C O D x s ( g C O D x h d)\" 1 K N H 4 Saturation constant for ammonium for X H g N m\"3 K02 Saturation constant for S02 g0 2 m\" 3 Kpo4 Saturation constant for ortho-phosphate for X H gPm\" 3 K S Saturation constant for substrate Ss g C O D S s m\"3 ksTO Storage rate constant g C O D s s (g C O D X H d)\" 1 K S T O Saturation constant for X S T O g C O D x s t o (g C O D x h ) \" 1 K X Hydrolysis saturation constant g C O D x s (g C O D X H ) \" 1 L C Lethal concentration gm\"3 O U R Oxygen uptake rate g 0 2 m\"3 d_1 pH pH , -PHA Poly-hydroxy-alkanoates -P04\" Ortho-phosphate gPm\" 3 Qeff Effluent flow nvV Q i n Influent flow m 3d- ! Qrec Recycle flow (from secondary clarifier) m3d\"' Qwaste Wastage flow m3d-' RMP Refiner mechanical pulp -S Soluble component of the model gm\"3 SGW Stone groundwood -S H Rapidly hydrolysable organic substrates g C O D m 3 SI Inert soluble organic matter g C O D m\"3 SMP Soluble microbial products -S02 Dissolved oxygen g 0 2 m\"3 S O U R Specific oxygen uptake rate g O z ^ C O D x h d ) \" 1 SRT Solids retention time d 156 Ss Readily biodegradable organic substrate g COD m\"3 SS Suspended solids g SS m\"3 T Temperature °C TMP Thermomechanical pulp -TOC Total organic carbon gm\" 3 TSS Total suspended solids g SS m\"3 UBC University of British Columbia VOC Volatile organic compounds -Vol Volume m 3 VSS Volatile suspended solids g SS m\"3 WWTP Wastewater treatment plant -X Particulate component of the model gm\" 3 X H Heterotrophic biomass g C O D X H m\"3 X I Inert particulate organic matter g C O D x i m \" 3 X S Slowly biodegradable substrate g CODxs m\"3 XsTO Organics stored by heterotrophs g CODxsto m\"3 Y H Aerobic yield of heterotrophic biomass g C O D x h (g CODxsto)\"1 Y S T O Aerobic yield of stored product per Ss g CODxsto (g CODss) - 1 Y O B S Observed yield of heterotrophic biomass g C O D x h (g CODss)\"1 P H Maximum heterotrophic growth rate d\"1 M-H.ASM1 Maximum heterotrophic growth rate as used in ASM1 d\"1 o Standard deviation -e Temperature coefficient for 'traditional' Arrhenius equation 0T Temperature coefficient for ASM3 Arrhenius equation -157 10. Glossary Autotrophic bacteria: Bacteria able to utilize CO2 as a sole source of carbon. Bleaching: Chemical treatment of a pulp to alter the colouring matter so the pulp has a higher brightness. Calender: Assembly of rolls at the dry end of the paper machine imparting a finish to paper, reducing bulk and roughness. Cellulose: A structural polysaccharide of cell walls, composed of glucose. It is the main constituent of plant fibres. Electron acceptor: A substance that accepts electron during an oxidation-reduction reaction. In aerobic systems the electron acceptor is the oxygen (O2), while in anaerobic or anoxic systems it can be NO/ , SO4\", CO3\" or C. Electron donor: A substance that donates electron during an oxidation-reduction reaction. In activated sludge systems the electron donor is the substrate (food supplier). Exocellular: Outside the cell, extracellular. Extractives: Generally, any material that can be dissolved in a solvent. In wood, it specifically refers to the tall oil or turpentine precursors that amount to 1 to 5 of the wood substance. Fibrils: Thread-like elements unravelled from the walls of native cellulose fibres. Floe: Flake formed by microorganisms within the activated sludge process. F / M : Food to microorganisms ratio. Hardwood: Wood from trees of the angiosperm class, usually with broad leaves and deciduous in temperate zones, which lose their leaves in winter. Hardwood fibres are short in relation to softwood fibres. Heterotrophic bacteria: Bacteria that obtain their energy from the oxidation of organic compounds. Hydrolysis: Decomposition of a complex molecule and reaction with water to form at least two smaller molecules. It is usually accomplished by the microorganisms with the aid of extracellular enzymes. Lignin: A highly polymerized substance which principal role is to cement the fibres together. The chemistry of lignin is extremely complex. Its structure consists primarily of phenyl propane units linked together in three dimensions. 158 MATLAB®: MATLAB® is a high-performance language for technical computing. It integrates computation, visualization, and programming in an environment where problems and solutions are expressed in mathematical notation. Mesophilic bacteria: Bacteria with a growth temperature optimum between 25 and 40°C. Readability: In a balance, readability is the value of the finest division of the scale. This term is often used interchangeably with accuracy to indicate how well a scale displays the correct results. Redox: related to the oxidation-reduction reaction. Respirometer: Instrument for measurement of the respiration rate, that is, the mass of oxygen consumed per unit of volume and unit of time. Shives: Small bundles of fibres that have not been separated completely in the pulping operations. Softwood: Wood from cone-bearing trees (also called evergreens). Softwood fibres are longer in relation to hardwood fibres. Thermophilic bacteria: Bacteria with a growth temperature optimum between 45 and 80°C. Via l : Small glass or plastic container. Xenobiotic: A chemical foreign to the body that is not endogenous or present in the body under normal conditions. 159 11. References Ali , M. , and Sreekrishnan, T. R. (2001). Aquatic toxicity from pulp and paper mill effluents: a review. Adv. Environ. Res., 5(2), 175-196. Almemark, M. , and Frostell, B. (1988). Treatment of CTMP effluents in Scandinavia. Treatment of Chemi-Thermomechanical Pulping Effluents, CPPA Environmental Conference, Vancouver, Canada. American Public Health Association, American Water Works Association, and Water Environemntal Federation. (1998). Standard Methods for the Examination of Water and Wastewater, 20th edn. Washington D.C, USA. Archibald, F., and Young, F. (2003). Common stresses affecting activated sludge health and performance - What the four-assay can tell us. 7th International Water Association Symposium in Forest Industry Wastewaters, IWA, Seatlle. Argaman, Y. (1995). A steady-state model for the single sludge activated-sludge system. 1. Model description. Water Res., 29(1), 137-145. Artan, N. , and Orhon, D. (1989). The effect of reactor hydraulics on the performance of activated-sludge systems. 2. The formation of microbial products. Water Res., 23(12), 1519-1525. Babuna, F. G., Ince, O., Orhon, D., and Simsek, A. (1998). Assessment of inert COD in pulp and paper mill wastewater under anaerobic conditions. Water Res., 32(11), 3490-3494. Barker, P. S., and Dold, P. L. (1997). General model for biological nutrient removal activated-sludge systems: model presentation. Water Environ. Res., 69(5), 969-984. Barr, T. A., Taylor, J. M. , and Duff, S. J. B. (1996). Effect of HRT, SRT and temperature on the performance of activated sludge reactors treating bleached kraft mill effluent. Water Res., 30(4), 799-810. Beccari, M. , Dionisi, D., Giuliani, A., Majone, M. , and Ramadori, R. (2002). Effect of different carbon sources on aerobic storage by activated sludge. Wat. Sci.- Tech., 45(6), 157-168. 160 Berube, P. R., and Hall, E. R. (2000). Effects of elevated operating temperatures on methanol removal kinetics from synthetic kraft pulp mill condensate using a membrane bioreactor. Water Res., 34(18), 4359-4366. Beun, J. J., Dircks, K., Van Loosdrecht, M . C. M. , and Heijnen, J. J. (2002). Poly-beta-hydroxybutyrate metabolism in dynamically fed mixed microbial cultures. Water Res., 36(5), 1167-1180. Billing, A. E., and Dold, P. (1988). Modelling techniques for biological reaction eystems. 1. Mathematical description and model presentation. Water SA, 14(4), 185-192. Box, G. E., Jenkis, G. M. , and Reinsel, G. C. (1994). Time Series Analysis: Forecasting and Control, 3rd edn. Prentice-Hall International, London. Brdjanovic, D., Van Loosdrecht, M . C. M. , Hooijmans, C. M. , Alaerts, G. J., and Heijnen, J. J. (1997). Temperature effects on physiology of biological phosphorus removal. J. Environ. Eng.-ASCE, 123(2), 144-153. Brdjanovic, D., Van Loosdrecht, M . C. M. , Versteeg, P., Hooijmans, C. M. , Alaerts, G. J., and Heijnen, J. J. (2000). Modeling COD, N and P removal in a full-scale WWTP Haarlem Waarderpolder. Water Res., 34(3), 846-858. Brouwer, H., Klapwijk, A., and Keesman, K. J. (1998). Identification of activated sludge and wastewater characteristics using respirometric batch-experiments. Water Res., 32(4), 1240-1254. Brun, R., Kuhni, M. , Siegrist, H., Gujer, W., and Reichert, P. (2002). Practical identifiability of ASM2d parameters - systematic selection and tuning of parameter subsets. Water Res., 36(16), 4113-4127. Carstensen, J., Harremoes, F., and Madsen, H. (1995). Statistical identification of Monod-kinetic parameters from online measurements. Wat. Sci. Tech., 31(2), 125-133. Carucci, A., Dionisi, D., Majone, M. , Rolle, E., and Smurra, P. (2001). Aerobic storage by activated sludge on real wastewater. Water Res., 35(16), 3833-3844. Cech, J. S., and Chudoba, J. (1983). Influence of accumulation capacity of activated sludge microorganisms on kinetics of glucose removal. Water Res., 17(6), 659-666. 161 Christie, E., and McEacher, G. (2000). Pulping the Law, How Pulp Mills are Ruining Canadian Waters, Sierra Legal Defence Fundation, Toronto. Chudoba, P., Capdeville, B., and Chudoba, J. (1992). Explanation of biological meaning of the So/Xo ratio in batch cultivation. Wat. Sci. Tech., 26(3-4), 743-751. Chui, C. K., and Chen, G. (1989). Linear Systems and Optimal Control, Springer-Verlag, New York. Chunsheng, F., and Poch, M . (1998). Fuzzy model and decision of COD control for an activated sludge process. Fuzzy Sets and Systems, 93(3), 281-292. Confer, D. R., and Logan, B. E. (1997). Molecular weight distribution of hydrolysis products during biodegradation of model macromolecules in suspended and biofilm cultures. 1. Bovine serum albumin. Water Res., 31(9), 2127-2136. Cornacchio, L. A., and Hall, E. R. (1988). Characteristics of CTMP wastewaters. Treatment of Chemi-Thermomechanical Pulping Effluents, CPPA Environmental Conference, Vancouver. Curtis, T. P., Craine, N . G., and Milner, M . G. (1998). The comparison of biological diversity in wastewater and mixed liquor in activated sludge. Fourth Kollekolle Seminar on Activated Sludge Modelling, Copenhagen, Denmark. Daigger, G. t., and Grady, C. P. L. (1982). The dynamics of microbial growth on soluble substrates. Water Res., 16, 365-382. Dalentoft, E., and Thulin, P. (1997). The use of aerobic selectors in activated sludge systems for treatment of wastewater from the pulp and paper industry. Wat. Sci. Tech., 35(2-3), 181-188. Dalzell, D. J. B., Alte, S., Aspichueta, E., de la Sota, A., Etxebarria, J., Gutierrez, M. , Hoffmann, C. C , Sales, D., Obst, U., and Christofi, N. (2002). A comparison of five rapid direct toxicity assessment methods to determine toxicity of pollutants to activated sludge. Chemosphere, 47(5), 535-545. Dircks, K., Beun, J. J., Van Loosdrecht, M. , Heijnen, J. J., and Henze, M . (2001a). Glycogen metabolism in aerobic mixed cultures. Biotechnol. Bioeng., 73(2), 85-94. 162 Dircks, K., Henze, M. , Van Loosdrecht, M . C. M. , Mosbaek, H., and Aspegren, H. (2001b). Storage and degradation of poly-beta-hydroxybutyrate in activated sludge under aerobic conditions. Water Res., 35(9), 2277-2285. Dochain, D., and Vanrolleghem, P. (2001). Dynamical Modelling and Estimation in Wastewater Treatment Processes, IWA Publishing, London, UK. Dochain, D., Vanrolleghem, P. A., and Vandaele, M . (1995). Structural identifiability of biokinetic models of activated- sludge respiration. Water Res., 29(11), 2571-2578. Dold, P., Ekama, G. A., and Marais, G. R. (1980). A general model for the activated sludge process. Prog. Water Technoi, 12, 47-77. Dold, P., and Marais, G. R. (1986). Evaluation of the general activated sludge model propose by the IAWPRC task group. Wat. Sci. Tech., 18, 63-89. Dubeski, C. V. (1993). Aerobic Treatment of CTMP Wastewater in Sequencing Batch Reactors. M.A.Sc. Thesis, Department of Bio-Resources Engineering, University of British Columbia, Vancouver. Dupont, R., and Sinkjaer, O. (1994). Optimization of waste-water treatment plants by means of computer-models. Wat. Sci. Tech., 30(4), 181-190. Eckenfelder, W. W., and Grau, P. (1992). Activated Sludge Process Design and Control-Theory and Practice, Technomic Pub. Co., Lancaster, PA, USA. Eckenfelder, W. W., Jr. (2000). Industrial Water Pollution Control, 3rd edn. McGraw-Hill, Boston. Eis, B. J., Scheffler, K. T., and Simmons, R. (1990). The use of activated sludge process models to evaluate wastewater treatment alternatives at a pulp and paper mill. TAPPI Environmental Conference, Tappi Press, Seattle, 449-464. Eliosov, B., and Argaman, Y. (1995). Hydrolysis of particulate organics in activated-sludge systems. Water Res., 29(1), 155-163. Franta, J., Helmreich, B., Pribyl, M. , Adamietz, E., and Wilderer, P. A. (1994a). Advanced biological treatment of paper-mill wastewaters - Effects of operation conditions on COD removal and production of soluble organic-compounds activated-sludge systems. Wat. Sci. Tech., 30(3), 199-207. 163 Franta, J., Wilderer, P. A., Miksch, K., and Sykora, V. (1994b). Effects of operation conditions on advanced COD removal in activated-sludge systems. Wat. Sci. Tech., 29(7), 189-192. Gapes, D. J., Frost, N. M. , Clark, T. A., Dare, P. H., Hunter, R. G., and Slade, A. H. (1999). Nitrogen fixation in the treatment of pulp and paper wastewaters. Wat. Sci. Tech., 40(11-12), 85-92. Gaudy, A. F. J., and Gaudy, E. T. (1971). Biological Concepts for Design and Operation of the Activated Sludge Process. Report No. 17090, FQJ, 09/71. US EPA, U.S. Environmental Protection Agency, Washington, D.C. Germirli, F., Orhon, D., and Artan, N . (1991). Assessment of the initial inert soluble COD in industrial wastewaters. Wat. Sci. Tech., 23(4-6), 1077-1086. Gernaey, K., Maffei, D., Vanrolleghem, P., and Verstraete, W. (1999). A new pH-based procedure to model toxic effects on nitrifiers in activated sludge. J: Chem. Technoi. Biotechnol, 74(7), 679-687. Gill, M . D., and Ross, D. M . (1990). Nutrients and temperature affect effluent suspended solids. TAPPIEnvironmental Conference, Tappi Press, Seattle, 515-518. Ginestet, P., Maisonnier, A., and Sperandio, M . (2002). Wastewater COD characterization: biodegradability of physico- chemical fractions. Wat. Sci. Tech., 45(6), 89-97. Goel, R., Mino, T., Satoh, H., and Matsuo, T. (1998a). Comparison of hydrolytic enzyme systems in pure culture and activated sludge under different electron acceptor conditions. Wat. Sci. Tech., 37(4-5), 335-343. Goel, R., Mino, T., Satoh, H., and Matsuo, T. (1998b). Intracellular storage compounds, oxygen uptake rates and biomass yield with readily and slowly degradable substrates. Wat. Sci. Tech., 38(8-9), 85-93. Goel, R., Mino, T., Satoh, H., and Matsuo, T. (1999). Modeling hydrolysis processes considering intracellular storage. Wat. Sci. Tech., 39(1), 97-105. Gostick, N . A. (1990). The nutrient requirements in biological effluent treatment. Paper Technology, August, 33-35. 164 Grady, C. P. L., Daigger, G. T., and Lim, H. (1999). Biological Wastewater Treatment, 2nd. edn. New York. Grady, C. P. L., Smets, B. F., and Barbeau, D. S. (1996). Variability in kinetic parameter estimates: a review of possible causes and a proposed terminology. Water Res., 30(3), 742-748. Greenbaum, P. J. (2002). Making progress in water reduction. Pulp Pap.-Can., 103(2), 10-13. Gujer, W., Henze, M. , Mino, T., Matsuo, T., Wentzel, M. C , and Vonmarais, G. (1995). The Activated-Sludge Model No. 2 - Biological phosphorus removal. Wat. Sci. Tech., 31(2), 1-11. Gujer, W., Henze, M. , Mino, T., and Van Loosdrecht, M . (1999). Activated Sludge Model No. 3. Wat. Sci. Tech., 39(1), 183-193. Hanada, S., Satoh, H., and Mino, T. (2002). Measurement of microorganisms with PHA production capability in activated sludge and its implication in Activated Sludge Model No. 3. Wat. Sci. Tech., 45(6), 107-113. Hangos, K. M. , and Cameron, I. T. (2001). Process Modelling and Model Analysis, 1st. Academic Press, London. Helle, S. (1999). A Respirometric Investigation of the Activated Sludge Treatment of BKME During Steady State and Transient Operating Conditions. PhD. Thesis, Department of Chemical Engineering, University of British Columbia, Vancouver, Canada. Helle, S., and Duff, S. J. B. (1997). Effect of shock loads on the activated sludge treatment of BKME. 83rd Annual Meeting, Technical Section CPPA. Helle, S., and Duff, S. J. B. (2003). Multi-component kinetics of activated sludge treatment of bleached kraft mill effluent. 7th International Water Association Symposium on Forest Industry Wastewaters, Seattle. Henze, M . (1992). Characterization of wastewater for modeling of activated sludge processes. Wat. Sci. Tech., 25(6), 1-15. Henze, M. , Grady, C. P. L., Gujer, W., Marais, G. R., and Matsuo, T. (1987). Activated Sludge Model No. 1. Sci. and Tech. Rep. No. 1, Int. Assoc. Water Qual., London, U.K. 165 Henze, M. , Gujer, W., Mino, T., Matsuo, T., Wentzel, M . C., Marais, G. V. R., and Van Loosdrecht, M . C. M . (1999). Activated Sludge Model No. 2d, ASM2d. Wat. Sci. Tech., 39(1), 165-182. Henze, M. , Gujer, W., Mino, T., and van Loosdrecht, M . (2000). Activated Sludge Models ASM1, ASM2, ASM2d and ASM3. Scientific and Technical Report No. 9, IWA publishing, London. Hulsbeek, J. J. W., Kruit, J., Roeleveld, P. J., and Van Loosdrecht, M . C M . (2002). A practical protocol for dynamic modelling of activated sludge systems. Wat. Sci. Tech., 45(6), 127-136. Hynninen, P. (1998). Environmental Control, Finnish Paper Engineers' Association and TAPPI, Jyvaskyla, Finland. Hynninen, P., and Viljakainen, E. (1995). Nutrient dosage in biological treatment of wastewaters. TappiJ., 78(5), 105-108. India Paper. (2003). History of Paper, http://www.indiapapermarket.com/history.asp. Insel, G., Gui, O. K., Orhon, D., Vanrolleghem, P. A., and Henze, M . (2002). Important limitations in the modeling of activated sludge: biased calibration of the hydrolysis process. Wat. Sci. Tech., 45(12), 23-36. Janning, K. F., Le Tallec, X., and Harremoes, P. (1998). Hydrolysis of organic wastewater particles in laboratory scale and pilot scale biofilm reactors under anoxic and aerobic conditions. Wat. Sci. Tech., 38(8-9), 179-188. Jeppsson, U., and Olsson, G. (1993). Reduced-order models for online parameter identification of the activated sludge process. Wat. Sci. Tech., 28(11-12), 173-183. Kantardjieff, A., Jones, J. P., and Germain, M . (1997). The Canadian pulp and paper industry and the new ISO 14000 environmental standards. Pulp Pap.-Can., 98(1), 28-30. Kappeler, J., and Gujer, W. (1992). Estimation of kinetic parameters of heterotrophic biomass under aerobic conditions and characterization of wastewater for activated sludge modeling. Wat. Sci. Tech., 25(6), 125-139. Kaprelyants, A. S., Gottschal, J. C , and Kell, D. B. (1993). Dormancy in non-sporulating bacteria. Ferns Microbiol. Rev., 104(3-4), 271-286. 166 Kaprelyants, A. S., and Kell, D. B. (1996). Do bacteria need to communicate with each other for growth? Trends Microbiol, 4(6), 237-242. Karahan-Gul, O., Artan, N. , Orhon, D., Henze, M. , and Van Loosdrecht, M . C. M . (2002a). Experimental assessment of bacterial storage yield. J. Environ. Eng.-ASCE, 128(11), 1030-1035. Karahan-Gul, O., Artan, N. , Orhon, D., Henze, M. , and Van Loosdrecht, M . C. M . (2002b). Respirometric assessment of storage yield for different substrates. Wat. Sci. Tech., 46(1-2), 345-352. Keesman, K. J., and Spanjers, H. (2000). Endogenous model state and parameter estimation from an extensive batch experiment. Biotechnol. Bioeng., 68(4), 422-429. Keesman, K. J., Spanjers, H., and Van Straten, G. (1998a). Analysis of endogenous process behavior in activated sludge. Biotechnol Bioeng., 57(2), 155-163. Keesman, K. J., Spanjers, H., and Van Straten, G. (1998b). Analysis of endogenous process behaviour. Wat. Sci. Tech., 37(12), 227-235. Kell, D. B., Kaprelyants, A. S., Weichart, D. H., Harwood, C. R., and Barer, M . R. (1998). Viability and activity in readily culturable bacteria: a review and discussion of the practical issues. Antonie Van Leeuwenhoek, 73(2), 169-187. Koch, G., Kuhni, M. , Gujer, W., and Siegrist, H. (2000). Calibration and validation of Activated Sludge Model No. 3 for Swiss municipal wastewater. Water Res., 34(14), 3580-3590. Kong, Z., Vanrolleghem, P., and Verstraete, W. (1994). Automated respiration inhibition-kinetics analysis (ARIKA) with a respirographic biosensor. Wat. Sci. Tech., 30(4), 275-284. Kong, Z., Vanrolleghem, P., Willems, P., and Verstraete, W. (1996). Simultaneous determination of inhibition kinetics of carbon oxidation and nitrification with a respirometer. Water Res., 30(4), 825-836. Koning, A. F., Cross, L., Sikes, J. E. G., and Turk, O. (1994). From conception to resolution: effluent treatment upgrade at Alberni pulp & paper to meet special quality regulations. Pulp Pap.-Can., 95(12), 506-510. 167 Kostamo, A., Holmbom, B., and Kukkonen, J. V. K. (2003). Removal of wood extractives from pulp mill effluents by activated sludge treatment. 7th International Water Association Symposium on Forest Industry Wastewaters, Seattle, USA. Krishna, C , and Van Loosdrecht, M . C. M . (1999). Substrate flux into storage and growth in relation to activated sludge modeling. Water Res., 33(14), 3149-3161. Larisch, B. C , and Duff, S. J. B. (1996). Activated sludge treatment of TCF bleached kraft mill effluent. 5th IA WQ Symposium on Forest Industry Wastewaters, Vancouver, Canada. Larrea, L., Garciaheras, J. L., Ayesa, E., and Florez, J. (1992). Designing experiments to determine the coefficients of activated-sludge models by identification algorithms. Wat. Sci. Tech., 25(6), 149-165. Lishman, L. A., and Murphy, K. L. (1994). The significance of hydrolysis in microbial death and decay. Water Res., 28(11), 2417-2419. Lukasse, L. J. S., Keesman, K. J., and Van Straten, G. (1997). Estimation of short-term BOD, respiration rate and kinetics of activated sludge. Water Res., 31(9), 2278-2286. Madigan, M . T., Martinko, J. M. , and Parker, J. (1997). Brock Biology of Microorganisms, 8th. edn. Prentice-Hall, Inc., Upper Saddle River, New Jersey. Makinia, J., and Wells, S. A. (2000). A general model of the activated sludge reactor with dispersive flow -1. Model development and parameter estimation. Water Res., 34(16), 3987-3996. Mamais, D., Jenkins, D., and Pitt, P. (1993). A rapid physical-chemical method for the determination of readily biodegradable soluble COD in municipal wastewater. Water Res., 27(1), 195-197. Mao, H. Z., and Smith, D. W. (1995). A mechanistic model for assessing biodegradability of complex wastewaters. Water Res., 29(8), 1957-1975. Marsili-Libelli, S., and Tabani, F. (2002). Accuracy analysis of a respirometer for activated sludge dynamic modelling. Water Res., 36(5), 1181-1192. Mathys, R. G. (1991). Aerobic Treatment of a CTMP Wastewater Using a Rotating Biological Contactor. M.A.Sc. Thesis, Department of Chemical Engineering, University of British Columbia, Vancouver. 168 McAllen, V. E. (1988). Treatment of chemi-thermomechanical pulping effluents. Treatment of chemi-thermomechanical pulping effluents, CPPA Environmental Conference, Vancouver, Canada. Meijer, S. C. F., Van Loosdrecht, M . C. M. , and Heijnen, J. J. (2001). Metabolic modelling of full-scale biological nitrogen and phosphorus removing WWTP's. Water Res., 35(11), 2711-2723. Metcalf & Eddy. (2003). Wastewater Engineering, Treatment and Reuse, 4th edn. McGraw Hill, Boston. Mino, T., San Pedro, D. C , and Matsuo, T. (1995). Estimation of the rate of slowly biodegradable COD (sbCOD) hydrolysis under anaerobic, anoxic and aerobic conditions by experiments using starch as model substrate. Wat. Sci. Tech., 31(2), 95-103. Mobius, C. H. (1989). Biological treatment of papermill waste-water in an activated- sludge cascade reactor. Wat. Sci. Tech., 21(12), 1825-1828. Morgenroth, E., Kommedal, R., and Harremoes, P. (2002). Processes and modeling of hydrolysis of particulate organic matter in aerobic wastewater treatment - a review. Wat. Sci. Tech., 45(6), 25-40. Mujunen, S. P., Minkkinen, P., Teppola, P., and Wirkkala, R. S. (1998). Modeling of activated sludge plants treatment efficiency with PLSR: a process analytical case study. Chemometrics Intell. Lab. Syst., 41(1), 83-94. NCASI. (1986). Simulation of Organic Compound Removal in Biological Wastewater Treatment Processes. Technical Bulletin N° 511, New York. NCASI. (1990). Field Verification of the NCASI Predictive Model of Organic Compound Removal by Biological Wastewater Treatment Processes. Technical Bulletin N° 582, New York. NCASI. (2000). Literature Review on Micronutrients in Aerobic Biological Wasteater Treatment. Technical Bulletin N°818, New York. Nowak, O., and Svardal, K. (1993). Observations on the kinetics of nitrification under inhibiting conditions caused by industrial waste-water compounds. Wat. Sci. Tech., 28(2), 115-123. Nowak, O., Svardal, K., and Schweighofer, P. (1995). The dynamic behavior of nitrifying activated-sludge systems influenced by inhibiting waste-water compounds. Wat. Sci. Tech., 31(2), 115-124. Olsson, G., and Newell, R. B. (1999). Wastewater Treatment Systems: Modelling, Diagnosis and Control, IWA publishing, London. Orhon, D., Cokgor, E. U., and Sozen, S. (1998). Dual hydrolysis model of the slowly biodegradable substrate in activated sludge systems. Biotechnol. Tech., 12(10), 737-741. Orhon, D., Cokgor, E. U., and Sozen, S. (1999a). Experimental basis for the hydrolysis of slowly biodegradable substrate in different wastewaters. Wat. Sci. Tech., 39(1), 87-95. Orhon, D., Karahan, O., and Sozen, S. (1999b). The effect of residual microbial products on the experimental assessment of the particulate inert COD in wastewaters. Water Res., 33(14), 3191-3203. Pedersen, J. (1992). Controlling activated-sludge process using EFOR. Wat. Sci. Tech., 26(3-4), 783-790. Pedersen, J., and Sinkjaer, O. (1992). Test of the activated-sludge models capabilities as a prognostic tool on a pilot scale waste-water treatment-plant. Wat. Sci. Tech., 25(6), 185-194. Petersen, B., Gernaey, K., and Vanrolleghem, P. A. (2001). Practical identifiability of model parameters by combined respirometric-titrimetric measurements. Wat. Sci. Tech., 43(7), 347-355. Pollard, P. C , and Greenfield, P. F. (1997). Measuring in situ bacterial specific growth rates and population dynamics in wastewater. Water Res., 31(5), 1074-1082. Pulp & Paper International. (2001). Annual Review. 43(7). Ralston, M. , and Jennrich, R. (1978). Dud, a derivative-free algorithm for non linear least squares. Technometrics, 20(1), 7-14. Randall, E. W., Wilkinson, A., and Ekama, G. A. (1991). An instrument for the direct determination of oxygen utilization rate. Water SA, 17(1), 11-18. 170 Reichert, P. (1994). Concepts Uderlying a Computer Program for the Identification and Simulation of Aquatic Systems. PhD. Thesis, Swiss Federal Institute of Environmental Science and Technology (EAWAG), Dubendorf, Switzerland. Rieger, L., Koch, G., Kuhni, M. , Gujer, W., and Siegrist, H. (2001). The EAWAG Bio-P module for Activated Sludge Model No. 3. Water Res., 35(16), 3887-3903. Roeleveld, P. J., and Van Loosdrecht, M . C. M . (2002). Experience with guidelines for wastewater characterisation in The Netherlands. Wat. Sci. Tech., 45(6), 77-87. Sawyer, C. N. , McCarty, P. L., and Parkin, G. F. (1994). Chemistry for Environmental Engineering, 4th. edn. McGraw-Hill, Inc. Schroderus, S., Ma, H., Munger, L., and Peters, N . (2002). Practical aspects of energy conservation in pulp and paper mills. Pulp Pap.-Can., 103(4), 36-39. Sheffer, M . S., Hiraoka, M. , and Tsumura, K. (1984). Flexible modelling of the activated sludge system - Theoretical and practical aspects. Wat. Sci. Tech., 17, 247-258. Siegrist, H., Rieger, L., Koch, G., Kuhni, M. , and Gujer, W. (2002). The EAWAG Bio-P module for Activated Sludge Model No. 3. Wat. Sci. Tech., 45(6), 61-76. Siegrist, H., and Tschui, M . (1992). Interpretation of experimental-data with regard to the Activated-Sludge Model No 1 and calibration of the model for municipal waste-water treatment plants. Wat. Sci. Tech., 25(6), 167-183. Slade, A. H. (2003). Results from research published in 1992. Personal communication: e-mail to P. Baranao. Slade, A. H., Dare, P. H., and Leonard, A. M . (1991). Modelling aerated lagoons treating bleached kraft mill effluents. Determination of model inputs. TAPPI Environmental Conference, Tappi Press, San Antonio, Texas, 223-234. Smook, G. A. (2002). Handbook for Pulp & Paper Technologists, 3rd. edn. Angus Wilde Publications Inc., Vancouver. Sollfrank, U., and Gujer, W. (1991). Characterization of domestic wastewater for mathematical modeling of the activated sludge process. Wat. Sci. Tech., 23(4-6), 1057-1066. 171 Sollfrank, U., Kappeler, J., and Gujer, W. (1992). Temperature effects on waste water characterization and the release of soluble inert organic material. Wat. Sci. Tech., 25(6), 33-41. Sozen, S., Cokgor, E. U., Orhon, D., and Henze, M . (1998). Respirometric analysis of activated sludge behaviour - II. Heterotrophic growth under aerobic and anoxic conditions. Water Res., 32(2), 476-488. Spanjers, H., and Keesman, K. (1994). Identification of wastewater degradation kinetics. Proceedings of the Third IEEE Conference on Control Alplications, IEEE, Glasgow, Scotland, 1011-1016. Spanjers, H., Olsson, G., and Klapwijk, A. (1994). Determining short-term biochemical oxygen-demand and respiration rate in an aeration tank by using respirometry and estimation. Water Res., 28(7), 1571-1583. Spanjers, H., and Vanrolleghem, P. (1995). Respirometry as a tool for rapid characterization of wastewater and activated sludge. Wat. Sci. Tech., 31(2), 105-114. Springer, A. M . (2000). Industrial Environmental Control, Pulp and Paper Industry, 3rd edn. Tappi Press, New York. Sreckovic, G. (2001). Modelling Activated Sludge Treatment of Pulp and Paper Wastewater. Ph.D. Thesis, Department of Civil Engineering, University of British Columbia, Vancouver. Stanyer, D. J. (1997). Modelling Carbon Oxidation in Pulp Mill Activated Sludge Systems: Determining Model Parameters. M.A.Sc. Thesis, Department of Civil Engineering, University of British Columbia, Vancouver. Steffens, M . A., Lant, P. A., and Newell, R. B. (1997). A systematic approach for reducing complex biological wastewater treatment models. Water Res., 31(3), 590-606. Stratton, S., Gleadow, P., and Johnson, A. P. (2003). Pulp mill process closure: A review of global technology developments and mill experiences in the 1990s. 7th International Water Association Symposium on Forest Industry Wastewaters, Seattle. Sundholm, J. (1999). Mechanical Pulping, Finninsh Paper Engineers' Association and TAPPI, Helsinky, Finland. 172 Torrijos, M. , Cerro, R. M. , Capdeville, B., Zeghal, S., Payraudeau, M. , and Lesouef, A. (1994). Sequencing batch reactor: a tool for wastewater characterization for the IAWPRC model. Wat. Sci. Tech., 29(7), 81-90. Tulleken, H. (1992). Grey-box Modelling and Identification Topics. PhD, Delft University of Technology, Holland. Turk, S. (1988). Options for treatment of CTMP effluents. Treatment of chemi-thermomechanical pulping effluents, CPPA Environmental Conference, Vancouver. Ubisi, M . F., Jood, T. W., Wentzel, M . C , and Ekama, G. A. (1997). Activated sludge mixed liquor heterotrophic active biomass. Water SA, 23(3), 239-248. Van Aalast-Van Leeuwen, M . A., Pot, M . A., Van Loosdrecht, M . C. M. , and Heijnen, J. J. (1997). Kinetic modeling of poly(beta-hydroxybutyrate) production and consumption by Paracoccus pantotrophus under dynamic substrate supply. Biotechnol. Bioeng., 55(5), 773-782. Van Loosdrecht, M . C. M. , and Heijnen, J. J. (2002). Modelling of activated sludge processes with structured biomass. Wat. Sci. Tech., 45(6), 13-23. Van Loosdrecht, M . C. M. , and Henze, M . (1999). Maintenance, endogeneous respiration, lysis, decay and predation. Wat. Sci. Tech., 39(1), 107-117. Van Loosdrecht, M . C. M. , Pot, M . A., and Heijnen, J. J. (1997). Importance of bacterial storage polymers in bioprocesses. Wat. Sci. Tech., 35(1), 41-47. Van Veldhuizen, H. M. , Van Loosdrecht, M . C. M. , and Heijnen, J. J. (1999). Modelling biological phosphorus and nitrogen removal in a full scale activated sludge process. Water Res., 33(16), 3459-3468. Vanrolleghem, P., and Vandaele, M . (1994). Optimal experimental design for structure characterization of biodegradation models - online implementation in a respirographic biosensor. Wat. Sci. Tech., 30(4), 243-253. Vanrolleghem, P. A., Kong, Z. D., and Coen, F. (1996). Full-scale on-line assessment of toxic wastewaters causing change in biodegradation model structure and parameters. Wat. Sci. Tech., 33(2), 163-175. 173 Vanrolleghem, P. A., Schilling, W., Rauch, W., Krebs, P., and Aalderink, H. (1999a). Setting up measuring campaigns for integrated wastewater modelling. Wat. Sci. Tech., 39(4), 257-268. Vanrolleghem, P. A., Spanjers, H., Petersen, B., Ginestet, P., and Takacs, I. (1999b). Estimating (combinations of) Activated Sludge Model No. 1 parameters and components by respirometry. Wat. Sci. Tech., 39(1), 195-214. Vanrolleghem, P. A., Vandaele, M. , and Dochain, D. (1995). Practical identifiability of a biokinetic model of activated sludge respiration. Water Res., 29(11), 2561-2570. Volskay, V. T., and Grady, C. P. L. (1988). Toxicity of selected RCRA compounds to activated sludge microorganisms. Journal of the Water Pollution Control Federation, 60, 1850-1856. Volskay, V. T., and Grady, C P . L. (1990). Respiration inhibition kinetic analysis. Water Res., 24(7), 863-874. Wanner, O., Kappeler, J., and Gujer, W. (1992). Calibration of an activated-sludge model based on human expertise and on a mathematical optimization technique - A comparison. Wat. Sci. Tech., 25(6), 141-148. Weijers, S. R., Kok, J. J., Preisig, H. A., Buunen, A., and Wouda, T. W. M . (1996). Parameter identifiability in the IAWQ model No. 1 for modelling activated sludge plants for enhanced nitrogen removal. Comput. Chem. Eng., 20, S1455-S1460. Weijers, S. R., and Vanrolleghem, P. A. (1997). A procedure for selecting best identifiable . parameters in calibrating Activated Sludge Model No.l to full-scale plant data. Wat. Sci. Tech., 36(5), 69-79. Wentzel, M . C , and Ekama, G. A. (1997). Principles in the design of single-sludge activated sludge systems for biological removal of carbon, nitrogen, and phosphorus. Water Environ. Res., 69(7), 1222-1231. Wentzel, M . C , Ekama, G. A., Loewenthal, R. E., Dold, P. L. , and Marais, G. R. (1989). Enhanced polyphosphate organism cultures in activated-sludge systems. 2. Experimental behavior. Water SA, 15(2), 71-88. 174 Wentzel, M . C , Mbewe, A., Lakay, M . T., and Ekama, G. A. (1999). Batch test for characterisation of the carbonaceous materials in municipal wastewaters. Water SA, 25(3), 327-335. Wilson, R. W., and Frenette, E. G. (1988). Development of anaerobic/aerobic treatment at consolidated-Bathurst Inc. Treatment of Chemi-Thermomechanical Pulping Effluents, CPPA Environmental Conference, Vancouver, Canada. Winkler, S., Muller-Rechberger, H., Nowak, O., Svardal, K., and Wandl, G. (2001). A new approach towards modelling of the carbon degradation cycle at two-stage activated sludge plants. Wat. Sci. Tech., 43(7), 19-27. 175 Appendix 1: Relationship between MOBS and u H The dynamic oxygen uptake rate in a batch tests with neither substrate nor oxygen limitation is, according to ASM3: OUR (t) = -[(l-YsTo)-ksTO + ((l-YH)/YH)-pH-FsTo(t) + bH]-XH(t) - bSTo-XsTo(t) [ALI] Being FsTo(t) equal to XSTo(t)/XH(t) / ( K S T o +XS To(t)/XH(t)) On the other hand, the mass balance for the heterotrophic biomass with neither substrate nor oxygen limitation is, according to ASM3: dXH/dt = (pH-FsTo(t)- bH) -XH(t) [A1.2] Assuming FSTO is constant, which is a good assumption in Region 2 of Figure A 1.1, integration of Equation Al.2 leads to: XH(t) = X H , o - e ( ( ^ F s t o - b h ) * t [A1.3] 400 1 T3 300 -CO E CM o 200 -OJ 100 -Figure A l . l : Respirograph simulated with high F/M Assuming b H = b S T 0 , being X H E T = X H + X S T O ; replacing Equation A 1.3 into Equation A l . l ; and rearranging, Equation A 1.4 is obtained. 176 OUR (t) = - [ ( l-YsTo)-ksTO + ((l-YH)/YH)-HH-FsTO(t))]- X H ,o• e - bH-XHETbs-bh).t. bH-XHET(t) [A1.5] Comparing Equation A1.4 with the equivalent obtained based on ASM1 (Equation A1.5) is noticed that the term bH\"XHET(t) in Equation A 1.4 is common to both equations, so subtracting it and dividing OUR(t) by OUT(t=0), it is obtained: OUR(t)/OUR(t0)ASM3 = e [A1.6] OUR(t)/OUR(to)ASMi = e «»«>bs-bh>,t [Al.7] Making Equations A1.6 and A1.7 equal leads to: PH'FsTO -bH = POBS-bH [Al .8] Since FSTO is always lower than 1, it is concluded that PH>POBS [A1.9] The value of FSTO depends mainly on two factors: the ksTo/ PH ratio, and the F/M ratio. The higher the ksTo/ PH ratio the higher FSTO; and the ; and also the higher the F/M ratio the higher FSTO-Figure A l . l presents a simulated respirograph, which can be divided into five stages. In stagel XSTO is being generated, until a relatively 'steady-state' is achieved. During stage 2, the rate of storage is relatively equal to the rate of growth, and the value of FSTO is relatively constant. During stage 3, the lack of Ss makes the storage rate decrease, until Ss is depleted at the point when OUR is maximum. During stage 4 the XSTO is being used for growth until it is depleted, and endogenous respiration state is finally achieved with stage 5. 177 Appendix 2: Sensitivity Analysis Appendix 2.1 Sensitivity function for MATLAB® used in analysis % SENSITIVITY OF ASM3 MODEL % % S e n s i t i v i t y Calculations % % Pablo Baranao, December 2002, based on Bob Newell, 1996 % % K i n e t i c Parameters % par(l) = kh; par(2) Kx; par(3) = ksto % par(4) Ko; par(5) Ks; par(6) = Ksto; % par(7) = muh; par(8) = Knh4; par(9) = bh; % par(10) = bsto; p a r ( l l ) = mua; par(12) = Kanh4; % par(13) = Kao; par(14) = ba; % 2-Stoichiometric Parameters % par(15) = f s i ; par(16) = Ysto; par(17) = Yh; % par(18) = Ya; par(19) = f x i ; par(2 0) = i n S i ; % par(21) = inSs ; par(22) = in X i ; par(23) = inXs; % par(24) = inBM; par(25) = i s s X i ; par(26) = issXs % par(27) = issBM; par(28) = issSto; % 3-Arrhenius thetas (based on 20°C) T> 0. \"5 par(29) = theta kh; par( 30) = theta ksto; > % par(31) = theta muh; par( 32) = theta bh; % par(33) = theta mua; par(34) = theta ba,-% % Oxygen concentration (only f o r s e n s i t i v i t y analysis) % par(35) = So % CTMP estimated parameters values par(l:6) = [ 3 1 5 0.2 1 10] ; par(7:14) = [6 0.01 0.1 0.1 1 1.5 0.5 0.15]; par(15:20) = [0.1 0.85 0.63 0.24 0.2 0.01]; 178 par(21:28) = [0.03 0.02 0.04 0.07 0.75 0.75 0.90 0.60]; par(29:35) = [0.045 0.045 0.045 0.045 0.105 0.105 1]; par=par'; parO = par; % dynamic s e n s i t i v i t y c a l c u l a t i o n s % smlA = [] ; h i = 1 Tank IA ' ; disp( [ h i ] ) ; h i = 1 Si Ss Snh4 Snox X i Xs Xh Xsto Xa 1 ; disp( [ h i ] ) , for i = 1:35, smp = [ i ] ; parsave = par(i) ; par(i) = 1.01 * parsave ; x0 = [90 50 2.7 6 650 150 3500 900 10]; % i n i t i a l concentrations smp = [ smp ASM3_sens( xO, par, 33 )/10 - ASM3_sens( xO, parO, 33 )/10 ] par(i) = parsave ; smlA = [ smlA; smp ] ; end ; disp( smlA ) % % TANK IB smlB = [] '; h i = ' Tank IB 1 ; disp( [ h i ] ) ; h i = ' Si Ss Snh4 Snox X i Xs Xh Xsto Xa' ,-disp( [ h i ] ) for i = 1:35, smp = [ i ] ,-parsave = par(i) ; par(i) = 1.01 * parsave ; xO = [90 0 0.5 8 650 120 3500 600 10]; % i n i t i a l concentrations 179 smp = [ smp ASM3_sens( xO, par, 3 3 )/10 - ASM3_sens( xO, parO, 33 )/10 ] f par(i) = parsave ; sralB = [ smlB; smp ] ; end ; disp( smlB ) % % TANK 2 sm2 = [ ] ; h i = ' Tank 2 ' ; disp( [ h i ] ) ; h i = ' S i Ss Snh4 Snox X i Xs Xh Xsto Xa 1 ; disp( [ h i ] ) , for i = 1:35, smp = [ i ] ; parsave = par(i) ; par(i) = 1.01 * parsave ; xO = [90 0 0.2 10 650 100 3500 300 10]; % i n i t i a l concentrations smp = [ smp ASM3_sens( xO, par, 33 )/l0 - ASM3_sens( xO, parO, 33 )/10 ] par(i) = parsave ; sm2 = [ sm2; smp ] ; end ; disp( sm2 ) % . % TANK 3 sm3 = [] ; h i = 1 Tank 3 ' ; disp( [ h i ] ) ; h i = ' S i Ss Snh4 Snox X i Xs Xh Xsto Xa' ; disp( [ h i ] ) , for i = 1:35, smp = [ i ] ; parsave = par(i) ; par(i) = 1.01 * parsave ; 180 xO = [90 0 0.1 10 650 80 3500 100 10]; % i n i t i a l concentrations smp = [ smp ASM3_sens( xO, par, 33 )/10 - ASM3_sens(. xO, parO, 33 )/10 ] par(i) = parsave ; sm3 = [ sm3; smp ] ; end ; disp( sm3 ) % 1 % TANK 4 sm4 = [] ; h i = 1 Tank 4 1 ; disp( [ h i ] ) ; h i = ' S i Ss Snh4 Snox X i Xs Xh Xsto Xa' ; disp( [ h i ] ) , for i = 1:35, smp = [ i ] ; parsave = par(i) ; par(i) = 1.01 * parsave ; xO = [90 0 0.1 10 650 75 3500 0 10]; % i n i t i a l concentrations smp = [ smp ASM3_sens( xO, par, 33 )/10 - ASM3_sens( xO, parO, 33 )/10 ] par(i) = parsave ; sm4 = [ sm4,- smp ] ; end ; disp( sm4 ) 181 function dx=ASM3(x, par, T); % ASM3 i s an ASM3 d i a l e c t that only take i n account the aerobic processes % The output, dx, are the derivatives of a l l the states. % Syntax: dx=pbdmodell(x, xin, g, par, v o l , T) % % Pablo Baranao, UBC, November 2 002 % _ _ % Arrhenius kh = par(l) * exp(par(2 9) * (T- 20) ) ; ksto = • par(3) * exp(par(3 0) * (T- 20).) ; muh = par(7) * exp(par(31) * (T- 20) ) ; bh = par(9) * exp(par(32) * (T- 20) ) ; bsto = par(10) * exp(par(32) * (T- 20) ) ; mua = p a r ( l l ) * exp(par(33) * (T- 20) ) ; ba = par(14) * exp(par(34) * (T-20) ) ; % processes rates % PI Hydrolysis p i = kh * x(6)/x(7) / ( par(2) + x(6)/x(7) ) * x(7); % P2 Aerobic storage of Ss p2 = ksto * (par(35) / (par(4) + par(35))) * (x(2) / (x(2)+par(5))) * x(7); % P3 Aerobic growth of Xh p3 = muh * (par(35) / (par(4) + par(35))) * (x(3) / (par(8) +x(3))) ... * (x(8)/x(7) / (par(6) + x(8)/x(7))) * x(7); % % P4 Aerobic endog. r e s p i r a t i o n p4 = bh * (par(35) / (par(4) + par(35))) * x(7); % P5 Aerobic r e s p i r a t i o n of Xsto p5 = bsto * (par(35) / (par(4) + par(35))) * x(8); % P6 Aerobic growth of Xa p6 = mua * (par(35) / (par(13) + par(35))) * (x(3) / (par(12)+x(3))) * x(9) % P7 Aerobic endog. r e s p i r a t i o n 182 p7 = ba * (par(35) / (par(13) + par(35))) * x(9); % d e f f e r e n t i a l equations dx(l)= par(15)*pl; % S i dx(2)= p i - p2 ; %Ss dx(3)= (par(23)-par(21))*pl + par(21)*p2 - par(24)*p3 + (par(24)-par(19)*par(22))*p4 ... - (par(24) + 1/par(18))*p6 + (par(24)-par(19)*par(22))*p7; %Snh dx(4)= p6/par(18); %Snox dx(5)= par(19)*p4 + par(19)*p7; %Xi dx(6)= p i ; %Xs dx(7)= p3 - p4; %Xh dx(8)= par(16)*p2 - p3/par(17) - p5; %Xsto dx(9)= p6 - p7; %Xa dxdt=[dx(l) dx(2) dx(3) dx(4) dx(5) dx(6) dx(7) dx(8) dx(9)]; % END 183 CO D > 3 (ZI a o \"O 0) co at X> co O cd \"3 T3 O o > N co C (D 00 < H o H • M X X X oo 0 3 6 3 X Q O U 60 s s Q O 60 X Q O CJ 60 Q O CJ 60 Q O CJ 60 Q O CJ 60 60 U 3 > B r t 00 1/1 NO r f IT) r f l/-> in NO CN r f ro O 5 9 t-- ^ l - H oo NO o NO NO m r f o o o o r f IT) o o o 00 r f ON CN O ~ * l/-) 00 NO tN ro CN ON CN CN r f ON ON t--CN o NO I O NO X Q O CJ 60 O 60 Q O CJ 60 1—H o (N 00 1-H 1 m NO r f r~- CN m *~—t ro ON in NO O ON ON 00 O CN O CN CN J - H r f CN fN. O \"O r f OO O o 00 r f m t—I O ON 00 00 • 00 in ON ro ro o • o • O (N O NO r f in oo ro NO r f O o o o X Q O U x Q O U CN 60 60 X Q O CJ 60 a o CJ 60 a o u 60 X Q O U 60 o CD x X X X 00 ICQ CO o ICO 00 v. H-» I O H 6 a x Q O U 60 X Q O U bo X Q O V bO Q O U. bO Q O U bD Q O U bO l o bO 4) > I 5 o OS CN d d O ^ H Os CN ^ d cn ro o - H r -H O OO T T CN 00 cn CN 00 T T CN 00 cn CN CN d CN d 8 q ^ oo cn vo vov r-CN o T T vo vd cn d .—i un 00 Os 00 CN i un oo VO 2 00 un oo CN un o m . r i un x Q O U bo bo Q O U bO ac x> o x o CN ^ r-J CN cn d un x Q O bO OJO o X Q v O un Os CN un CN d oo o CN d o CM o d un Os CN un CN d oo o ^ H O r-Os o X Q O U bo Q O U bO x Q O U bo x Q O U bO CN O VO VO © 2 2 9 cn un Os Os VD un 00 cn vo CN r-un - H CN o un o CN O o i-54 ac © «D Appendix 3: Sampling Campaigns Descriptions Three measurement campaigns were undertaken in order to calibrate the model, which are described in detail in the following sections. Appendix 3.1: Sampling campaign N°1 The first sampling campaign was undertaken between January 20 th and January 24th, 2002. The first campaign was completed over a five days period, to perform a screening of many parameters along the complete WWTP together with the measurement of some model parameters. Six sampling points were selected and samples were taken twice a day at every point. The samples were taken considering the HRT of the different compartments of the WWTP in order to group different samples into sequences. Each sequence was characterized based on the time of the first sample, which corresponded to the influent. Table A3.1 presents the chronology of this sampling campaign, indicating the sequence that corresponds to each sample together based on its time and sampling location. Table A3.1: First sampling campaign chronology Influent Cell IA Cell 4 Recycle Wastage Effluent Day Time Sequences 21 9:00 1 1 21 15:00 2 2 1 21 20:45 2 1 1 1 22 9:00 3 3 2 2 2 22 14:15 4 4 3 22 22:30 4 3 3 3 23 8:30 5 5 4 4 4 23 14:30 6 6 5 • 23 21:00 6 5 5 5 24 9:00 7 7 6 6 6 24 14:10 7 24 18:15 7 186 Table A3.2 presents the analytical measurements performed to each sample, depending on its location source and whether or not filtration was performed. The tests included chemical oxygen demand (COD), ammonia (NH3), total Kjeldahl nitrogen (TKN), total phosphorus (TP), alkalinity, nitrite (NO2\"), nitrate (NO3\"), ortho phosphate (P04), total suspended solids (TSS), and volatile suspended solids (VSS). Table A3.2: Daily measurement matrix Source Filtration COD NH 3 TKN TP Alkalinity NOx P0 4 TSS VSS Influent Yes No X X X X X X X X X X X Cell IA Yes No X X X X X X X X X X X Cell 4 Yes No X X X X X X X X X X Recycle Yes No X X X X X X X X X X Wastage Yes No X X X X X X X X X X Effluent Yes No X X X X X X X X X X X Appendix 3.2: Sampling campaign N°2 The second sampling campaign was undertaken between February 24 th and February 26th, 2002. The major objective of this campaign was to take fresh samples of activated sludge and influent in order to perform respirometric batch tests. Only samples from Cell 4 and influent were used for this purpose. In addition, samples from influent and effluent were taken to UBC in order to perform long-term BOD tests. COD, COD s , TSS, and VSS were also measured for all these samples. 187 Appendix 3.3: Sampling campaign N°3 The third sampling campaign was undertaken by Ing-Wei Lo on April 20 th. It consisted only on three samples (influent, Cell 4, and effluent), which were used to perform additional respirometric tests, a repetition of the long-term BOD test, and to estimate the readily biodegradable COD based on the method proposed by Mamais et al. (1993). COD, COD s , TSS, and VSS were also measured for all these samples. 188 Appendix 4: Summary of respirometric tests Appendix 4.1 Data from batch tests performed mixing wastewater with sludge Tables A4.1 to A4.10 present the respirometric information obtained from batch tests performed mixing wastewater with sludge, as explained in Section 3.7.1.2. The row in bold represents the breakpoint on that respirograph. Table A4.1: Respirometric information obtained from Test 2-1 Time Time OUR Oxygen consumed (JOUR dt) (d) (min) g0 2(m 3 d)\"1 g 0 2 m\"3 Total Growth Storage Endog. Resp. 0 0.0 369 - - - -0.0003 0.4 3513 0.58 0.005 0.47 0.11 0.0004 0.6 2828 0.32 0.004 0.28 0.04 0.0011 1.6 1379 1.47 0.06 1.16 0.26 0.0013 1.9 1065 0.24 0.03 0.14 0.07 0.0029 4.2 689 1.40 0.37 0.44 0.59 0.0045 6.5 627 1.05 0.46 0.00 0.59 0.0077 11.1 511 1.82 0.64 0.00 1.18 0.0114 16.4 474 1.82 0.46 0.00 1.36 0.0146 21.0 442 1.47 0.29 0.00 1.18 0.0198 28.5 457 2.34 0.42 0.00 1.92 0.0261 37.6 450 2.86 0.53 0.00 2.32 0.0325 46.8 435 2.83 0.47 0.00 2.36 Total 18.21 3.78 2.49 11.94 189 Table A4.2: Respirometric information obtained from Test 4-1 Time (d) Time (min) OUR g02(m3d)-' Oxygen consumed (JOUR dt) g 0 2 m'3 Total Growth Storage Endog. Resp. 0.0000 0.0 275 - - - -0.0004 0.6 4381 0.93 0.06 0.76 0.11 0.0008 1.2 4156 1.71 0.19 1.41 0.11 0.0014 2.0 3502 2.30 0.51 1.62 0.17 0.0021 3.0 1900 1.89 0.95 0.75 0.19 0.0027 3.9 1718 1.09 0.92 0.00 0.17 0.0036 5.2 1422 1.41 1.16 0.00 0.25 0.0049 7.1 1171 1.69 1.33 0.00 0.36 0.0062 8.9 1007 1.42 1.06 0.00 0.36 0.0085 12.2 685 1.95 1.31 0.00 0.63 0.0110 15.8 633 1.65 0.96 0.00 0.69 0.0137 19.7 430 1.44 0.69 0.00 0.74 0.0163 23.5 579 1.31 0.59 0.00 0.72 0.0213 30.7 405 2.46 1.08 0.00 1.38 0.0289 41.6 337 2.82 0.72 0.00 2.10 0.0340 49.0 340 1.73 0.32 0.00 1.41 0.0406 58.5 306 2.13 0.31 0.00 1.82 0.0531 76.5 276 3.64 0.19 0.00 3.45 Total 30.62 12.34 3.78 14.49 190 Table A4.3: Respirometric information obtained from Test 4-3 Time Time OUR Oxygen consumed (JOUR dt) (d) (min) g02(m3d)-' g 0 2 m\"3 Total Growth + Storage Endog. Resp. 0.0000 0.0 103 - - -0.0154 22.2 142 1.88 1.86 0.02 0.0448 64.5 167 4.54 4.43 0.12 0.1502 216.3 375 28.58 27.20 1.38 0.1700 244.8 339 7.06 6.77 0.29 0.1801 259.3 255 3.00 2.84 0.16 0.2190 315.4 88 6.67 5.92 0.74 0.2930 421.9 87 6.46 4.57 1.90 0.3139 452.0 91 1.86 1.29 0.57 0.3408 490.8 31 1.64 0.84 0.80 0.3890 560.2 38 1.66 0.02 1.64 191 Table A4.4: Respirometric information obtained from Test 5-1 Time (d) Time (min) OUR g0 2 (m3 d)\"1 Oxygen consumed (JOUR dt) g 0 2 m\"3 0.0000 0.0 790 Total Growth Storage Endog. Resp. 0.0005 0.7 2867 0.91 0.02 0.61 0.29 0.0013 1.9 1302 1.67 0.11 1.10 0.46 0.0020 2.9 1137 0.85 0.17 0.28 0.40 0.0030 4.3 1029 1.08 0.38 0.13 0.58 0.0046 6.6 983 1.61 0.69 0.00 0.92 0.0059 8.5 836 1.18 0.43 0.00 0.75 0.0084 12.1 781 2.02 0.58 0.00 1.44 0.0145 20.9 669 4.42 0.91 0.00 3.51 0.0213 30.7 674 4.57 0.65 0.00 3.92 0.0280 40.3 643 4.41 0.55 0.00 3.86 0.0508 73.2 610 14.28 1.15 0.00 13.13 0.0611 88.0 610 6.29 0.35 0.00 5.93 0.0665 95.8 609 3.29 0.18 0.00 3.11 Total 46.59 6.18 2.11 38.30 192 Table A 4 . 5 : Respirometric information obtained from Test 5-2 Time (d) Time (min) OUR g 0 2 (m3 d)\"1 Oxygen consumed (JOUR dt) g0 2 rn 3 0.0000 0.0 427 Total Growth Storage Endog. Resp. 0.0003 0.4 3943 0.66 0.02 0.50 0.13 0.0007 1.0 4628 1.71 0.11 1.44 0.17 0.0011 1.6 4483 1.82 0.19 1.46 0.17 0.0017 2.4 4181 2.60 0.45 1.89 0.26 0.0030 4.3 2029 4.04 1.63 1.85 0.56 0.0039 5.6 1949 1.79 1.41 0.00 0.38 0.0050 7.2 1413 1.85 1.38 0.00 0.47 0.0065 9.4 1275 2.02 1.37 0.00 0.64 0.0089 12.8 971 2.69 1.67 0.00 1.03 0.0117 16.8 856 2.56 1.36 0.00 1.20 0.0155 22.3 779 3.11 1.48 0.00 1.62 0.0203 29.2 709 3.57 1.52 0.00 2.05 0.0245 35.3 681 2.92 1.12 0.00 1.80 0.0305 43.9 618 3.90 1.33 0.00 2.56 0.0352 50.7 582 2.82 0.81 0.00 2.01 0.0443 63.8 553 5.16 1.27 0.00 3.89 0.0509 73.3 513 3.52 0.70 0.00 2.82 0.0716 103.1 444 9.90 1.06 0.00 8.85 Total 55.97 18.86 6.64 30.47 193 Table A4.6: Respirometric information obtained from Test 7-1 Time (d) Time (min) OUR g 0 2 (m3 d)\"' Oxygen consumed (JOUR dt) g 0 2 m\"3 0.0000 0.0 68 Total Growth Storage Endog. Resp. 0.0009 1.3 1250 . 0.59 0.00 0.53 0.06 0.0066 9.5 1038 6.52 0.14 6.00 0.38 0.0101 14.5 1057 3.67 0.19 3.24 0.24 0.0155 22.3 1071 5.74 0.45 4.93 0.36 0.0215 31.0 1035 6.32 0.72 5.19 0.41 0.0277 39.9 946 6.14 0.99 4.73 0.42 0.0403 58.0 752 10.70 2.79 7.07 0.85 0.0495 71.3 580 6.13 2.69 2.82 0.62 0.0585 84.2 448 4.63 3.16 0.86 0.61 0.0674 97.1 430 3.91 3.31 0.00 0.60 0.0763 109.9 465 3.98 3.38 0.00 0.60 0.0852 122.7 427 3.97 3.37 0.00 0.60 0.0947 136.4 402 3.93 3.29 0.00 0.64 0.1050 151.2 403 4.14 3.45 0.00 0.70 0.1149 165^ 5 345 3.70 3.03 0.00 0.67 0.1253 180.4 343 3.58 2.87 0.00 0.70 0.1342 193.2 339 3.03 2.43 0.00 0.60 0.1433 206.4 321 3.00 2.39 0.00 0.61 0.1527 219.9 315 2.99 2.35 0.00 0.63 0.1623 233.7 313 3.01 2.37 0.00 0.65 0.1692 243.6 309 2.15 1.68 0.00 0.47 Total 91.83 45.05 35.36 11.42 194 Table A4.7: Respirometric information obtained from Test 7-3 Time (d) Time (min) OUR g0 2 (m3 d)\"1 Oxygen consumed (JOUR dt) g 0 2 m\"3 Total Growth Storage Endog. Resp. 0.0000 0.0 362.0 - - - -0.0027 3.9 754.7 1.51 0.03 1.37 0.11 0.0079 11.4 724.4 3.85 0.21 3.42 0.21 0.0125 18.0 721.3 3.33 0.36 2.78 0.18 0.0169 24.3 698.1 3.12 0.50 2.44 0.18 0.0210 30.2 698.6 2.86 0.60 2.10 0.16 0.0250 36.0 685.1 2.77 0.71 1.89 0.16 0.0292 42.0 680.2 2.87 0.88 1.82 .0.17 0.0336 48.4 653.8 2.93 1.07 1.69 0.18 0.0378 54.4 664.5 2.77 1.16 1.44 0.17 0.0487 70.1 417.9 5.90 3.66 1.80 0.44 0.0588 84.7 396.8 4.11 3.71 0.00 0.41 0.0661 95.2 383.7 2.85 2.56 0.00 0.29 0.0719 103.5 362.8 2.16 1.93 0.00 0.23 Total 41.03 17.39 20.75 2.89 195 Table A4.8: Respirometric information obtained from Test 8-1 Time (d) Time (min) OUR g 0 2 (m3 d)\"1 Oxygen consumed (JOUR dt) g 0 2 m'3 0.0000 0.0 518 Total Growth Storage Endog. Resp. 0.0013 1.9 1027 1.00 0.01 0.91 0.09 0.0058 8.4 1109 4.80 0.17 4.34 0.30 0.0089 12.8 993 3.26 0.24 2.81 0.20 0.0121 17.4 1073 3.31 0.35 2.74 0.21 0.0154 22.2 1062 3.52 0.48 2.83 0.22 0.0188 27.1 1099 3.67 0.61 2.84 0.22 0.0221 31.8 1060 3.56 0.71 2.63 0.22 0.0250 36.0 956 2.92 0.72 2.01 0.19 0.0282 40.6 915 2.99 0.90 1.89 0.21 0.0315 45.4 919 3.03 1.04 1.77 0.22 0.0349 50.3 726 2.80 1.19 1.38 0.22 0.0384 55.3 645 2.40 1.35 0.82 0.23 0.0424 61.1 660 2.61 1.70 0.64 0.26 0.0465 67.0 556 2.49 1.92 0.30 0.27 0.0561 80.8 505 5.09 4.46 0.00 0.63 0.0649 93.5 481 4.34 3.76 0.00 0.58 0.0781 112.5 450 6.14 5.27 0.00 0.87 0.1041 149.9 407 11.14 9.43 0.00 1.71 Total 69.08 34.30 27.92 6.85 196 Table A4 .9 : Respirometric information obtained from Test 8-3 Time (d) Time (min) OUR g 0 2 (m3 d)\"1 Oxygen consumed (JOUR dt) g 0 2 m\"3 0.0000 0.0 477 Total Growth Storage Endog. Resp. 0.0007 1.0 1652 0.75 0.01 0.69 0.04 0.0035 5.0 1205 4.00 0.32 3.51 0.17 0.0055 7.9 1207 2.41 0.49 1.80 0.12 0.0073 10.5 1101 2.08 0.63 1.34 0.11 0.0093 13.4 1127 2.23 0.91 1.20 0.12 0.0113 16.3 1287 2.41 1.12 1.17 0.12 0.0138 19.9 1262 3.19 1.71 1.32 0.15 0.0168 24.2 977 3.36 2.50 0.67 0.18 0.0214 30.8 990 4.52. 4.25 0.00 0.28 0.0236 34.0 832 2.00 1.87 0.00 0.13 0.0263 37.9 885 2.32 2.15 0.00 0.16 0.0293 42.2 787 2.51 2.33 0.00 0.18 0.0323 46.5 772 2.34 2.16 0.00 0.18 0.0353 50.8 713 2.23 2.05 0.00 0.18 0.0440 63.4 619 5.80 5.27 0.00 0.52 0.0484 69.7 550 2.57 2.31 0.00 0.27 0.0543 78.2 532 3.19 2.84 0.00 0.36 0.0602 86.7 500 3.04 2.69 0.00 0.36 0.0659 94.9 477 2.78 2.44 0.00 0.34 Total 53.73 38.05 11.71 3.97 197 Table A4.10: Respirometric information obtained from Test 9-1 Time Time OUR Oxygen consumed (JOUR dt) (d) (min) g02(m3d)-1 g 0 2 m\"3 Total Growth Storage Endog. Resp. 0.0000 0.0 41 - - - -0.0012 1.7 777 0.49 0.00 0.44 0.05 0.0022 3.2 882 0.83 0.01 0.78 0.04 0.0036 5.2 904 1.25 0.02 , 1.17 0.06 0.0083 12.0 880 4.19 0.17 3.83 0.19 0.0139 20.0 897 4.98 0.38 4.37 0.23 0.0189 27.2 864 4.40 0.50 3.70 0.21 0.0233 33.6 879 3.83 0.56 3.09 0.18 0.0271 39.0 884 3.35 0.58 2.62 0.16 0.0311 44.8 877 3.52 0.70 2.65 0.16 0.0359 51.7 871 4.19 . 0.97 3.03 0.20 0.0415 59.8 883 4.91 1.31 3.37 0.23 0.0473 68.1 873 5.09 1.55 3.30 0.24 0.0523 75.3 854 4.32 1.50 2.61 0.21 0.0566 81.5 843 3.65 1.41 2.06 0.18 0.0611 88.0 852 3.81 1.60 2.03 0.18 0.0650 93.6 854 3.33 1.48 1.68 0.16 0.0698 100.5 812 4.00 1.95 1.85 0.20 0.0755 108.7 497 3.73 2.50 1.00 0.23 0.0818 117.8 471 3.05 2.79 0.00 0.26 0.0884 127.3 439 3.00 2.73 0.00 0.27 0.1040 149.8 358 6.21 5.57 0.00 0.64 Total 76.1 28.3 43.6 4.3 198 Appendix 4.2 Summary of batch tests information Table A4.11-A: Summary of batch tests information Summary information 7-1 7-3 •8-1 8-3 9-1 Location P. Alb. P. Alb. P. Alb. P. Alb. UBC Wastewater added (mL) 1500 1500 2000 2000 2000 Sludge added (mL) 500 300 400 400 400 Water added (mL) 0 500 0 0 0 Sludge cell 4 4 4 4 4 Filtration Filtered Unfiltered Filtered Unfiltered Unfilter Temperature (°C) 24.5 22.3 29.7 31.5 21.6 ATU addition yes yes no yes yes COD wastewater (g m\"3) 673 775 604 664 845 COD ww original (g m\"3) 775 775 664 664 845 COD s wastewater (g m\"3) 673 673 604 604 774 COD sludge (g in 3) 5748 5748 6544 5539 5294 COD s sludge (g m\"3) 125 125 159 118 164 COD mix (g m\"3) 1942 1604 1594 1477 1587 COD s mix (g in 3) 536 582 530 523 672 COD end (g m\"3) 1446 1422 1275 1422 1451 CODs end (gm\"3) 208 297 267 297 377 TSS wastewater (g m\"3) 0 40 .0 43 45 VSS wastewater (g m\"3) 0 35 0 33 35 TSS sludge (g m\"3) 3496 3496 3396 3510 3716 VSS sludge (g m\"3) 3064 3064 2986 3110 3216 TSS mix (g m\"3) 874 616 566 621 657 VSS mix (g m\"3) 766 540 498 546 565 Endogenous OUR (g 0 2 d\"1) 68 40 66 60 41 Endogenous area (g 0 2) 11.4 2.9 6.8 4.0 4.3 Storage area (g O2) 35.4 20.7 27.9 11.7 43.6 Growth area (g O2) 45.0 17.4 34.3 38.0 28.3 F/M 0.537 0.846 0.745 0.742 0.899 199 Table A4.11-B: Summary of batch tests information Summary information 2 4-1 4-3 5-1 5-2 Location P. Alb. P. Alb. P. Alb. P. Alb. P. Alb. Wastewater added (mL) 75 400 2000 70 500 Sludge added (mL) 2000 2000 3 2000 2070 Water added (mL) 0 0 0 0 0 Sludge cell 4 IA IA IA IA Filtration Unfiltered Filtered Unfiltered Filtered Unfiltered Temperature (°C) 31.5 24.5 26 41.6 40.7 ATU addition no yes no yes yes COD wastewater (g m\"3) 726 512 635 611 709 COD ww original (g m\"3) 726 635 635 709 709 COD s wastewater (g m\"3) 541 512 512 611 611 COD sludge (g m-3) 6281 6912 6912 6520 6520 CODs sludge (g m\"3) 134 262 262 316 316 COD mix (g m\"3) 6080 5845 644 6320 5389 COD s mix (g m\"3) 149 304 512 326 373 COD end (g in 3) 4387 - 511 - -COD s end (g in 3) 154 - 400 - -TSS wastewater (g m\"3) 41 0 27 0 37 VSS wastewater (g m\"3) 25 0 22 0 25 TSS sludge (g rn3) 3960 4340 4340 3532 3532 VSS sludge (g m'3) 3425 3765 3765 3117 3117 TSS mix (g m\"3) 3818 3617 3617 3413 2852 VSS mix (g in 3) 3302 3138 3138 3012 2515 Endogenous OUR (g 0 2 d\"1) 36.9 276 - 576 427 Endogenous area (g 02) 12.0 14.5 7.6 38.3 30.5 Storage area (g 02) 2.5 3.8 - 2.1 6.6 Growth area (g 02) 3.7 12.3 - 6.2 18.9 F/M 0.004 0.013 0.080 0.003 0.023 200 Appendix 4.3 Data from long-term batch tests Table A4.12: Endogenous respiration Test N° 1 (T =24.4 °C) Day OUR Temperature OUR Temp. Corr. (0T=O.O4) ln(OUR) d g 0 2 (nr3 d)\"1 °C gO z (m3 d)\"1 -1.4 151 23.0 157 5.06 1.7 132 22.4 141 4.95 2.4 117 23.8 118 4.77 3.4 103 24.5 101 4.61 3.7 105 24.6 102 4.63 4.7 116 24.2 115 4.74 7.4 72 23.6 73 4.29 8.4 54 23.5 55 4.00 9.7 63 24.3 62 4.12 Table A4.13: Endogenous respiration Test N° 2 (T = 30.2°C) Day OUR Temperature OUR Temp. Corr. (9T=0.04) ln(OUR) d g 0 2 (m3 d)\"1 °C g0 2 (m3 d)\"1 -1.3 250 29.2 261 5.59 1.8 214 27.9 234 5.53 2.4 171 25.2 209 5.49 2.8 214 29.3 222 5.43 3.4 206 30.4 205 5.32 3.8 126 27.9 139 5.00 4.4 197 29.8 200 5.31 5.4 167 31.4 160 5.04 201 Appendix 4.4 Respirometric measurements for estimating POBS Table A4.14: Respirometric data from batch test 9-2 Time OUR ln(OUR) d gm-3d-' -0.48 7.1 1.96 0.49 11.6 2.45 0.50 9.4 2.24 0.52 10.7 2.37 0.53 12.2 2.50 0.54 10.1 2.31 0.62 22.8 3.13 0.64 22.7 3.12 . 0.65 27.5 3.31 0.66 31.1 3.44 0.67 30.1 3.40 0.69 32.2 3.47 0.70 29.7 3.39 0.71 40.6 3.70 0.72 47.5 3.86 1.03 166.7 5.12 1.04 176.6 5.17 1.05 191.4 5.25 1.06 216.9 5.38 1.09 280.4 5.64 1.11 287.8 5.66 1.11 290.9 5.67 1.12 320.0 5.77 1.13 355.0 5.87 1.13 327.0 5.79 1.14 365.0 5.90 202 1.15 396.9 5.98 1.15 420.3 6.04 1.16 405.3 6.00 1.16 387.3 5.96 1.18 402.8 6.00 1.24 123.5 4.82 1.26 122.3 4.81 1.27 124.0 4.82 1.30 124.1 4.82 1.32 113.9 4.74 203 Appendix 4.5 Respirometric information for estimating KQ2 Table A4.15 Decrease of OUR due to a decrease on DO in the presence of substrate Dissolved Oxygen g 0 2 m\"3 Temperature °C OUR g 0 2 m\"3 d\"1 OUR temp. corr. g O a m\"3 d'1 OUR theor. g 0 2 m'3 d;1 OUR/OUR the0r 4.02 24.4 968 968 968 1.00 3.41 24.4 1077 1077 950 1.13 2.85 24.4 940 940 932 1.01 2.19 24.4 851 851 913 0.93 1.69 24.5 822 825 895 0.92 1.26 24.5 733 736 877 0.84 0.84 24.5 467 469 859 0.55 0.55 24.5 410 411 840 0.49 0.36 24.5 245 246 822 0.30 0.23 24.5 152 153 804 0.19 0.17 24.5 88 88 783 0.11 Table A4.16 Decrease of OUR due to a decrease on DO in the absence of substrate, Test 1 Dissolved Oxygen Temperature OUR OUR temp. corr. OUR/OURi„i g 0 2 m\"3 °C g 0 2 i n 3 d 4 g 0 2 m\"3 d\"1 -4.50 33.0 327 327 0.96 3.82 33.1 324 322 0.95 3.16 33.2 289 287 0.84 2.60 33.2 279 277 0.81 2.06 33.3 282 279 0.82 1.53 33.4 291 286 0.84 1.02 33.4 260 256 0.75 0.55 33.5 243 238 0.70 0.17 33.6 134 131 0.38 0.08 33.6 1 1 0.00 204 Table A4.17 Decrease of OUR due to a decrease on D O in the absence of substrate, Test 2 Dissolved Oxygen Temperature OUR OUR temp. corr. OUR/OURini g 0 2 m\"3 ° C g 0 2 rn3 d\"1 g 0 2 m'3 d\"11 -4.89 24.9 238 238 0.96 4.32 24.9 239 239 0.97 3.73 25.0 219 220 0.88 2.73 25.0 216 217 0.87 2.21 25.0 193 194 0.78 1.70 25.1 206 208 0.83 1.18 25.1 185 187 0.75 0.70 25.1 . 179 180 0.72 0.25 25.2 129 131 0.52 0.14 25.2 0 0 0.00 Table A4.18 Decrease of OUR due to a decrease on DO in the absence of substrate, Test 3 Dissolved Oxygen g 0 2 m\"3 Temperature ° C OUR g 0 2 m\"3 d\"1 OUR temp. corr. g 0 2 m\"3 d\"1 O U R / O U R n i 5.56 31.0 524 524 0.95 4.88 31.5 541 530 0.96 4.20 32.0 527 506 0.92 3.52 32.4 535 506 0.92 2.85 32.9 527 488 0.89 2.17* 33.3 531 484 0.88 1.50 33.8 530 474 0.86 0.82 34.2 520 457 0.83 0.22 34.6 366 317 0.58 0.06 35.0 1 1 -205 Appendix 4.6 Respirometric information for estimating 8T,bH Table A4.19 OUR during a rapid increase in temperature to endogenous sludg Time Temperature OUR min °C g O2 m\"3 min\"1 103 27.8 0.34 138 27.5 0.31 153 27.4 0.29 204 27.1 0.30 223 31.3 0.37 290 41.9 0.55 206 Appendix 5: Summary of sampling campaign N°1 results Tables A5.1 to A5.13 presents the results of the analytical measurements for the samples taken during the first sampling campaign. Table A5.1: COD results for sequences 1 to 7 at different plant locations Sequence COD (g m\"3) Influent Cell IA Cell 4 Recycle Wastage Effluent 1 729 8039 6642 16324 55245 114 2 488 4877 6961 10993 41176 86 3 483 5539 4817 18309 50441 102 4 383 5025 4804 16544 91 5 635 6912 4277 13750 123 6 647 5760 3922 17721 134 7 709 6520 3701 137 Table A5.2: Soluble COD results for sequences 1 to 7 at different plant locations Sequence COD (gm 3) Influent Cell IA Cell 4 Recycle Wastage Effluent 1 544 201 113 137 306 91 2 331 167 87 91 167 54 3 414 138 110 125 213 74 4 295 134 76 66 5 512 262 191 86 6 564 225 201 103 7 611 316 105 100 207 Table A5.3: TSS results for sequences 1 to 7 at different plant locations Sequence COD (gm\"3) Influent Cell IA Cell 4 Recycle Wastage Effluent 1 4604 15804 39200 2 49 3948 4060 36050 . 49 3 41 3824 3960 12850 39300 41 4 28 3805 3744 14060 39100 28 5 27 4340 3932 12560 27 6 40 4286 3804 12720 36735 40 7 3532 3780 Table A5.4: VSS results for sequences 1 to 7 at different plant locations Sequence COD (gm\"3) Influent Cell IA Cell 4 Recycle Wastage Effluent 1 13933 33797 2 32 3401 3509 31447 32 3 25 3305 3425 11197 34322 25 4 13 3303 3249 14257 34297 13 5 22 3765 3421 12197 22 6 33 3729 3326 11117 32177 33 7 3117 3277 11177 Table A5.5: NOV + N 0 3 results for sequences 1 to 7 at different plant locations Sequence COD (gm 3) Influent Cell IA Cell 4 Recycle Wastage Effluent 1 0.009 0.005 0.852 0.028 0.03 1.037 2 0.007 0.060 0.306 0.049 0.01 2.360 3 0.000 0,000 0.111 0.069 0.00 0.903 4 0.000 0.084 0.199 0.044 0.552 5 0.000 0.029 0.023 0.120 0.010 6 0.011 0.022 0.047 0.116 0.000 7 0.000 0.018 0.000 0.062 208 Table A5.6: NO2\" results for sequences 2, 4, and 5 at different plant locations Sequence COD (gm\"3) Influent Cell IA Cell 4 Recycle Wastage Effluent 2 0.067 0.150 0.054 0.306 4 0.015 0.104 0.021 0.035 5 0.014 Table A5.7: Ammonia results for sequences 1 to 7 at different plant locations Sequence COD (g m\"3) Influent Cell IA Cell 4 Recycle Wastage Effluent 1 0.01 0.41 0.19 3.47 22.17 0.14 2 0.00 0.47 1.02 2.41 12.39 0.16 3 0.01 1.01 0.99 2.54 23.50 0.05 4 0.00 1.63 0.61 2.45 0.03 5 0.00 0.74 0.01 1.91 0.05 6 0.01 0.13 0.24 0.97 0.01 7 0.01 0.62 0.00 0.01 Table A5.8: TKN results for sequences 1 to 6 at different plant locations Sequence COD (gm\"3) Influent Cell IA Cell 4 Recycle Wastage Effluent 1 3.9 57 57 182 580, 3.2 2 2.4 58 48 107 481 2.6 3 3.5 35 53 162 602 2.7 4 3.6 47 46 167 617 3.2 5 4.9 52 44 147 519 3.4 6 5.0 46 66 170 637 3.6 209 Table A5.9: Soluble TKN results for sequences 1 to 5 at different plant locations Sequence COD (gm\"3) Influent Cell IA Cell 4 Recycle Wastage Effluent 1 3.5 4.2 3.0 7.8 63 2.8 2 2.3 2.8 2.8 5.0 44 2.0 3 3.0 3.7 3.3 5.9 50 2.4 4 6.2 3.2 8.2 2.7 5 4.6 3.0 9.3 2.8 Table A5.10: PO/ results for sequences 1 to 7 at different plant locations Sequence COD (gm\"3) Influent Cell IA Cell 4 Recycle Wastage Effluent 1 4.1 4.7 3.8 8.8 46 3.7 2 3.0 3.6 3.6 5.9 38 2.9 3 3.9 4.6 4.1 5.3 43 3.1 4 3.9 7.5 4.1 9.2 3.5 5 5.4 5.4 3.9 11.2 3.7 6 5.2 5.2 4.7 6.3 3.8 7 5.3 6.1 4.0 3.7 Table A5 . l l : Total P results for sequences 1 to 6 at different plant locations Sequence COD (gm\"3) Influent Cell IA Cell 4 Recycle Wastage Effluent 1 1.4 397 474 1485 3785 2.1 2 0.7 433 374 891 3340 1.2 3 0.7 258 431 1377 2869 1.9 4 0.7 429 340 1432 2972 0.5 5 1.4 447 341 1279 3453 1.5 6 2.7 320 396 1496 4099 1.0 210 Table A5.12: Soluble total P results for sequences 1 to 5 at different plant locations Sequence COD (gm'3) Influent Cell IA Cell 4 Recycle Wastage Effluent 1 0.9 1.3 0.7 3.8 23 0.7 2 0.2 1.3 1.3 2.6 17 0.6 3 0.1 1.6 0.9 2.9 27 1.3 4 2.6 1.0 3.2 0.4 5 1.4 0.2 2.6 0.6 211 Appendix 6: Data used from historical database Due to the large amount of information used from the Port Alberni historical database, the data are presented in two files available in a floppy disk attached to this thesis. Both files are compressed in the file \"Appendix 6 Disk.zip\". The excel file \"Year 2002.xls\" presents the data from all year 2002 with a frequency of three data per day. The raw data, as they were obtained from the database, are available from column A N to column FC. However, the same data is presented in a more friendly and useful manner between columns B and AK. In a similar way, the excel file \"Week.xls\" presents the data of the week corresponding to the first measuring campaign (January 20 th to January 24th, 2003). 212 Appendix 7: MATLAB® file used for full-scale simulation % Model for plant with 5 aerobic tanks using ASM3 model % Main c h a r a c t e r i s t i c s : % - Variable i n f l u e n t concentrations % - Incorporates temperature % % Pablo Baranao, UBC, March 2003 function dxdt=plantASM3(t,X) % Kinetic Parameters % estimated parameters values par(l:6) = [3 1 24 0.3 30 0.2]; par(7:14) = [3.5 0.01 0.13 0.13 1 1.5 0.5 0.15]; par(15:20) = [0 0.91 0.73 0.24 0.2 0.00]; par(21:28) = [0.00 0.01 0.00 0.07 0.75 0.75 0.90 0.60]; par(29:34) = [0.04 0.06 0.07 0.04 0.105 0.105]; par=par'; %Influent concentrations %X(1) = S i ; X(2) = Ss; X(3) = Snh4; %X(4) = Snox; X(5) = X i ; X(6) = Xs; %X(7) = Xh; X(8) = Xsto; X(9) = Xa; %Influent values from S e r i e s l 4 . t x t % S e r i e s l 4 . t x t contains Seriesl4 for many days, 1 value per day % Seriesl4 i n f i l e % 1st column: time % 2nd column: COD i n % 3rd column: Nutrient flow % 4th column: TSS i n % 5th column: TSS rec % 6th column: Q i n 213 % 7th column: Q rec % 8th column: Q waste % 9th-13th column: Temperatures (1A-4) % 14th-18th column: Oxygen concentrations (1A-4) global S e r i e s l 4 ; % make data a v a i l a b l e within the solver function load S e r i e s l 4 . t x t ; % used for load data into Matlab i=max(find(Seriesl4(:,1)<=t)); % f i n d index to current data % COD f r a c t i o n s f S i = 0.14 ; fSs 0.52 ; f X i = 0.10 ; fXs 0 .24 % Nutrient concentrations NH3conc = 106000 + 140000; %(mg/l); UreaConc from urea to NH4 i s assumed NOxconc = 70000; %(mg/l); changed from 70000 14000; Rapid transformation %Influent flow rates qin = S e r i e s l 4 ( i , 6 ) qret = S e r i e s l 4 ( i , 7 ) qw = S e r i e s l 4 ( i , 8 ) qnutrient = S e r i e s l 4 ( i , 3 ) % Input concentrations S i _ i n = Seriesl4 (i,2) * f s i ; %mg/l S e r i e s l 4 ( i , 2 ) * fSs; %mg/l qnutrient * NH3conc / (qnutrient + qin + q r e t ) ; %mg/l S e r i e s l 4 ( i , 2 ) * f x i ; %mg/l Seriesl4 (i,2) * fXs; %mg/l qnutrient * NOxconc / (qnutrient + qin + q r e t ) ; %mg/l 0; Ss_in Snh_in X i _ i n Xs_in Sno_in Xh_in Xsto_in Xa i n x i n Xa in] [Si i n Ss i n Snh i n Sno i n X i i n Xs i n Xh i n Xsto i n %Recycle c h a r a c t e r i z a t i o n 214 TSSrec = Se r i e s l 4 ( i , 5 ) ; Xi rec = TSSrec * 1.1 * 0.018; COD p a r t i c u l a t e = 1.1 TSS (based on previos simulation) Xs rec = TSSrec * 1.1 * 0.058; % Percentages are also based on previous l i m i t a t i o n Xh rec = TSSrec * 1.1 * 0.822 % Xsto rec TSSrec * 1.1 * 0.099 Xa rec = TSSrec * 1.1 * 0.005 % xrec_sol = (X(37:45) .* [1 1 1 1 0 0 0 0 0]'); xrec_part = [ 0 0 0 0 Xi_rec Xs_rec Xh_rec Xsto_rec Xa_rec] 1; xrec = xrec_sol + xrec_part; % Temperatures T1A = week(i,9) ; TIB = week(i,10); T2 = week(i,11) ; T3 = week(i,12) ; T4 = week(i,13) ; % Oxygen concentrations 01A = week(i,14); 01B = week(i,15); 02 = week(i,16); 03 = week(i,17); 04 = week(i,18); %Volumes VIA = 5250 / 2; %m3 V1B = 5250 / 2; %m3 V2 = 5250; %m3 V3 = 5250; %m3 V4 = 5250; %m3 % I n i t i a l concentrations (xO) % S i Ss Snh4 Snox Xi Xs Xh Xsto Xa 215 % x0(l:9) [130 50 1 5 6 50 150 3200 900 100] ; % x0(10:18) = [130 0 0 5 8 50 120 3200 700 100] ; % x0(19:27) ' = [130 0 0 2 10 50 100 3200 500 100] ; % x0(28:36) = [130 0 0 1 10 50 80 3200 400 100] ; % x0(37:45) = [130 0 0 1 10 50 75 3200 300 100] ; % Tanks dxdtlA = ASM3(X(1:9), [xin xrec], [qin qr e t ] , par, VIA, T1A, 01A); % Tank IA dxdtlB = ASM3 (X (10:18) , [X(l:9)], [qin + qr e t ] , par, V1B, TIB, OIB) ; % Tank IB dxdt2 = ASM3(X(19:27), [X(10:18)], [qin + qr e t ] , par, V2, T2, 02); % Tank 2 dxdt3 = ASM3 (X(28:36) , [X(19:27)], [qin + qr e t ] , par, V3, T3, 03); % Tank 3 dxdt4 = ASM3(X(37:45), [X(28:36)], [qin + qr e t ] , par, V4, T4, 04); % Tank 4 % Return Derivatives %Derivatives a l l tanks dxdt=[dxdtlA dxdtlB dxdt2 dxdt3 dxdt4]' ; % END function dx=ASM3 (x, par, T); % ASM3 i s an ASM3 d i a l e c t that only take i n account the aerobic processes % % The output, dx, are the derivatives of a l l the states. % % Syntax: % dx=pbdmodell(x, xin, q, par, v o l , T) % % Pablo Baranao, UBC, November 2 002 % _ % Model components 216 % X(l) = S i % X(2) = Ss % X(3) = So2 % X(4) = Xso % X(5) = Xi % X(6) = Xs % X(7) = Xh % K i n e t i c Parameters % par(l) = kh; par(2) = Kx; par(3) = ksto % par (4) = Ko; par (5) = Ks; par (6) = Ksto; % par(7) = muh; par(9) = bh; par(10) = bsto; % Stoichiometric Parameters % par(15) % par(18) f s i ; f x i ; par(16) = Ysto; par(17) Yh; % Arrhenius thetas (based on 20°C) % % par(19) = theta kh; par(20) = theta ksto; % par(21) = theta muh; par(22) = theta bh; % Arrhenius kh = par(l) ksto = par(3) muh = par(7) bh = par(9) * exp(par(19)*(T-30) * exp(par(20)*(T-30) * exp(par(21)*(T-30) * exp(par(22)*(T-30) bsto = par(10) * exp(par(22)*(T-30) % processes rates % PI Hydrolysis p i = kh * x(6)/x(7) / ( par(2) + x(6)/x(7) ) * x(7) 217 % P2 Aerobic storage of Ss p2 = ksto * (x(3) / (par(4) + x(3))) * (x{2) / (x(2)+par(5))) * x(7); % P3 Aerobic growth of Xh p3 = muh * (x(3) / (par(4) + x(3))) * (x(4)/x(7) / (par(6) + x(4)/x(7) ) ) * x(7) ; % % P4 Aerobic endog. r e s p i r a t i o n p4 = bh * (x(3) / (par(4) + x(3))) * x(7); % P5 Aerobic r e s p i r a t i o n of Xsto p5 = bsto * (x(3) / (par(4) + x(3))) * x(4); % d e f f e r e n t i a l equations dx(l)= par(15)*pl; % S i dx(2)= p i - p2 ; %Ss dx(3)= -(1-par(16))*p2 - (1-par(17))*p3/par(17) -(1-par(18))*p4 -p5; %So2 dx(4)= par(16)*p2 - p3/par(17) - p5; %Xsto dx(5)= par(18)*p4; %Xi dx(6)= p i ; %Xs dx(7)= p3 - p4; %Xh dxdt=[dx(l) dx(2) dx(3) dx(4) dx(5) dx(6) dx(7)]; % END 218 Appendix 8: Photos Figure A8.3: Respirometer with temperature-controlled water recirculation system 220 Figure A8.5: Computer set-up for recording the dissolved oxygen concentration Appendix 9: Analytical results of experiments for model parameters estimation Appendix 9.1 COD results from long-term batch test for estimating bH Table A9.1: Particulate COD results and X H estimation from a long-term batch test Time Particulate COD X H ln(XH) d g COD m\"3 g COD rn 3 -0 4760 3570 0.00 2 4403 3124 -0.13 4 4100 2745 -0.26 5 4015 2639 -0.30 7 3501 1997 -0.58 Appendix 9.2 COD results for estimating the production of SI during endogenous respiration Table A9.2: COD and CODs results from a long-term batch test without the addition of substrate Time COD CODs d g COD m\"3 g COD m - 3 0 4934 174 2 4534 131 4 4212 112 5 4124 109 7 3596 95 8 3934 88 18 3615 83 25 3290 80 35 3284 68 42 3051 73 222 Appendix 9.3 C O D results for estimating Y 0 B S Table A9.3: COD and particulate COD results from a high F/M ratio batch test (9-2) Time COD CODs Part COD hr g COD rrf3 g COD m\"3 g COD rn 3 0.0 652 578 74 16.6 657 547 110 18.3 699 564 135 20.7 610 449 162 Appendix 9.4 Long-term B O D tests Table A9.4: Long-term BOD Test N° 1 with filtered and unfiltered influent and effluent Time BOD Influent BOD Effluent d g O2 m 3 g 0 2 nf3 Total Filtered Total Filtered 0.3 50 0.4 1.0 148 0.9 1.0 145 0.9 1.3 209 1.3 1.9 278 1.8 1.9 274 1.6 2.3 356 3.0 456 3.0 455 2.8 4.0 533 4.0 529 5.0 568 385 4.5 1.2 5.0 529 385 4.5 1.1 5.9 583 8.0 7.3 223 Table A9.5: Long-term BOD Test N° 2 with filtered and unfiltered influent and effluent Time BOD Influent BOD Effluent d g 0 2 m 3 g 0 2 m 3 Total Filtered Total Filtered 1.0 1 2 3 1.2 3 .0 2 1 4 1 4 5 4 . 0 3 .0 221 3.6 7.0 3 7 0 2 2 4 7.2 2 .3 7.0 3 7 0 2 2 4 7 . 1 . 2 .0 7.0 3 5 4 8.0 3 8 0 2 4 5 9 .0 3 9 8 2 7 5 8 .7 2 .4 1 1 . 0 4 2 8 2 4 7 9 .9 1 1 . 0 4 4 7 10.1 1 5 . 0 4 6 5 3 3 8 12 .5 4 . 0 2 1 . 0 4 8 8 15 .2 3 0 . 0 17 .0 224 "@en ; edm:hasType "Thesis/Dissertation"@en ; vivo:dateIssued "2003-11"@en ; edm:isShownAt "10.14288/1.0063907"@en ; dcterms:language "eng"@en ; ns0:degreeDiscipline "Civil Engineering"@en ; edm:provider "Vancouver : University of British Columbia Library"@en ; dcterms:publisher "University of British Columbia"@en ; dcterms:rights "For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use."@en ; ns0:scholarLevel "Graduate"@en ; dcterms:title "Modelling carbon oxidation in pulp mill activated sludge systems : calibration of ASM3"@en ; dcterms:type "Text"@en ; ns0:identifierURI "http://hdl.handle.net/2429/14560"@en .