UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The effect of pH on microbial activity and community structure in the biological removal of resin acids.. 1998

You don't seem to have a PDF reader installed, try download the pdf

Item Metadata


ubc_1999-390004.pdf [ 27.41MB ]
JSON: 1.0063788.json
JSON-LD: 1.0063788+ld.json
RDF/XML (Pretty): 1.0063788.xml
RDF/JSON: 1.0063788+rdf.json
Turtle: 1.0063788+rdf-turtle.txt
N-Triples: 1.0063788+rdf-ntriples.txt

Full Text

THE EFFECT OF pH ON MICROBIAL ACTIVITY A N D C O M M U N I T Y STRUCTURE IN THE BIOLOGICAL R E M O V A L OF RESIN ACIDS F R O M WASTEWATER by A L A N GIDEON W E R K E R B.A.Sc , The University of Waterloo, 1985 M.Eng., The Unversity of Toronto, 1988 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES Department of Civil Engineering (Environmental Engineering) We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH C O L U M B I A September 1998 © Alan Gideon Werker, 1998 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, 1 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 C\\J V— €Ka^-\kjECe . vV jC r The University of British Columbia Vancouver, Canada Date ^ ^ Z - J c £ V V s € ^ g - V~l | DE-6 (2/88) Abstract Pulp mills in Canada rely on biological treatment systems for the removal of resin acids that are released from wood during pulping and bleaching. These are priority contaminants for the pulping industry, since they have been frequently associated with events of toxicity breakthrough. Although, tighter mill control has helped to minimize the losses of resin acids in wastewater, acute toxicity removal in downstream biological treatment systems may still be insufficient, particularly under dynamic loading conditions. The mechanisms responsible for these limitations are not fully understood. This dissertation documents a fundamental study into the fate of resin acids during biological treatment. The objective was to quantify the influence of pH on the resin acid bioavailability, metabolism, and retention time during biological treatment. A progression of two batch and two continuous flow bioreactor investigations were undertaken to consider the interrelated issues of pH-dependent microbial activity and resin acid hydrophobicity. Changes in pH, within the typical range for biological treatment, significantly altered the bioavailability of resin acids and the community of microorganisms responsible for resin acid biodegradation. A sudden input of resin acids promoted an elevated level of community change during continuous treatment of bleached kraft mill effluent. The capacity of a treatment system to remove resin acids was found to be a function of the contaminant loading history. Time lag before biological removal in response to a shift-up in resin acid loading was significant and was also affected by the treatment system pH. Hence, the prevailing bioreactor pH operating condition, in conjunction with the period and amplitude of loading transients were shown to be key aspects controlling the microbial community structure and physiological state, which in turn determine the rate and extent of biological removal of resin acids. Through the course of the four investigations, novel contributions have been made in the areas of surface tension and microbial fatty acid measurements that have engineering application in future modelling and monitoring of the behaviour of microbial wastewater treatment processes. i i Table of Contents Abstract ii Table of Contents iii List of Figures iv List of Tables ix Acknowledgement xi Dedication xii Chapter 1 - Resin Acids and Toxicity Breakthrough in Pulp M i l l Effluent Treatment 1 Chapter 2 - Surfactancy and Solubility in the Biological Removal of Resin Acids 19 Chapter 3 - Microbial Community Structure in the Kinetics of Resin Acid Removal 80 Chapter 4 - A Resin Acid Shock Load During Continuous Biological Treatment 184 Chapter 5 - Transient Resin Acid Loading During Continuous Biological Treatment 263 Chapter 6 - Recommendations and Future Applications 336 Appendix A - Experimental Data 350 ii i List of Figures Figure 1 -1. Common pine resin acids of the abietane skeletal class 9 Figure 1 -2. Common pine resin acids of the pimarane and isopimarane skeletal class 10 Figure 1 -3. Chlorinated resin acid derivatives formed during pulp bleaching 10 Figure 2-1. A small (elemental) section of an arbitrarily curved gas/liquid interface 27 Figure 2-2. Sample data from Hua and Rosen (1988) illustrating the effect of surfactant concentration (c) on dynamic surface tension for N-dodecyl-N-methylglycine at pH 9.0 and 25 °C 32 Figure 2-3. The maximum bubble pressure method 49 Figure 2-4. Experimental setup for the maximum bubble pressure method 50 Figure 2-5. A typical maximum bubble pressure signal 51 Figure 2-6. Detail of the critical points in the pressure signal 51 Figure 2-7. Batch growth resin acid depletion curves for pH 6,7 and 8 from top to bottom 57 Figure 2-8. Dynamic surface tension during biological removal of resin acids for pH 6,7 and 8 from top to bottom 59 Figure 2-9. Derived surface pressure changes with time in the absence of resin acids that were believed to be due to the presence of kraft lignin in the media 60 Figure 2-10. The quasi-static Langmuir concentration K (equation (2-20)) as a function of surface age and pH ! 60 Figure 2-11. Normalized surface loading (y) as a function of surface age (t) and resin acid concentration 62 Figure 2-12. The derived initial flux of resin acid to a freshly generated gas/liquid interface at pH 8 63 Figure 2-13. The derived initial flux of resin acid to a freshly generated gas/liquid interface at pH 6 and 7 63 Figure 2-14. The experimentally derived dependency of the sticking factor <X> on the squared compliment of normalized surface loading based on the initial surface flux (equation (2-48) and (2-51)) 65 Figure 3-1. Chemical structure of abietic (left) versus pimaric acid (right) 84 Figure 3-2. Flow diagram for biochemical lipid analysis of natural microbial communities 97 Figure 3-3. A schematic drawing depicting a segment of the classical lipid bilayer containing intrinsic (I) proteins 99 Figure 3-4. Typical retention time data for a homologous series of saturated straight chain fatty acids (10:0 to 24:0) by GC/FED (HP Series II) with a DB5 (30 m, 0.32 mm ID and 0.25 pm film thickness) 133 Figure 3-5. Equivalent chain length (ECL) template for monoenoic, branched and hydroxy fatty acids from a DB-5 column calibration (Figure 3-4) made with standard mixtures of fatty acid methyl esters (Supelco FAME and BAME component mixtures) 133 Figure 3-6. Relative peak positions as a function of double bond location for homologous cis and trans monoenoic fatty acids 134 iv Figure 3-7. Mean logratio fatty acid compositional data (y. = log (x./ x16:0)) by PLFA versus WCFA analysis from five replicate samples taken during log growth in Lauryl Tryptose Broth. . 137 Figure 3-8. Whole-cell fatty acid (TFA) increase with time representing biomass production during mixed culture batch growth in Lauryl Tryptose Broth 138 Figure 3-9. Whole-cell fatty acid (TFA) extracts versus biomass measured as dry weight (•) or as Coomassie protein (A) during mixed culture batch growth on Lauryl Tryptose Broth 138 Figure 3-10. Relative WCFA composition similarity (Is from equation (3-36)) with respect to the terminal time point during batch mixed culture growth on Lauryl Tryptose Broth 139 Figure 3-11. Relative proportion (logratio) of the monoenoic fatty acid 16:1 co7c with respect to 16:0 for DhA-35 grown at pH 6.0, 7.0 and 7.5 on either dehydroabietic acid (A) or pyruvate (T) as sole carbon sources 142 Figure 3-12. The derived soluble resin acid concentration as a function of pH for un-inoculated medium (Section 3.3.1) containing 39.8±1.9 pg/mL dehydroabietic acid (B-DHA) and 18.8±1.0 pg/mL abietic acid (»-AB A) 144 Figure 3-13. A typical total fatty acid (TFA) biomass growth curve on resin acid with an acclimated enrichment mixed culture, in this case, at pH 7.5 for experiment A 145 Figure 3-14. Maximum specific growth rates for enrichment cultures of abietane resin acid degraders as a function of pH from three separate experiments (A, B, C from top to bottom) 147 Figure 3-15. Comparison of maximum specific growth rates (Q) for mixed cultures enriched for abietane resin acid degraders at 5 different pH levels inoculated into media containing either sodium acetate (A) or abietane resin acids (•) as the sole carbon sources 148 Figure 3-16. A typical linear yield approximation (YT F A= - A[TFA]/A[TRA]) for biomass production, as total fatty acid (TFA), versus total resin acid (TRA) consumption during exponential (balanced) growth (t,< t < t, in equation (3-57)) for an acclimated mixed culture 149 Figure 3-17. Average molar total fatty acid (TFA) or biomass yield from three replicate experiments (A, B, C) monitoring microbial growth for total resin acid (TRA) depletion (left graph) as a function of pH 149 Figure 3-18. Initial (T) versus residuum (A) resin acid concentrations obtained for replicated mixed culture batch experiments as a function of pH 151 Figure 3-19. Experiment A model substrate depletion curves for pH conditions 6, 7 and 8 (Top to Bottom) 154 Figure 3-20. A more detailed consideration of total fatty acid TFA yield during mixed culture exponential growth (tL < t < t2 in Figure 3-13) 155 Figure 3-21. A second example of a nearly linear biomass or total fatty acid TFA yield with respect to total resin acid removal (•) during mixed culture exponential growth (t{ < t < t2 in Figure 3-13) 155 Figure 3-22. Peak metabolite concentrations observed during mixed culture batch growth experiment A 158 Figure 3-23. Two pathways for the initial biodegradation of dehydroabietic acid 158 Figure 3-24. Relative proportions (logratios) of the monoenoic fatty acid 18: l(o7c with respect to 16:0 for resin acid enrichment cultures grown up on abietane resin acids (•) or sodium acetate . „ (A) as a function of pH Figure 3-25. Logratio fatty acid compositions for Experiment B as a function of pH 161 Figure 3-26. The first logcontrast canonical component (LCC-1) for Experiment B summarising 80.4 percent of the enrichment culture between-group chemotype variability as a function of pH. 162 Figure 3-27. The normalized eigenvector for the first canonical component for experiment B. .. 163 v Figure 3-28. Dendrogram from the similarity matrix for the mean WCFA compositions describing the enrichment cultures in Experiment B as a function of pH. 164 Figure 3-29. The first and second logcontrast canonical components in Experiment C for enrichment cultures grown on resin acids (right) and sodium acetate (left) as a function of pH (as labelled) 165 Figure 3-30. Normalized eigen vectors for the first and second logcontrast canonical components 166 Figure 3-31. Dendrogram from the similarity matrix for the mean WCFA compositions describing the enrichment cultures in Experiment C as a function of pH 166 Figure 3-32. Simulation of biphasic batch growth on two substrates using the system of equations given in Table 3-10 175 Figure 4-1. Schematic of an ideally mixed chemostat with liquid volume, V. 193 Figure 4-2. Model transient substrate concentrations during ten hydraulic turn over periods in a well mixed chemostat 195 Figure 4-3. Hypothetical Langmuir type sorption/desorption (V) progression during a changing surface concentration (C0) with time (Top Left) 200 Figure 4-4. Idealized schematic of the experimental bench scale moving bed bioreactor (MBBR). 206 Figure 4-5. Photograph of the actual bench scale MBBR 207 Figure 4-6. A sessile drop of ethylene glycol on an optically smooth surface of the ANOX carrier plastic 217 Figure 4-7. The digitized ethylene glycol drop shown in Figure 4-6 217 Figure 4-8. Experimental data (•) for the lithium washout curves following a spike load at time zero for reactor A (Top) and reactor B (Bottom) 219 Figure 4-9. Statistics for the surrogate (O-methylpodocarpic acid) extraction recoveries for biofilm and aqueous samples from the two reactors as well as for the aqueous feed samples 221 Figure 4-10. Residence time distribution data and the functional fit (fa(t) in equation (4-28) and Table 4-6) for the aqueous concentration of total resin acids (TRA) in Reactors A and B 223 Figure 4-11. Residence time distribution data and the functional fit (/it) in equation (4-28) and Table 4-6) for the aqueous concentration of chloro-dehydroabietic acid (Cl-DHA) in Reactors A andB 223 Figure 4-12. Residence time distribution data and the functional fit (/b(t) in equation (4-28) and Table 4-6) for the biofilm loading of total resin acids (TRA) in Reactors A and B 224 Figure 4-13. Residence time distribution data and the functional fit (fb(t) in equation (4-28) and Table 4-6) for the biofilm loading of chloro-dehydroabietic acid (Cl-DHA) in Reactors A and B. 224 Figure 4-14. Estimated total resin acid recovery from reactors A and B by the comparison of mass loss from the system (AMTS) to mass loss in the effluent (AMTe) for sample times (T) less than and equal to 12 hours 227 Figure 4-15. Estimated chloro-dehydroabietic acid recovery from reactors A and B by the comparison of mass loss from the system (AM/) to mass loss in the effluent (AMTe) for sample times (x) less than and equal to 12 hours 227 Figure 4-16. Resin acids in the surface scum that accumulated by 2.7 hours (t/9H = 1) after the spike load 228 Figure 4-17. Dendrogram from the similarity matrix for the mean whole cell fatty acid (WCFA) compositions for the suspended and biofilm from reactors A and B, as well as the WCFA composition of the BKME feed 231 vi Figure 4-18. Average bioreactor biomass levels before (-10 < r| = t/9H < 0 ) and after (0 < r\ < 10 ) a resin acid shock load 232 Figure 4-19. Logcontrast canonical components (LCC) of the fatty acid spectra for suspended and biofilm extracts 234 Figure 4-20. The calculated and sorted probabilities for the logratio radius distributions (Chapter 3; equation (4-25)) for fatty acid compositions from Reactor B plotted against the uniform order statistic 234 Figure 4-21. The trajectory describing the population dynamics of the suspended biomass in reactor A 239 Figure 4-22. Data smoothing and differentiation of the first weighted canonical component from logcontrast canonical component analysis (LCCA) of the suspended biomass samples from reactor B 242 Figure 4-23. A comparison of the average state speed (equation (4-42)) estimated from the indicated number of observations (n) over the 96 hour period of observation 243 Figure 4-24. Population dynamics expressed as state speed (equation (4-42)) from fatty acid logcontrast canonical component analysis 244 Figure 4-25. The derived correspondence between the state speed (equation (4-42)) and the total resin acid (TRA) levels after the shock load 245 Figure 4-26. Pooled reactor TOC measurements taken before, during and after a spike load transient of resin acids 247 Figure 4-27. Comparison of the estimated (equation (4-44)) versus the predicted (equation (4-17) and Table 4-16) flux of total resin acids to the carrier biofilms in reactors A and B 249 Figure 4-28. Predicted equilibrium Langmuir isotherms for biofilm sorption of total resin acids (TRA) based on the experimental parameter estimates for equation (4-17) listed in Table 4-16... 249 Figure 4-29. Simulated TRA fate data for the set of ordinary differential equations (4-17) and (4-48) 252 Figure 5-1. The three classes of periodic bioreactor operation (adapted from Pickett (1982)) 270 Figure 5-2. Kinetics of sorption of resin acids to a biofilm based on the results of Chapter 4 273 Figure 5-3. Idealized schematic of the experimental bench scale moving bed bioreactor (MBBR). 277 Figure 5-4. Photograph of the bench scale MBBR 278 Figure 5-5. Tracer feed selection (Top) and sampling (Bottom) during the experiment 279 Figure 5-6. Lithium input (top) and the measured lithium effluent levels in reactor A (middle) and B (bottom) during the coarse of the experiment 284 Figure 5-7. Average TOC levels in the BKME feed and the two bioreactors from 84 samples taken over the 21 day experiment 286 Figure 5-8. Measured TOC levels over time for the BKME feed (•), reactor A (•) and reactor B (•) 286 Figure 5-9. Progress of TOC removal for reactors A and B over time 288 Figure 5-10. The residual errors between the models plotted in Figure 5-9 and the measured TOC removal efficiencies 289 Figure 5-11. Biomass levels as total fatty acid (TFA) 292 Figure 5-12. Amounts of suspended and dissolved total resin acids (TRA) and total fatty acids (TFA) expressed as a percent of the total concentration 294 vii Figure 5-13. Dendrogram relating the similarities of fatty acid spectra from whole, filter and filtrate samples drawn from reactor A (A), reactor B (B) and the BKME feed (F) 294 Figure 5-14. Dendrogram relating the similarities of fatty acid spectra from biofilm and suspended samples from reactor A (A) and reactor B (B) for the experiment reported in Chapter 4(1) and the current investigation (2) 296 Figure 5-15. Dendrogram relating the similarities of fatty acid spectra from biofilm, suspended and effluent samples from reactor A (A) and reactor B (B) for the current investigation 296 Figure 5-16 Discrimination between the suspended and effluent microbial communities based on their fatty acid spectra 297 Figure 5-17. Logcontrast canonical components (LCC) of the fatty acid spectra for suspended and biofilm extracts 298 Figure 5-18. Reactor A suspended biomass community state and population dynamics 301 Figure 5-19. Reactor B suspended biomass community state and population dynamics 301 Figure 5-20. Reactor A biofilm biomass community state and population dynamics 302 Figure 5-21. Reactor B biofilm biomass community state and population dynamics 302 Figure 5-22. Retention (Reactor A) and selective withdrawal (Reactor B) of total resin acids (TRA) 304 Figure 5-23. The time course of the estimated total resin acid (TRA) specific uptake rate (•) in response to the step changes of the influent loading 306 Figure 5-24. The transient response to a shift-up in total resin acids (TRA) was found to differ significantly with pH 308 Figure 5-25. Logcontrast canonical component analysis of fatty acid compositions for the suspended biomass in reactor A (Top) and reactor B (Bottom) over selected time periods 310 Figure 5-26. Logcontrast canonical component analysis of fatty acid compositions for the biofilm biomass in reactor A (Top) and reactor B (Bottom) over selected time periods 311 Figure 5-27. The history of total resin acid (TRA) sorption to the carrier biofilm for the acidic conditions (pH 6) of reactor A 313 Figure 5-28. The history of total resin acid (TRA) sorption to the carrier biofilm for the alkaline conditions (pH 8) of reactor B * Figure 5-29. Shake flask B l data illustrating the results of batch growth on BKME spiked with resin acids at pH 8. Biomass was measured as total fatty acid (TFA) 316 Figure 5-30. A comparison of effluent TOC levels from the moving bed bioreactors and after 30 hours of batch growth 318 Figure 5-31. Mixed-culture batch growth rates estimated for replicate shake flasks at pH 6 (Top) and pH 8 (Bottom) 319 Figure 5-32. The estimated substrate specific growth rates for replicate shake flasks at pH 6. ... 321 Figure 5-33. The estimated substrate specific growth rates for replicate shake flasks at pH 8. ... 322 Figure 5-34. A comparison of lag times for growth (TFA) and for the specific removal of TOC and TRA. 323 viii List of Tables Table 2-1. Weights of phosphate salts for a 100 mM pH buffered medium 44 Table 2-2. Parameter values for the Monod model with a threshold substrate concentration (equation (2-32)) used to model the substrate depletion curves presented in Figure 2-7 56 Table 3-1. Generalized chemical structures for the typical groups of bacterial fatty acids 96 Table 3-2. Nomenclature for the common groups of fatty acids 96 Table 3-3. The general structure and types of the important phospholipids found in microbial membranes (adapted from Harwood (1984)) 99 Table 3-4. Similarity measures for compositions x A and x B containing D fatty acids (Bousfield et al. 1983; Eerola and Lehtonen 1988) 111 Table 3-5. Marker phospholipid fatty acids in various cell types (adapted from (Basile et al. 1995; Kates 1964; Vestal and White 1989)) 114 Table 3-6. Weights of phosphate salts for a 100 mM pH buffered medium 124 Table 3-7. Cellular fatty acid composition of the resin acid-degrading isolate DhA-35 (Mohn et al. 1997) ". 140 Table 3-8. Least squares estimates of biomass growth kinetic parameters (equation (3-57)) as illustrated in Figure 3-13 146 Table 3-9. Comparison of average total fatty acid (TFA) or biomass yield coefficients for abietane resin acid enrichment culture growth on resin acids or sodium acetate as sole carbon sources Table 3-10. Reduced Synthetic Chemostat Model for batch growth on multiple carbon sources. 173 Table 4-1. Contact angle and solid-vapour surface energies for high density polyethylene, ANOX biofilm carriers and Teflon 216 Table 4-2. MBBR volume estimated from measurements of effluent flow 219 Table 4-3. Average nutrient levels (phosphorus and nitrogen) for the two MBBR reactors and the BKME feed 220 Table 4-4. Estimated bioreactor volumetric inflow rates based on the conservation of mass balance for phosphorous and nitrogen and flow continuity 220 Table 4-5. Values used to determine the total biofilm area 220 Table 4-6. Least squares parameter values for equation (4-28) describing the experimental aqueous, /a(t), and biofilm, fb(t), residence time distributions 225 Table 4-7. Average extracted fatty acid liquid concentrations and surface loadings for suspended and biofilm samples 230 Table 4-8. Estimated total TFA biomass levels (pmoles) for Reactors A and B 231 Table 4-9. Calculated values for the Anderson-Darling test statistic for the total fatty acid compositional data sampled over 96 hours 235 Table 4-10. Critical values for the Anderson-Darling test statistic for assessing logistic normality from the d-dimensional radius distribution (adapted from Aitchison (1986)) 235 Table 4-11. Results of logcontrast canonical component analysis on the time series of suspended biomass samples taken from Reactor A 239 ix Table 4-12. Results of logcontrast canonical component analysis on the time series of suspended biomass samples taken from Reactor B 239 Table 4-13. Results of logcontrast canonical component analysis on the time series of biofilm biomass samples taken from Reactor A 240 Table 4-14. Results of logcontrast canonical component analysis on the time series of biofilm biomass samples taken from Reactor B 240 Table 4-15. Results of logcontrast canonical component analysis on the time series of fatty acid compositions taken from the BKME feed ; 240 Table 4-16. Least-squares estimates for the model parameters of equation (4-17) 248 Table 5-1. Chemical amounts used to mix 1 L of concentrated tracer stock solutions 276 Table 5-2. MBBR volume estimated from measurements of effluent flow 283 Table 5-3. Average nutrient levels (phosphorus and nitrogen) for the two MBBR reactors and the BKME feed 283 Table 5-4. Estimated bioreactor volumetric inflow rates based on the conservation of mass balance for lithium, phosphorus and nitrogen and flow continuity. 285 Table 5-5. Average extracted fatty acid liquid concentrations and surface loadings for suspended and biofilm samples 291 Table 5-6. Estimated TFA biomass levels (pmoles) for Reactors A and B 292 Table 5-7. Values used to determine the total biofilm area 292 Table 5-8. Results of logcontrast canonical component analysis on the time series of suspended biomass samples taken from Reactor A 299 Table 5-9. Results of logcontrast canonical component analysis on the time series of suspended biomass samples taken from Reactor B 299 Table 5-10. Results of logcontrast canonical component analysis on the time series of biofilm biomass samples taken from Reactor A 300 Table 5-11. Results of logcontrast canonical component analysis on the time series of biofilm biomass samples taken from Reactor B 300 Table 5-12. Best fit parameters for a logistic function (equation (5-13)) empirically describing biomass production as total fatty acids (TFA) during replicate (1 and 2) batch growth as a function of pH on BKME effluent spiked with resin acids 317 Table 5-13. Best fit parameters for a logistic function (equation (5-13)) empirically describing acid soluble total organic carbon (TOC) depletion during replicate (1 and 2) batch growth on BKME effluent as a function of pH 317 Table 5-14. Best fit parameters for a Boltzman function (equation (5-14)) empirically describing total resin acid (TRA) depletion during replicate (1 and 2) batch growth on BKME effluent as a function of pH 317 x Acknowledgement A friend once commented, that a doctoral degree is a measure of one's tenacity more than anything else. My own experience seems to support that hypothesis. As my father likes to remind me, life is what happens when we make plans. My four year plan expanded into six years, and led me in exciting directions that could not have been anticipated. However, the ability to hold on tightly through those valleys of despair, that I imagine engulf all Ph.D. candidates at some point (or points), is very much influenced by the support of friends, colleagues, and family. These individuals have a tremendous power to make or break one's resolve. In this regard I feel that I have been very fortunate. Foremost, Rebecca has been my rock through my deepest valleys and my highest peaks. For my older friends, the "boys" (David, Michael, Karen and Joey), plus Jeff and Megan, my prolonged tenure as a student has been a constant source of amusement. It is a great comfort to have friends who know my history and help me to see the humour in myself. Learning is my secret for eternal youth. So there is no doubt that my activities will continue to be a source of amusement. What occupies my head most of the day, is so foreign to the "boys". If we all thought the same way there would be nothing to share, and nothing to laugh about. I toast to continued laughter, and learning. Years from now I think that 1992-93 will be seen as one of the golden years for Environmental Engineering at UBC. At least from my perspective, I enrolled into the program along with a number of well motivated and seasoned students who added an unexpected richness to my time at UBC. Ron, Don and Jean are wonderful friends with whom I often regained perspective from "reality checks". They are also the kind of people who bring people together. If not for them, I would never have been treated to the friendships of Farahad, Loralee, Serena, Claire, Zarina, Emi and Erin, whose companionship, I similarly treasure and hope to enjoy for years to come. Another pearl of wisdom from my father, is that one cannot teach experience. Since I entered the program from a disparate background in engineering, I was lucky to have had the technical support of Susan Harper, Paula Parkinson and Jufang Zhou. Additional technical and logistical help was always kindly provided by the staff at the UBC Pulp and Paper Centre, especially, Tim Paterson, Peter Taylor, Rita Penco and Lisa Brandly. Lina Madilao was very helpful on the initial mass spectrometry. Ron Dolling and Scott Jackson were always ready to share their ideas and talents in electronics. Sally Finora helped to kick start the surface tension work. For me, these people were the sine qua non for my research activities and progress. They are unsung heros that see so many graduate students through to a degree. The assistance of Ann Wilson was vital to the pure culture experiments. Similarly, Stephanie Ebelt worked as a temporary laboratory assistant, and was extremely meticulous and helpful. Dr. Paul Bicho was instrumental in getting me started with the microbiology and chemistry surrounding resin acids. Dr. Bill Mohn and Vincent Martin have been two other great resources in the microbiology. I found that a little encouragement can go a long way. Andy Chan and Paulo Lin were high school students, who I had the pleasure to mentor with a science project in consecutive years, respectively. Rather than busy them with a make-work project, they were involved with and contributed directly to the goals of my own research. Andy helped to trouble shoot the method of analysis for the lithium tracer work, and Paulo worked on measuring biomass by a number of conventional techniques for the mixed culture control experiments. Both Andy and Paulo used an aspect of their projects for the annual Greater Vancouver Science Fair. It was nice to learn that in the fair they both won prizes. I hope that they continue in Science. I feel indebted to Dr. Aat Voskamp, my friend and supervisor, during my employment in the Netherlands. Aat's compassion for research and truth, recharged my fire that was quenched after some former disillusionment with academia. I am also grateful to my current supervisor, Dr. Eric Hall whose commitment and support of my project continued unabated even when my direction may have seemed a little obtuse. It has been a privilege to have had his encouragement to explore new ideas. New ideas come from experience with things that do not work. A healthy research environment permits one to learn from, and look for opportumty in mistakes or unexpected results. I am also grateful to my supervisory committee, and especially Dr. Eric Hall for their guidance in research and in preparing this dissertation. Dr. Hall's attention to detail, and command of the english language are two of his many qualities that I aspire to someday attain. Research costs time and money. For my time, I have been fortunate to have been supported from a number of sources. In kind, I have received much assistance from Jeanne Taylor and Western Pulp Limited Partnership in Squamish BC. I am extremely grateful to NSERC, PAPRICAN, the NSERC/COFI Industrial Research Chair in Forest Products Waste Management and the Sustainable Forest Management Network, for sustaining my existence as a graduate student and for funding my laboratory overhead. xi Dedication I would like to dedicate this dissertation to my parents, Fred and June Werker. The nature of my worries pale in comparison to those during times that they have lived through. For this luxury of an education that they have valued and helped to provide, I am eternally grateful. xii Chapter 1 Resin Acids and Toxicity Breakthrough in Pulp Mill Effluent Treatment Summary Resin acids are hydrophobic carboxylic acids that are released from wood during pulping. These hydrophobic contaminants have been related to events of aquatic toxicity breakthrough from biological treatment systems treating pulp mill wastewaters. They are also one of a number of problematic contaminants that impede the reuse of process waters. Thus, although resin acids need to be consistently removed from mill effluents, biological treatment has been found to be periodically inadequate. Since hydrogen ion concentration strongly influences the nature of resin acids in solution, pH was thought to be the principal parameter controlling the fate of these compounds during biological treatment. A set of five research objectives was defined for assessing the mechanisms of pH influence on the biological removal of resin acids. Four experimental investigations were undertaken to address the stated objectives, and, thereby, answer fundamental questions of practical importance for the biological treatment of pulp mill effluent. Table of Contents 1.1 Introduction 2 1.1.1 Pulp Mill Effluent and Toxicity Breakthrough 3 1.1.2 Wood, Resin Acids and Toxicity Breakthrough 6 1.1.3 Resin acids 9 1.2 Dissertation Question, Objectives and Overview 12 1.2.1 Dissertation Question and Objectives 12 1.2.2 Dissertation Overview 14 1.3 References 17 1 1.1 Introduction This dissertation documents a fundamental study of physicochemical and biological interactions that influence the fate of resin acids during secondary biological treatment. The purpose of this first chapter is three fold. First, a general introduction will be given to pulp mill effluent, in order to identify resin acids as deleterious contaminants in discharged effluent and in retained process waters. This discussion provides the motivation for this research project, namely, a better understanding of the limitations in resin acid biodegradation that could lead to optimized or novel contaminant removal strategies. After providing this background information, the dissertation question is stated along with the hypothesis that defined the experimentation that needed to be undertaken. Finally, an outline of the dissertation is presented. While the focus of the dissertation is resin acid removal from pulp mill effluent, many of the issues, concepts and experimental results are applicable to biological wastewater treatment in general. For instance, resin acids, like many other aquatic toxicants, are hydrophobic. Frequently, hydrophobic contaminants that cause wastewater to be acutely lethal, represent just a small fraction of the overall wastewater organic content. In addition, pulp mill effluents, like other sewered wastewaters, are inherently variable in quantity and quality. Most wastewater treatment plants rely on a microbiological unit process to remove dissolved and suspended organic contaminants. Transient or non-steady state influent conditions are a challenge to biological treatment systems. Thus the topic of this dissertation applies also to the more general issue of trace, hydrophobic organic contaminant biological removal, from complex wastewater under non-steady state conditions. It should be noted that over the six years of this research project, the intended application of the results has changed. In 1993, resin acids and events of toxicity breakthrough were topics of recurring concern in the pulping industry. It was this concern that motivated the project. As a result of improved in-mill practices, events of discharged toxic effluent are becoming rare. However, although upstream improvements have been made at pulp mills, a parallel increased understanding of the fate of these contaminants in the downstream biological treatment process is 2 still lacking. M y effort has been to improve that understanding. Thus, the contribution of this study should not be viewed in the light of a problem that has been mitigated, but rather, in terms of the potential utility of the ideas, methods and quantitative results towards better monitoring, modelling and design of biological wastewater treatment processes. The following background discussion is organized into three sections. First, a brief overview is given of pulp mill effluent and the occurrence of toxicity breakthrough from secondary biological treatment systems. Toxicity breakthrough at pulp mills is frequently caused by the inadvertent release of resin acids. The observed variability in the release and removal of resin acids during pulping is then briefly discussed. Finally, chemical properties of resin acids and their role as useful or problematic chemical by-products, and as aquatic toxicants, are summarized. 1.1.1 Pulp Mill Effluent and Toxicity Breakthrough Production of pulp and paper requires extensive use of energy, chemicals and water. The large volumes of process effluents (in the order of two million cubic meters per day) that are discharged by the pulping industries have the potential for environmental impact (Hebron et al. 1997). In the United States, the pulp industry represents the third largest fresh water user (Beckett et al. 1992). Canada ranks second in the world for wood pulp production and first for export of this commodity (Gaston 1993). For the province of British Columbia in 1995, domestic and export revenue of pulp and paper products accounted for 5.5 billion Canadian dollars of the provincial economy (Hebron et al. 1997). The current pulp and paper industry direction is towards progressive systems closure, whereby process waters would be reused and the fresh water intake would only replace losses due to evaporation (Ramamurthy and Wearing 1998). This industry objective is a considerable technical and economic challenge. The treated water quality requirements for water reuse are more stringent than those imposed by current environmental regulations for effluent discharge. More than likely the industry will move away from cornmon sewers towards separate water circuits with specialized treatment kidneys refreshing selected process streams. Specialized contaminant 3 removal systems are best devised from a fundamental understanding of contaminant fate. Thus, while the prime motivation of this work was the issue of aquatic toxicity in externally discharged mill effluent, in view of the industry's current agenda, the results may have more direct engineering application in the design of process water treatment kidneys. It should not be surprising that material that is harmful to the environment can also be problematic in pulp and paper processing. Waste from the pulp and paper industry was the major industrially-related environmental concern for British Columbia and New Brunswick in 1987 (Gaston 1993) and continues to raise environmental concerns that yet unidentified or unregulated compounds may exert subtle but quite deleterious chronic effects (Colborn et al. 1997; Servos et al. 1996). M i l l wastewater discharged to the environment carries inorganic and organic compounds in both dissolved and particulate form, giving rise to environmental problems associated with colour, turbidity, biochemical oxygen demand (BOD) and aquatic toxicity (Colodey and Wells 1992). Canadian Federal Pulp and Paper Effluent Regulations (PPER) controlling the discharge of BOD, total suspended solids (TSS) and acutely lethal effluent (ALE) came into full effect in 1996 (Hebron et al. 1997). In 1996 there were ten days of recorded A L E discharge by mills in British Columbia as compared to 3,935 days in 1990, out of a total possible 8,395 discharge days per year. This improvement of the A L E discharge record is likely due to tighter in-process control and the almost universal use of secondary biological treatment. The federal regulations require pulp and paper mills to report operating conditions and effluent quality. Federal or provincial inspectors work to verify the compliance of all pulp and paper mills by on-site inspections, audits, electronic reporting, and environmental impact assessment. Mills can apply to the government for authorization to temporarily exceed the normal limits for discharge. Noncompliance is met with warnings and can eventually lead to legal action. In 1996, one British Columbia pulp mill was convicted and fined 100 thousand Canadian dollars for discharge of toxic effluent (Hebron et al. 1997). Thus, while events of A L E are becoming rare, the legal ramifications for any one single event are becoming more significant. 4 Aquatic toxicity is commonly assessed with bioassays on juvenile fish (McLeay 1987). Fish are most frequently used because of their commercial significance, their role in the food chain and their sensitivity (Walden and Howard 1981). The standard acutely toxic threshold level is defined by the lethal concentration for 50 percent of the population after 96 hours, or the 96 hour L C . 50 Sublethal concentrations of toxicants can induce changes in fish arterial tension, plasma glucose, blood cell counts, liver glycogen and muscle glucose levels. Studies (Walden and Howard 1981) indicate that a threshold response level for the sublethal effects is approximately 5 to 10 percent of the 96 hour L C level. 50 Exposure offish to sublethal concentrations of pulp and paper mill effluents elicits a response suggesting interference to respiration. Chronic effects of mill effluent, such as reduced fecundity and other physical or behavioural maladies due to endocrine disruption, are of current concern not just for fish, but for organisms further up the food chain. Thus, closure of mill aqueous discharges would certainly be helpful by translating many external environmental concerns into internal technological issues. In any pulp mill or bleach plant, increasing water recycling will promote the build up of extractives, either dispersed in the process water or deposited onto fibres or equipment (Jonasson et al. 1997). In spite of the technological hurdles to be overcome in contending with the problems of removing extractives and other elements or compounds from recycled water, the approach of closure as a means of environmental protection is a sound goal for pulp mills and for industry as a whole. So long as wastewater is discharged to the environment, the short and long term impacts on aquatic life must be considered. There are many factors affecting the survival of fish that are subject to contact with pulp and paper mill effluent (Brouzes 1976; Ng 1977). Although BOD is the classical parameter used to measure wastewater strength, the removal of BOD cannot be equated with removal of toxicity (Mueller and Walden 1976). Toxic organic contaminants typically represent a small fraction of the overall influent BOD. Resin acids are a significant portion of the toxicity in mill effluent (McLeay 1987). In an Ontario Ministry of the Environment report (Water Resources Branch 1992) summarising acute lethality data for Ontario's pulp and 5 paper sector effluents over the winter and spring of 1990, resin acids were cited as one of the primary causes of acute lethality in the samples taken. Resin acids continue to be associated with issues of toxicity in the discharge of pulp mill effluent (Baley, B. (1998). Personal Communication, Environment Canada; Wasylenchuk, E.J. (1998). Personal Communication, Alberta Environment Protection; Funke 1996). 1.1.2 Wood, Resin Acids and Toxicity Breakthrough The four major components of wood are polysaccharides, lignin, extractives and ash- forming minerals (Fengel and Wegener 1984). The polysaccharides, cellulose and hemicellulose, make up the wood structure. Lignin is the glue that interpenetrates, binds and strengthens this structure. The extractives, representing a small part of wood, contribute to the properties of colour, odour, taste and resistance to decay (U.S. Department of Agriculture 1955). Extractives can be removed from wood by solvents such as water, alcohol, acetone, benzene and ether. The volatile extractive fraction, such as the terpenes, can be isolated by steam distillation. Fatty acids, resin acids, tannins and colouring matter are the major substances that are extractable by non- polar solvents. The water-soluble components extracted from wood include carbohydrates, proteins and inorganic salts. Tree species differ widely in the type and amount of extractives, which can range from 0 to 10 percent of the wood composition. Extractive distribution throughout wood is also heterogeneous (Wangaard 1981). Resin acids form part of the extractive content of wood. A natural wood component, these compounds are primarily located in the resin canals and parenchyma cells of trees, but are also to be found in sapwood, parenchyma tissue, heartwood, needles and bark (McCubbin 1983). The wood resin content is reported to be approximately the same for both coniferous and deciduous trees (Brouzes 1976), ranging from 1.0 to 2.3 percent of the dry wood weight. Softwoods, or conifers (gymnosperms) represent about 80 percent of available forests in Canada, with jack pine, spruce, fir and hemlock being the most important species for the forest industry (McCubbin 1983). During the pulping process, whether kraft, sulphite or mechanical, these wood extracts are washed away from the pulp fibres and form part of the wastewater stream. 6 Awareness of the acute toxicity of resin acids dates from as early as 1931 (Brouzes 1976). In 1936, Hagman (Hemingway and Greaves 1973) showed that resin acid concentrations in wastewater exceeding 1 ppm were toxic to fish. In 1950, Van Horn, Anderson and Katz (Hemingway and Greaves 1973) found that sodium salts of resin acids were toxic to minnows at 1 ppm and to Daphnia at 3 ppm. Motivated by periodic instances of inadequate detoxification of kraft mill effluent in the late 1960s, Rogers (1973) deduced the main toxic components to be two resin acids, namely, dehydroabietic and isopimaric acid. Rogers also noted the apparent relationship between resin acid content of the wastewater and its toxicity. Therefore, resin acids must be prevented from being discharged to receiving waters. Differences are to be found in the relative proportions of resin acids present in different types of wood. The softwoods produce high resin acid yields. In spruce and pine, 27 and 29 percent of the wood resins are resin acids, respectively. In hardwoods, a significantly smaller proportion of the wood resins are present as resin acids. For example, in birch and aspen only 0.30 and 0.75 percent of the wood resins are present as resin acid, respectively. Therefore, changes to the wood furnish feeding the pulp mill will cause fluctuations in the resin acid concentrations of the raw mill effluent. Reduction in resin acid content of wood can be accomplished by pre-treatment of wood prior to pulping. Seasoning of the wood attenuates toxicity levels in the effluent (Brouzes 1976). Chip aging decreases resin acid levels by natural oxidation augmented by the action of bacteria or fungi (Leach and Thakore 1976). Resin levels in the effluent are a function of wood chip age and also time of year, since resin acid content in trees is seasonal. Aging or "seasoning" of felled trees to reduce resin content takes 12 months in log form and 2 months in chip form. The capital costs and human resources required to manage wood storage are appreciable. Seasoning also reduces pulp brightness and yield (Rocheleau et al. 1998). Thus, the practical extent of pre-treatment of wood by aging for resin acid removal may be restricted due to the additional costs and the potential for the reduction in pulp quality. In the absence of extensive pre-treatment, the task of resin acid removal must be consistently accomplished during wastewater treatment. 7 Perhaps the most comprehensive studies of resin acid toxicity and wastewater removal were performed during the 1970s under the auspices of the Canadian Forestry Service, Department of the Environment, Canadian Pollution Abatement Research (CPAR) program. Dr. John Leach, one of the principal investigators under the CPAR program, studied the biodegradation of wood extractives important to bleached kraft mill effluent (Leach et al. 1977). Degradation rates, in order of most to least quickly degradable, were dehydroabietic, pimaric, monochloro- dehydroabietic, and dichloro-dehydroabietic acid. Thus, naturally-occurring wood extractives were more readily biodegradable than their chlorinated analogues. Pilot scale fermentation studies (Leach et al. 1978) demonstrated that when biotreatment was adequate to reduce the concentrations of resin and unsaturated fatty acids to less then 1 ppm, the resultant effluent demonstrated no acute toxicity. Degradation of dehydroabietic acid and pimaric acid went to completion in 1 to 3 days. Much of the research in the 1970's regarding biological detoxification was driven by a concern for the wide temporal variations in the outfall toxicity. While the major toxicants were shown to be readily oxidized, results from 5-day aerated lagoons at mills in British Columbia indicated failure to eliminate toxicity up to 20 percent of the time (Leach et al. 1978). The current trend for biological detoxification in Coastal British Columbia has been to use the U N O X activated sludge system. Where topography does not restrict land usage, aerated lagoons remain the most often used secondary treatment system in Canada (Gaston 1993). Incidents of periodic A L E discharge due to elevated outfall resin acid concentrations continue to occur today (Baley, B. (1998). Personal Communication, Environment Canada; Wasylenchuk, E.J. (1998). Personal Communication, Alberta Environment Protection). The fact that incidents of toxicity breakthrough are periodic, would suggest that biological wastewater systems cope poorly with transient contaminant loads. While mills work to tighten process control and to close process water circuits, they would benefit from an improved fundamental understanding of the limitations of biological wastewater treatment systems for removing transient loads of specific contaminants. Since the extent of plant wastewater monitoring is usually limited, it is often 8 impossible to trace the causes of toxicity breakthrough at pulp mills after the fact. Surprisingly, there have been no literature reports of studies in which laboratory scale bioreactors were subjected to resin acid load fluctuations that were carefully monitored. The existing literature indicates that previous laboratory bioreactor research experiments were typically steady state investigations, in which research success was linked to high removal efficiency. Laboratory research with bioreactors would be more informative i f efforts were directed towards challenging biological systems with non-steady influent conditions in order to explore and understand the inherent limitations. The laboratory is the ideal setting for measuring bioreactor response to well controlled stimuli. Monitoring bioreactor response to dynamic conditions was one of the experimental approaches used for the present investigation. 1.1.3 Resin acids Resin acids are weak tricyclic monocarboxylic acids of limited solubility (Soltes and Zinkel 1989; Taylor et al. 1988). Resin acids commonly found in pulp mill effluent are either abietic-type (Figure 1-1) or pimaric-type (Figure 1-2). Abietic-type acids have an isopropyl side chain at Figure 1-1. Common pine resin acids of the abietane skeletal class. Top row (left to right): Abietic, neoabietic and palustric acid. Bottom row (left to right): Levopimaric and dehydroabietic acid (Soltes and Zinkel 1989). 9 Figure 1-2. Common pine resin acids of the pimarane and isopimarane skeletal class. From left to right: Pimaric, isopimaric and sandaracopimaric acid (Soltes and Zinkel 1989). Figure 1-3. Chlorinated resin acid derivatives formed during pulp bleaching. From left to right: 12-C1- dehydroabietic acid, 14-Cl-dehydroabietic acid, and 12,14-Cl2-dehydroabietic acid. postion 13, while pimaric-type acids have methyl and vinyl substiruents. Bleaching of pulp can result in chlorine substitution (Figure 1-3). Variations within each skeletal class arise due to the number and position of the double bonds. Dehydroabietic acid (DHA) exhibits an ionization constant (pKa) of 5.7 and a total solubility of 6.6 mg/L (Nyren and Back 1958). The solubility of the unionized acid is 4.9 mg/L. D H A is the most soluble diterpene resin acid because it possesses the highest number of double bonds (Bruun 1952). Salts of the resin acids can readily be formed with sodium, calcium, zinc, magnesium and aluminium (Soltes and Zinkel 1989). Salts of resin acids, in contrast to the acids themselves, are orders of magnitude more soluble in water and exhibit amphiphilic properties that make them useful as soaps. The critical micelle concentration of potassium dehydroabietate is approximately 2.5-3.0 x 10 2 molar (Corrin et al. 1946; Kolthoff and Stricks 1948). One of the features of resin acids that make them interesting model compounds for research, is that they become less hydrophobic under alkaline conditions. Experiments examining the fate of 10 resin acids as a function of pH reveal general principles about the influence of contaminant hydrophobicity on biological removal. Rosin soaps, which are refined saponified diterpene resin acids, find application in paper sizing (Strazdins 1989) and as emulsifiers for polymerization reactions (Davis 1989). Rosin is readily refined from tall oil, which comes initially from the soap skimmings off black liquor. The kraft pulping process acts to solubilize natural resinous material in trees by saponification. Therefore, resin acids in the effluent stream are initially in their resinate or soap form. The soaps can be converted back to their acids in the presence of free hydrogen ions. Carbon dioxide can precipitate resin acids from alkaline solutions (Soltes and Zinkel 1989). The pH changes following the combination of acidic or alkaline sewer lines will influence the dissolved resinate. If these soaps are converted back into acids, they may be precipitated, but they can also be held in solution by association with other soaps in solution (Nyren and Back 1958; Soltes and Zinkel 1989) or with other dissolved organic matter such as lignin breakdown products (kraft lignin) (Kulovaara et al. 1987). Resin acids may be a component of compounds referred to by the paper industry as "stickies" (Doshi and Dyer 1997) that form pitch deposits. During paper making, pitch deposits form on rolls, felts and fabrics. Their deposition can impede water removal, impact runnability and cause sheet defects. The reuse of pulping process water requires that "stickies" be effectively controlled. Thus resin acids are among the problematic compounds that must be contended with on the road of progressive systems closure. As long as mills continue to discharge effluent into marine or riparian waters, environmental impact and toxicity will continue to be issues of investigation. More than seventy percent of the acute toxicity found in pulp mill effluent is attributable to the hydrophobic fraction (Ng et al. 1974). Within the hydrophobic fraction, resin acids are the primary toxicants, frequently related to incidents of toxicity breakthrough (Kovacs and O'Connor 1996; McLeay 1987; Taylor et al. 1988). Being natural wood extractives, resin acids are waste by-products endemic to the pulping process (Allen et al. 1993). They are of environmental concern primarily because they exhibit 11 acute toxicity (LC ) towards aquatic life at concentrations in the order of 1 ppm (Taylor et al. 50 1988). While aerobic secondary biological treatment can remove resin acids, periodic toxicity breakthrough continues to force mill shut downs (Kovacs and O'Connor 1996). Furthermore, even if sublethal concentrations at the outfall are achieved, these compounds can persist (Brownlee and Strachan 1977; Fox 1976) and bioaccumulate (Niimi and Lee 1992), posing a chronic threat to aquatic life (Oikari et al. 1988; Tana 1988). Local changes in environmental conditions can redistribute resin acids between aqueous, particulate and sedimentary phases (Volkman et al. 1993). A better understanding of the fate of resin acids in wastewater treatment systems should provide insight towards more reliable methods of removal, thereby reducing the risks of exposure for these, or similar contaminants in aquatic environments. Such advancement has been the underlying goal of this research project. 1.2 Dissertation Question, Objectives and Overview 1.2.1 Dissertation Question and Objectives There is little doubt that hydrogen ion concentration is an important parameter in a biological treatment process. Most bacteria cannot proliferate at pH levels below 4.0 or above 9.5. Generally, the optimum pH for bacterial growth is between 6.5 and 7.5. The typical operating pH range for secondary treatment of pulp mill effluent is between 6.0 and 8.0. As will be discussed, a change in pH, over this same typical range for secondary biological treatment, significantly alters resin acid solubility and hydrophobicity. Therefore, the following dissertation question was posed. What will be the dominant mechanism of pH influence on the biological removal of resin acids in kraft mill effluent? The stated thesis question is based on the premise that: 12 change in pH, within the typical narrow operating range for secondary treatment of pulp mill effluent, will significantly influence the probability of resin acid biological removal. The probability of contaminant removal in a bioreactor depends on the contaminant bioavailability, the microbial state, and the contaminant retention time. Thus, the hypothesized impact of pH on the likelihood of resin acid removal during secondary treatment, may be attributed to pH-dependent variations in resin acid bioavailability, fluctuations in metabolic activity or microbial community structure, and changes in the contaminant retention time. Therefore, an investigation was required to consider each of the following important factors. These will also be discussed later in more detail. i. Reduced contaminant aqueous solubility can serve to restrict the contaminant bioavailability. Resin acid solubility is a function of pH. While it is intuitively obvious that alkaline conditions, which increase resin acid solubility, are likely to enhance the extent of biological removal, the sensitivity of resin acid biodegradation to solubility has not been clearly quantified in the research literature. ii. Microbial communities in a bioreactor are shaped by reactor environmental conditions, such as pH. The influence of pH on communtiy selection and on growth kinetics for microorganisms that can degrade resin acids, has not been established in the scientific literature. Hydrophobic compounds can also inhibit microbial activity. The possibility of pH- dependent changes in microbial inhibition, due to pH-dependent changes in resin acid hydrophobicity, has also not been previously studied. Further, studies of the time lags in microbial response to fluctuating resin acid loadings during continuous treatment, are lacking. iii. Contaminant retention time during secondary treatment can be influenced by the kinetics and the extent of contaminant sorption on biomass, which are a function of hydrophobicity. Resin acid hydrophobicity is a function of pH. The influence of pH on resin acid hydrophobicity and any associated change in resin acid sorption kinetics have not been 13 adequately quantified in the research literature. Consequently, the following five research objectives were defined. A . To determine the influence of resin acid solubility on contaminant bioavailability. B. To determine the influence of pH on the microbial community structure and the uptake rate of resin acids. C. To determine the influence of resin acid hydrophobicity on microbial activity. D. To determine the influence of resin acid hydrophobicity on contaminant sorption to biomass and, consequently, retention time. E. To determine the influence of pH on the time lag in increased metabolic activity for resin acid biodegradation. 1.2.2 Dissertation Overview In order to consider the stated objectives, it was recognized that pH must be the principal independent parameter for experiments designed to elucidate the importance of bioavailability, metabolic activity, microbial community structure, and adsorption kinetics in the biological removal of resin acids from pulp mill effluent. Little attention has been given to pH-sensitivity in the removal of resin acids. It will be seen that in this regard the scientific literature is also in apparent contradiction. A progression of two batch and two continuous flow microbiological investigations were undertaken in order to clearly assess the role of pH in resin acid biodegradation. Due to the specific aspects of each investigation, they have been organized chronologically into separate chapters as follows: Chapter 2. Surfactancy and Solubility in the Biological Removal of Resin Acids Batch growth experiments were performed to determine the extent to which resin acid solubility influences bioavailability during biological treatment (Objective A). In these 14 experiments, changes in adsorption kinetics were monitored using dynamic surface tension measurements. Deriving the adsorption kinetics from surface tension data required the development of a novel quasi-static approach to isotherm modelling. From the isotherm modelling, it was possible to quantify pH-induced changes in resin acid diffusivity. This measure of diffusivity provided a means to explain observed changes in resin acid bioavailability. Chapter 3. Microbial Community Structure in the Kinetics of Resin Acid Removal In a second series of batch growth experiments, it was necessary to consider the influence of microbial community structure on the hypothesized pH-dependent rate of resin acid removal (Objective B). Microbial communities were distinguished by comparing compositions of fatty acids extracted from culture samples. The ability to assess changes in community structure was a key contributor to the success of this investigation. Measurements of mixed culture growth kinetics are notoriously variable. Valuable insight was gained by using community structure as the basis of comparison to interpret the observed variability within and between experiments. Control experiments were performed to assist in the interpretation of the observed changes in the microbial fatty acid spectra. Chapter 4. A Resin Acid Shock Load During Continuous Biological Treatment Based on the results from the batch experimentation, the first continuous flow experiment was then undertaken. Two parallel, bench scale, single stage, completely mixed moving bed bioreactors, treating whole bleached kraft mill effluent, were operated and compared. One of the bioreactors was operated under constant acidic conditions (pH 6) while the other was run under constant alkaline conditions (pH 8). Resin acid adsorption onto the biomass and the possibility for resin acid-induced growth inhibition were investigated (Objectives C and D). Analytical techniques developed for comparing microbial communities and quantifying population dynamics from microbial fatty acid compositions were advanced and became principal tools that were applied for data interpretation. 15 Chapter 5. Transient Resin Acid Loading During Continuous Biological Treatment Biological response to transient loads of resin acids was monitored in a second similar continuous flow experiment. It was of interest to know the extent to which the acclimation of a microbial community to a change in contaminant loading, was pH-dependent (Objective E). A n acclimation period or lag time is a well known phenomenon for bacteria responding to changing environmental conditions. Since pulp mill effluents are known to have variable organic contents, the dynamic response of communities is of practical importance. In this dissertation, background, concepts, theories, and methods are presented in the chapters in which they were first used. This format was chosen, in part, to help in the readability of a dissertation that has delved into a number of rather specialized areas of mathematics and science that were unique to each aspect of the investigation. Much of the mathematical detail is included because these developments were integral to the data analysis and interpretation. Such detail is important to those wishing to further test and pursue the ideas presented. The second objective of the thesis format was to preserve the development of ideas, since the earlier research strongly influenced the approach taken in later experiments. The scope of the research unfortunately resulted in an unexpectedly long treatise. To assist in the readability, each experimental chapter was written as a relatively self-contained extended publication. Experimental methods, results and ideas that become established in the earlier chapters form points of reference later in the thesis. 16 1.3 References Allen, S. L., Allen, L. H., and Flaherty, T. H. (1993). Defoaming in the pulp and paper industry. Defoaming, P. R. Garret, ed., Marcel Dekker, Inc., Surfactant Science Series, 151-175. Beckett, R., Wood, J. W, and Dixon, D. R. (1992). Size and chemical characterization of pulp and paper mill effluents by flow field fractionation and resin adsorption. Environmental Technology, 13, 1129- 1140. Brouzes, R. J. P. (1976). Fish toxicity with specific reference to the pulp and paper industry. EPS 3-WP- 76-4, Environment Canada. Brownlee, B., and Strachan, W. M . J. (1977). Distribution of some organic compounds in the receiving waters of a kraft pulp and paper mill. Journal of the Fisheries Research Board of Canada, 34, 830- 837. Bruun, H. (1952). Properties of monolayers of rosin acids. Acta Chemica Scandinavica, 6, 494-501. Colborn, T., Dumanoski, D., and Myers, J. P. (1997). Our Stolen Future, Plume/Penguin. Colodey, A. G., and Wells, P. G. (1992). Effects of pulp and paper mill effluents on esraarine and marine ecosystems in Canada: a review. Journal of Aquatic Ecosystem Health, 1, 201-226. Corrin, M . L., Klevens, H. B., and Harkins, W. D. (1946). The determination of critical concentrations for the formation of soap micelles by the spectral behavior of pinacyanol chloride. Journal of Chemical Physics, 14(8), 480-486. Davis, C. B. (1989). Rosin soaps as polymerization emulsifiers. Navel Stores, D. F. Zinkel and J. Russell, eds., Pulp Chemicals Association, New York, 625-642. Doshi, M . R., and Dyer, J. M . (1997). Paper Recycling Challenge, Volume I, Stickies., Doshi & Associates. Fengel, D. and Wegener, G. (1984). Wood: Chemistry, Ultrastructure, Reactions, Walter de Gruyter. Fox, M . E. (1976). Fate of selected organic compounds in the discharge of kraft paper mills into lake superior. Identification and analysis of organic pollutants in water, L. H. Keith, ed., Ann Arbor Science, 641-659. Funke, L. (1996). Order issued to pulp mill for exceeding effluent limits. 96-034, Government of Alberta. Gaston, C. (1993). Pulp and paper industry compliance costs. 11-528E, Statistics Canada. Hebron, K., Krahn, P., and Bannister, G. (1997). 1996 Annual compliance report for pulp and paper effluent regulations in British Columbia. Regional Program Report 97-27, Environment Canada. Hemingway, R. W, and Greaves, H. (1973). Biodegradation of resin acid sodium salts. Tappi Journal, 56(12), 189-192. Jonasson, R. G., Tosto, F., Holloway, L., Rawluk, M. , Goulet, C , Fuhr, B., and Leary, G. Comparison of extractives in pulps from mechanical and chemical pulp mills: Composition and impact. Spring Conference, Whistler, British Columbia. Kolthoff, I. M. , and Stricks, W. (1948). Solubilization of dimethylaminoazobenzene in solutions of detergents -1. The effect of temperature on the solubilization and upon the critical concentration. Journal of Physical and Colloidal Chemistry, 52, 915-941. Kovacs, T., and O'Connor, B. (1996). Insights for toxicity-free pulp and paper mill effluents. MR 331, Paprican. Kulovaara, M. , Kronberg, L., and Pensar, G. (1987). Recoveries of some chlorophenolics and resin acids from humic water. The Science of the Total Environment, 62, 291-296. Leach, J. M. , Mueller, J. C , and Walden, C. C. (1977). Biodegradability of toxic compounds in pulp mill effluents. Transactions of the technical section CPPA, 3(4), TR126-TR130. Leach, J. M. , Mueller, J. C , and Walden, C. C. (1978). Biological detoxification of pulp mill effluents. Process Biochemistry, 13(1), 18-21. Leach, J. M. , and Thakore, A. N . (1976). Toxic constituents in mechanical pulping effluents. Tappi Journal, 59(2), 129-132. 17 McCubbin, N. (1983). The basic technology of the pulp and paper industry and its environmental protection practices. EPS 6-EP-83-1, Environment Canada. McLeay, D. (1987). Aquatic Toxicity of Pulp and Paper Mill Effluents: A Review. EPS Report 4/PF/l, Environment Canada. Mueller, J. C , and Walden, C. C. (1976). Detoxification of bleached kraft mill effluents. Journal of the Water Pollution Control Federation, 48(3), 502-510. Ng, K. S. (1977). Detoxification of bleached kraft mill effluents by foam separation, PhD, University of British Columbia, Vancouver. Ng, K. S., Mueller, J. C , and Walden, C. C. (1974). Study of foam separation as a means of detoxifying bleached kraft mill effluents, removing suspended solids and enhancing biotreatability. CPAR Project Report 233-1, Ottawa. Niimi, A. J., and Lee, H. B. (1992). Free and conjugated concentrations of nine resin acids in rainbow trout (Oncorhynchus mykiss) following waterbome exposure. Environmental Toxicology and Chemistry, 11, 1403-1407. Nyren, V , and Back, E. (1958). The ionization constant, solubility product and solubility of abietic and dehydroabietic acid. Acta Chemica Scandinavica, 12(7), 1516-1520. Oikari, A., Lindstrom-Seppa, P., and Kukkonen, J. (1988). Subchronic metabolic effects and toxicity of a simulated pulp mill effluent on junvenile lake trout, Salmo trutta m. lacustris. Ecotoxicology and Environmental Safety, 16, 202-218. Ramamurthy, P., and Wearing, J. T. System closure: A Canadian perspective. 84th Annual Meeting, Technical Section of the CPPA, Montreal, A215-A222. Rocheleau, M. J., Sithole, B. B., Allen, L. H., Iverson, S., Farrell, R., and Noel, Y. (1998). Fungal treatment of Aspen chips for wood resin reduction: A laboratory evaluation. Journal of Pulp and Paper Science, 24(2), 37-42. Rogers, I. H. (1973). Isolation and chemical identification of toxic components of kraft mill wastes. Pulp & Paper Magazine of Canada, 74(9), 303-308. Servos, M. R., Munkittrick, K. R., Carey, J. H., and van der Kraak, G. J. (1996). Environmental fate and effects of pulp and paper mill effluents., St. Lucie Press. Soltes, E. J., and Zinkel, D. F. (1989). Chemistry of rosin. Navel Stores, D. F. Zinkel and J. Russel, eds., Pulp Chemicals Association, 261-345. Strazdins, E. (1989). Paper sizes and sizing. Navel Stores, D. F. Zinkel and J. Russell, eds., Pulp Chemicals Association, New York, 575-624. Tana, J. J. (1988). Sublethal effects of chlorinated phenols and resin acids on rainbow trout (Salmo Gairdneri). Water Science and Technology, 20(2), 77-85. Taylor, B. R., Yeager, K. L., Abernethy, S. G., and Westlake, G. F. (1988). Resin Acids, Queen's Printer for Ontario. U.S. Department of Agriculture. (1955). Wood Handbook, U.S. Government Printing Office. Volkman, J. K., Holdsworth, D. G., and Richardson, D. E. (1993). Determination of resin acids by gas chromatography and high-performance liquid chromatography in paper mill effluent, river waters and sediments from the upper Derwent Estuary, Tasmania. Journal of Chromatography, 643,209-219. Walden, C. C , and Howard, T. E. (1981). Toxicity of pulp and paper mill effluents - a review. Pulp & Paper Canada, 82(4), T143-T148. Wangaard, F. F. (1981). Wood: Its Structure and Properties. Clark C. Heritage Memorial Series on Wood, Pennsylvania State University. Water Resources Branch. (1992). Acute lethality data for Ontario's pulp and paper sector effluents covering the period from January 1990 to June 1990. ISBN 0-7729-8926-5, Ontario Ministry of the Environment. 18 Chapter 2 Surfactancy and Solubility in the Biological Removal of Resin Acids Summary The objective of this investigation was to determine the influence of resin acid solubility on the contaminant bioavailability (Objective A - Chapter 1). The solubility and surfactancy of resin acids are strongly influenced by pH within the range typically used for biological treatment. Changes in solubility and surfactancy for hydrophobic contaminants can be sensed by dynamic surface tension measurements. Relative changes in the contaminant diffusivity can also be derived from dynamic surface tension measurements. Such changes impact on the bioavailability of resin acids during batch treatment. Under acidic conditions, resin acids form associations with other dissolved organic matter contained in pulp mill effluent, while under alkaline conditions, they behave as relatively soluble surfactants. A residuum of a contaminant at the end of a batch treatment represents an important limitation for biological removal. The resin acid residuum, or threshold concentration, was found to increase under acidic growth conditions. This residuum increase corresponded to an inferred reduction in resin acid bioavailability. Table of Contents 2.1 Introduction 20 2.1.1 The Fate of Resin Acids 21 2.1.2 Dynamic Surface Tension Measurement 25 2.1.3 Modelling Adsorption by Dynamic Surface Tension Measurements 28 2.1.4 Quasi-static Langmuir Adsorption 33 2.1.5 Modelling Kinetics of Batch Growth with Threshold Concentrations 39 2.2 Experimental Methods and Materials 43 2.2.1 Mixed Culture Batch Growth 43 2.2.2 Dynamic Surface Tension Measurement 48 2.2.3 Nonlinear Regression Analysis 52 2.3 Results 55 2.3.1 Mixed Culture Batch Growth 55 2.3.2 Dynamic Surface Tension Measurements 58 2.4 Discussion 66 2.5 Conclusions 73 2.6 References 75 19 2.1 Introduction The purpose of this study was to determine i f pH could alter the bioavailability of resin acids (Objective A - Chapter 1). It was felt that a fundamental study into the fate of resin acids was justifiable i f this effect could be shown to be valid from a very limited preliminary investigation. Hence, batch growth experiments were performed as a function of pH, using a representative medium of kraft mill effluent. The experimental objective was to monitor the influence of pH on biological removal in relation to the resin acid chemical properties of surfactancy and solubility. Of particular interest was the issue of bioavailability. Surfactancy was considered by isotherm modelling of resin acid adsorption kinetics determined from dynamic surface tension measurements. Solubility was estimated according to the Standard Methods (Clesceri et al. 1989) definition for suspended solids. Necessary background literature and theory used for the interpretation of the experimental results are presented in the introduction. First, a review of the relevant literature is presented, indicating that adsorptive properties of resin acids are linked to pH and that adsorption is an important aspect of the contaminant fate. This review also describes how previous literature data on the influence of pH on resin acid degradation are incomplete and in apparent contradiction. The ability of surface tension measurements to follow resin acid adsorption is then explained with references to a relevant body of literature. Since surface tension is an indirect measurement of adsorption, the relationships that couple the kinetics of surface tension change to the kinetics of adsorption are described. Thermodynamic justification is then given for the Quasi-static Langmuir Adsorption model that was developed to interpret the experimental data. Along with the physico-chemical properties of resin acids, quantifying microbiological activity was an important part of this study. Therefore, some important issues surrounding the modelling of batch growth kinetics are also covered in the introduction. Due to their toxicity, some organic pollutants may be of concern even at trace levels. Thus, the contaminant concentration at which microbial growth can no longer be sustained is an important parameter to be derived from batch growth experiments. This concentration level is often referred to as a threshold for microbial 20 growth. The required modification to the Monod model to include such a threshold resin acid concentration is also presented in the introduction. 2.1.1 The Fate of Resin Acids The hydrophobicity of resin acids makes them surface active compounds that will tend to gravitate towards interfaces in aqueous environments. This property is manifested in conventional secondary biological treatment systems by resin acid accumulation in surface foams. Resin acid concentrations in surface foams above biobasins can attain levels that are orders of magnitude higher than those in the underlying liquid (Chandrasekaran et al. 1978; Fein et al. 1992; Servizi et al. 1975; Zitko and Carson 1971). Collapse of the foam layer is a probable cause for some of the treatment detoxification failures that have been reported in the past (Leach et al. 1977; Servizi etal. 1975). In a more recent example of contaminant concentration in surface foam (Kruzynski 1995), foam sampled from the discharge pond at a British Columbia pulp mill exhibited an L C of 0.75 50 percent for rainbow trout (van Agglin 1995). The resin acid concentration in the collapsed foam was in the order of 5 ppm, which was a thousand fold greater than that in the underlying wastewater (Bicho 1995). Foam sampled downstream of this pulp mill on the Fraser river also contained elevated levels of resin acids (Duncan 1993). While only resin acids were assayed for these samples, they would not have been the only contaminant in the foam causing toxicity. In spite of its toxic contents, foam from pulp mill lagoons has been reduced to that of an aesthetic problem that is a (public) nuisance, especially i f it is blown to the surrounding community (Palermo and Holzer 1992). Although foam is problematic, the tendency for resin acids to adsorb onto the gas bubbles that generate foams has been of some benefit for their selective oxidation by ozone treatment (Roy-Arcand and Archibald 1996). Hence, resin acids are readily collected at gas/liquid interfaces of bubbles and foams. Adsorption at gas/liquid interfaces can be sensed by measuring surface tension reduction (Burcik 1950; Burcik 1953). The rate and extent of surface tension lowering should be related to the 21 contaminant bulk liquid concentration. Therefore, i f resin acid adsorption is an important controlling aspect of the contaminant fate during secondary treatment, then it should be possible to characterize this transport mechanism from surface tension measurements. The fate of these chemicals in wastewater treatment systems mirrors their behaviour in the open environment. One can consider the partitioning of a pollutant from an effluent stream into biological (lipid), solid (suspended or settled particulate matter), aqueous, or gaseous environmental compartments. Partitioning is dependent on the physico-chemical properties of a pollutant. The fate of pulp mill effluent contaminants in treatment systems (Mackay and Southwood 1992; Mackay et al. 1996) and in the environment (Kolset and Heiberg 1988) has been modelled using fugacity concepts. Fugacity measures the thermodynamic driving force for partitioning into air, soil, water, biota, suspended solids and sediment. From such considerations, predictions are made about where most of the solute will partition and where the highest concentrations should occur (Mackay and Paterson 1981). Each compartment can have associated rate constants for physical or biological degradation. A chemical will persist in the environment, or in a treatment system, if it is partitioned away from the more rapid degradative processes. In biotreatment systems, resin acids could exhibit limited bioavailability because they are likely to be transported into surface foams (Leach et al. 1977) as surfactants. The bioavailability of resin acids can also be limited as a result of close association with particulate organic matter due to their hydrophobicity (Alexander 1994). For instance, Brownlee (1977) estimated that the half-life of dehydroabietic acid discharged into Lake Superior was 21 years in the sediment, as compared to 0.12 years in the water column. It is unclear i f the prolonged half-life in the sediment was due to reduced bioavailability or due to persistence under anaerobic conditions. It has been stated that the major dispersal mechanism for resin acids at the final effluent outfall is by adsorption onto, or sedimentation with, suspended particulates (Carlberg and Stuthridge 1996). Therefore, an important part of understanding the fate of trace hydrophobic compounds in treatment systems or the environment, is in the determination of the governing thermodynamic 22 potentials driving contaminant partitioning behaviour. Interfaces, either gas/liquid or solid/liquid, between environmental compartments, are regions where significant amounts of hydrophobic chemicals collect and concentrate (Hoff et al. 1993; Valsaraj 1994). In this respect, surface tension measurements should also be able to indicate the extent to which the surface phase should be considered as a distinct compartment for fugacity type fate models. The literature suggests that pH is the dominant parameter determining whether resin acids partition as a soluble surfactant or as an insoluble particulate-bound contaminant. Ng (1974a) found that an alkaline pH resulted in more consistent effluent detoxification by foam fractionation for the same duration of treatment. Hence under alkaline conditions, resin acids appear to be better surfactants. The dispersive power of resin acids has also been shown to be much greater under alkaline conditions (Drobosyuk et al. 1982). Conversely, the extraction recovery of resin acids from slightly acidic humic water was found to decrease over time (Kulovaara et al. 1987). This reduction in extraction recovery indicates that under acidic conditions resin acids become bound to other organic matter in solution. Reduced extraction recovery over time was probably due to "aging" interactions with dissolved organic matter (Hassett and Anderson 1979; Kukkonen 1992). Aging refers to the observed reduction of the bioavailability of a sequestered compound over time (Alexander 1994). Hence, the fate of resin acids changes in the transition from acidic to alkaline environments. This observed change with pH can readily be understood i f one considers that the pKa value for resin acids predicts a significant change in the dissociated fraction in the pH range from 6 to 8 (Nyren and Back 1958). The direct effect of an increased dissociated fraction of resin acid in solution with increased pH is a parallel increase in total solubility. For instance, in the pH range from 6 to 8, the total solubility of abietic acid increases from a few micrograms per millilitre to over 100 pg/mL (Nyren and Back 1958). The pH range from 6 to 8 needs to be further emphasized because it also defines the normal operating range of a mill secondary biological treatment process. Therefore, within the normal operating pH conditions of pulp mill secondary biological treatment systems, resin acids, which 23 need to be metabolized to well below 1 u.g/mL, undergo a dramatic change in nature. There have been only two studies that explicitly considered the effect of pH on biodegradation of resin acids. One investigation utilized batch growth conditions (Hemingway and Greaves 1973) and the other, twenty years later, yielded conflicting results with a bench scale continuous system (Liu et al. 1993). Hemingway and Greaves (1973) monitored pH during batch growth on resin acids (40 u,g/mL) in a Dubos buffered medium containing 10% bisulfite liquor. With slightly acidic starting conditions, it was found that no significant resin acid removal occurred until the pH exceeded 7.3. Unfortunately, little can be derived from this aspect of their investigation. Without pH control, it is unclear i f the significance of pH 7.3 had more to do with diauxic growth on other organic acids in the bisulfite liquor, than a supposed limiting pH for biological removal. Liu et al. (1993) operated bench scale, continuous flow stirred-tank reactors with a 3 day hydraulic retention time (HRT) on chemi-thermomechanical pulping (CTMP) effluent containing 45 p.g/mL resin acid. In contradiction to Hemingway and Greaves' conclusion regarding the effect of pH, this study demonstrated that under steady state conditions, a consistent 98% removal of influent resin acid could be achieved in the pH range from 5 to 8. Unfortunately, the possibility of changes in biodegradation kinetics with pH was not considered, perhaps due to the relatively long HRT used for this part of the study. Had a shorter HRT been used it might have been possible to observe different steady state effluent concentrations which could have been related to pH-dependent removal kinetics. Additionally, while the observed 98% removal efficiency is significant, it should be noted that, depending on pH (McLeay et al. 1979a; McLeay et al. 1979b; Zanella 1983), the residual 2% resin acid in the effluent could still be considered acutely toxic to aquatic life. If this observed 2% residual persisted due to a limitation in bioavailability, it would likely not be reduced by increasing the HRT further. In other words, it remains unclear i f the observed contaminant residual concentration was an inherent limit for removal or a function of the removal kinetics during continuous treatment. Therefore, a shortcoming exists when contaminant removal efficiencies are used to indicate efficacy in the reduction of toxicants in wastewater, when the goal is detoxification. 24 There are five major conclusions to be drawn from this discussion of the literature. Surfactancy and solubility are important determinants of the fate of resin acids in kraft pulp mill effluents. Dissociation of resin acids as a function of pH strongly influences their surfactancy and solubility. The significance of surfactancy and solubility with respect to biological removal of these contaminants has not been well established in the literature. The two published investigations that explicitly examined the effect of pH on biodegradation are in apparent contradiction. Finally, the properties of resin acids significantly change within the typical pH range used for secondary treatment of pulp mill effluents. In order to clarify the results and the apparent contradictions between the published findings about the influence of pH on resin acid biodegradation kinetics, batch growth experiments were conducted with abietic and dehydroabietic acid in a well buffered matrix of treated and filtered bleached kraft mill effluent. Growth curves were generated at constant pH in the range from 6 to 8 in order to determine whether changes in total resin acid solubility could either alter the uptake kinetics or affect bioavailability. Resin acids that are mobile enough to become adsorbed at a gas/ liquid interface, may be mobile enough to diffuse into a microbe to be metabolized. Hence, availability for adsorption was considered to be a potential indicator for relative differences in bioavailability. Kinetics of adsorption were calculated from dynamic surface tension measurements. 2.1.2 Dynamic Surface Tension Measurement Since adsorption of solute at the gas/liquid interface will act to lower the liquid surface tension, measurements of surface tension as a function of solute concentration and pH would provide information about differences in the mobility of resin acids in solution. Dynamic surface tension measurements can indicate both the extent and the rate of surfactant adsorption at a gas/ liquid interface (Burcik 1950; Burcik 1953). There are a few reports in the literature of the use of surface tension measurement in conjunction with pulp mill effluent or its constituents. Brasch (1974) measured the equilibrium surface tension 25 of black liquor, producing a surface tension versus concentration relationship similar to those of common surfactants. Berk et al. (1979) reported a strong correlation between Daphnia magna median survival time and surface tension, for foam fractionated spent sulphite liquor. In that study, foam fractionation reduced toxicity by 40%, by concentrating the toxicants in 20% of the original volume. Keirstead (1978) reported the influence of fungal treatment on the surface tension of a 1% (w/v) sodium lignosulfonate solution. Werker et al. (1996) followed surface tension changes during the hydroxylation of dehydroabietic acid by the fungus Mortierella isabellina. These studies indicate that surface tension measurements of wastewater can be used as a method to quickly screen for the presence of hydrophobic contaminants. In pulp mill effluent, where toxicity is, to a large extent, associated with such hydrophobic compounds as resin acids, surface tension measurement offers promise as a rapid monitoring tool in effluent treatment quality control. The kinetics of surfactant adsorption at the gas/liquid interface can be monitored by dynamic surface tension measurements. This approach depends on the surface aging phenomenon (Adamson 1982) whereby a fresh gas/liquid interface is experimentally generated and the progress of surface tension lowering due to solute adsorption is monitored with time. By the application of a suitable thermodynamic model, the reduction in surface tension can be related to solute loading at the interface (Kohler 1993). Measurement of dynamic surface tension is especially relevant for dynamic engineering processes in which equilibrium conditions may not be attained. For example, dynamic surface tension measurements have been used to explain the lack of correlation between equilibrium surface tensions (by contact angle measurements) and the flotation recovery of mining ores (Finch and Smith 1972). This fundamental information has engineering application in modelling of the kinetics of adsorption for the purpose of industrial process design (Bing et al. 1988; Clarke and Wilson 1983). Dynamic surface tension measurement has also been used to study the effects of surfactant adsorption on oxygen transfer in aerobic biological treatment systems (Masutani and Stenstrom 1991). A number of experimental techniques that cover a range of adsorption time scales have been 26 devised to assess the kinetics of surface aging by indirectly monitoring gas/liquid surface tension changes in time. Whether one considers surface aging over tens of milliseconds by the oscillating jet method (Thomas and Potter 1975), over tens of seconds by the maximum bubble pressure method (Mysels 1990), or over minutes by the pendant drop or drop volume method (Adamson 1982), the theory of capillarity (Adamson 1982) is the important theoretical starting point. The theory of capillarity centres on Laplace's equation relating the radii of curvature (R and R ) 1 2 across an interface (Figure 2-1) to the differential pressure (Ap) and surface tension ( a ) of a gas/ t liquid interface aged for the time period t (Adamson 1982): Ap = cr, ^2 Measuring surface tension lowering provides an indirect indication of the progress of surfactant adsorption at a gas/liquid interface with time. However, it is usually of interest to estimate the actual progression of the contaminant surface loading. Since the contaminant surface loading cannot be measured directly, it is necessary to apply a theoretical model in order to translate the surface tension o for an interface aged over the time, t, to a corresponding surface loading, T. t t Contaminant surface loading is expressed in units of moles per unit area. Figure 2-1. A small (elemental) section of an arbitrarily curved gas/liquid interface. In general it is necessary to use two radii of curvature to describe the curved surface. The two radii would be equal if the section was an element of a sphere. There will be a pressure difference Ap across the surface which will be balanced by the interfacial surface tension, a, according to the equation of Young and Laplace (Adamson 1982). 27 2.1.3 Modelling Adsorption by Dynamic Surface Tension Measurements Much of the theory surrounding dynamic surface tension measurement seems to be directed towards deriving the solute diffusivity. Estimating changes in resin acid diffusivity was felt to be important for the present investigation. A change in diffusivity with pH would indicate a relative change in mobility which could be related to differences in bioavailability. In contrast to the current investigation, the motivation for previous theoretical research on modelling dynamic surface tension measurements arose due to the relationship between solute diffusivity and foaming (Burcik 1950). The transport of surfactant from the bulk liquid solution to a perturbed or non- equilibrium gas/liquid interface represents an important destabilising effect for the thin films that make up foams (Bendure 1975). In the case of foam stability, the rate of surfactant adsorption to the film surface needs to be relatively low (Davies and Rideal 1963b). The problem of relating solute diffusivity to adsorption at a gas/liquid interface was solved originally by Ward and Tordai (1946). Their derivation required the definition of a near surface, being the bulk liquid layer directly adjacent to the interface. By assuming that the kinetics of attainment of adsorption equilibrium between the near surface and the interface were instantaneous, they asserted that the process of adsorption was diffusion-limited by solute transport from bulk solution to the near surface. Mathematically, the problem then became analogous to the temperature as a function of the distance from a surface of a semi-infinite solid with surface temperature varying in time. Under this construct, surface loading, T, as a function where c is the liquid solute concentration, D the diffusion coefficient and c the near surface b s concentration as a function of time. Note that while the liquid concentrations (c and c) are b s expressed in moles per unit volume, the surface loading (T) is given in moles per unit area. If the Ward and Tordai model (equation (2-2)) is restricted to just the initial short time interval of adsorption, it can be simplified to: of time, t, was modelled to be: (2-2) 28 Thus Ward and Tordai predicted a linear dependence of the initial adsorption rate on the square root of time. Since the surface loading (T) could not be measured directly, a model relating t surface loading to surface tension was required in order to test the experimental validity of equation (2-3) and to estimate a surfactant diffusivity (D). The theory of diffusion-controlled adsorption kinetics for systems obeying Langmuir's adsorption isotherm was developed by Hansen (1960). By adopting a specific adsorption isotherm relationship, the measured quantity of surface tension (a) could be used to determine the model parameter of diffusivity (D). Hansen made short and long term time approximations of the Ward and Tordai model and arrived at the following equation based on Gibbs adsorption theorem: < T , - < r e = - i y ? r i n for small times and: \__ TL 1 (2-4) TfRT (2-5) < J ! - ° e = - ± f = CbylTtDt for long times. In these equations, cj is the surface tension as a function of time, a is the t e equilibrium surface tension, T is the limiting surface loading, R is the universal gas constant and L T is the temperature. Hansen concluded that a diffusion-controlled adsorption mechanism was not a good representation of the initial dependence of surface tension aging with time. Bendure (1971) also noted that diffusion-limited adsorption modelling did not fit the measured data over all time scales. Joos and Rillaerts (1981) incorporated convection into the Ward and Tordai model to improve the estimation of solute diffusivity from dynamic surface tension measurements. However, i f convection is indeed significant, one would expect that the near surface would maintain essentially the bulk liquid concentration thereby contradicting the underlying premise of the original Ward and Tordai theory. Without a near surface deficit of solute, no diffusion barrier 29 would exist for the adsorption process. In any event, the necessity to define short and long time scales (equations (2-4) and (2-5)), whose delineation would be a function of bulk solute concentration, severely limits the practical application of the Ward and Tordai representation. With reference to traditional continuum mechanics mass transport modelling (Incropera and de Witt 1981), there is more practical gain in relating dynamic surface tension measurements to a relationship describing the flux of solute to the gas/liquid interface while ascribing some probability for the adsorption of any impinging solute molecule: dY , r , (2-6) dt where T is the solute loading at the interface, J is the flux of solute to the interface and O is a sticking coefficient (Ertl 1983) specifying the probability that a particle striking the surface becomes adsorbed. The flux is a function of solute mobility, B , times some function of the bulk solute concentration, c . The solute mobility will be a constant characteristic of the solute for the b particular conditions of the solvent. The sticking probability is conceivably related to both the degree of coverage (Ertl 1983) and also to the spreading pressure. Spreading pressure (IT) is the surface analogue to pressure in a closed container. It represents a depression in the solvent surface tension (c ) due to solute surface loading (T). The adsorbed solute monolayer behaves 0 similarly to a two dimensional gas. In its simplest form, the equation of state for the adsorbed phase is an analogue to the ideal gas law: (2-7) U = (70-<j = rRT  v ' Thus, for the sticking probability, the degree of coverage (T) relates to the chances of finding an empty site and the spreading pressure (IT) to the surface forces that must be overcome in squeezing yet another molecule into the surface monolayer. There are two possible models that could describe the flux of solute to the interface. Flux of solute to the interface could be thought of as being a passive process that is dependent on random occurrences from Brownian motion. Flux in this case would be directly related to bulk solute 30 concentration. The higher the solute concentration, the greater the chances of solute collisions with the interface would be. Conversely, based on a thermodynamic model, adsorption could be seen to be driven by solute chemical potential differences between the bulk solution and the surface (Ruthven 1984). In this case, the flux would be dependent on the natural logarithm of bulk solute concentration. The question is, which, i f either, of these two models provides the best representation. Modelling single-stage removal of a surfactant from a water column by bubble fractionation (solvent sublation), Clarke and Wilson (1983) assumed, on the basis of intuitive justification, that the kinetics of adsorption are governed by an equation of form: dT dt W. <2"8> where T is the equilibrium surface loading of a Langmuir isotherm. The rate of mass transfer to e the interface is proportional to the difference between the actual surface loading and the equilibrium surface loading. Equating equation (2-8) to equation (2-6), the sticking coefficient is given by the expression in parenthesis and the function of bulk concentration is contained within T , the equilibrium surface loading. No experimental results have been found in the literature that e confirm the validity of equation (2-8). Hua and Rosen (1988) have shown (Figure 2-2) that the process of surface aging for a constant surfactant concentration can be generalized empirically by a sigmoid on a logarithmic time scale: <ya-<ym (2-9) 1+ ('/**)" where the surface tension o, falls with time from the level of the freshly generated interface (a) t o to a so-called meso-equilibrium surface tension, a , for the aged surface. Meso-equilibrium is m differentiated from the equilibrium surface tension because over much longer time scales, surfactant conformational changes at the interface can act to lower the surface tension even further. For the present investigation, meso-equilibrium was considered as the operational 31 a Figure 2-2. Sample data from Hua and Rosen (1988) illustrating the effect of surfactant concentration (c) on dynamic surface tension for N-dodecyl-N-methylglycine at pH 9.0 and 25 °C. The data show surface tension lowering (o"t) in terms of surface age or log(t) for a family of surfactant concentrations. The log-surfactant concentrations, log(c), in water were equal to: -2.992 (•), -3.224 (O), -3.410 (•»), -3.525 (A), -3.701 (•), -3.826 (0), and -4.108 (•). Solid lines were calculated by Hua and Rosen using best fit parameters for equation (2-9). Note the change in the time lag before the onset of the log-linear decrease in the surface tension. definition for equilibrium. The Hua and Rosen (1988) empirical curve parameters {rj , n, t*} are functions of ionic strength, m temperature and solute bulk liquid concentration (Figure 2-2). Systematic changes in surfactant concentration, with all other things held constant, produces a family of similar curves, each of which is described by equation (2-9). The duration of the initial plateau, or the time before significant surface tension lowering, is referred to as an induction period. In a subsequent publication (Gao and Rosen 1995), the induction period, which is defined by the parameters {n, t*}, was shown to be related to solute diffusivity as defined by the Ward and Tordai model (Ward and Tordai 1946) for diffusion-controlled adsorption. Upon reviewing the literature, it became apparent that modelling of adsorption kinetics (Clarke and Wilson 1983) requires a "time implicit" mathematical description (equation (2-8)). Time implicit refers to the fact that surface age is not explicitly considered in determining the flux of surfactant to the interface. Each increment in surface loading depends on the current state of the interface. In contrast, experimental theorists appear, since the work of Ward and Tordai, to be 32 locked onto a time explicit description of adsorption at a gas/liquid interface. A time explicit description of adsorption, such as equation (2-4) or equation (2-9), means that the surface age must be known explicitly before the increment in surface loading can be calculated. It is probably for this reason that, other than providing qualitative insight, theoretical developments in dynamic surface tension measurements seem to have had limited engineering application. That the parameters, {a , n, t*}, were found to be systematic functions of solute concentration (Hua and m Rosen 1988) means that, in principle, it should be possible to similarly obtain an expression describing the family of iso-time dynamic surface tension curves relating the dependence of surfactant concentration on a particular surface age. Cast in this form, it will be shown that such an expression would help to justify or improve upon the model relationships assumed by Clarke and Wilson (1983). It was found that, for the purpose of the present investigation, such a time implicit approach became the only means with which to obtain useful information from dynamic surface tension measurements. The method applied was to idealize the adsorption process as a locus of quasi-static equilibrium states (Callen 1985). 2.1.4 Quasi-static Langmuir Adsorption Real adsorption processes involve considerations of rate, making them a temporal succession of non-equilibrium states leading to some terminal equilibrium. In contrast, a quasi- static process is defined in terms of a dense, ordered succession of equilibrium states. A quasi- static adsorption process would not involve consideration of rate or time per se although it may be possible to contrive a temporal relationship to the quasi-static process. The quasi-static process can be understood from a conceptual example. Consider a closed system containing a freshly formed gas/liquid interface in an aqueous solution with a surfactant solute. Conceptually the surface aging process is seen to move from its initial state, A , through a locus of points {B, C, D,...} in thermodynamic configuration space by the successive relaxation of some constraint. With the relaxation of the constraint, more solute adsorbs at the interface. Each state given by the locus of points defines equilibrium conditions of adsorbed solute. In the initial relaxation of the constraint, the system is permitted to proceed from A to B but no further. The 33 system disappears from A and subsequently appears at B after passing through non-representable non-equilibrium states. Relaxing the constraint further, the system proceeds to C, and so forth. At each point a greater level of solute is adsorbed at the interface. The quasi-static locus of states can be approximated by a real process in a closed system only i f the entropy is monotonically increasing along the quasi-static locus of points (Callen 1985). The correspondence between each state on the locus of points to experimentally-measured steps of increasing surface age provides the temporal relationship of the real process to the quasi-static one. Theoretical justification for modelling adsorption as a quasi-static process can be found in the thermodynamic relations for non-equilibrium surfaces presented by Kohler (1993). This justification starts with the definition of Gibbs free energy of adsorption (AG a): AG°=G--G> ( 2 " 1 0 ) where G e is the actual free energy at equilibrium and & is a hypothetical reference free energy i f adsorption did not take place. For simplicity, but with no loss of generality, the system being considered for this discussion is restricted to two chemical components. Component 1 is the solvent and component 2 the solute surfactant. The location of the (dividing) surface is defined by the plane where the "surface excess" of the solvent (component 1) is zero. The term surface excess comes as a result of the need to mathematically define a surface plane between two phases that in reality do not change sharply from one to the other (Adamson 1982). If one considers the interface between two bulk phases (gas and liquid), the mass balance for any chemical component across the interface will include a term for the mass located in the phase transition region. The surface excess per unit area refers to the amount of any component in this transition region. Since it is convenient to suppose that the two bulk phases are uniform up to some arbitrary dividing plane, the location of this plane must be clearly defined. With the definition for the dividing surface given above, surface loading (F) refers unambiguously to the solute surface excess (component 2). In addition, any references to chemical potential (u.) are also made with respect to the solute. 34 For an ideal dilute solution, where the bulk liquid phase is large with respect to the surface phase, the surface tension of the reference state can be approximated by the solvent surface tension and the chemical potential within the bulk solution will be negligibly affected by adsorption. Hence, the above mentioned Gibbs reference state (Gr) is approximated by the free energy of the system when no solute is adsorbed at the interface. Under these conditions the Gibbs free energy of adsorption simplifies to: AG" =(o-(J0)A = -UA ( 2 _ U ) where a is the surface tension, a is the surface tension for zero solute surface excess, A is the o surface area and IT is the surface pressure (equation (2-7)). From equation (2-11), the specific Gibbs free energy of adsorption is AG a normalized by the surface area (A): 4rf=*-«r.=-n < 2 " 1 2 ) Since surface pressure (IT) relates to surface loading or surface excess (Equation (2-7)), Gibbs free energy is minimized by the exchange of solute from solution to the surface phase. Equation (2-12) can be combined with Gibbs adsorption equation at constant temperature: (2-13) da = -Tdpi v ' and this combination yields: Ag^ =(o--<70)= )dd = -)tdU (2" 1 4 ) where the tilde is used to specify integration variables and p.b is the chemical potential for the solute in the bulk solution which is approximately constant. Integrating equation (2-14) by parts expresses the Gibbs free energy change occurring when the solute amount T is exchanged between the bulk liquid phase (with constant chemical potential pb) and the surface phase: 35 Ag:=pdo = f\jl-Lib)df J&n JO (2-15) Again the tilde is used to specify integration variables. At the beginning of the exchange process of solute from solution to the surface, surface loading (IT) is zero. Ultimately the surface phase reaches equilibrium with the bulk liquid phase which, at constant pressure, is defined by: ji(T,r.-) = nb(T,cb) ( 2 " 1 6 ) Although the adsorption process passes through non-equilibrium states, from equation (2-15) the chemical potentials of the surface can still be identified with equilibrium potentials of the form: P - m r ) ( 2 " 1 7 ) A surface phase cannot exist without the existence of two neighbouring extended bulk phases, namely, gas and liquid. Hence, the intensive surface state thermodynamic variables depend not only on T and T but also in general on the bulk solute concentration. If equation (2-17) represents an equilibrium potential then a quasi-static equilibrium condition can be defined by: * ( 2" 1 8) U'(T,Tt) = ilbCT,cb) where T is the surface loading at a surface age of t. With reference to the locus of points t describing the quasi-static adsorption process, at constant temperature and pressure, the corresponding states in thermodynamic configuration space are defined by the sequence of surface loadings {T , T , T , Y , . . . } . To be valid, the states must be along a path of decreasing A B C D Gibbs free energy of adsorption, thereby moving the system along the route towards a minimum Gibbs free energy. According to the quasi-static equilibrium condition of equation (2-18), there exists an isotherm type relationship of the form: (2-19) r,=r,(T,cb) describing the surface loading for a surface age at a given temperature and bulk solute concentration. The experimental validity of a quasi-static equilibrium condition is demonstrated, 36 for instance, by the ability of dynamic surface tension measurements to determine critical micelle concentrations (Brown et al. 1952). Resin acids have been shown to adsorb to biosolids according to the Langmuir isotherm (Liu et al. 1996a). The Langmuir isotherm describes adsorption to identical independent sites and, in keeping with the notation of equation (2-19), is given by: 6 =T> = c» ( 2 _ 2 0 ) ' r £ cb+Kt The limiting surface loading (T ) is a constant, dependent on stearic and charge interactions that L influence the packing of the solute at the interface. For instance, a monolayer of protonated resin o acids will collapse (buckle) at a surface coverage of 42.0 and 40.4 A per molecule for abietic and dehydroabietic acid respectively (Bruun 1952b). The difference between the two abietanes is due to the higher polarity of dehydroabietic acid because of its extra double bond. Therefore, T is a physical constant for the family of isotherms described by equation (2-20). The limiting surface loading will , however, be a function of the condition of the aqueous matrix. In the case of resin acids, ionization, with increased pH, influences the molecular configuration at the interface which in turn affects the degree of molecular surface coverage and interaction (Bruun 1952a). The presence of inorganic salts or the formation of molecular complexes also influences solute interactions in the monolayer (Davies and Rideal 1963a). The parameter K measures the standard state parameters of the surfactant (Lucassen-Reynders t and van den Tempel 1964): K, = Krt exp 1Ahad\ (2-21) where Ah a d is the enthalpy of adsorption (Kohler 1993) and K r is an isotherm constant. With the t quasi-static equilibrium condition of equation (2-18) written in terms of the partial molar adsorption enthalpy and entropy: 37 Aflad =Ahad -TAsad =0 (2-22) the adsorption entropy can be written in terms of the Langmuir parameters: Asad =RM (2-23) Therefore, K r has entropic meaning and the standard adsorption entropy increases i f K r t t decreases. For the family of isotherms describing the quasi-static adsorption along a locus of points in thermodynamic configuration space, K is the relaxation constraint, where by definition: must be true. Combining Gibbs' adsorption equation (2-13) for an ideally dilute solution with the quasi-static Langmuir isotherm of equation (2-20) provides the link between surface loading and the measured quantity of surface tension: Therefore, from isothermal dynamic surface tension data that can be modelled by a Langmuir isotherm, it should be possible to relate the surface tension at particular surface age (t) to a corresponding bulk surfactant concentration. In defining this relationship, one should be able to estimate the limiting surface loading which is independent of surface age. The kinetics of adsorption are related to this quasi-static adsorption model, by the manner in which the parameter K decreases with surface age. For the data to be considered thermodynamically admissible, K t t must be continuously decreasing with surface age. KA>KB>KC>... (2-24) (jt=a0-RTTL\n \ + ̂ ~ (2-25) J 38 2.1.5 Modelling Kinetics of Batch Growth with Threshold Concentrations In batch growth experiments, beyond answering the basic question of biodegradability of a target contaminant, one is often interested in describing the data in terms of the parameters specified by the equation originally used for enzymology by Michaelis-Menton and for microbial growth by Monod (Bailey and Ollis 1986; Panikov 1995): QS (2-26) q ks+S where, q is the specific growth rate, S is the limiting substrate concentration, Q is the maximum specific growth rate when S » k , and k is the limiting nutrient concentration at which the s s specific growth rate is half its maximum value. These two constants can be derived from the period of exponential or balanced growth during batch cultivation. The constants are typically used for model predictions in the design of biological treatment systems (Liu et al. 1996b). Balanced microbial growth can be expressed by, dX_ = x = QSX (2-27) dt ks+S and growth proceeds with a parallel substrate depletion, dS _ QSX (2-28) dt~ Y(ks+S) where, X is the concentration of biomass and Y is the biomass yield coefficient. If the biomass yield coefficient is assumed to be constant, equation (2-28) can be integrated to give the non- linear Monod equation (Simkins and Alexander 1984) describing substrate concentration as a function of time: ks ln| rf\=(S0+X0/Y + ks UXJYY+?;~S)- + XJY)Q • t ( 2 " 2 9 ) where X and S are the initial biomass and substrate concentrations respectively. 0 0 39 The attraction of equation (2-29) is that, given experimental data for X and S(t), estimates of Y, o Q, and k can be obtained from non-linear least squares curve fitting of just the substrate s depletion curve or, with S and X(t), just the biomass production curve (Dang et al. 1989; o Robinson and Tiedje 1983; Simkins and Alexander 1985; Templeton and Grady (1988). However, because the parameters Q and k have similar sensitivity functions, their identifiability is s susceptible to measurement noise. The sensitivity functions express the influence, of a small perturbation in the model parameters being estimated, on the state variables of the system. The state variables in this case are biomass and substrate concentrations. Both measurement noise and sampling frequency influence the accuracy of the estimates. Simulation studies have shown that the correct parameters cannot generally be obtained from stochastic data (Holmberg 1982). The objective function to be minimized in modelling a data set will vary only slightly over wide ranges of parameter values, all giving a good fit. Therefore, rate constants obtained from just the substrate depletion curve must be viewed with some degree of scepticism. Linearisation techniques are possible (Levenspiel 1984), but these introduce statistical bias that may influence the experimentally-determined parameters (Berthouex and Brown 1994; Robinson 1985). However, the linearized equations are useful for obtaining initial guesses required for non-linear parameter estimation methods (Panikov 1995; Robinson 1985). The Monod model suffers from systematic errors at modestly small values of S (Panikov 1995). Monod kinetics predict the complete removal of a contaminant with the depletion rate becoming first order (Simkins and Alexander 1984) when S becomes much less than k . However, complete s removal may not be achievable and this is important i f the residuum remains acutely toxic or i f the contaminant tends to bioconcentrate or worse, is biomagnified (Muir and Servos 1996). Therefore, in a study of toxicity removal kinetics from wastewater, threshold levels for removal need to be addressed. The threshold contaminant concentration is defined by the level at which microbial growth on that contaminant can no longer be supported. Typical threshold concentrations for contaminants in aerobic systems have been reported in the 0.1 to 1.0 mg/L concentration range (Kobayashi and Rittmann 1982). 40 One probable cause for the limitation in the biological removal of hydrophobic contaminants is substrate availability (Providenti et al. 1993; Sikkema et al. 1995). Mass transfer from the particulate phase to the aqueous phase can be rate limiting. For example, the delivery rate of polycyclic aromatic hydrocarbons to competent bacteria, rather than the kinetics of oxidation, has been shown to be a limiting step (Stuck and Alexander 1987; Volkering et al. 1992; Volkering et al. 1995). Optimising environmental parameters, such as temperature or pH, which alter the physico-chemical properties of targeted wastewater contaminants so as to increase their bioavailability, should be a primary engineering design strategy. With reference to resin acid solubility, the tolerance range for biotreatment pH control may need to be tighter than the limits currently defined by the discharge regulations. The general opinion is that microbial uptake of cyclic hydrocarbons is a passive process (Sikkema et al. 1995). With reference to dynamic surface tension measurements, dissolution and passive transport of a contaminant to a microbe is akin to dissolution and adsorption at a gas/liquid interface. Thus, surface activity could, in principle, provide an indirect indication of bioavailability. The other aspect of limitation for biological removal is related to microbial energetics. A respiration threshold exists when the carbon requirements for cell maintenance fall below the diffusion rate of the chemical to the cell surface (Alexander 1994). This threshold can be lowered by the presence of alternate carbon sources. A threshold substrate concentration, S", below which growth is impossible, is therefore, a biologically justified modification of Monod's equation For the purpose of this investigation, the operational definition made for the threshold S* is a residuum at the cessation of observable growth. For an assumed constant yield, the rate of change in substrate concentration can again be derived based on the modified Monod equation (2- 30). However, for a numerical solution method, it is best to first normalize the equations. The non-dimensional variables and parameters that were chosen are as follows: (Panikov 1995): q = Q S-S* Ks+S (2-30) 41 Similar to equation (2-28), the non-dimensional Monod rate of substrate depletion with a threshold concentration and a constant yield is equal to: g(ri) = — = dr\ K + S ds s-s * ( V (2-32) There are two approaches to solving equation (2-32). The first approach entails analytical integration giving an implicit function similar to equation (2-29) which can be solved by numerical iteration. The second approach is to solve equation (2-32) directly by numerical integration. Since both methods are numerical and integration algorithms are easy to program and have more general usage, the latter approach was chosen. 42 2.2 Experimental Methods and Materials 2.2.1 Mixed Culture Batch Growth Medium and Batch Culture Preparation For these experiments, both microbiological inoculum and treated effluent were kindly supplied by Western Pulp Limited Partnership in Squamish, British Columbia. In Squamish, Western Pulp Limited Partnership has a coastal bleached kraft mill. At the time samples were obtained, the mill was pulping a fir:cedar:hemlock furnish and bleaching with 75% chlorine dioxide substitution at a 700 ADt/day production rate. Whole mill effluent undergoes secondary biological treatment by a high rate U N O X process that, at the time, was operating with an 8 hour HRT. Further references made to Squamish effluent refer to biologically (secondary) treated and filtered (Whatman 40) whole bleached kraft mill effluent (BKME). Abietane resin acid was recrystallized from commercial rosin (Hercules Pamite 79) following the method of Halbrook (1966). A 30:70 mixture of abietic (ABA) and dehydroabietic (DHA) acid was obtained. A concentrated solution of resin acid was prepared by reflux boiling 2.5 grams abietane resin acid stock, in 25 mL of HPLC grade methanol (MeOH). Parallel shake flask cultures were maintained at three buffered pH conditions. A l l glassware was thoroughly washed and fired at 475°C to remove any trace organic residue prior to use. Medium was prepared by adding a 1  lA mL aliquot of concentrated resin acid to 1.5 L of heated (60°C) Squamish effluent adjusted to pH 10 with 2 N NaOH. From this bulk solution, aliquots of 500 mL were decanted into 2 litre Erlenmeyer flasks containing weighed mixtures of Na HPO and 2 4 NaH PO H O salts to yield a 100 m M phosphate buffer at pH 6, 7 or 8 (Table 2-1). 2 4 2 This method of adding resin acid to a hot effluent at an alkaline pH, before pH adjustment, was chosen to produce an aqueous matrix that would resemble kraft mill effluent. Methanol helped to disperse the resin acid in the matrix and also supplied a carbon source typical of pulp mill effluent (McCubbin 1983; Springer 1986). Just prior to inoculation, 5 mL of 10% sterilized N H CI was 4 added to each flask and the pH was trimmed, i f necessary, dropwise, with 50% H SO or 2 N 2 4 43 Table 2-1. Weights of phosphate salts for a 100 mM pH buffered medium. pH NaH 2P0 4 (GMW= 137.99) Na 2HP0 4 (GMW= 141.96) g/500 mL g/500 mL 6 6.05 0.87 7 2.69 4.33 8 0.37 6.72 NaOH. In preliminary batch tests, a 5 mL aliquot of a sterilized 7.5% stock solution of sodium acetate was also added to the medium at this time to provide a representative readily biodegradable primary carbon source (McCubbin 1983). The carbon source combination of resin acid, methanol and sodium acetate in the kraft mill secondary treated effluent matrix formed a synthetic kraft mill effluent having a well defined BOD content for controlled batch experimentation on resin acids. A 1% inoculum was taken from mature cultures acclimated at the respective pHs. A l l cultures were initiated with a common inoculum of mixed liquor from the Western Pulp U N O X biobasin. Flasks were agitated on a reciprocal shaker table (40 rpm) in the dark within a 37°C thermostated room. Cultures were acclimated by two batch growth cycles prior to an experimental run. Resin Acid Sampling and Analysis During the time course of an experimental run, whole samples and filtered samples were taken for resin acid analysis. Filtered samples were used to compare the time course for depletion of the suspended versus the dissolved resin acid fraction. For filter samples, broth aliquots of 1 mL were drawn through 2.5 cm diameter, 0.45 um cellulose acetate membrane filters. The filters were rinsed with 2 mL physiological saline (8.5 g/L NaCl) and transferred to 10 mL glass tubes with Teflon caps. For whole broth samples, 1 mL aliquots drawn from the shake flasks were transferred to similar 10 mL glass tubes. To these extraction samples, 1 mL of 1 N NaOH spiked with O-methylpodocarpic acid (O-MPCA) was added immediately. The 1 N NaOH acted to arrest microbial activity while also solubilizing resin acid and microbial constituents alike. Alkaline storage was chosen to prevent time-dependent loss in extraction recovery (Kulovaara et al. 1987). The O-MPCA served as a sample extraction surrogate. Samples were stored frozen at minus 10°C pending extraction. 44 Prior to extraction, sealed sample tubes were brought to 90°C for 30 minutes to fully solubilize resin acids. Samples were extracted twice with 2 mL methyl-tert-butyl ether (MTBE) with intense vortex mixing followed by centrifugation (20 minutes at 2060 x g) for complete phase separation. With the addition of the first 2 mL of M T B E , 1 mL of 2 N H SO was used to acidify 2 4 the sample in order to reduce the aqueous solubility of the resin acids for improved extraction. For each extraction, the solvent was transferred directly to 2 mL GC vials and dried under vacuum. Resin acid recovery from B K M E by solvent extraction under acidic conditions has been shown to be relatively poor (Voss and Rapsomatiotis 1985). Voss and Rapsomatiotis report optimum extraction efficiency around pH 9. However, spiked sample recoveries have been shown to be optimal at low pH (Li et al. 1997). The method of alkaline saponification followed by acidic extraction was chosen to create the conditions of spiked sample extraction. Solvent emulsions could always be collapsed by centrifugation. Recoveries with respect to O-MPCA were typically greater than 90 percent. Gas chromatographic analysis of underivatized resin acids (Gref 1988) was found to be possible. However, consistent results could only be obtained by the formation of the resin acid methyl esters prior to chromatographic analysis. Methylation was accomplished by dispensing and vortexing, in the dried GC vials, 500 pL of chilled (0 to 4°C) M T B E and HPLC grade methanol (80:20) containing excess dissolved diazomethane, and spiked with heneicosanoic acid methyl ester (HCA-ME) and tricosanoic acid (TCA). Diazomethane gas was dissolved in the M T B E / methanol mixture within a closed vessel (Pierce 28131). The diazomethane gas was generated by reacting N-methyl-N-nitroso-N'-nitroguanidine with 5 N NaOH. Extracted resin acids and cellular fatty acids were identified and measured by GC/FID (HP5890 Series II with a DB5 column of 30 m having a 0.32 mm ID and 0.25 pm film thickness). Samples of 1 pL were injected at 280°C with an initial column temperature of 100°C. After a 5 minute holding time, the temperature was taken up to 200°C at 20°C/min and followed by a ramp of 1°C/ min to 230°C. A post-run ramp to 290°C at 20°C/min was used to clean out the column. A helium carrier gas was used with a column head pressure of 10 psi. Resin acids were quantified 45 using the response factor of 99% pure dehydroabietic acid (Helix) standards prepared from a 1000 U-g/mL stock solution in MTBE. Internal standards and extraction surrogates were quantified with their respective 99% pure standards as 1000 pig/mL stock solutions in M T B E With a 1 uX injection, the GC/FID detection limit for D H A was found to be approximately 0.10 p,g/mL. Protein Sampling and Analysis For this investigation, biomass was measured in terms of protein using the Coomassie Brilliant Blue G dye binding assay (Bradford 1976; Gogstad and Krutnes 1982; Read and Northcote 1981; Stoscheck 1990). The method was chosen for its simplicity and sensitivity. Time-varying inert suspended solids in the medium precluded using the more traditional optical density or gravimetric techniques for biomass. Filter samples were taken for protein biomass measurements as bovine serum albumin (BSA). Broth aliquots of 1 mL were drawn through 2.5 cm diameter, 0.45 um cellulose acetate membrane filters. The filters were rinsed with 2 mL physiological saline (8.5 g/L NaCl) and transferred to 10 mL glass tubes with Teflon caps and stored at minus 10°C. Protein was extracted from the filter-harvested cells by adding 2 mL 0.1 N filter-sterilized NaOH and incubating the filters in the sealed tubes at 100°C for 5 minutes. A Coomassie dye reagent was made by dissolving 100 mg of Sigma (B-0770) Brilliant Blue G in a mixture of 100 mL 85% phosphoric acid and 50 mL 95% ethanol. Once the dye was dissolved the final volume was brought up to 1 litre with deionized water. Colour change at a wave length of 595 nm was calibrated against Bovine Serum Albumin (Sigma) standards. To accommodate the expected range of protein in the samples, a number of sample and standard volumes (75, 100 and 150 u,L) were combined with a fixed reagent volume (150 pJL) producing a range of sensitivities in terms of colour change versus protein concentration. Adsorption was measured with Corning disposable sterile ELISA 96 well (300 pL) flat bottom polystyrene plates in a Molecular Devices Thermo Max Microplate reader. The ELISA plate wells were first referenced with just the 150 pL reagent aliquot. Replicate standards and sample volumes were added and the plate promptly 46 measured again for colour development. A l l calibration and sample readings for any one growth curve were made within one ELISA plate. The method detection limit was found to be approximately 0.5 mg/L protein as Bovine Serum Albumin. Methanol Sampling and Analysis For methanol, 1 mL samples were taken and stored frozen in 10 mL glass tubes with Teflon caps pending analysis. Methanol in the samples was quantified by GC/FID. Sealed sample tubes were vortexed and then incubated at 90°C for 10 minutes in order to reduce the potential for bioactivity during analysis. After centrifuging (2060 x g) for 30 minutes, a 200 pX sub-sample was diluted with a 200 uX aliquot of deionized water spiked with iso-propyl alcohol as an internal standard. Methanol in the samples was quantified against a dilution series from a 1000 ppm stock solution of H P L C grade methanol in distilled and deionized water prepared in the same manner. The gas chromatogram was generated on an HP 5890 Series II with a DB624 column (30 m with a 0.32 mm ID and 3.0 pm film thickness) and a column head pressure of 5.5 psi. One microlitre aqueous injections were made with injection and detection temperatures of 110 and 200°C respectively. The column temperature was held at 40°C for 3 minutes after injection and then ramped to 120°C at 20°C/min. Sodium Acetate Sampling and Analysis For the analysis of sodium acetate, 1 mL samples were drawn and immediately acidified with 200 U.L of 10 % phosphoric acid and then frozen at minus 10°C in 2 mL crimped-top G C autosampler vials pending analysis. Sodium acetate was measured by GC/FID (HP 5880A Series) as acetic acid following the method outlined in the Supelco GC Bulletin 751G (Supelco 1982). pH Sampling and Analysis To confirm that the medium buffering capacity was not exceeded, 1 mL samples were dispensed into standard glass disposable test tubes and measured directly with a calibrated Accumet pH electrode (No. 69488) with a Cole Palmer CP Chemcadet (5986-40) pH meter. 47 2.2.2 Dynamic Surface Tension Measurement Aliquots of 1 mL broth samples were drawn from the shake flasks and transferred to 10 mL glass tubes. To inhibit biological activity, 10 u i of 10% sodium azide were added. A l l glassware had been thoroughly washed and rinsed with deionized water and then fired at 475°C in a muffle furnace for over an hour to remove any organic material. Dynamic surface tension was measured directly after sample temperature equilibration to 20±1°C. After the surface tension measurement, the sample pH was taken. Dynamic surface tension measurements were made using the maximum bubble pressure (MBP) method. Since this technique was developed by Sugden (1922; 1924), it has been become a readily applied tool for experimental research in surface thermodynamics (Bendure 1971; Bing et al. 1988; Brown et al. 1952; Finch and Smith 1972; Hirt 1990; Hua and Rosen 1988; Iliev and Dushkin 1992; Kuffner 1961; Masutani and Stenstrom 1991; Posner et al. 1952). Mysels (1990) has written a relatively recent review of the history and experimental considerations for the method. With this method, gas is forced out of a capillary tube of radius, r, immersed at a depth, h (Figure 2-3). From Laplace's equation (2-1), the pressure, p, required to produce an assumed spherical bubble reaches a maximum (p ) when the emerging bubble becomes a hemisphere with the capillary's radius: CT, , (2-33) P M =P-Pa=2— + Pgh r where p is the atmospheric pressure, p is the liquid density (0.99821 g/cm3 at 20°C), g is the a acceleration due to gravity (980.98 cm/s2) and a is the dynamic surface tension as a function of t the bubble formation time, t. For a pure solvent, p is theoretically independent of the rate of M bubble production. However, i f surfactants are present in solution, a longer time for bubble formation permits a greater extent of surfactant adsorption at the gas/liquid interface. Increased surfactant concentration at the gas/liquid interface lowers the surface tension, as observed by a 48 a i r / l i q u i d i n t e r f a c e J b u b b l e f o r m a t i o n glass c a p i l l a r y Figure 2-3. The maximum bubble pressure method. Gas bubbles are blown through a glass capillary tube of radius, r, that is immersed to a depth, h, into a liquid sample. The pressure needed to generate the bubbles is monitored and interpreted. corresponding change in the measured p . M Equation (2-33) is not exact. The existence of a pressure gradient with liquid depth results in a formed half bubble that is not exactly hemispherical. The M B P surface tension is calculated precisely from the formula: r{pM ~ PSh)fc (2-34) where / is a correction factor for non-sphericity. Sugden (1922) tabulated the correction factor c as a function of the dimensionless ratio (r/a), where a is the capillary constant defined as: a = 2 ^ Pg (2-35) Bendure (1971) approximated the tabulated values by an empirical power series: / = a0+al(r/a) + a2(r/a) + a3(r/a) + ... (2-36) a0 = 0.99951 a2 =-0.69498 a 3 =-0.11133 a,= 0.01359 a4 = 0.56447 a5 =-0.20156 The shape correction calculation was refined by Johnson and Lane (1974). However, the correction factor differs from unity only in the third decimal place when the capillary radius is around 0.02 cm. 49 v//y//////////////////////A h Figure 2-4. Experimental set-up for the maximum bubble pressure method. Regulated flow (b) of distilled water from a constant head tank (a) pressurizes a closed reservoir (c). Transmission of the reservoir pressure to the rest of the system is damped by a fixed orifice (d). Pressure at the capillary tube (e) produces gas bubbles in the sample contained within a test tube (f). The test tube temperature is regulated through a thermostated jacketed beaker (g). Gage pressure at the capillary is measured by an electronic differential pressure transducer (j) whose signal is recorded (1). Capillary immersion depth is set by the stage height (h) whose level is referenced by a needle gauge micrometer (i). Bubbles were produced with the sample tube immersed in a thermostated bath at 20±1°C (Figure 2-4). Bubble formation proceeded at constant pressure supplied by an air-charged reservoir. The air reservoir was pressurized by displacement with deionized water. Bubbling frequency decreased as the reservoir pressure decreased with gas lost from the system in each bubble. Bubbling frequencies from about 10 Hz down to 0.04 Hz were monitored. Micropipet capillaries (10 uX) having a radius in the order of 0.02 cm were used. The effective capillary radius was assessed by measurement and calibration against the surface tension of deionized water (a=72.75 dyne/cm at 20°C). Although the capillary lumen radius could be measured with a microscope, in practice the anchoring radius for the bubble can be slightly outside the lumen. Since surface tension determination is sensitive to capillary radius, the radius was determined indirectly by calibration against the reference liquid. The radius calculated in this manner implicitly included the correction factor, / , which was approximated to be constant for all other measurements. Capillary immersion depth, h, was measured to the closest 0.025 mm by a needle gauge micrometer. The change in liquid level due to liquid displacement by the capillary or by bubble 50 7.6 7.4 7.2 7.0' 6.8 O.6.6. 1 6 A °" 6.2 6.0 5.8 5.6 1 1 1 ' 1 ' t - f 1 1 1 t ' • ii :p 1 4 7 r ; •: ; * .' - ip ii i if :T ? : i ? - m • ? • ; * i ! » : 80 85 90 time (seconds) 95 100 7.6- 7.4- 7.2- 7.0- 6.8- 6.6- 6.4- 6.2- 6.0- 5.8- 5.6- 87.25 f bubble release t new bubble formation 87.50 87.75 88.00 time (seconds) 3.25 Figure 2-5. A typical maximum bubble pressure signal (Left). This signal contains four bubble events corresponding to the four transient pressure dips. The plateau pressure decreases slightly with gas lost from the reservoir in each bubble. Bubble life or surface age (t) is defined from a point of minimum pressure to the next point of maximum pressure. Figure 2-6. Detail of the critical points in the pressure signal (Right). Surfactant adsorption at constant pressure (pM) terminates at the critical point of a hemispherical geometry. A bubble is explosively produced leading to a fresh gas-liquid interface that forms the next bubble. The dead time is the period between maximum pressure and bubble release. formation affects the MBP. The influence of this type of error was minimized by the use of a micro-capillary tube and was standardized by maintaining the same immersion depth of0.508 cm for all sample and reference measurements. Pressure values during bubble formation were logged at 50 Hz directly onto a computer (PC- LabCard Series PCL-812G 12 bit A/D driven by Labtech Notebook L E software on an IBM-PC clone with a 486 D L C 40 M H z CPU with math co-processor) through a calibrated differential pressure transducer (0-20 cm H O Celesco V R Series Differential Pressure Transducer with a 2 LCCD-110 Carrier Demodulator) producing a 1 volt/cm H O signal. 2 In real time, the digitized pressure signal was analyzed by applying Savitzky and Golay (1964) least squares smoothing and differentiation. Figure 2-5 illustrates a typical segment from a pressure-time trace showing the release of four bubbles. A slight decrease in the (plateau) reservoir pressure level can be seen with the release of each bubble. Figure 2-6 is a close-up of the pressure data during one bubble release. First and second time derivatives of the raw pressure trace helped to define the distinct calm and explosive phases of bubble production permitting the 51 discretization of the pressure-time signal into single bubble events. Identification of bubble events by pressure gradients enabled automatic triggering of data acquisition, minimizing the amount of information that needed to be stored to disk. Since a measurement sequence could take as long as 30 minutes, data files would have become unmanageably large had a triggering scheme not been used. Two data files were created. One stored the start and stop times of each bubble event and the other logged the time series of pressure values at 50 Hz during each event typified by Figure 2-6. For each bubble produced, the formation time (or bubble life) and the corresponding maximum pressure were ascertained by numerical data smoothing, cubic spline interpolation and extremum search algorithms (Hornbeck 1975; Press et al. 1989). Bubble life or surface age (t) is defined by the period from the point of bubble release of one event to the point of maximum pressure (p ) in the next (Figure 2-5). The bubble production dead time defined by the period of M explosive bubble release (Figure 2-6) was therefore accounted for. A sample measurement was defined by a recorded log of around 100 bubble events distributed over a range of surface ages. Bubble life represents the time available for surfactant adsorption at the bubble's gas/liquid interface. A longer bubble life allows more surfactant to be adsorbed at the bubble surface with a correlated lower p and surface tension. The incremental extent of adsorption for a longer bubble M life depends on the surfactant's bulk concentration, interfacial concentration, molecular diffusivity and chemical structure. In order to analyse the compiled adsorption data as a function of concentration for a constant surface age, the derived experimental o data, obtained over a range t of random t values for each sample measurement, were interpolated by a cubic spline on a log time scale. In this manner, all sample measurements were aligned to a set of common times or surface ages. 2.2.3 Nonlinear Regression Analysis Nonlinear regression was required to estimate the parameters of the equation (2-32) that best fit the experimental substrate depletion curve. Parameter values {S , S*, Q, k , *F} were O S estimated by non-linear least squares regression by Taylor series linearization (Draper and Smith 1966) modified by a logarithmic data transformation for variance stabilisation. Since the substrate 52 depletion curve involved experimental concentrations over four orders of magnitude, a lack of variance constancy would unduly influence the numerical results (Berthouex and Brown 1994). There are two kinds of random errors that combine to affect experimental data. Background noise is omnipresent and is assumed to have a constant variance. Measurement error is assumed to be proportional to the level of signal. Least squares regression on the untransformed concentration data would be biased by larger absolute experimental errors for the data at higher concentration. This bias resulted in a poor estimate of the threshold concentration. In order to give all data in the depletion curve equal weight, log-transformed concentrations were used. The objective function was the sum of square errors, SS(0), between the natural logarithm of the measured and estimated concentration: SS(®) = ^(uk-Uk)2 (2-37) where 0 is the vector of p parameters (the five, {S , S\ Q, k , x ¥} , in this case), N is the number o s of observations of the transformed experimental concentrations (u ) and U is the natural logarithm of the integral of equation (2-32): Uk=U(rlk,0) = Mg(rlk,Q)) = \n uk=\n(sk) r k s -K  s + s s — 1+— (2-38) Given the j t h estimate of the parameter set, 0 j , the Taylor series method predicts the next best estimate, 0 j + 1 in minimising the objective function as follows: e>+i = ej + ( G T G ) ~ 1 (GTz) (2"39) where Z is the discrepancy vector and G is the matrix of partial derivatives with respect to the parameters: 53 z = ux-U(rivQj) uN-U(j]N,&) G = 30, 30, °-U(Tl'N,®J) 30 „ S* 0 ' = Q K y (2-40) Since the partial derivatives in the matrix G could not be determined analytically, they were numerically calculated by a second order finite difference approximation (Hornbeck 1975): 3*7(^,0') 1 -3g (77„0O + 4g(r7 i ,0>+A0,.)-g(7 7 „0 y +2A0,.) (2-41) 30,. g(Vk,®J) 2A0,. The numerical solution is doubly iterative. It is iterative in the first instance due to the successive integration of equation (2-32) for calculated values of U as a function of the parameter set. The k second level of iteration is in the search for numerical convergence to a minimum objective function. Numerical integration was accomplished with a fourth order Runge-Kutta (Hornbeck 1975) algorithm with self-limiting step-size halving optimisation. A n approximate 100(1-a) percent confidence contour for the best parameter estimate <0> is defined by (Draper and Smith 1966): SS(6) = SS((e)) f 1 + - -F(p,N-p,\-a) (2-42) N-p, where F(p, N - p , 1-cc) is the 1 - a point of the F(p, N - p ) distribution. The precision of the estimated parameters was considered by generating the approximate 1-a joint confidence regions for logical pairs of the parameters. This was done by calculating SS(0) in the neighbourhood of the minimum by varying the parameters in pairs, such as {Q, k} or {S , S*}, while holding all s 0 other parameters constant. From the objective function mapping as a function of the parameter pair, a contour plot was then generated with the help of commercial software (Surfer Version 6.01, Golden Software Incorporated). 54 2.3 Results The experimental results have been organized in two sections. In the first section, the experimental results and considerations for assessing the mixed culture batch growth data are presented. Diauxic growth was observed for the medium containing the combined carbon sources of resin acids and sodium acetate. Due to the experienced inconsistency of resin acid biodegradation in the presence of sodium acetate, the additions of sodium acetate were abandoned. Unfortunately, biomass measurements using the protein assay were also unsuccessful due to matrix interferences. However, the substrate depletion data could still be modelled by Monod kinetics with the incorporation of a threshold concentration. The threshold concentration appeared to be greatest under acidic growth conditions. The results of the dynamic surface tension measurements are subsequently reported. The surface tension data were found to be well represented by the model of quasi-static Langmuir adsorption. Differences in pH greatly influenced the flux and adsorption of resin acids at a gas/liquid interface. It was interesting to observe a lower bound or threshold concentration for resin acid flux in solution. The increase in this flux threshold corresponded qualitatively with the observed threshold concentration for biodegradation. 2.3.1 Mixed Culture Batch Growth Initial batch experiments with the three carbon sources (sodium acetate, methanol, and resin acids) in B K M E effluent gave unpredictable results. While acetate was always consumed over a twenty-four hour period, the resin acids were often left essentially untouched. Under these mixed substrate conditions, resin acids were depleted most consistently at pH 6. After the first few experiments, sodium acetate was no longer added to the medium, due to the uncertainty that it introduced. The remainder of the results describe batch growth studies with the Squamish effluent media containing only resin acids and associated methanol as carbon sources. It should be noted, however, that methanol was also not consistently metabolized. Figure 2-7 shows total and suspended resin acid depletion data for pH 6, 7 and 8. Regression 55 Table 2-2. Parameter values for the Monod model with a threshold substrate concentration (equation (2-32)) used to model the substrate depletion curves presented in Figure 2-7. Total Resin Acid Depletion pH So uM S* uM k s UM Q Hr 1 l / ¥ 6 7 8 171 179 180 2.7 0.8 0.7 292 66 13 3.0 1.0 0.2 0.00003 0.00007 0.03226 Suspended Resin Acid Depletion pH So UM S* uM k s uM Q Hr 1 6 7 8 94 32 5 2.3 0.7 0.5 155 8 0 3.0 1.0 0.2 0.00005 0.00004 0.10489 curves were generated according to the best fit parameters obtained for the model of Monod growth with a substrate threshold (equation (2-32)). The optimal parameter values used to fit the data are reported in Table 2-2. Unfortunately, biomass measurements as protein (BSA), that were quite precise during method trials, were found to suffer from an interference that increased with decreasing pH. Without a reliable measurement for biomass, neither the batch growth lag time, or the biomass concentrations, could be reliably estimated. Therefore, the constants *P, Q and k s served solely as empirical curve fit parameters. However, from a visual inspection of the substrate data, the rate of substrate removal was influenced by pH and this difference was reflected by differences in Q and k . s A threshold resin acid concentration (S*) was also observed (Table 2-2). Error in the estimation of the threshold concentration based on the approximate 95% joint confidence interval for (S , o S*} was found to be no more than 15 percent of the absolute value. The residuum can be seen to be essentially suspended and was highest (2.7 U.M) for the pH 6 condition (Figure 2-7). A resin acid concentration of 2.7 pJVI translates to approximately 0.8 mg/L which is still a level that could be acutely lethal towards fish. Therefore, in this case, biological treatment alone did not appear to be sufficient for detoxification. 56 Figure 2-7. Batch growth resin acid depletion curves for pH 6,7 and 8 from top to bottom. The data show the time courses of removal for both total (•) and particulate (•) resin acid modelled by best fit curves from equation (2- 32). See Table 2-2 for model parameter values. 57 2.3.2 Dynamic Surface Tension Measurements Results of the corresponding dynamic surface tension measurements made during batch growth are shown in Figure 2-8. The data are presented for a number of surface ages, forming a family of quasi-static Langmuir isotherms. The group of isotherms was fitted simultaneously to equation (2-25) by generalized reduced gradient non-linear optimisation (Microsoft Excel 97 SR- 1 Solver) with the constraint of a common T . The resultant limiting surface loading, T , is noted L L on the graph for each pH. Kraft mill effluent contains a significant amount of amorphous high molecular weight poly-electrolytes which are the recalcitrant breakdown products of lignin. These compounds impart the characteristic colour to the effluent. Because of their high molecular weight, kraft lignins can also be surface active. Therefore it was not surprising to find a changing reference surface tension, a (equation (2-25)) corresponding to a background matrix o surfactancy, varying both with pH and surface age. Inherent surface aging that was believed to be due to kraft lignin adsorption is shown in Figure 2-9 in terms of the derived matrix surface pressure change with age. The trend in the isotherm data, shown in Figure 2-8 for the condition of pH 7, suggested the presence of a critical micelle resin acid concentration in the neighbourhood of 100 pM. In this case, isotherm modelling was restricted to concentrations below 100 pM. The parameter K was e estimated by fitting respective K values to an exponential decay curve on a logarithmic time scale t (Figure 2-10). The extrapolated value, K is a minimum representing the point of maximum e entropy or minimum Gibbs free energy. From the values of K and T the surface loading T was calculated as a function of solute t L t concentration (equation (2-20)). Based on the extrapolated equilibrium value, K , the normalized e surface loading (y) with time as a function of solute concentration was calculated from: t = T L = c k ± K ± (2-43) r e Cb+K, The y data at constant pH were empirically found to be well represented by an equation of form: t 58 50 • 0.1 10 100 1000 Resin Acids (u,M) Figure 2-8. Dynamic surface tension during biological removal of resin acids for pH 6,7 and 8 from top to bottom. Experimental data at the indicated surface ages is shown along with the family of quasi-static Langmuir isotherms (equation (2-25)) that were fitted to the data. The isotherms generated for the data at pH 7 suggest the presence of a critical micelle concentration in the neighbourhood of 100 uM. 59 3.0 0 5 10 15 20 25 30 35 Surface Age t (seconds) Figure 2-9. Derived surface pressure changes with surface age in the absence of resin acids that were believed to be due to the presence of kraft lignin in the media. 700 600 500 400 100 300 -4 200 Surface Age t (seconds) Figure 2-10. The quasi-static Langmuir concentration K (equation (2-20)) as a function of surface age and pH. Equilibrium values (K) for an infinite surface age were estimated from regression analysis to an exponential decay function. e 60 y - P-' B = ^ ( 2 " 4 4 ) Equation (2-44) describes a sigmoid on a logarithmic time scale, where q is the rate constant for a adsorption and i is the lag time to reach a normalized surface loading of 50 percent. a The rate constant (q) was found to be independent of resin acid concentration. The lag time (T ) a a taken to reach 50 percent surface loading increased with decreasing solute concentration (Figure 2-11). Equation (2-44) is analogous to equation (2-9) used by Hua and Rosen (1988), except in this case adsorption is expressed directly in terms of surface loading. Figure 2-11 shows that both the rate of surface loading (q) and the lag time (x) increases with decreasing pH. a a From equation (2-44), the rate change in surface loading can be calculated explicitly in time and this expression can also be equated to the general surface loading rate equation (2-6): dr_rdy_ rja (2-45) dt  e dt (l+pj)2 If the sticking coefficient (O) is assumed to be initially equal to one, then the initial solute flux should be related to the solute mobility B times some function of solute concentration (equation (2-6)): Jo=TePa=Bf(cb) (2"46) The initial solute flux from the adsorption data at pH 8 was linearly dependent on resin acid concentration (Figure 2-12). However, the data at pH 6 and 7 were both better represented by a linear dependence on the logarithm of resin acid concentration (Figure 2-13). In Figure 2-13 it can be seen that the adsorption data for pH 6 and 7 both displayed a knee or threshold concentration below which solute flux became independent of resin acid concentration. The initial solute flux at pH 8 is an order of magnitude higher than the fluxes at pH 6 and 7. The flux at pH 7 is greater than at pH 6. If the surface flux J is assumed to be constant and equal to the initial flux of solute to the interface 61 1.0 • l-i 0.5 • II Resin Acid Concentration 93 / 42 .* 5 , ' 0 . 6 | i M / > I'/ 'jsr o.o • 1.0 < I- 0.5 • II * * r .83 ' ? — ; — Resin Acid Concentration 169 .'135 81 XT' 5JIM r a o.o • 10 0.1 100 Surface Age t (seconds) Figure 2-11. Normalized surface loading (y) as a function of surface age (t) and resin acid concentration. Experimental data are shown for pH 6, 7 antl 8 (top to bottom) along with best fit model curves (equation (2-44)) obtained by generalized reduced gradient nonlinear optimisation (Microsoft Excel SR-1 Solver) for the whole data set at each pH condition respectively. With reference to equation (2-44) the q, values are 3.07, 1.72 and 1.18 for pH 6, 7 and 8. The xa values can be read from the curves at a surface loading y, = 1/2. Not all data have been shown to allow for better visual clarity. 62 u Vi 0.8 0.7 0.6 o 0.5 S a s 13 0.4 0.3 0.2 0.1 0.0 A , ^ J H 8 | K 20 40 60 80 100 120 140 Resin Acids (|j,M) 160 180 200 Figure 2-12. The derived initial flux of resin acid to a freshly generated gas/liquid interface at pH 8. The initial flux is directly proportional to resin acid concentration. 0.035 0.030 -I 0.025 u Vi S 0.020 0.015 4 — 0.010 JS 0.005 0.000 0.1 cb* = 2.7 jiM ^ cb* = 11.9 jiM 10 100 1000 Resin Acids (ji,M) Figure 2-13. The derived initial flux of resin acid to a freshly generated gas/liquid interface at pH 6 and 7. The initial flux is directly proportional to the logarithm of resin acid concentration up to a threshold resin acid concentration, cb*. 63 then from equation (2-45) and (2-46): 0 + / V ) 2 By manipulating equation (2-45), it can be shown that the assumption of a constant surface flux leads to a time-implicit expression of the sticking factor in terms of the squared compliment of the normalized surface loading: (l + A,<)2 Accordingly, a time implicit expression for the sticking factor, <1>, was sought in the form of: * = *(d-rj) <2-49> The results plotted in Figure 2-14 however, show that equation (2-48) is not strictly valid, meaning that the surface flux is not constant but is also dependent on surface loading. However, it is still possible to write a time-implicit relationship for the rate change of surface loading in terms of the initial surface flux as follows: £ = 0 = |>, .(l-7,) 2 ' , 5><̂  ( 2"50) dr ,=i ,=i The crudest approximation of equation (2-50), which works reasonably well at pH 8 but progressively poorer at pH 7 and 6, is a linear representation as shown in Figure 2-14: f-•O-r.JV., < 2 - 5 1 ) The sticking probability is highest for pH 8 and decreases with increasing surface coverage. The experimental data (Figure 2-12 and Figure 2-13) suggest that, depending on the pH condition, the initial flux J is related to concentration by an equation of form: o 64 j \jQ+H{cb-cb\^ncb-\ncb) pH <7 (52) ° 1 Bcb pH~8 where H(c b - cb*) is Heaviside's unit step function defining the threshold concentration for the adsorption dependency on resin acid concentration and j 0 is a background adsorption flux. 65 2.4 Discussion For the present investigation, batch removal of resin acids was found to be greater than 98%, however the residuum or threshold concentration was still well within the resin acid concentration range reported for sublethal effects to aquatic organisms (McLeay 1987). In the research literature on the biological removal of resin acids, to the author's knowledge, no mention has previously been made of such threshold concentrations. Thus, microbial oxidation can achieve significant depletion of a contaminant, but final absolute concentrations cannot be neglected in the consideration of toxicity removal. The residuum was greatest at pH 6 indicating the potential of reduced bioavailability of resin acids under acidic conditions. This result corresponds to the fact that resin acid extractability has been observed to diminish with time after storage under slightly acidic conditions (Kulovaara et al. 1987). A good fit to the experimental substrate depletion data was obtained with a Monod model that was modified to include a threshold concentration (equation (2-32)). Thus, conventional Monod kinetics (equation (2-28)), that are commonly used to design wastewater treatment systems, may not give the best representation for toxicity removal. While a compound may be readily biodegradable, a residuum concentration, that is not predicted by traditional Monod kinetics, can persist. Even i f this residuum is below the level of acute toxicity, its discharge can still pose an environmental threat i f the contaminant is known to bioaccumulate. Resin acids have been reported to concentrate in tissues of fish (Niimi and Lee 1992). Therefore, although biological treatment may provide an economic means to remove the bulk of a toxic contaminant from a wastewater stream, the presence of a threshold concentration after biological treatment may necessitate subsequent effluent polishing. Literature reports of high removal efficiencies for a particular contaminant can be misleading. Typically, literature data are obtained for the removal of a given contaminant, under steady state conditions, for a particular hydraulic retention time. Such work is incomplete since it cannot be known i f the observed effluent concentration is either an inherent threshold due to a restriction in bioavailability, or whether it is a function of the removal kinetics. The wastewater treatment design strategy to contend with limitations in 66 contaminant bioavailability would need to be quite different. For instance it might be necessary to introduce surfactants to help solubilize the hydrophobic contaminants. For ionogenic contaminants like resin acids, pH control appears to be an important factor for controlling solubility and associated bioavailability. While these results support the findings of Liu (1993) for resin acid biodegradation under either acidic or alkaline media conditions, the findings of Hemingway and Greaves' (1973) were also reproduced in the sense of a reduced extent of removal below pH 7. The observed preferential microbial growth on sodium acetate when this carbon source was included in the medium further suggests that diauxic growth was indeed a factor in the results presented by Hemingway and Greaves. The collected dynamic surface tension data were modelled well by a family of quasi-static adsorption isotherms. Similar to a previous investigation of resin acid adsorption (Liu et al. 1996a), a Langmuir isotherm provided a good representation of the data. The method of data reduction that was developed for the present study has a far greater scope of application, since the theory is independent of the isotherm chosen. For example, solute interaction at the surface can be accounted for by Temkin adsorption (Kohler 1993). Solute interaction and molar surface area would be incorporated by a Van der Waals adsorption model. However, a practical complexity may arise with the use of more comprehensive isotherm models due to the increased number of parameters that would need to be determined. The observed qualitative differences in the isotherms at different pH conditions reflect significant changes to the properties of resin acids in the B K M E . These changes do not occur in isolation, but are probably influenced by the presence of other dissolved organic compounds in the effluent. Lignin is the second most abundant component of wood, representing 25-30 percent of its weight. The ubiquitous high molecular weight kraft lignin released during wood fibre pulping is a poly- electrolyte with surfactancy that was also found to diminish with pH. Marton (1964) has shown that, like resin acids, the solubility of kraft lignin decreases with decreasing pH. The increased colloidal nature of resin acids at lower pH is evident from the higher suspended fraction observed. 67 Therefore, as pH decreases, both resin acids and kraft lignin exhibit reduced solubility. As a result, is likely that, under acidic conditions, resin acids will form associations with other resin acids in solution, and also with the kraft lignin. Decreased diffusivity with increased effective molecular weight (size) was reflected by the lag time for adsorption that increased with decreasing pH. Surface tension lowering, under acidic conditions, indicated that resin acids, complexed with other organic matter in solution (kraft lignin), were still surface active. However, the larger effective size for the hydrophobic complex, should mean a less efficient packing arrangement at the interface. This notion was confirmed by the fact that the limiting resin acid loading (T ) at pH 6 was roughly half that at pH 8. The limiting surface loading (T ) was L L observed to be 0.13 nmoles/cm2 at pH 6, versus 0.24 nmoles/cm2 at pH 8. The solubility of kraft lignin is influenced by ionic strength, however, up to a 0.10 molar salt concentration, the change in solubility is most pronounced below pH 7 (Marton 1964). Therefore, association between kraft lignin and resin acids in solution is likely to become more significant only below pH 7. The adsorption data at pH 7 suggest that resin acid self-association dominated the resin acid interactions at this pH. Tendency towards self-association at pH 7 was observed in the first instance by the presence of a critical micelle concentration. The critical micelle concentration (CMC) of potassium dehydroabietate is approximately 2.5-3.0 x 10 2 molar (Corrin et al. 1946; Kolthoff and Stricks 1948). A mixture of dehydroabietic acid with dehydroabietate should have a lower C M C . Further, the lag time for adsorption (x in Figure 2-11) was longer than at pH 8, indicating a a higher effective molecular weight of the adsorbing solute. In contrast to the limiting surface loading at pH 6, T at pH 7 was higher than at pH 8. The derived value for T at pH 7 was 0.30 L L nmoles/cm2, versus 0.24 nmoles/cm2 at pH 8. Therefore, the adsorbed complex at pH 7 likely consisted of an arrangement of the more hydrophobic protonated resin acids sequestered by the more soluble dissociated resin acids. Changes in the limiting surface loading with pH therefore indicate the presence of a maximum in the absorbable amount of resin acid, depending on the extent of molecular association induced by changing solubility with pH. Hence, a wastewater 68 treatment unit process, that makes use of contaminant adsorption, must consider both the potential extent for adsorption (T ) in addition to the kinetics of adsorption ((3 in equation (2- L a 44)). The potential for micelle formation is another practical consideration. Thus, although the potential surface loading at pH 7 was greater than at pH 8, these gains may be offset by the observed increased lag time and also micelle formation above a 100 micromolar concentration. As part of the CPAR program, Ng (1977; 1974b) investigated the feasibility of foam separation as a stand-alone physical unit process to detoxify bleached kraft mill effluent. Foam separation removed up to 65 percent of the resin acid content of the effluent and a pH greater than 7 was necessary for consistent detoxification. Branion (1992) more recently made preliminary tests into foam fractionation for thermomechanical pulping wastewater. Similarly to Ng et al. 's results on B K M E , no significant removal was noted at pH 5, while a 60 percent reduction occurred at pH 9. The surface tension data and accompanying adsorption theory of the present investigation have provided the model framework to predict and explain the empirical observations of these previous investigations. Ng et al. experimentally tested a number of parameters in order to optimize their unit process. Such experimental investigations of trial and error may not be the most efficient for determining optimal design parameters. From the results of this experiment it should be apparent that a number of systematic dynamic surface tension measurements can yield more fundamental information about the adsorption process. Armed with kinetic parameters for adsorption, a sensitivity analysis for the factors that most influence the adsorption unit process can be readily made numerically by computer simulation (Clarke and Wilson 1983). Once the important process parameters have been identified, an engineer would have better guidelines with which to base the process prototype design. That limiting bioavailability is a factor behind the observed increased Monod threshold concentration (S*) at pH 6 is supported by the corresponding adsorption flux threshold concentration (c *) that increased from pH 7 to pH 6 (Figure 2-13). The observed flux threshold, b c *, indicated a limiting concentration at which the resin acids could no longer be considered as b free solutes in solution. Below c *, what adsorbed was likely an amorphous complex of high b 69 molecular weight material into which some resin acid was dissolved. Adsorption of the kraft lignin was expected since high molecular weight organochlorines in pulp mill effluent have been shown to adsorb to biomass (Yan and Allen 1994). The overall drop in solute flux with pH, measured by the adsorption lag time (T in Figure 2-11) can be related to an increased effective a size of the complexed solute molecule in solution. In other words, the observed decrease in resin acid diffusivity with pH is explained by an effective increase in molecular size of the adsorbing solute. The change in the functionality (equation (2-52)) relating solute flux to the concentration also reflects the change in the nature of resin acids in solution over the pH range tested. Therefore, much insight has been gained by studying the surfactancy of resin acids during their biological oxidation. Since compounds like resin acids are believed to be degraded intracellularly (Sikkema et al. 1995), the transport and adsorption to biomass are important steps in their removal. A fundamental understanding of the conditions that can strongly influence the delivery of a contaminant to a microorganism can only assist in identifying causes of limitations and routes for optimisation in treatment processes. Limitations in contaminant transport become increasingly relevant when trace residual concentrations are sufficient to cause acute or chronic toxicity. The approach taken in the interpretation of the dynamic surface tension measurements provided model time-implicit relationships and parameter values that have engineering application in computer simulation for the design of new or optimized treatment processes. The derived rate expression for surface loading (equation (2-51)) was not the same as the equation (2-8) assumed by Clarke and Wilson (1983). Therefore, modelling the fate of a surfactant solute requires parameters and relationships that should first be determined experimentally. While the attention of this study has been directed towards resin acids in pulp mill effluent, it should be noted that toxic organic contaminants are, in general, weakly soluble and hydrophobic. Thus the technique used for this investigation should be equally informative for a similar study on pentachlorophenol for example. In addition to the application of dynamic surface tension measurements to expose the nature of hydrophobic contaminants in wastewater, the measurement technique also has application as rapid method for effluent monitoring. Resin acids are one of a 70 plethora of hydrophobic contaminants in pulp mill effluent (Sunito et al. 1988). X A D resin adsorption experiments (Dixon et al. 1992) have demonstrated that 70-75 percent of the total organic carbon content from whole mill effluent is composed of hydrophobic molecules. A reduction in effluent foaming tendency, which is related to the concentration of these hydrophobic compounds, has also been shown to parallel BOD removal (Carpenter and Gellman 1966). Therefore, BOD and toxicity removal from pulp mill effluent by biological treatment can, in principle, be calibrated to dynamic surface tension measurements. However, the factors affecting the survival of fish exposed to pulp and paper mill effluent also include ammonia, carbon dioxide, chlorine, dissolved oxygen, dissolved solids, heavy metals, pH, sulphur compounds, suspended solids and temperature (Brouzes 1976). Thus dynamic surface tension monitoring would only test for the onset of one probable cause for toxicity breakthrough. The acute toxicity of B K M E varies widely between mills and between effluents from the same mill discharged at different times (Walden and Howard 1981). Toxicity variation can occur over periods as short as 15 minutes. At the other extreme, fluctuations in bleachery effluent toxicity occur on the time scale of one or two hours. Further, some pulp and paper mills are equipped to run more than one kind of pulping process simultaneously. Dixon (1992), characterising the molecular weight distribution of contaminants in mill effluents, found from successive samples that, while colour and total organic carbon remained similar, the molecular weight distributions shifted. These shifts represent changes in the organic components, which may in turn influence biological treatability. It may be possible to relate shifts in the organic content of pulp mill wastewater prior to biological treatment to trends in dynamic surface tension. While inherent variability seems to be a natural aspect of pulp mill effluent, limitations in the frequency and extent of treatment plant monitoring mean that it is often difficult to retrace the causes of toxic breakthrough after the fact. Occurrences of toxic discharges are not highly publicized and data are not readily available. Toxicity monitoring is typically performed by weekly L C bioassays on Daphnia magna and monthly L C bioassays on fish. Knowledge of a 50 50 toxicity breakthrough event may therefore lag days or weeks behind its occurrence. Periodic 71 events would go unnoticed. Therefore, a reliable technique that can flag the onset of an upset has significant practical application for environmental protection. Dynamic surface tension measurements made as a regular process control measurement could be used to signal the alarm of a potential toxicity event due to compounds like resin acids i f a trend of increasing surfactancy is monitored. The technique is amenable to automation. 72 2.5 Conclusions The pH-dependent resin acid flux to a gas/liquid interface suggests that acidic conditions promote the association of resin acids with other dissolved organic matter. Under alkaline conditions, resin acids are relatively mobile surfactants in solution. Under acidic conditions, a significant threshold concentration exists, at which the flux of resin acids to a gas/liquid interface becomes independent of the resin acid concentration. In parallel to the surface tension measurements, after batch growth treatment an elevated threshold concentration for resin acid removal was observed under acidic growth conditions. This threshold or residuum was modelled as the level at which microbial growth could no longer be supported. The residuum was observed to be present primarily in the suspended resin acid fraction of the medium. Thus, although resin acids were shown to be readily removed under acidic or alkaline conditions, the observed residuum in conjuction with the solubility and surface tension data indicate a pH-dependence for the contaminant bioavailability. Resin acids become less mobile under acidic conditions, resulting in a reduced potential extent of biological removal. It is important to note that the removal of resin acids during the reported batch treatments was assumed to be due to biological activity. Although biomass changes were monitored, the resultant microbial protein data were unreliable. Therefore, for the subsequent investigations, a better monitor for biomass was required. Therefore, with respect to objective A in Chapter 1, the data from this preliminary investigation indicate that pH-dependent resin acid solubility significantly influences the contaminant bioavailability under acidic pH conditions. Differences in the growth rates were observed. However, without reliable biomass data, the kinetic coefficient values could not be interpreted with confidence. Therefore, a second batch treatment investigation was required to reaffirm the present conclusions (Objective A - Chapter 1) and to better consider any pH-dependence for the microbial growth kinetics (Objective B - Chapter 1). The second batch treatment investigation is reported in Chapter 3, Microbial Community Structure in the Kinetics of Resin Acid Removal. 73 Dynamic surface tension measurements made by the maximum bubble pressure method during batch microbial growth were instrumental in interpreting the interactions of resin acids in solution. The quasi-static adsorption theory that was developed for the purpose of this investigation, facilitated the formulation of mathematical models that were useful for explaining both the current results and the related literature. The development of such adsorption models has application in computer simulation for process design. Further, the results suggest that dynamic surface tension measurements could be utilized as an informative treatment process monitor. Such a process monitor would serve as an early warning indicator of a mill or bioreactor upset that would have a high potential of leading to an event of toxicity breakthrough. 74 2.6 References Adamson, A. W. (1982). Physical Chemistry of Surfaces, John Wiley & Sons. Alexander, M . (1994). Biodegradation and bioremediation, Academic Press. Bailey, J. E., and Ollis, D. F. (1986). Biochemical Engineering Fundamentals, McGraw-Hill. Bendure, R. L. (1971). Dynamic surface tension determination with the maximum bubble pressure method. Journal of Colloid and Interface Science, 35(2), 238-248. Bendure, R. L. (1975). Introduction to foams and their physical-chemical properties. Tappi Journal, 58(2), 83-87. Berk, D., Zajic, J. E., and Behie, L. A. (1979). Foam Fractionation of Spent Sulphite Liquor Part II: Separation of Toxic Components. The Canadian Journal of Chemical Engineering, 57, 327-332. Berthouex, P. M. , and Brown, L. C. (1994). Statistics for Environmental Engineers, Lewis Publishers. Bicho, P. (1995). Personal communication., Forest Products Biotechnology, University of British Columbia, Vancouver, BC. Bing, Y., Xiu, D., and Yajun, S. (1988). Foam fractionation technique II. Design method about foam fractionation column. Journal of East China Institute of Chemical Technology, 14(5), 579-587. Branion, R. (1992). Treatment of thermomechanical pulping wastewaters. #13(RC 18), #13(RC 20), STDF-AGAR 13(SA-2), Science Council of British Columbia Research Grants, Vancouver. Brasch, D. J. (1974). The foam fractionation of spent kraft pulping liquor. Appita, 28(1), 29-32. Brouzes, R. J. P. (1976). Fish toxicity with specific reference to the pulp and paper industry. EPS 3-WP- 76-4, Environment Canada. Brown, A. S., Robinson, R. U., Sirois, E. H., Thibault, H. G., McNeill, W, and Tofias, A. (1952). Critical micelle concentrations by a bubble pressure method. Journal of Chemical Physics, 56, 701-705. Brownlee, B., Fox, M . E., Strachan, M . J., and Joshi, S. R. (1977). Distribution of dehydroabietic acid in sediments adjacent to a kraft pulp and paper mill. Journal of the Fisheries Research Board of Canada, 34, 838-843. Bruun, H. (1952a). An interpretation of the effect of ionization upon monolayer properties of rosin acids. Acta Chemica Scandinavica, 6, 955-957. Bruun, H. (1952b). Properties of monolayers of rosin acids. Acta Chemica Scandinavica, 6, 494-501. Burcik, E. J. (1950). The rate of surface tension lowering and its role in foaming. Journal of Colloid Science, 5, 421-436. Burcik, E. J. (1953). Effect of electrolytes on the rate of surface tension lowering; the rate of surface equilibrium attainment as a factor in detergency. Journal of Colloid Science, 8, 520-528. Callen, H. B. (1985). Thermodynamics and an introduction to thermostatics, John Wiley & Sons. Carlberg, G. E., and Stuthridge, T. R. (1996). Environmental fate and distribution of substances. Environmental Fate and Effects of Pulp and Paper Mil l Effluents, M . R. Servos, K. R. Munkittrick, J. H. Carey, and G. J. van der Kraak, eds., St. Lucie Press, 169-178. Carpenter, W. L., and Gellman, I. Measurement, control and changes in foaming characteristics of pulping wastes during biological treatment. Industrial Waste Conference, Purdue, 203-213. Chandrasekaran, K., Reis, R., Tanner, G., and Rogers, H. (1978). Removing toxicity in an aerated stabilization basin. Pulp & Paper Canada, 79(10), T304-309. Clarke, A. N., and Wilson, D. J. (1983). Foam flotation: theory and applications, Marcel Dekker. Clesceri, L. S., Geenberg, A. E., and Trussell, R. R. (1989). Standard Methods for the Examination of Water and Wastewater., M . A. H. Franson, ed., APHA, AWWA, WPCF. Corrin, M . L., Klevens, H. B., and Harkins, W. D. (1946). The determination of critical concentrations for the formation of soap micelles by the spectral behavior of pinacyanol chloride. Journal of Chemical Physics, 14(8), 480-486. Dang, J. S., Harvey, D. M. , Jobbagy, A., and Leslie Grady Jr., C. P. (1989). Evaluation of biodegradation 75 kinetics with respirometric data. Journal of the Water Pollution Control Federation, 61(11/12), 1711- 1721. Davies, J. T., and Rideal, E. K. (1963a). Chapter 5 - Properties of monolayers. Interfacial Phenomena, Academic Press, 217-281. Davies, J. T., and Rideal, E. K. (1963b). Chapter 8 - Disperse Systems and Adhesion. Interfacial Phenomena, Academic Press, New York and London, 343-451. Dixon, D. R., Wood, F. J., and Beckett, R. (1992). Characterization of organics in pulp and paper mill effluents before and after physicochemical treatment. Environmental Technology, 13, 111 7-1127. Draper, N . R., and Smith, H. (1966). Applied Regression Analysis, John Wiley & Sons. Drobosyuk, V. M. , Borodulina, M . R., Krylatov, Y. A., and Talmud, S. L. (1982). The dispersive power of resin acids and fatty acids. Journal of Applied Chemistry of the USSR, 55(9), 2105-2107. Duncan, B. (1993). Foam samples collected from the Fraser River., BC Environment. Ertl, G. (1983). Kinetics of chemical processes on well-defined surfaces. Catalysis Science and Technology, J. R. Anderson and M. Boudart, eds., Springer-Verlag, 209-282. Fein, J., Beavan, M. , Effio, A., Gray, N. , Moubayed, N. , and Cline, P. (1992). Comprehensive study of a bleached kraft mill aerobic secondary treatment system. Water Pollution Research Journal of Canada, 27(3), 575-599. Finch, J. A., and Smith, G. W. (1972). Dynamic surface tension of alkaline dodecylamine acetate solutions in oxide flotation. Mining andMetallurgy(\2), C213-218. Gao, T., and Rosen, M . J. (1995). Dynamic surface tension of aqueous surfactant solutions 7. Physical significance of dynamic parameters and the induction period. Journal of Colloid and Interface Science, 172, 242-248. Gref, R. (1988). Gas chromatographic analysis of underivatized resin acids. Journal of Chromatography, 448, 428-432. Halbrook, N . J., and Lawrence, R. V. (1966). The isolation of dehydroabietic acid from disproportionated rosin. J. Org. Chem., 31, 4246-4247. Hansen, R. S. (1960). The theory of diffusion controlled absorption kinetics with accompanying evaporation. Journal of Chemical Physics, 64, 637-641. Hassett, J. P., and Anderson, M . A. (1979). Association of hydrophobic organic compounds with dissolved organic matter in aquatic systems. Environmental Science and Technology, 13(12), 1526-1529. Hemingway, R. W, and Greaves, H. (1973). Biodegradation of resin acid sodium salts. Tappi Journal, 56(12), 189-192. Hirt, D. E. (1990). Dynamic surface tension of hydrocarbon and fluorocarbon surfactant solutions using the maximum bubble pressure method. Colloids and Surfaces, 44, 101-117. Hoff, J. T., Mackay, D., Gillham, R., and Shiu, W. Y. (1993). Partitioning of organic chemicals at the air- water interface in environmental systems. Environmental Science and Technology, 27,2174-2180. Holmberg, A. (1982). On the practical identifiability of microbial growth models incorporating Michaelis- Menten type nonlinearities. Mathematical Biosciences, 62, 23-43. Hornbeck, R. W. (1975). Numerical Methods, Quantum Publishers. Hua, X. , Y , and Rosen, M . J. (1988). Dynamic surface tension of aqueous surfactant solutions 1. Basic parameters. Journal of Colloid and Interface Science, 124(2), 652-659. Iliev, T. H., and Dushkin, C. D. (1992). Dynamic surface tension of micellar solutions studied by the maximum bubble pressure method. Colloid & Polymer Science, 270, 370-376. Incropera, F. P., and de Witt, D. P. (1981). Fundamentals of Heat Transfer, John Wiley & Sons. Johnson, C. H. J., and Lane, J. E. (1974). Surface shape and the calculation of surface tension from maximum bubble pressure. Journal of Colloid and Interface Science, 47( 1), 117-121. Joos, P., and Rillaerts, E. (1981). Theory on the determination of the dynamic surface tension with the drop volume and maximum bubble pressure methods. Journal of Colloid and Interface Science, 79(1), 96- 100. 76 Keirstead, K. F. (1978). Foam fractionation and surface tension method of characterizing soluble lignins. Pulp & Paper Magazine of Canada, 4, T6-9. Kobayashi, H., and Rittmann, B. E. (1982). Microbial removal of hazardous organic compounds. Environmental Science and Technology, 15(3), 170-183. Kohler, H.-H. (1993). Thermodynamics of adsorption from solution. Coagulation and Flocculation: theory and applications, B. Dobias, ed., Marcel Dekker, 1-36. Kolset, K., and Heiberg, A. (1988). Evaluation of the "FUGACITY" (FEQUM) and the "EXAMS" chemical fate and transport models: A case study on the pollution of the Norrsundet Bay (Sweden). Water Science and Technology, 20(2), 1-12. Kolthoff, I. M., and Stricks, W. (1948). Solubilization of dimethylaminoazobenzene in solutions of detergents -1. The effect of temperature on the solubilization and upon the critical concentration. Journal of Physical and Colloidal Chemistry, 52, 915-941. Kruzynski, G. (1995). Personal communication., Department of fisheries and Oceans, Vancouver, BC. Kuffner, R. J. (1961). The measurement of dynamic surface tensions of solutions of slowly diffusing molecules by the maximum bubble pressure method. Journal of Colloid Science, 16,497-500. Kukkonen, J. (1992). Effects of lignin and chlorolignin in pulp mill effluents on the binding and bioavailability of hydrophobic organic pollutants. Water Research, 26(11), 1523-1532. Kulovaara, M., Kronberg, L., and Pensar, G. (1987). Recoveries of some chlorophenolics and resin acids from humic water. The Science of the Total Environment, 62, 291-296. Leach, J. M., Mueller, J. C , and Walden, C. C. (1977). Biodegradability of toxic compounds in pulp mill effluents. Transactions of the technical section CPPA, 3(4), TR126-TR130. Levenspiel, O. (1984). The Chemical Reactor Omnibook, OSU Book Stores, Inc. Li, K., Chen, T , Bicho, P., Breuil, C , and Saddler, J. N. (1997). Factors affecting gas chromatographic analysis of resin acids present in pulp mill effluents.. Liu, H. W, Lo, S. N., and Lavallee, H. C. (1993). Study of the performance and kinetics of aerobic biological treatment of a CTMP effluent. Tappi Journal, 94(12), 172-176.. Liu, H. W, Lo, S. N., and Lavallee, H. C. (1996a). Mechanisms of removing resin and fatty acids in CTMP effluent during aerobic biological treatment. Pulp & Paper Canada, 79(5), 145-154. Liu, H. W, Lo, S. N., and Lavallee, H. C. (1996b). Theoretical study on two-stage anaerobic-aerobic biological treatment of a CTMP effluent. Water Quality Research Journal of Canada, 31(1), 1-35. Lucassen-Reynders, E. H., and van den Tempel, M. Surface equation of state for adsorbed surfactants. Chemistry, Physics and Application of Surface Active Substances, Brussels, 779-791. Mackay, D., and Paterson, S. (1981). Calculating fugacity. Environmental Science and Technology, 15(9), 1006-1014. Mackay, D., and Southwood, J. M. (1992). Modelling the fate of Organochlorine chemicals in pulp mill effluents. Water Pollution Research Journal of Canada, 27(3), 509-537. Mackay, D., Southwood, J. M., Kukkonen, J., Shiu, W. Y , Tarn, D. D., Varhanickova, D., and Lun, R. (1996). Modelling the fate of 2,4,6-trichlorophenol in pulp and paper mill effluent in Lake Saimaa, Finland. Environmental Fate and Effects of Pulp and Paper Mill Effluents, M. R. Servos, K. R. Munkittrick, J. H. Carey, and G. J. van der Kraak, eds., St. Lucie Press, 219-228. Marton, J. (1964). On the structure of Kraft lignin. Tappi Journal, 47(11), 713-719. Masutani, G. K , and Stenstrom, M. K. (1991). Dynamic surface tension effects on oxygen transfer. Journal of Environmental Engineering, 117(1), 126-142. McCubbin, N. (1983). The basic technology of the pulp and paper industry and its environmental protection practices. EPS 6-EP-83-1, Environment Canada. McLeay, D. (1987). Aquatic Toxicity of Pulp and Paper Mill Effluents: A Review. EPS Report 4/PF/l, Environment Canada. McLeay, D. J., Walden, C. C , and Munro, J. R. (1979a). Effect of pH on toxicity of kraft pulp and paper mill effluent to salmonid fish in fresh and seawater. Water Research, 13, 249-254. 77 McLeay, D. J., Walden, C. C , and Munro, J. R. (1979b). Influence of dilution water on the toxicity of kraft pulp and paper mill effluent, including mechanisms of effect. Water Research, 13, 151-158. Muir, D. C. G., and Servos, M . R. (1996). Bioaccumulation of bleached kraft pulp mill related organic chemicals by fish. Environmental Fate and Effects of Pulp and Paper Mill Effluents, M . R. Servos, K. R. Munkittrick, J. H. Carey, and G. J. van der Kraak, eds., St. Lucie Press, 283-296. Mysels, K. J. (1990). The maximum bubble pressure method of measuring surface tension, revisited. Colloids and Surfaces, 43, 241-262. Ng, K. S. (1977). Detoxification of bleached kraft mill effluents by foam separation, PhD, University of British Columbia, Vancouver. Ng, K. S., Mueller, J. C , and Walden, C. C. (1974a). Process parameters of foam detoxification of kraft effluent. Pulp & Paper Magazine of Canada, 75(7), T263-268. Ng, K. S., Mueller, J. C , and Walden, C. C. (1974b). Study of foam separation as a means of detoxifying bleached kraft mill effluents, removing suspended solids and enhancing biotreatability. 233-1, Canadian Forestry Service, Ottawa. Niimi, A. J., and Lee, H. B. (1992). Free and conjugated concentrations of nine resin acids in rainbow trout (Oncorhynchus mykiss) following waterborne exposure. Environmental Toxicology and Chemistry, 11, 1403-1407. Nyren, V , and Back, E. (1958). The ionization constant, solubility product and solubility of abietic and dehydroabietic acid. Acta Chemica Scandinavica, 12(7), 1516-1520. Palermo, D. R., and Holzer, K. A. (1992). Degradation of BOD and foam in a paper mill aerated lagoon by biological methods. Tappi Environmental Conference Proceedings, 811-819. Panikov, N . S. (1995). Microbial Growth Kinetics, Chapman & Hall. Posner, A. M. , Anderson, J. R., and Alexander, A. E. (1952). The surface tension and surface potential of aqueous solutions of normal aliphatic alcohols. Journal of Colloid Science, 7, 623-644. Press, W. H., Flannery, B. P., Teukolsky, S. A., and Vetterling, W. T. (1989). Numerical Recipes in Pascal, Cambridge University Press. Providenti, M . A., Lee, H., and Trevors, J. T. (1993). Selected factors limiting the microbial degradation of recalcitrant compounds. Journal of Industrial Microbiology, 12, 379-395. Robinson, J. A. (1985). Determining microbial kinetic parameters using nonlinear regression analysis. Advances in Microbial Ecology, K. C. Marshall, ed., Plenum Press, 61-114. Robinson, J. A., and Tiedje, J. M . (1983). Nonlinear estimation of Monod growth kinetic parameters from a single substrate depletion curve. Applied and Environmental Microbiology, 45(5), 1453-1458. Roy-Arcand, L., and Archibald, F. (1996). Selective removal of resin and fatty acids from mechanical pulp effluents by ozone. Water Research, 30(5), 1269-1279. Ruthven, D. M . (1984). Principles of adsorption and absorption processes, John Wiley & Sons. Savitzky, A., and Golay, M . J. E. (1964). Smoothing and Differentiation of Data by Simplified Least Squares Procedures. Anal. Chem., 36(8), 1627-1639. Servizi, J. A., Gordon, R. W., Rogers, I. H., AND, and Mahood, H. W. (1975). Chemical Characteristics, acute toxicity and detoxification of foam on two aerated lagoons. Environment Improvement Conference, 45-52. Sikkema, J., De Bont, J. A. M. , and Poolman, B. (1995). Mechanisms of membrane toxicity of hydrocarbons. Microbiological Reviews, 59(2), 201-222. Simkins, S., and Alexander, M . (1984). Models for mineralization kinetics with the variables substrate concentration and population density. Applied and Environmental Microbiology, 47(6), 1299-1306. Simkins, S., and Alexander, M . (1985). Nonlinear estimation of the parameters of Monod kinetics that best describe mineralization of several substrate concentrations by dissimilar bacterial densities. Applied and Environmental Microbiology, 50(10), 81—824. Springer, A. M . (1986). Industrial environmental control - Pulp and Paper Industry, John Wiley & Sons. Stuck, G., and Alexander, M . (1987). Role of dissolution rate and solubility in biodegradation of aromatic compounds. Applied and Environmental Microbiology, 53(2), 292-297. 78 Sugden, S. (1922). The determination of surface tension from the maximum pressure in bubbles. Journal of the Chemical Society - Transactions Part 1, 121, 858-866. Sugden, S. (1924). The determination of surface tension from the maximum pressure in bubbles - Part II. Journal of the Chemical Society - Transactions Part 1, 125, 27-31. Sunito, L. R., Shiu, W. Y., and Mackay, D. (1988). A review of the nature and properties of chemicals present in pulp mill effluents. Chemosphere, 17(7), 1249-1290. Supelco. (1982). Separating aqueous free carboxylic acids (C2-C5) at ppm concentrations. GC Bulletin 751G, Supelco Inc. Templeton, L. L., and Leslie Grady Jr., C. P. (1988). Effect of culture history on the determination of biodegradation kinetics by batch and fed-batch techniques. Journal of the Water Pollution Control Federation, 60(5), 651-658. Thomas, W. D. E., and Potter, L. (1975). Solution/Air Interfaces 1. An oscillating jet relative method for determining dynamic surface tensions. Journal of Colloid and Interface Science, 50(3), 397-412. Valsaraj, K. T. (1994). Hydrophobic compounds in the environment: adsorption equilibrium at the air- water interface. Water Research, 28(4). van Agglin, G. (1995). Personal communication., Ministry of the Environment, Vancouver, BC. Volkering, E , Breure, A. M. , Sterkenburg, A., and van Andel, J. G. (1992). Microbial degradation of polycyclic aromatic hydrocarbons: effect of substrate availability on bacterial growth kinetics. Microbiology and Biotechnology, 36, 548-552. Volkering, F., Breure, A. M. , van Andel, J. G., and Rulkens, W. H. (1995). Influence of nonionic surfactants on bioavailability and biodegradation of polycyclic aromatic hydrocarbons. Applied and Environmental Microbiology, 61(5), 1699-1705. Voss, R. H., and Rapsomatiotis, A. (1985). An improved solvent-extraction based procedure for the gas chromatographic analysis of resin and fatty acids in pulp mill effluents. Journal of Chromatography, 346, 205-214. Walden, C. C., and Howard, T. E. (1981). Toxicity of pulp and paper mill effluents - a review. Pulp & Paper Canada, 82(4), T143-T148. Ward, A. F. H., and Tordai, L. (1946). Time-dependence of boundary tensions of solutions - 1. The role of diffusion in time-effects. Journal of Chemical Physics, 14(7), 453-461. Werker, A. G., Bicho, P. A., Saddler, J. N. , and Hall, E. R. (1996). Surface tension changes related to the biotransformation of dehydroabietic acid by Mortierella Isabellina. Environmental Fate and Effects of Pulp and Paper Mil l Effluents, M . R. Servos, K. R. Munkittrick, J. H. Carey, and G. J. van der Kraak, eds., St. Lucie Press, 139-150. Yan, G., and Allen, D. G. (1994). Biosorption of high molecular weight organochlorines in pulp mill effluent. Water Research, 28(9), 1933-1941. Zanella, E. (1983). Effect of pH on acute toxicity of dehydroabietic acid and chlorinated dehydroabietic acid to fish and Daphnia. Bulletin for Environmental Contamination and Toxicology, 30, 133-140. Zitko, V , and Carson, W. V. (1971). Resin acids and other organic compounds in groundwood and sulfate mill effluents and foams. Manuscript Report Series No. 1134, Fisheries Research Board of Canada. 79 Chapter 3 Microbial Community Structure in the Kinetics of Resin Acid Removal Summary The objective of this investigation was to determine the influence of pH on the microbial community structure and the uptake rate of resin acids (Objective B - Chapter 1). The microbial community structures of enrichment cultures degrading resin acids were compared by the statistical analysis of microbial fatty acid compositions. The data indicate that there exists a diversity of communities that can degrade resin acids in nature. Different communities of resin acid-degrading microorganisms exhibited different growth kinetics. For any one experiment, alkaline pH conditions were observed to exhibit optimal removal kinetics. Acidic conditions again resulted in an elevated residuum or threshold concentration for biological removal (Objective A - Chapter 1). While the community of resin acid-degrading organisms varied with the time of sampling from a full scale treatment system, the enrichment culture selected from a biosolids sample was sensitive to pH. A comparison of pure versus mixed culture experimental results, indicated that changes in fatty acid compositions of mixed cultures reflect a change in community structure, more than an adaptative membrane response of individual species. Table of Contents 3.1 Introduction 81 3.1.1 Understanding Biological Removal of Resin Acids 83 3.1.2 Moving Beyond Black Box Wastewater Treatment 88 3.1.3 Biomass and Community Structure from Microbial Fatty Acid Analysis 92 3.1.4 Microbial Lipids, Fatty Acids and Their Analysis 95 3.1.5 A Review of Fatty Acid Compositional Analysis 101 3.1.6 The Principal Objective and Method Specific Issues 121 3.2 Experimental Methods and Materials 123 3.2.1 Mixed Culture Batch Growth Experiments 123 3.2.2 Pure Culture Batch Growth Experiments 127 3.2.3 Mixed Culture Method Control Experiments 128 3.3 Results 132 3.3.1 Equivalent Chain Length (ECL) Methodology 132 3.3.2 Mixed Culture Method Control Experiments 136 3.3.3 Pure Culture Batch Growth Experiments 139 3.3.4 Mixed Culture Batch Growth Experiments 142 3.4 Discussion 167 3.5 Conclusions 177 3.6 References 179 80 3.1 Introduction The objective of this phase of the overall study was to consider how changes in pH may affect the microbial growth kinetics in resin acid removal (Objective B- Chapter 1). Secondary wastewater treatment plants are designed to reduce effluent biochemical oxygen demand (BOD) and total suspended solids (TSS) (Metcalf & Eddy 1991), but are relied upon for consistent effluent detoxification. The removal of specific organic contaminants is one aspect of detoxification. Toxic organic contaminants can represent just a small fraction of the overall influent BOD while being a significant proportion of the effluent toxicity towards aquatic life in the receiving water. The removal of BOD cannot necessarily be equated with the removal of toxicity (Mueller and Walden 1976). A challenge in biological detoxification is in the consistent removal of these specific pollutants down to non-toxic levels. Organic pollutants that are toxic to aquatic life at low concentrations, in the order of one part per million, tend to be relatively stable and hydrophobic. Resin acids are such a group of aquatic toxicants specific to pulp mill effluent (Chapter 1). The previous study (Chapter 2) served to characterize important pH-dependent physico-chemical properties of resin acids that affect the contaminant bioavailability. The nature of resin acid associations in solution can impact the contaminant transport by passive diffusion, which is a required step for microbial uptake. In Chapter 2, a threshold level for biological resin acid removal was observed to increase with decreasing pH. Based on solubility and dynamic surface tension data, the increase in residuum was related to a decrease in the solute mobility. Resin acid uptake kinetics were also observed to change with pH but could not be completely characterized due to the absence of a reliable biomass measure. The previous difficulties encountered with attempts at quantifying biomass by more conventional non-specific assays, were overcome in the present stage of the investigation by extracting, identifying and measuring specific microbial constituents during batch growth. Microbial fatty acids hydrolyzed and extracted from cellular lipid were used to quantify and interpret the growth kinetics of mixed cultures on resin acids as a function of pH. The added benefit of analysing microbial fatty acids 81 was that relative differences in the structure of microbial communities cultured at different pH levels could be compared. This added information became essential to the data interpretation. Since solution pH, in the 6 to 8 range, that defines typical treatment operating limits, strongly affected the mobility of resin acids in solution (Chapter 2), it was of interest to know i f an optimum growth-linked resin acid removal rate existed in this pH range. Bacteria in a biological treatment system are sensitive to selective pressures like pH. Consequently, the question of the extent to which the community of resin acid degraders is altered by pH has implications for process control. Since pH and the form of resin acids in solution are separate but interdependent effects, it was necessary to try to distinguish between the direct effects of pH on microbial activity and the indirect pH-related effect of a change in resin acid hydrophobicity. Experiments were therefore undertaken to consider the mechanism of pH-dependence that most influenced the kinetics of resin acid removal. The batch growth kinetics and ecology of mixed cultures enriched for resin acid degraders and acclimated to a number of different pH levels were compared. Pure culture experiments were performed to help in the interpretation of the observations made on the mixed cultures. In addition, control experiments were conducted to help in the interpretation of the results of microbial lipid analysis from mixed cultures. Research to assess the community structure, as well as the quantity of biomass in a biological treatment system, moves beyond the more traditional black box engineering approach of wastewater treatment. A better understanding of microbial ecology in secondary treatment systems has been pivotal for success in understanding and controlling sludge bulking phenomena (Blackall et al. 1991; Fahmy and Hao 1990; Khan and Forster 1991; Lemmer 1986; Lemmer and Baumann 1988; Richards et al. 1990). A n analogous contribution was seen as necessary for understanding and optimising organic toxicant removal in biological systems. The development of this understanding required a readily measurable fingerprint for complex microbial communities that could be used to monitor the kinetics of change and adaptation in response to a variety of environmental stresses. By establishing cause and effect relationships, a fingerprint for the system microbial community could serve ultimately in treatment process control. A 82 microbial fingerprinting assay that can be adapted to on-line measurement has significant engineering application. Fatty acids extracted from microbial lipids were found to have the potential to be used as such a fingerprint. It is a measurement that is also amenable to some degree of automation. Therefore, microbial fatty acid analysis and its interpretation became a major part of this investigation. 3. J. 1 Understanding Biological Removal of Resin Acids The applied experimental methodology of mixed culture batch growth, following steps of enrichment (Brock and Madigan 1991) with selected resin acids as the sole carbon source, is a simplification of a full scale secondary treatment system. The scientific approach of reduction of complicated processes can provide data that have little bearing on the system of interest. Therefore, it is important to consider the engineering relevance of this experimental approach to full scale biological wastewater treatment. The conundrum in laboratory wastewater experiments is in the sufficient simplification of a complex system to yield general and meaningful results that have engineering application for the real process. By oversimplification, the important features of the real system may be lost. The disadvantage of attempting to include too much of the system complexity is that the relationships underlying the results may be difficult to clearly interpret. Fundamental understanding of microbial activity toward specific organic compounds comes from basic research on pure culture isolates. It is on this footing that some experimental simplification has been made. As noted in Chapter 1, resin acids can be classed into two chemical groups, abietanes and pimaranes (Figure 3-1). Bacterial isolates have been cultured that are quite specific in their ability to degrade either abietanes or pimaranes (Mohn 1995; Wilson et al. 1996). Furthermore, the ability to metabolize the naturally-occurring resin acids does not necessarily confer an ability to metabolize the chlorinated forms (Mohn et al. 1997) which are degraded more slowly (Leach et al. 1977). Pure culture research into the metabolic pathways for the catabolism of dehydroabietic acid indicates that different organisms have 83 Figure 3-1. Chemical structure of abietic (left) versus pimaric acid (right). Both abietanes and pimaranes are based on the same diterpene hydrocarbon skeleton. Abietic-type acids have an isopropyl side chain, while pimaric-type acids have the methyl and vinyl substituents at this site. Abietanes have greater isomerization reactivity than the pimaranes due to their two conjugated double bonds. Some microorganisms are specific in their abilities to degrade either of these two forms of resin acid. developed similar, but distinct strategies of enzymatic attack (Biellmann et al. 1973a; Biellmann et al. 1973b). Therefore, results from mixed culture laboratory experiments using a cocktail of resin acids would be clouded by the uncertainty of community variability due to the complex feed composition. In other words, i f each individual resin acid can support the growth of a number of individual microbial species, then the number of possible distinct communities of resin acid degrading microorganisms increases with the number of resin acids composing the substrate mix. To constrain the number of possible community permutations, only abietane resin acids were considered. A study using only these selected resin acids is without any loss in generality because, based on the literature, enrichment cultures for pimaranes, chlorinated resin acids or mixtures thereof, could have also been generated. The decision to use only one resin acid type, was to more clearly isolate the influence that pH and pH-dependent resin acid solubility have on biodegradation. Although pimaranes and abietanes are a part of the same chemical family and will have similar pH-dependent solubility characteristics, they are, from a microbial perspective, distinct compounds. Generally, the effect of pH on the removal of a particular compound should decrease with increased diversity of the group of competent microorganisms (Alexander 1994). By including only abietane resin acids in the medium, it was expected that the role of microbial diversity on the contaminant uptake as a function of pH could be assessed. Alexander also comments that the effect of pH on biodegradation of polluting chemicals has received little attention. A lack of 84 literature experience on pH effects was felt to be further impetus for avoiding any unnecessary experimental complication. Pure culture microbial research is an extreme, but necessary level of experimental reductionism for obtaining fundamental scientific knowledge about the nature of microorganisms. However, natural or engineered microbial systems contain many interacting species. Neutralism, or the null response of one organism due to the presence of another is apparently rare (Bailey and Ollis 1986). Pure culture studies are inadequate for predicting the fate of substances in the environment if, in practice, removal enlists many interactions between different organisms (Kobayashi and Rittmann 1982). Therefore, pure culture kinetics have limited engineering application. Applied research in wastewater treatment must allow for the potential, i f only in a restricted sense, for interactions between microorganisms. Mutualism, the ability of species to grow faster together, than in isolation, is more common than neutralism (Bailey and Ollis 1986). Mutualism can involve a symbiotic exchange of growth factors, such as dissolved gas exchange between aerobic bacteria and photosynthetic algae. Commensalism is also a form of symbiosis where one species enjoys benefits, such as moderated environmental conditions, provided by the activity of the first. Competition between microorganisms can produce amensalism, or the suffering of a second species as a result of its interaction with the first. Amensalism could be due to the production of toxic compounds (antibiotics) or the removal of essential nutrients. Panikov (1995) discusses a survival or life strategy triad characteristic of microbes, that shape the resultant communities in natural ecosystems. Survival of individual populations within a community will depend on their ability to (a) compete with other populations, (b) recover from environmental perturbations and (c) survive periods of stress. Real microbial systems involve many interacting populations, operating on a number of trophic levels, that make up a complicated food web of predator and prey (Bailey and Ollis 1986). Meaningful simplification would, therefore, appear to be an oxymoron. However, the experimental method of mixed culture, batch enrichment for resin acid degraders does represent 85 an important facet of the complete system that can be clearly defined. There are three conceivable mechanisms for the biological removal of a selected organic compound, namely, growth-linked metabolism, uncoupled metabolism and cometabolism (Alexander 1994; Brock and Madigan 1991; Kobayashi and Rittmann 1982). In growth-linked metabolism, culture biomass increases at a rate that parallels the disappearance of the limiting substrate (Alexander 1994). As a rule, organisms convert organic substrate to CO and water for energy to proliferate. In the process, leaked products of catabolism may 2 similarly support the growth of other organisms. Oxygen serves not only as the terminal electron acceptor in aerobic respiration, but also as a substrate, in oxygenase-catalyzed reactions necessary for ring cleavage and hydroxylation (Providenti et al. 1993). The cellular efficiency in the conversion or yield of biomass from organic carbon varies with metabolic pathway, organism, substrate, concentration and environmental conditions. For growth to occur, the compound must first be transformed enzymatically to chemical intermediates that are components of the major metabolic pathways such as the tricarboxylic acid cycle (Brock and Madigan 1991), that are common to heterotrophic microorganisms. The initial catabolic steps could be accomplished as a consorted effort of a number of species in the cornmunity. In some situations, organic mineralization is uncoupled from growth and acts only to maintain rather than proliferate populations (Alexander 1994). Catabolism for maintenance energy sustains motility, intracellular and extracellular concentration gradients, and the replenishment of hydrolysed macromolecules (Sherrard and Schroeder 1973). Another form of non-growth related respiration is in the cellular response to abrupt environmental transitions such as swings of 'famine to feast' (Panikov 1995). Thresholds of removal exist when slow substrate utilisation kinetics that occur with very low concentrations, provide insufficient energy flux for maintenance (Kobayashi and Rittmann 1982). Contaminants may also persist i f their concentrations are too low to induce the production of the necessary catabolic enzymes. Co-metabolism occurs when other organic material is present as the primary energy source for growth and a secondary breakdown of the target contaminant occurs (Brock and Madigan 1991; 86 Kobayashi and Pvittmann 1982). For example, pentachlorophenol degradation is enhanced with the addition of glutamate to the growth medium (Providenti et al. 1993). However, diauxic growth, or the preferential metabolism of alternate carbon sources may limit biodegradation of the target contaminant. Diauxic growth was observed in the present study for batch growth on a mixture of sodium acetate and resin acids (Chapter 2). It is also possible that the target contaminant can be gratuitously metabolized by the production of enzymes induced by the presence of another substrate. If the microbial community removing the target substrate is small and is governed solely by co-metabolism, then an increase in the contaminant loading will not increase the community size and the conversion rate will be maintained (Alexander 1994). In a secondary biological treatment process, the contribution of growth-linked, uncoupled and cometabolic mechanisms will sum to yield the net observed removal rate. Respirometric methods have been developed for determining the extant or in situ biodegradation kinetics of specific organic compounds (Ellis et al. 1998). However, such a measurement cannot distinguish the relative contributions of the three separate mechanisms. Measurement of extant kinetics provides information of an immediate nature that is specific to the history of the biomass sample. Determining extant kinetics relies on an assumption about the biomass fraction metabolising the test compound. The respective biomass fraction is usually assumed to be proportional to the fraction of the influent biodegradable chemical oxygen demand (COD) contributed by the target compound. However, this estimate of the biomass fraction implicitly assumes only growth-linked metabolism. Instead of assuming growth-linked metabolism, the approach of the current investigation was to explicitly measure growth-linked mixed culture metabolism. This was seen as a starting point from which, in future work, the other possible mechanisms of biological resin acid removal could eventually be considered. Selection for growth-linked metabolism can be ensured by culture enrichment using the target contaminant as the sole carbon source. Culture enrichment is the first step in the protocol used by microbiologists to obtain isolates that grow on one selected carbon source (Alexander 1994; Brock and Madigan 1991). Small 87 inoculum transfers from sequential batch growth cycles will dilute out of the system those organisms that do not grow. Contributions of cometabolism and uncoupled growth are thereby washed out. By the competitive exclusion principle (Bailey and Ollis 1986), when two species compete in a common environment, one species eventually disappears. Competitive exclusion through culture enrichment selects for the fastest growing community of bacteria. Some members of the community may grow quickly on the substrate, others may be able to maintain their numbers by mutualism or commensalism. Therefore, while enrichment on a sole carbon source may procure isolates, the procedure selects for the fittest growth-linked community that will entail both isolates and partnerships. Prey dynamics are also not ruled out, but these will only become important once growth on the primary carbon source slows down during the decline phase (Alexander 1994). The existence of only growth-linked substrate removal by the enrichment culture can be verified from experimental data of substrate depletion and biomass production. The specific rate of substrate removal should correspond directly to the specific rate of biomass production. Microbial diversity can be assessed by the variability between the ecology of one enrichment culture and another. Assessment of the microbial community structure, moves beyond the more traditional approach of black box wastewater treatment. 3.1.2 Moving Beyond Black Box Wastewater Treatment Biological treatment processes are complex, controlled microbial ecosystems containing a mixed community of microorganisms interacting with each other, and their local environment to affect the concentrations of individual pollutants. There are many examples in the literature that demonstrate how activated sludge can acclimate to degrade many organic contaminants under steady state conditions, with a variety of bioreactor configurations. Demonstration of biodegradability is necessary for determining the feasibility of using a biological system to remove toxic organic contaminants, versus other physical, or chemical treatment methods. However, the requirement for microbial acclimation does limit the ability of biological systems to accommodate process transients. This limitation has often been implicitly neglected in the 88 research reported in the literature, since steady state removal is typically investigated. Experimental studies into biological treatment generally utilize steady state bioreactor operation, in which the success of the process is equated to a high removal efficiency. In contrast, real systems continually experience both hydraulic and contaminant loading transients. The rate at which microorganisms acclimate to these transients is, therefore, an important facet of treatment reliability. If a microbiological wastewater treatment approach has been shown to be feasible, then representative rates of biodegradation are needed to assess its practical application. For instance, tank volume and aeration cost are important design constraints. Conservative mathematical modelling used to design and predict the expected performance of full scale biological treatment systems requires intrinsic rate constants (Grady 1990). Rate constants must be obtained experimentally with bench or pilot scale biological treatment processes. Generally, the mixed microbial culture in a continuous flow treatment system is considered to be a single, so-called "pseudo-species" (Bailey and Ollis 1986). Effective kinetics for any given pseudo-species can be determined from batch growth experiments inoculated with biomass from continuous flow systems (Ellis et al. 1998; Klecka and Maier 1985; Shamat and Maier 1980). Computer simulations of a full scale process can then be made assuming that those kinetic parameters are valid and that the fraction of the total biomass acting on the xenobiotic compound is proportional to its contribution to the total biodegradable chemical oxygen demand (COD). However, unqualified use of these experimentally-determined microbiological rate constants is tenuous, because the history of the mixed culture in a batch growth experiment influences the resulting kinetic parameters (Templeton and Grady 1988). Mixed microbial populations are dynamic entities that are strongly dependent on the culture history and the selective pressures created by the growth conditions (Chiu et al. 1972). Changes in selective pressures on mixed populations by altering such factors as cell recycle rate, inlet flow rate, feed composition, temperature and pH will produce markedly changed microbiological communities or so called pseudo-species. 89 Degradation experiments using a pure microbiological culture will provide kinetic parameters which are intrinsic to that particular organism under the particular set of growth conditions chosen. Unfortunately, due to the inherent variability of mixed-microbial populations, similar kinetic characterization of mixed culture systems cannot be intrinsic without a means to distinguish one mixed population from another. Hence, i f the black box or pseudo-species approach is applied to experimentation with mixed microbial populations, experimentally- derived kinetic parameters should be taken as extrinsic and, strictly speaking, locally and temporally limited in their scope of application. A black box approach to mixed populations makes it impossible to be sure of the source of the variability in the growth parameters estimated with repeated experimentation. Comparison of kinetic constants by researchers has little value unless it can be confirmed that the substance in the black box is of a similar or dissimilar nature. If similar rate constants are consistently found with a variety of pseudo-species, then more confidence is generated regarding the robustness of the biological process with respect to the target contaminant. Conversely, i f a variety of pseudo-species produce a range of rate constants, then a large factor of safety must be employed in designing the treatment process. If pseudo- species changes are sensitive to particular bioreactor conditions then tighter process control may be required. Therefore, significant advancement could be made in the interpretation and application of mixed population kinetic parameters, i f some form of fingerprint could be obtained to characterize a degree of similarity between microbial communities. One attractive method to compare mixed communities of microorganisms originates from the study of lipid extracts from biomass samples (Vestal and White 1989). The attraction of this approach for the present study is related to the ability of the method to quantify low concentrations of biomass during batch growth in a pulp mill effluent matrix. Since bacterial lipid analysis also enables the measurement of the community structure (quality) as well as the quantity of the biomass, a significant step can be made beyond the limitations of more conventional environmental engineering black box microbiology. In the present study, a measure for biomass was needed for determining growth yields and specific growth rates. A measure for biomass community structure was useful for the interpretation of variability in 90 growth yields and rates for different experimental conditions. The need for a non-conventional measure of microbial biomass arose due to problems with the more conventional methods. Resin acids were a component of the medium suspended solids, whose interference in gravimetric or optical density methods changed with pH and time. Interference from the background high molecular weight dissolved organic matter (HMW DOM) or kraft lignin in the medium was also pH-dependent because these constituents become increasingly colloidal with decreasing pH (Marton 1964). While protein assays were found to be relatively sensitive, the Coomassie Brilliant Blue G dye protein-binding dye assay on membrane filter-harvested cells (Chapter 2) suffered from interference, which increased with decreasing pH. Protein assay interference was attributed to colloidal kraft lignin and resin acid. The dye binding protein assay depends on Van der Waals forces and hydrophobic interactions (Compton and Jones 1985). Aromatic structures show a strong affinity for the dye. Native lignin contains many aromatic functional groups which makes its concentration quantifiable by a related colourimetric technique (see Standard Methods section 5550 (Clesceri et al. 1989)) that is also used for protein assays (Deutscher 1990; Peterson 1977). Time restrictions and other practical limitations during sampling ruled out the use of direct measures of biological activity such as oxygen uptake respirometry or microbial adenosine triphosphate (ATP) extraction. As part of the method development for the present study, attempts at ATP measurement by cold acid extraction (Karl and Craven 1980) and photometric quantification by firefly bioluminescence (Karl 1980) were found to be laborious and imprecise. The extreme labile nature of ATP was a factor in the difficulty of using ATP as a biomass measure. The indicator of viable biomass required for the present experiments needed to be an easily preservable analyte and a conserved microbial constituent with a relatively rapid in situ turnover rate. Microbial fatty acids were selected both as a biomass measurement and a means to differentiate the communities cultured. The assay for microbial fatty acids (see Section 3.2) also offered significant practical advantage since the substrate resin acids could be quantified from the same measurement. Reduction of the sampling and analytical burden permitted the testing of many 91 more experimental conditions. 3.1.3 Biomass and Community Structure from Microbial Fatty Acid Analysis Lipid analysis has been developed by microbial ecologists to measure microbial community biomass, structure and activity under in situ conditions (Vestal and White 1989). Microorganisms rarely exist in nature as monocultures. The culturing of microorganisms is a process of elimination that allows only certain microbes with specific metabolic properties to grow under the selected conditions of incubation. Growth of microbes on artificial media may reveal only a small fraction of the total microorganisms present in the original system. In principle, isolation techniques may filter out those organisms that are the key members of the communities from the environment under study. Advancement in biological wastewater treatment research requires understanding of the behaviour of complete communities of microorganisms in treatment systems. Phospholipid fatty acids are a relatively conserved constituent of every cellular membrane. They are not found in storage lipids and furthermore, they undergo a relatively rapid turnover rate in both living cells and non-viable cells added to the environment (White 1983). Cell lipid extracts contain ester-linked fatty acids, typically with carbon chain lengths of 12 to 20 atoms (Harwood and Russell 1984). Gas chromatographic identification of phospholipid fatty acid (PLFA) compositions of pure microbial cultures has application in taxonomic classification of different bacterial and fungal species (Bousfield et al. 1983; Eerola and Lehtonen 1988; Lechevalier and Lechevalier 1988; Moss and Dees 1976; Stahl and Klug 1996). The fatty acid spectra for bacterial members of the same family are qualitatively similar and, for different families, distinguishably different (Kates 1964). While quantitative variations in the relative abundance of fatty acids in a given profile have been observed with changes in the composition of the growth medium and with the age of the culture, the characteristic family pattern remains distinctive (Haack et al. 1994; Kates 1964). Thus, pure bacterial cultures grown under controlled growth conditions have unique whole cell fatty acid profiles which can be used to differentiate even closely related organisms (Miller and Berger 1985). Microorganism genera 92 and species can be distinguished by the presence or absence of a particular fatty acid and, for well controlled conditions, their relative abundance. Since changes in the relative abundance of phospholipid fatty acids coincide with physical, chemical or temporal changes in the growth conditions, fatty acid analysis can be used to follow the kinetics of microbial adaptations to transient conditions in biological treatment systems. Cells alter the fatty acids in their lipids as they adapt to environmental conditions to maintain membrane fluidity. Cell viability depends on membrane fluidity necessary for sustaining membrane protein function (Harwood and Russell 1984; Rose 1989). For example, a decrease in growth temperature has been found to result in a increase in the proportion of unsaturated fatty acids (Marr and Ingraham 1962). Hydrophobic organic contaminants can also influence microbial membrane lipid composition. The general opinion is that: (a) uptake of cyclic hydrocarbons by microorganisms is a passive process, (b) partitioning of these lipophilic compounds into the cytoplasmic membrane is integral to their uptake, and (c) these substrates must enter the cell prior to their metabolism (Sikkema et al. 1994; Sikkema et al. 1995). Since hydrophobic, organic contaminants partitioning to the cell membrane can disrupt membrane lipid-protein interactions (fluidity), they can exert a toxic effect, requiring microbial adaptation before resumption of normal metabolic activity. Analogous to the observed temperature response noted above, fatty acid composition alterations have been induced by alcohols or organic solvents in the growth medium (Ingram 1976; Ingram 1977; Pinkart et al. 1996). Phospholipid fatty acids have been found useful to track the biomass and community structure in terrestrial (Bruggemann et al. 1995; Cavigelli et al. 1995; Franzmann et al. 1996; Frostegard et al. 1997; Reichardt et al. 1997; Sundh et al. 1997; Tunlid and White 1992; Zelles et al. 1992; Zelles et al. 1994) and marine (Bobbie and White 1980; Gillan and Hogg 1984; Khandekar and Johns 1990; Reemtsma and Ittekkot 1992; White et al. 1979) environments. Recently, lipid analysis has been utilized for monitoring bench scale biofilters treating hydrogen sulfide and volatile organic compounds at a wastewater treatment plant (de Castro et al. 1997; Webster et al. 93 1997). Lipid-based biomass measures correlate well with ATP, direct cell enumeration, D N A synthesis and respirometric assays (Balkwill et al. 1988; White et al. 1979; Zelles et al. 1994). Some discrepancy in the correspondence of plate counts to fatty acid concentrations in a biofilter study by Webster et al. (1997), was reported to be due to a decrease in culturable organisms. Due to the presence of non-culturable organisms, plate counts are anticipated to yield lower values than the true numbers present in mixed culture systems (Lechevalier and Lechevalier 1988). Although phospholipid fatty acid (PLFA) concentration is, in itself, a biomass measure, just as dry weight, protein, ATP, or volatile suspended solids (VSS), conversion of PLFA to more traditional measures is often preferred. Escherichia coli has been used as a reference microorganism for conversion of PLFA to numbers of microorganisms. One pica mole of palmitic acid, a fatty acid that is ubiquitous in microbial lipids, is equivalent to about 5x l0 5 cells the size of E. coli (Baird and White 1985; White 1983). Balkwill et al. (1988) estimated that 100 pmol PLFA is equivalent to one gram dry weight of cells. The conversion by Balkwill et al. assumes a specific cell weight of 5x10 1 3 grams per cell. However, this value for E. coli has been reported to be 2 .8xl0 1 3 g/cell (Brock and Madigan 1991). Thus, there exists some discrepancy in the attempts to estimate cell numbers from PLFA data. The phospholipid contents of various bacteria have further been shown to vary between 4 and 91 milligrams per gram dry weight (Kates 1964). Specific taxa of bacteria can yield significantly different fatty acid contents, ranging from 11 to 197 umol of PLFA per gram of dry weight (Haack et al. 1994). The lipid content of microorganisms is also known to vary with environmental conditions (Tunlid and WTiite 1992). Therefore, due to this well-documented variability, a reliance on literature "constants" for conversion of PLFA data is not advisable practice. The PLFA to dry weight ratio could itself be a monitor of community change. PLFA conversion should not be considered necessary since it is, by itself, a reference parameter to be compared and monitored in the same manner as VSS. 94 Conversion of PLFA data may have some utility when constructing a carbon balance of a treatment system for assessing microbial energetics. The carbon content of bacteria is typically 50% of the dry weight (Metcalf & Eddy 1991). From experimentally-derived PLFA to dry weight ratios, biomass carbon produced, can be estimated and compared to the corresponding organic carbon consumed. The monitoring of phospholipid fatty acids in bioreactors, as opposed to natural environments, has the advantage that engineered biological systems tend to be relatively well controlled and easily monitored systems. Repeated experiments with bioreactors can be used to establish links between process variables and community adaptations based on phospholipid fatty acid profiles. 3.1.4 Microbial Lipids, Fatty Acids and Their Analysis The analysis of lipids in microbial samples can entail a suite of assays (Figure 3-2), of which microbial fatty acid identification and quantification is one component (White 1983). The assays inform on biomass, community structure or nutritional status. This information has engineering application in the control of biological wastewater treatment systems (Riebel et al. 1997). Of the measurable constituents shown in Figure 3-2, only fatty acid analysis measures both biomass and community structure. There are a number of fatty acids of varying chain length commonly associated with microorganisms. The major fatty acid groupings are the saturated, unsaturated, branched, cyclic and hydroxy types (Table 3-1). Unsaturated fatty acids in bacteria are typically monoenoic (mono-unsaturated), indicating that the presence of polyenoic fatty acids in a sample can be attributed to contributions from plant, algae, fungi or cyanobacteria (Harwood and Russell 1984). The nomenclature of fatty acids (Ratledge and Wilkinson 1988a) needs some clarification since different formats appear in the literature, leading to possible confusion. Fatty acids always have a systematic name and, often, a trivial name. For instance, palmitic acid is the trivial name for the saturated sixteen carbon fatty acid, hexadecanoic acid. For clarity, only systematic names will be used in this text. Due to their compactness, shorthand designations will also be 95 Table 3-1. Generalized chemical structures for the typical groups of bacterial fatty acids. Bacterial Fatty Acid Generalized Chemical Structure Straight Chain Saturated C H 3 - ( C H 2 ) „ - C - O H O Straight Chain Unsaturated C H 3 - ( C H 2 ) 0 - C H = C H - ( C H 2 ) M - C — O H o Branched Saturated C H 3 - ( C H 2 ) „ X C H 2 - ( C H 2 ) M - C - O H CH/ 0 Iso-Branched Saturated C H 3 C H 2 - ( C H 2 ) „ - C - O H CH/ 0 Antei so-Branched Saturated CH3 CH2v C H 2 - ( C H 2 ) N - C - O H CH/ » Cyclopropane Saturated C H 3 - ( C H 2 ) „ - C H — C H - ( C H 2 ) M - C — O H \ / O C H 2 " oc-Hydroxy Saturated C H 3 - ( C H 2 ) N - C H - C - O H O H O P-Hydroxy Saturated C H 3 - ( C H 2 ) N - C H - C H 2 - C - O H 1 II O H O Table 3-2. Nomenclature for the common groups of fatty acids. Generally the shorthand notation is of the form N:x where N is the number of carbon atoms and x is the number of double bonds. The "con" indicates the n* carbon counting from the methyl end of the chain. Otherwise position of elements should be counted from the carboxyl end of the chain. Shorthand abbreviations are as follows: c-cis, t-trans, br-branched, i-iso, a-anteiso, Me-methyl, cy-cyclo, and OH-hydroxy. Shorthand designation Type Systematic Name Nonspecific Specific Alternate saturated octadecanoic acid - 18:0 - cis-monoenoic cis-octadec-11-enoic acid 18:1 18:1(1 lc) 18:lco7c trans-monoenoic trans-octadec-11-enoic acid 18:1 18:l(llt) 18:lco7t dienoic cis,cis-octadeca-9,12-dienoic acid 18:2 18:2(9cl2c) 18:2co6 polyenoic cis,cis,cis-octadeca-9,12,15-trienoic acid 18:3 18:3(9cl2cl5c) 18:3co3 branched 10-methyloctadecanoic acid brl9:0 10Mel8:0 - iso-branched 16-methylheptadecanoic acid brl8:0 il8:0 16Mel7:0 anteiso-branched 15-methylheptadecanoic acid brl8:0 al8:0 15Mel7:0 cyclopropane cis-11,12-methyleneoctadecanoic acid cyl9:0 cyl9:0ci)7 - a-hydroxy 2-hydroxyoctadecanoic acid OH-18:0 2OH-18:0 aOH-18:0 p-hydroxy 3-hydroxyoctadecanoic acid OH-18:0 3OH-18:0 POH-18:0 96 Sample Hydrolysis Lipid Extraction Poly-p-hydroxy acid polymers (nutritional status) Solvent Extraction Organic Fraction Silici c Acid Chromatography Aqueous Fraction Residue ATP (Microbial activity) Organic Fraction Whole Cell Fatty Acids (Viable biomass & Community Structure) Neutral Lipids 1 Phospholipids Triglycerides, waxes & steroids LPS lipid A (Gram -ve Biomass) Teichoic acids (Gram +ve Biomass) Muramic acid (Eubacterial Biomass) Hydrolysis Solvent Extraction Organic Fraction 1 Aqueous Fraction Phospholipid Fatty Acids (Viable biomass & Community Structure) Phosphate (Viable Biomass) Figure 3-2. Flow diagram for biochemical lipid analysis of natural microbial communities (adapted from Vestal and White (1989)). frequently used. However, shorthand designations are complicated by non-specific and multiple specific forms of notation. The shorthand notation conventions adopted here are summarized by example in Table 3-2. Generally, carbon locations in a fatty acid are referenced either to the co- end (methyl end) of the chain or to the carboxyl end. Thus "col" is the carbon atom furthest from the carboxyl end. With respect to the carboxyl end, the alpha position is the second carbon atom and the beta position is the third carbon atom in the chain. The specificity of some lipids to particular kinds of microorganisms makes it possible to identify certain biomass fractions. As shown in Figure 3-2, the residue fraction can be assayed for constituents specific to gram negative or gram positive bacteria. For bacterial fatty acid enumeration there are two methodologies in use, namely phospholipid fatty acid (PLFA) and 97 whole cell fatty acid (WCFA) analysis. Phospholipid fatty acid analysis by definition considers only those fatty acids linked to phospholipids. This form of lipid analysis requires careful separation of microbial fractions before fatty acids are released from their lipids by hydrolysis. The established method of lipid extraction (Tunlid and White 1992; White et al. 1979) is the procedure developed by Bligh and Dyer (1959). Exposing a sample to a single-phase mixture of chloroform, methanol, and water (1.0:2.0:0.8) dissolves lipids while arresting metabolic activity (Vestal and White 1989). Changing the proportions of water and chloroform generates a biphasic mixture and the lipids partition to the lower solvent phase. The extracted lipids can be purified by silicic acid chromatography and the ester-linked fatty acids converted to their methyl esters by mild alkaline methanolysis. The fatty acid methyl esters (FAMEs) are readily analysed by gas chromatography (GC) using a flame ionisation detector (FID). This extraction procedure has recently been used to simultaneously follow microbial community structure and the uptake of polyaromatic hydrocarbons (Fang and Findlay 1996). The advantage of a more detailed lipid analysis is that sometimes the source of the fatty acid can provide more information than the type of fatty acid obtained. Since fatty acids do not occur in the cells as free acids but are generally linked to other compounds, the taxonomic value of fatty acid analysis can be markedly improved i f the lipids are fractionated into at least polar and neutral lipids before they are released by saponification (Lechevalier and Lechevalier 1988). The most common phospholipids consist of two long chain fatty acids esterified to phosphoglycerol (Table 3-3). The typical phosphoglycerides are phosphatidylinositol (PI), phosphatidylglycerol (PG), phosphatidylcholine (PC), phosphatidylethanolamine (PE), phosphatidylserine (PS) and diphosphatidylglycerol (DPG). These lipids are associated with cytoplasmic membranes, with some being more common in some microbial groups than in others (Harwood and Russell 1984). The lipid content of a microorganism reflects its genetic make-up. The cytoplasmic membrane is a selectively permeable barrier that isolates the cell contents from 98 Table 3-3. The general structure and types of the important phospholipids found in microbial membranes (adapted from Harwood (1984)). General Formula Regions X-substituent phospholipid X O=P—O" Polar Head Group Phosphoglycerol backbone H phosphatide acid 1 o 1 H 2 C — H C — C H 2 1 1 0 o 1 1 c = o c = o serine phosphatidylserine ethanolamine phosphatidylemanolamine I < choline phosphatidylcholine < { Non-polar Tails Two long chain ester linked fatty acids glycerol phosphatidylglycerol inositol phosphatidylinositol \ ( phosphatidylglycerol diphosphatidylglycerol Figure 3-3. A schematic drawing depicting a segment of the classical lipid bilayer containing intrinsic (I) proteins. The bilayer is an arrangement of phospholipids with their polar head groups (P) facing out and their long chain hydrocarbon (fatty acid) tails (C) facing in. The membrane is said to be fluid in the plane of the layer. Protein function depends on membrane fluidity (adapted from Brock and Madigan (1991)). the surrounding environment (Figure 3-3). Membranes also provide a special hydrophobic environment where enzymes and special cellular processes can operate (Brock and Madigan 1991; Harwood and Russell 1984). The alkyl (fatty acid) chains of the lipids in the membrane core are in a liquid-crystalline phase and the membrane is said to be fluid. Fluidity allows both intrinsic proteins and lipids to move in the plane of the membrane. The presence of iso or anteiso branched-chain acids in membrane lipids increases fluidity as does the presence of cis and trans double bonds. Shifts in relative proportions of fatty acids in the cytoplasmic 99 membrane are a direct microbial response to environmental factors. The disadvantages of the Bligh and Dyer lipid extraction method are the number of analytical steps required and the fact that the organic (extract) phase ends up below the aqueous phase. Whole cell fatty acid analysis is a more rapid alternative to PLFA, that was developed for clinical microbiology use (Miller and Berger 1985). Fatty acids from whole cell hydrolysates are obtained without regard to their source from within the cell. Fatty acids are released from whole cell samples by hot methanolic alkaline saponification. The cellular solution is acidified and the liberated fatty acids are methylated by heating the solution in the presence of methyl alcohol. Fatty acid methyl esters can be extracted from the acidified aqueous solution, but the organic extract must be washed with a dilute base to remove acid carryover. Once again the fatty acids are measured and identified by GC/FID. Whole cell fatty acid (WCFA) analysis will contain fatty acids from all cellular sources. For example, the outer membrane of gram-negative bacteria consists of lipopolysaccharide or LPS which is known to yield a-hydroxy, P-hydroxy and non-hydroxy fatty acids (Wilkinson 1989). However, data acquired by WCFA and PLFA protocols have been found to be of a similar nature (Haack et al. 1994). Hence, for the comparison of fatty acid spectra obtained from mixed cultures, both methods appear to be equally valid. Quantification of fatty acid mixtures by GC/FID requires the use of standards for peak identification. Due to the number of possible fatty acids that could be present, a commercial standard mixture (Supelco Inc.) of common bacterial fatty acid methyl esters (FAMEs) may not contain all the FAMEs found in an environmental sample. Unknown peaks can be tentatively identified by equivalent chain length (ECL) calibration of the GC column for a specific temperature program. The E C L expresses the relative elution position of any peak with respect to the closest saturated fatty acid. Gas chromatograms of a standard mixture containing a homologous series of saturated fatty acid methyl esters are used to equate retention time to an equivalent chain length or E C L (Gillan 1983; Miller and Berger 1985). The E C L for a peak, x, is calculated by interpolation of the retention times (t) of the two neighbouring saturated fatty 100 acids (n:0 and (n+l):0): A family plot of carbon length (n) versus relative peak position (ECL - n) for the FAMEs in the x standard mixture serves as an identification template. Since members of a homologous series of fatty acids tend to be linear on the family plot, probable identification of F A M E peaks is made from their E C L and interpolation on the family plot. Increased confidence in the identity of a fatty acid peak can be made by a confirmatory result using the same analysis with a different column or by GC with mass spectrometry (MS) on selected samples. Commercial hardware and software is available for F A M E analysis (Microbial Identification System, MIDI, Inc. Newark, Delaware). Some other protocols for F A M E analysis that have been recently published are pyrolysis mass spectrometry (Basile et al. 1995), in situ supercritical fluid extraction and derivatization followed by GC/MS (Gharaibeh and Voorhees 1996), and sodium methoxide methylation (Rozes et al. 1993). For the purpose of the present investigation, whole cell fatty acids were monitored during mixed culture batch growth experiments on resin acids in kraft pulp mill effluent as a function of pH. A modified WCFA type protocol was used for quantitative extraction, derivatization and analysis of both resin and fatty acids. The WCFA method applied was verified by experimental comparison to the more traditional PLFA analysis. F A M E peak confirmation was accomplished by E C L calibration followed by GC/MS analysis on selected samples. 3.1.5 A Review of Fatty Acid Compositional Analysis Analysis of fatty acid compositions is a multivariate statistical problem that requires special consideration (Aitchison 1986). However, it became evident that, in the research literature, these important considerations have not always been given due regard. Valid 101 interpretation of experimental data requires some understanding of the appropriate statistical methods. Thus, the purpose of this section is to provide the reader with a concise technical review of the statistical methods required for fatty acid compositional data used to compare microbial communities. From this review the reader should understand a number of important concepts that will be essential in following the experimental results. This review begins with a mathematical definition of fatty acid compositional data. Each sample or observation is often referred to as a fatty acid profile, or spectrum. Due to the interdependence between elements of a composition, a valid statistical comparison of fatty acid spectra requires that a logratio data transformation be made. Logratio transformation of zero values necessitates some special data handling. Spectra can also be compared by an analysis of similarity. A similarity index was derived based on the research literature. Although this index of similarity can be used to discern compositional differences, specific patterns of change within the fatty acid spectra can not be identified by this approach. One method to interpret the fatty acid spectra is in the consideration of certain marker fatty acids. Principal component analysis (PCA) is a multivariate statistical tool that can also facilitate a greater level of data interpretation. Logcontrast principal component analysis, the method of PCA for compositional data, can be used to summarize the compositional variability for a number of observations from one sample population. A sample population is defined by a community of microorganisms contrained by a set of environmental and growth conditions. However, in the literature and for the present investigation, it is of interest to assess the variability between sample populations. Discrimination between two or more groups of sample populations of compositional data, can be accomplished by logcontrast canonical component analysis. In this section of the introduction, the above mentioned concepts are reviewed. This literature survey of fatty acid compositional analysis, has been divided into five sub-sections, namely, (1) Compositional Data and Logratio Transformations, (2) Analysis of Compositional Similarity, (3) A n Introduction to Marker Fatty Acids and Chemotypes, (4) Logcontrast Principal and 102 Canonical Component Analysis, and (5) Summary of Fatty Acid Compositional Analysis. Compositional Data and Logratio Transformations For this discussion, an experiment is considered to consist of N samples (observations) taken from any one of M culture conditions. Therefore, the fatty acid data are a collection of N fatty acid profile observations for M conditions where N is not necessarily the same for each condition. Statistical analysis is required to assess the intra and mter-condition variability. Each observation X is a vector of D molar fatty acid concentrations which sum to give the sample total fatty acid (TFA) concentration: D (3-2) TFA = Xl+X2+X3+--- + X D = ^ X j v J where TFA quantifies biomass. To remove the influence of biomass in comparing one sample composition to another, the vector X is first normalized by TFA to produce a compositional data vector, x, called a fatty acid profile or spectrum: _ r , J _ (3-3) x -jx,, x 2, x 3,..., xD j TFA Thus x is a vector of proportions with non-negative elements x ,.. .,x subject to the constraint: 1 D i>,.. <3-4> 7=1 The crude compositional covariance (a ) between elements k and 1 estimated from N kl observations of the vector x is defined as: 1 N 1 N <ru=cov{xk,xl}=——^{xkn-xkXxbi-xl\ Xj = — 2 A — 1 „=i iv „ = 1 XJn (3-5) However, elements of the composition are not mutually independent. By the compositional constraint of equation (3-4), the D-part composition is completely specified by a d-part subvector (x ,.. .,x ) where 1 d 103 J = D-1 (3"6) Mathematically expressed, the subvector forms a d-dimensional simplex, Sd, embedded in D- dimensional real space. The simplex is defined by: (3-7) Sd = {(xi,...,xD):xl >0,...,xD >0;x, + — + xD = l} A simplex is a restricted part of real space. The constraint imposed on the data by the simplex requires special statistical consideration due in part to the absence of an interpretable covariance structure based on the crude compositional data (equation (3-5)). Spurious correlation arises since i f one element of the composition increases, others by definition must decrease. Therefore, the correlation coefficient is not free to range from -1 to 1. Biologists have been warned of the prospect of spurious correlation between (normalized) indices since the turn of the century (Aitchison 1986). Aitchison addresses statistical analysis of compositional data with considerable mathematical rigour. Due to the particular requirements of compositional data, specialized computer programs were written (Microsoft Visual Basic in Excel 97 SR-1) to perform the statistical analysis in the present study. A summary of some of the important aspects of the mathematics underlying these algorithms is presented in the ensuing text. Since the study of compositions concerns the relative magnitudes of the elements rather than their absolute values, it is suggested that correlations be made on the covariance of element ratios (quotients). Mathematical awkwardness in manipulating ratios can be overcome by considering logarithms of the quotients. For a D-part composition x, the compositional variation array is defined by the expected value and variance for logratios of the composition: (3-8) h\i} = £{log(x,. fxj)\, T.. = var{log(x,. /x j ) } , i = l,...,d; j = i + l,...,D Estimates from the experimental data follow standard formulae for means and variances: P 1 V i ( I \ <3"9) 104 '.-jki^*').-*')1 <3"10) The estimated values can be presented in an ordered matrix from which obvious features of the compositional data are readily seen. The matrix lower triangle is used for the logratio means and the upper triangle reports the logratio variances as follows: • *12 • ^dD <32D (3-11) Information such as relative compositional magnitudes (cj > 0 or cj < 0 ) or the relative variations (x > or x < is gathered directly from the ordered array. The covariance structure of a D- part composition is completely determined by the VidD independent logratio (x ) variances. The U covariance structure is the set of all (3-12) °UM = cov{log(x, fxk),log(Xj lx,)\ as i , j , k and 1 run through the values 1,..., D. The variation matrix T = [x ] determines the u covariance structure by the relationships: (3-13) In addition to the variation matrix T, there are two other ways in which the logratio compositional covariance structure can be specified. The logratio covariance matrix is formed by choosing a common divisor, x , for all the logratios and by restricting the consideration of D compositional covariance to the set of covariances defined by the dxd logratio covariance matrix: s = \Pij.DD J = K J = cov{log(x,. lxD), log(x,. lxD)), i, j = l,...,d (3-14) 105 This covariance structure is contained by the set of relationships: (3-15) aijM  = C 7 i j + c r k l -Gil -Gjk The logratio covariance matrix, X, is the covariance matrix of a d-dimensional random vector of the logratio transformed compositional data: y; =\og{xi /xD), i = l,...,d The order of parts and the choice of component divisor should make no difference to the outcome of the statistical analyses performed. The third covariance matrix is the centred logratio covariance matrix, T. The common component divisor is the geometric mean, g, of the D components: «*)=<*•••><,>*> ( 3 " ' 7 ) The centred logratio covariance matrix is defined by: r = IrJ = cov{log(x,. /g(x)),log(x, /g(x))\, i,j = \,...,D and is similarly related to the covariance structure by the relationships: (3-18) (3-19) ® Ij.kl = Yij +Ykl - Yil ~ J'jk The matrix T is the covariance matrix of the random vector, z, of the transformed compositional data: , v (3-20) z, = log(x,. / g(x)), i = 1,..., D The three matrix specifications {T, X, T} are equivalent. The necessity for three separate specifications is one of mathematical flexibility, depending on the analysis performed. Each specification has disadvantages relative to the other. T is not a covariance matrix, X is not symmetric in the compositional parts and T is a singular matrix. 106 Zero values are an obstacle to logratio transformation (Aitchison 1986). One practical solution is to treat a zero as an analytical trace value, 8, below the method detection limit. For an observation x with D elements each zero value can be replaced by: D2 where O is the number of zero elements. The non-zero values must also be adjusted accordingly: t = x,-s°tepi- < 3 " 2 2 ) D2 Sensitivity analysis can be performed to test for any influence of the choice of 8 on the statistical outcome. Another strategy of zero replacement is to assign zero values a random number in the range (0, 8], where 8 is the compositional detection limit. The latter approach was used. Estimation of mean and covariance logratio transformed compositional data statistics follow standard definitions: r i 1 ̂  (3-23) — 1 „-i (3-24) A test for logistic normality can be made by considering the d-dimensional radius (Mahalanobis distance) distribution computed from the matrix equation: Pn =(yn-v)TZ-1(yn-Y), n = l,...,N ( 3 2 5 ) where the superscript "T" is the transpose operator. For the assumption of multivariate normality of the y observations, the radii p should be approximately distributed as %2(d). If the n n computed value of the radius p is r, then the sorted calculated probabilities Pr(p > r) plotted against the order statistic (2n-l)/2N should approximately follow the diagonal of the unit square 107 for logistic normality. Consider \|/ and E estimated from two sets of observations of fatty acid profiles identified as conditions A and B. Assuming that both are described by logistic normal distributions L d(\|/, E) A and L d(\|/, E), statistical tests are made of the likelihood (A) of the parameter 0 being valid for B the random logratio variable y. The likelihood that 0 is true for y is written "A(0 | y)". The logarithm of the generalized likelihood ratio test statistic (Q) contrasts the likelihood for the hypothesis (0 ) versus the model (0 ) parameter estimates: h (2 = 2 1 o g t o ) = 2 1 o g A(O j y) (3-26) The test statistic Q is approximately distributed as %2(c) where c is the constraint in calculating the statistical parameter 0. If the computed value of the test statistic Q is q, then the probability Pr(Q>q) is calculated from the incomplete gamma function as follows: ?r(Q>q)=l-^'y-^dW, T(B) = [W^e-dW, / * = | , , = f ( 3 " 2 7 ) The approximate generalized likelihood ratio test statistic (equation (3-26)), for examining hypotheses in the comparison of compositions, is of the form: Q = NA log(| XM\/\I,A\)+NB logfl Z „ | /1 | ) ( 3 " 2 8 ) where "I EI" is the determinant of the respective estimated hypothesis and model covariance matrices. Testing for equivalence of sample population variance (E = E ) requires the calculation of a A B pooled (E) sample population estimate: p 1 / \ (3-29) Z . « ( A . E . + A „ E „ ) v ' P NA +NB A A B B > The test statistic for the hypothesis E equal to E is calculated from equation (3-28) assigning A B 108 b o t h £ a n d £ to £ and the probability Pr(Q>q) is based on Y2(V2d(d+l)). A combined sample hA hB p estimate, £ , is required for the hypothesis of equal means and variances (\|/ = \|/ and £ = £ ): C A B A B In this case q is calculated by assigning both £ and £ to £ and the probability Pr(Q>q) is hA hB C based on %2(V2d(d+3)). Should the hypothesis (\|/ = \\f and £ = £ ) be valid, the combined A B A B mean fatty acid profile is estimated by the equation: ¥c " ( N A ¥ A + N B ¥ B ) ( 3 " 3 1 ) The hypothesis of equal means (\|/ = V|/) allowing for the possibility of £ different from £ A B A B does not have explicit forms of the likelihood estimates similar to the expressions of the pooled and combined estimate of £ given by equations (3-29) and (3-30). The likelihood estimates in this case are calculated following an iterative process: initialize: ^hA — ̂ A X a * = E * (3-32) iterate: ^M=^A+(¥A-¥HX¥A-¥HY Successive approximations of £ and £ are made until they change negligibly with further hA hB iteration. The test statistic q is then obtained from equation (3-28) and the probability Pr(Q>q) is based on X2(d). Therefore, with logratios of compositional data the standard battery of tests for the analysis of variance are readily made in order to compare groups of samples. In terms of comparisons of 109 microbial community structure from fatty acid analysis, one can ask whether two sample populations are the same. With a representative number of observations from a given sample population, it is also possible to test the likelihood of another sample being a member of the same community of microorganisms from the hypothesis of the population mean being equal to the unknown sample composition. Since fatty acid extracts from mixed cultures are an ensemble average of the condition and species of microorganisms within a sample, then a test of sample population similarity considers likeness in terms of both condition and content. Analysis of Compositional Similarity Microbial taxonomists use a different approach for identifying and categorising microbial isolates. In clinical microbiology, microorganisms are clustered according to an index of similarity (I) ranging from 0 to 1 for increasing likeness between profiles from two isolates s (Bousfield et al. 1983; Eerola and Lehtonen 1988). The similarity index is like a distance scale which can be used for cluster analysis (Aldenderfer and Blashfield 1984; Manly 1986; Spath 1980). Cluster analysis is a multivariate classification tool for applying a hierarchy to measurements by agglomerating the compositions into groups of nearest neighbours. Results of this analysis are illustrated with dendrograms. Although the index of similarity seems to have been exclusively used in clinical microbiology for discriminating between bacterial isolates, there is no reason why such an index could not also be a meaningful tool for assessing the relative likeness of mixed cultures. The idea is quite simple but it is complicated by the number of methods used to calculate I . A number of the s equations that have been used in clinical microbiology are reported in Table 3-4. Different equations of profile similarity show differences in the sensitivity of classification. Increased sensitivity to discriminate between two profiles is compromised by an increased probability of false differentiation due to experimental variability. Eerola and Lehtonen (1988) demonstrated the best results with an exponential function weighted by peak size (Table 3-4, row 6). From their data they concluded that for a mean peak area around 1%, the coefficient of variation (CV) was about 30%. As the mean peak area increased, the C V decreased to around 110 Table 3-4. Similarity measures for compositions xA and xB containing D fatty acids (Bousfield et al. 1983; Eerola and Lehtonen 1988). Similarity ranges from 0 (no likeness) to 1 (identical). Note that for the weighted exponential function, the parameters {k,, k2, k3}are empirical constants. Method Formula Correlation Coefficient 1+x ta" XA x*» ~XB)(VX ta ~ XA y Vxta ~XB)2 Angle of Separation of Vectors 2\i4xtA'Jxis Degree of Overlap 1 D Stack Method 1 D — > min-i I Weighted Stack Method min fx, A \x>» i. Weighted Exponential Function X ^ 4 ^ / M , « > = max fx, A ./(«>)= {e»-k2Y' -\ + k, 10%. Thus, the minor peaks of a spectrum tend to be more variable, adding merit to the weighted approach for calculating an index of similarity. Further, with an expected C V range from 0 to 30%, the cost in lost similarity for peak ratios between 1 and 1.3 should be small relative to the peak ratios greater than 1.3. Based on the observed within-species variability of fatty acid peaks, Eerola and Lehtonen adjusted the empirical constants {k , k , k } of their 1 2 3 similarity index (Table 3-4, row 6) to optimize their success in discrimination. However, the use of three empirical constants to adjust the width of a probability distribution function is excessive. The features of Eerola and Lehtonen's observations can be more simply expressed by defining a peak ratio random variable: v: = In , i = l,...,D (3-33) The variable v sampled from the same microbial species is assumed to be normally distributed i as N(0,O") over the range [ - ° o , +°°]. The standard deviation, a , represents the above mentioned i i coefficient of variability that increases with decreasing peak area. However, Eerola and Lehtonen were successful in using a globally representative CV of 30% which is the same as saying: 111 (T, ~ cr ~ ln(l.3), i = \,...,D  (3"34) Errors in this approximation are compensated for by making a weighted sum for the index of similarity. The hypothesis that two peaks, x and x , are equal can be tested for each v by iA iB i calculating the probability of finding another value even more removed from zero mean. If the observed value of v is <; then the probability Pr(v > | q |) is equal to: 2 -W F(s,,a) = Pr(v, > | ) = — = [ e2^Ur1 = l-erf (3-35) Based on equation (3-35) the weighted similarity index will be defined for the present investigation as follows: 1=1 ^ ^ IB J (3-36) Equation (3-36) preserves the basic features of the findings of Eerola and Lehtonen while presenting a more fundamentally-rooted calculation for a similarity index. The function is derived from purely statistical considerations and involves only one scaling term (a) versus the three empirical factors of the Eerola and Lehtonen equation. The standard deviation, a , could conceivably be used as a tuning parameter that adjusts the sensitivity of trend alarms in the application of microbial ecology assessment as a process control tool in wastewater treatment. While the statistics for compositions and similarity indexes discussed so far, are helpful in comparing microbial fatty acid compositions, they are not as informative when it comes to understanding any underlying patterns for differences in fatty acid profiles. It is therefore, of interest to identify trends in elements of the composition that relate to external causes. Hypothetically, fatty acid patterns that indicate early signs of sludge bulking or the onset of microbial stress would necessitate different operator interventions. Correct process control action depends on valid pattern recognition. The ability to identify deterministic factors influencing coherent patterns in the fatty acid composition could ultimately lead to meaningful 112 input signals for process control, using this approach of fingerprinting microbial communities. An Introduction to Marker Fatty Acids and Chemotypes Identification of a direct microbial response to changing bioreactor conditions can be interpreted from the identification of certain marker fatty acids. For example, direct isomerization of cis unsaturated fatty acids to their complementary trans configurations by aerobic bacteria is a reported, rapid response mechanism to growth inhibition due to a toxic shock (Keweloh and Heipieper 1996). Conversion from cis to trans configuration changes membrane fluidity and could be an important survival mechanism i f lipid synthesis is also inhibited by the adverse stimulus. The relative proportion of trans monoenoic (16: lQ)7t and 18:lco7t) to cis monoenoic fatty acids (16:lco7c and 18:lco7c) has been used as a so-called microbial stress ratio (Frostegard et al. 1997; Reichardt et al. 1997; Webster et al. 1997). Membrane fluidity is tuned to temperature, or to the adsorption of hydrophobic contaminants, by a change in the proportion of saturated to unsaturated fatty acids (Harwood and Russell 1984; Ingram 1976; Ingram 1977; Marr and Ingraham 1962; Pinkart et al. 1996; Rose 1989; Sikkema et al. 1994; Sikkema et al. 1995). Microbial stress levels can be monitored by following the ratio of poly-p-hydroxybutyrate (PHB) to phospholipid fatty acid concentrations (Nichols and White 1989; Zelles et al. 1994). PHB is a prokaryotic endogenous storage polymer that accumulates in cells when the absence of one or more essential nutrients prevents the complete oxidation of an otherwise available carbon source. Increased relative PHB levels indicate unbalanced growth. With the accumulation of research data on microbial lipids (Ratledge and Wilkinson 1988b), some patterns have emerged, although often with notable exceptions, identifying certain fatty acids with particular groups of microorganisms (Table 3-5). Most bacteria can be classified as either straight or branched-chain producers and chain-length specificity or geometry and position of unsaturation is a further distinction (Gillan and Hogg 1984). Therefore, instead of attempting to identify specific individual organisms from fatty acid compositions, groupings of marker fatty acids can be made to follow bacterial chemotypes or distinct chemical "families" of bacteria. This approach has been taken by a number of researchers, allowing distinctions to be made 113 Table 3-5. Marker phospholipid fatty acids in various cell types (adapted from (Basile et al. 1995; Kates 1964; Vestal and White 1989)). Eubacteria: Common markers Aerobes Anaerobes Stationary phase Gram-positive Gram-negative Sulphate-reducers Clostridia Barophilic, psychrophilic Cyanobacteria: Actinomycetes: Fungi: Protozoa: Plants and Green Algae: Diatoms Green algae Microalgae Higher plants 15:0, il5:0, 16:1(09, 16:lco5, il7:0, al7:0, 18:l(07t, 18:loo5, il9:0, al9:0 16:lG>7, 16:lco7t, 18:lco7 cyl7:0, cyl9:0 cyl7:0, cyl9:0 il5:0,al5:0 POH 14:0 il7:l©7, 10Me-16:0, 17:lco6 c y l 5 : l 20:5,22:6 same as eubacteria plus 18:2co6 same as eubacteria plus 10Me-18:0 16:0, 18:lco8, 18:2co6, 18:3co6, 18:3co3 20:3co6, 20:4co6 16:l(o3t, 20:5co5, 20:5(o3 16:lcol3t, 18:3co3, 18:lco9 16:3co6 18:1(09, 18:1(011, 18:3(03, 20:5ax3, 26:0 between communities in time and space (Baird and White 1985; Bobbie and White 1980; Cavigelli et al. 1995; Gillan and Hogg 1984; Sundh et al. 1997; Webster et al. 1997; Zelles et al. 1992; Zelles etal. 1994). While it is interesting that fatty acid composition can be used to discern community differences, the results beg answers to questions of "why" and of "what impact", that do not seem to have been addressed in the literature. Cause and effect relationships are difficult to establish in open environmental s