UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Kinetic response of the sequencing batch reactor Ross, Philip Robertson 1988

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


831-UBC_1988_A6_7 R64.pdf [ 6.46MB ]
JSON: 831-1.0097834.json
JSON-LD: 831-1.0097834-ld.json
RDF/XML (Pretty): 831-1.0097834-rdf.xml
RDF/JSON: 831-1.0097834-rdf.json
Turtle: 831-1.0097834-turtle.txt
N-Triples: 831-1.0097834-rdf-ntriples.txt
Original Record: 831-1.0097834-source.json
Full Text

Full Text

KINETIC RESPONSE OF T H E SEQUENCING BATCH REACTOR by Philip Robertson Ross B.Sc. (University of Toronto. 1984) A thesis submitted in partial fulfillment of the requirements for the degree of The Faculty of Graduate Studies (Bio Resource Engineering) We accept this thesis as conforming to the required standard. The University of British Columbia October 1988 (c)Philip Robertson Ross 1988 Master of Science in In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of The University of British Columbia Vancouver, Canada DE-6 (2/88) Abstract The kinetic response of the sequencing batch reactor was investigated using 9 lab scale (5 liter working volume) acrylic reactors. The reactors were operated simultaneously in a 3 X 3 factorial experiment using temperature and feed strength as the major independent variables. The experiment was replicated over time. Response was found to be quite variable. The growth response of the activated sludge in the reactors was found to change both with elapsed time during an experiment and between experiments. This implies two fundamental limits on the calculation and application of simple kinetic theory to the SBR reactor and possibly to continuous flow systems as well. Calculation of some standard kinetic constants was performed and the confidence interval in the results calculated. The confidence intervals are quite large, indicating that numerical precision cannot be had with any realistic degree of reliability. The significant difference between dates may create a serious fundamental limitation in this type of experimentation and limit the scope of the results. Similarly, the variation inherent in the sampling process creates an absolute requirement for proper statistical experimental designs to determine the reliability of the results. i i i T a b l e o f C o n t e n t s T i t l e i Abstract i i Table of Contents i i i L i s t of Tables v L i s t of Figures v i i L i s t of plates . . . . . v i i i Acknowledgements i x Introduction 1.1 Process fundamentals 1 1.2 K i n e t i c studies 2 1.3 Objectives of study .5 Literature Review 2.1 Defining the activated sludge process ......6 2.2 K i n e t i c s and mathematical models 9 2.3 On the measuring of b a c t e r i a l growth 17 2.4 Ecology 21 2.5 Temperature e f f e c t s 30 2.6 SBR systems 32 Materials and Methods 3.1 Experimental design 39 3.2 Problems with HRT as a fac t o r 40 3.3 Revised experimental design 40 3.4 Rep l i c a t i o n through time 41 3.5 Detectable sources of v a r i a t i o n 45 Equipment 3.6 Overview 46 3.7 Reactors 46 3.8 A i r stones 47 3.9 Feeding system 47 3.10 Draw system . 50 3.11 A i r system ' 51 3.12 Temperature control 51 3.13 Feed composition 53 3.14 Inoculum sludge ....53 3.15 S t a t i s t i c a l analysis 56 i v Table of Contents (continued) Results and Discussion 4.1 Sample v a r i a b i l i t y 57 4.2 Regression a n a l y s i s . . . . . 6 5 4.3 K i n e t i c parameters 114 4.4 The e f fect of temperature 116 Conclusions 119 Recommendations 121 Literature cited 122 V List of Tables Table 1 Typ i c a l values of Umax and Ks of sludge grown on sewage 11 Table 2 Typ i c a l values of Umax and Ks of sludge grown on glucose based media 11 Table 3 An example of sewage treatment plant performance 20 Table 4 Comparison of the c h a r a c t e r i s t i c s of batch and continuous reactors 34 Table 5 K i n e t i c r e s u l t s of Braha and Hafner 37 Table 6 Feed composition 54 Table 7 Sample v a r i a b i l i t y from the beginning of the November 16 experiment 59 Table 8 Sample v a r i a b i l i t y at the end of the November 16 experiment 60 Table 9 Overall sample v a r i a b i l i t y for the November 16 experiment 61 Table 10 Sample v a r i a b i l i t y f o r the October 20 experiment 62 Table 11 Sample v a r i a b i l i t y for the October 28 experiment 63 Table 12 Sample v a r i a b i l i t y for the combined October 20 and October 28 experiments 64 Table 13 Regression analysis of dry matter growth data 81 Table 14 Regression analysis of v o l a t i l e s o l i d s growth data 82 Table 15 Regression analysis of % transmittance growth data 90 Table 16 Regression analysis of dry matter growth data October 5 and September 16 I l l Table 17 Regression analysis of v o l a t i l e s o l i d s growth data October 5 and September 16 112 List of Tables (Continued) Table 18 Regression analysis of squared dry matter growth data 113 Table 19 Regression analysis of squared v o l a t i l e s o l i d s growth data 113 Table 20 Regression analysis of squared % transmittance density s o l i d s growth data...113 Table 21 K i n e t i c parameters for the growth of dry matter 115 Table 22 K i n e t i c parameters for the growth of v o l a t i l e s o l i d s 115 Table 23 K i n e t i c parameters for the growth of % transmittance 116 Table 24 The e f f e c t of temperature on growth rates for the October 20 and 28th experiments 117 Table 25 The e f f e c t of temperature on growth rates for the November 16 experiment 118 v i i List of Figures Figure 1 Equipment schematic layout 42 Figure 2 Con t r o l l e r schematic layout. 43 Figure 3 Stages i n the sequencing batch reactor operation 44 Figure 4 Growth of dry matter fed 1000 COD media 67 Figure 5 Growth of dry matter fed 2000 COD media 69 Figure 6 Growth of dry matter fed 3000 COD media 71 Figure 7 Growth of v o l a t i l e s o l i d s fed 1000 COD media 75 Figure 8 Growth of v o l a t i l e s o l i d s fed 2000 COD media 77 Figure 9 Growth of v o l a t i l e s o l i d s fed 3000 COD media 79 Figure 10 Growth of % transmittance fed 1000 COD media 85 Figure 11 Growth of % transmittance fed 2000 COD media 87 Figure 12 Growth of % transmittance fed 3000 COD media 89 Figure 13 Predicted growth of dry matter fed 1000 COD media . 93 Figure 14 Predicted growth of dry matter fed 2000 COD media ; 95 Figure 15 Predicted growth of dry matter fed 3000 COD media 97 Figure 16 Predicted growth of v o l a t i l e s o l i d s fed 1000 COD media 99 Figure 17 Predicted growth of v o l a t i l e s o l i d s fed 2000 COD media 101 Figure 18 Predicted growth of v o l a t i l e s o l i d s fed 3000 COD media .....103 List of Figures (continued) Figure 19 Predicted growth of % transmittance fed 1000 COD media 105 Figure 20 Predicted growth of % transmittance fed 2000 COD media 107 Figure 21 Predicted growth of % transmittance fed 3000 COD media 109 List of Plates Plate 1 View of experimental equipment showing the reactors immersed i n the warm water bath 48 Plate 2 View of experimental equipment i l l u s t r a t i n g the feeding cassettes and support framework ..49 Plate 3 View of experimental equipment showing the Julabo c i r c u l a t i n g bath and the a i r flow control manifold 49 i x Acknowledgements I would like to acknowledge the effort and support given to me by my wife and friend Elaine Frances Wright. Without her considerate attention I could not have considered a project of this sort. My daughter Alexandra did alot for my emotional wellbeing during this period, reminding me what really matters. Erma and Valerie have also been great supporters in times of stress. Of course, none of this would have happened without the support and guidance of my committee: Dr. Richard Branion, Dr. George W. Eaton and Dr. K.V. Lo. My sincere thanks for all of the help they gave me. I'd also like to thank the guys of BioE : Dr. Ping Liao, Neil Jackson and Juergan Pehlke. Thanks guys. Its the friends you make along thg way that makes it really worthwhile. 1 Introduction Aerobic waste treatment is a biologically complex process involving several different types of living organisms performing distinct functions, the sum of which is the conversion of organic wastes from soluble and dispersed forms to insoluble and flocculated forms. The functional importance of several groups of bacteria, fungi, protozoans and other organisms has been investigated but has never been completely and reliably defined. 1.1 Process fundamentals The primary objective of the treatment process is the conversion of soluble, difficult to remove materials from the aqueous phase and the conversion of these materials into insoluble easy to remove matter. Colloidal and suspended materials will also be removed. Certain materials are not consumed biologically. The concentration of such materials will not decrease from influent to effluent (and may increase if they are the products of metabolism). The actual destruction of wastes is thought to be directly related to the uptake of soluble material across the membranes of cells or adsorption onto the surface of flocculant particles. The process seems most economically applied to wastewaters with a total BOD between 50 and 4000 mg/L (Grady and Lim, 1981; Metcalf and Eddy, 1979). The transformations that occur in the reactors are: the removal of soluble organic materials, the stabilization of insoluble organic matter and the conversion of soluble inorganic matter. The mass balance on the reactor can be used to determine the reactor volume required to treat a given flow of wastewater or the flow that can be accomodated by a reactor of specific volume. For a batch reactor of specific volume the reaction rate is equal to the rate of change of reactant concentration in the reactor. 2 Regardless of reactor configuration, all activated sludge systems have the following similarities: 1) they employ aerobic flocculent slurries of micro organisms. 2) the floes uptake soluble wastes from the waters and are removed by sedimentation processes. 3) microbes are conserved in the system. 4) reactor performance is affected by the mean cell retention time (MCRT) in the treatment system. The importance of the MCRT is undoubtably related to observations of successional stages occurring within the floe, with some of the resulting communities being more desireable from a treatment point of view than others. 1 .2 Kinetic studies Arrhenius (1889) derived equations describing the effect of temperature on the enzymic degradation of sugar. The Arrhenius equation is based on principles of physical chemistry, and has been verified as an accurate model of the effect of temperature on most chemical reactions. The usual form of the equation is: where (-E/RT) k = reaction rate ko = frequency factor R = gas constant T - degrees Kelvin E = activation energy Modified for application to the waste treatment field by Phelps (1944), the effect of temperature on reaction rate is still essentially approximated for a community of micro organisms by the effects of temperature on simple enzyme systems. However, this approach has 3 withstood many intervening years of application and experimental work. The Phelps version of the Arrhenius equation is k T = k 2 0 O<T-20> where krp is the reaction rate at any temperature T. k 2 0 is the rate at 20°C. O is the temperature sensitivity coefficient. Essentially the same situation exists for the modelling of microbial growth in biological waste treatment. Mo nod (1942) published a mathematical description of the growth rate of a pure bacterial culture as a function of the substrate concentration. He noted that other models could fit the data, but that it was both convenient and logical to use the same form as the Michaelis Menten equation describing enzymic reaction rates as a function of substrate concentration. A specific growth response can be characterized in Monod's model by specifying two model parameters: Umax and Ks. Therefore both the growth and temperature response of the waste treatment biotic community is estimated as a single enzyme system rather than as a complex system of micro organisms. The literature provides several experimental determinations of the kinetic response parameters Umax, Ks and O but they span a fairly large range. Although values are recommended for various treatment configurations, the reasons for the wide range of values are rarely conjectured. This approach has been satisfactory for continuous culture applications in which conditions are relatively stable and the growth response of the biotic community is limited only by the food supply. There is a conceptual basis for assuming continuity from single enzymes to cells: the idea of a fundamental growth limiting enzyme based reaction. However in sequencing batch reactors the reactor environment constantly changes with 4 respect to substrate concentration, dissolved oxygen levels and the accumulation of the products of metabolism. It is not clear whether the conventional approach to determining the microbial kinetic response can be applied to a dynamic system such as the SBR as there is probably not a single fundamental reaction limiting growth of the microbial community. It was therefore proposed to perform an experimental investigation into the kinetic response of activated sludge under the sequencing batch reactor operational protocol. The kinetic response of the sludge community and its variability were judged to be the major points of interest. The prevailing approach to reactor kinetic studies rely on steady state assumptions that are inherently not applicable to the SBR. Nonetheless, it would be convenient if comparable measures could be made to assess the relative performance in terms having general application in the waste treatment literature. An assessment of the degree of inherent variability in the process would also be useful to make specific recommendations about sampling and experimental procedures concerning the SBR and the scale of contingency factors appropriate in design calculations. The kinetics for most standard reactor type have been well explored in the literature both in theory and by experimental work. However, the sequencing batch reactor is fundamentally different from most of these reactor configurations in that almost every major chemical aspect of the reactor contents (ie food concentration, waste product concentration, dissolved oxygen concentration) changes radically during each cycle. It is not clear if the standard kinetic approaches apply to the SBR reactor, as the steady state assumption surely does not apply. 5 1.3 OBJECTIVES OF STUDY There are several aspects about the kinetic response of the sequencing batch reactor that have not yet been adequately documented. The effects of temperature and feed strength on the kinetic response of sludge populations growing under the sequencing batch operational scheme might not be comparable to continuous culture reactors due to the lack of a steady state of operation. Whether these factors interact with each other to modify reactor population reponse is also not well documented in either continuous culture or sequencing batch reactors. The standard expressions of microbial growth kinetics form the conceptual foundation of the operational guidelines for continuous culture reactors. Determining these parameters under the sequencing batch regime may have unforseen difficulties due to the unique nature of the reaction process. Similarly, the variance and therefore the limits of accuracy in the experimental process as a whole is an unknown quantity. The value of this is to give a quantitative assessment of the scale of contingency factor required for the sizing of SBR reactors based on kinetic factors alone. 6 Literature reveiw 2.1 D e f i n i n g the activated sludge system The development of the activated sludge process is usually attributed to Edwin Ardern and W.L. Lockett in 1914 (Grady and Lim, 1981). The activated sludge on which the process depends was observed to be brown, flocculant and with an inoffensive odor. Bacteria imbedded in a gelatinous matrix were observed to be the dominant organisms but the presence of a large variety of protozoans was also noted. Although not sure if the protozoans had a direct role in the process, Ardern and Lockett (1914) conjectured that the numbers of specific types of protozoans may indicate the "health" of the sludge. In a later paper, Ardern and Lockett (1915) describe the development of sludge populations. Building up sludge was observed to take 6 months or more, so it was advised to use an inoculum from a working plant. A fill and draw program was applied in these initial activated sludge experiments, which is essentially the same as the sequencing batch reactor operational protocol. Ardern (1917) states the fill and draw process has produced good effluents through winter (in Wisconsin) despite sub zero air temperatures. Treatment required more air but process efficiency reached reached 90% waste removal in Febuary, March and April. Ardern (1917) stated that the short comings of the fill and draw system are high requirements for operator attention, the build up of sedimented solids in the tank and the likelihood of aerator clogging due to periodic burial in the sludge blanket. O'Shaughnessy (1923) noted that continuous flow waste treatment systems require twice as much time to effect a given degree of purification as "fill and draw" methods. It was hypothesized that the 7 reason for this was the short-circuiting and hydraulic streamlining that interferes with mixing. The net effect was thought to be a reduction in the rate of contact between micro organisms and substrate. It was realized that unicellular life forms probably absorb much of the soluble carbonaceous material while protozoa were observed to be ingesting particles in the sludge. O'Shaughnessy (1923) proposed some modifications to the activated sludge process. These included a constant ratio of influent to sludge (now called the F:M ratio) and a short aeration period (4 to 6 hours). This removes most of the BOD, the remainder taking much longer. This reduces the mass of sludge required and produced, reduces the oxygen requirement and avoids the "useless destruction of lower organisms by higher organisms during prolonged action". The latter is undoubtably important in acheiving the low bacterial effluents. The process was quite variable; the duration of aeration required for treatment ranged from 18 to 24 hours in one plant and as little as 1 hour at another (O'Shaughnessy, 1923) . Jenkins (1942) also noted the variability in the biological activity of different sludges. He cites Reynoldson (1942) who related the activity of a sludge directly to the population of Vorticella species. Pillai and Subrahmanyan (1942) claimed Epistylis species to be the dominant ciliate for flocculation and flocculation accounts for up to 40% of purification of the waste water. Jenkins (1942) doubted both claims on the basis that protozoan populations differ radically without a concommittant effect on effluent quality. He speculated that bacterial activity may be affected by protozoan grazing but notes this is not confirmed. 8 Pillai and Subrahmanyan (1944) found that protozoans were absolutely required for proper functioning of the activated sludge process. Although they apparently believed that the role of bacteria was negligible, they did note that unsatisfactory treatment consistently occurred in sludge devoid of a healthy protozoan population. This agrees with the experimental results of Curds and Fey (1969). Many of the early biological studies of the activated sludge process name Zooglea ramigira as the organism primarily responsible for the characteristics of the sludge. Benedict and Carlson (1971) report a taxonomic look at sludge. Several genera were found to predominate, however Zooglea ramigira was not among them. McKinney and Horwood (1952) observed that several other bacteria besides Zooglea ramigira produced floes capable of rapid settling. The BOD removal rates of pure cultures of different bacteria was much less than sewage mixed cultures. Curtis and Curds (1971) state that Zooglea ramigira may in fact be an assemblage of different types of bacteria. They cite Friedman and Dugan (1968) and Prakasam and Dondero (1967) both of whom doubt Zooglea ramigira is in fact a proper species. Cluster analysis was used to determine the degree of association between species. Surprisingly, one of the closest associations was between Zooglea ramigira and Sphaerotilus natans, the organism often blamed for bulking in sludge. Heukelekian and Schulhoff (1938) investigated several types of bacteria isolated from activated sludge for their ability to clarify suspensions. None exceeded the rate at which unsterilized sewage settled. Unsterilized sewage sludge was likewise found to remove the greatest proportion of the suspended and soluble carbonaceous material. 9 Heukelekian and Littman (1939) note that although sewage bacteria are strict aerobes they are surprisingly resistant to anaerobic periods, surviving over 250 hours at 30°C in anaerobic containers. Growth was reported as slow. They tried without success to induce bulking in the pure cultures. The pure cultures settled faster than sewage. Lloyd (1944) noted the similarities between the sludge community and swamp or mud flat biota. 2.2 Kinetics and mathematical models Mathematical models provide a conceptual framework as well as some ability to predict responses in given situations. Monod (1949) published the most widely applied mathematical model of bacterial growth that is used in waste treatment. The Monod model is u= Umax * [S]/(Ks + [S]) Monod (1949) where u is the rate of growth. Umax is the maximum rate of growth. [S] is the concentration of substrate available for growth. Ks is the concentration of substrate at which growth proceeds at half the maximum rate. Monod (1949) noted that batch culture experiments involve a changing medium concentration and qualitative medium change. His choice of model is almost arbitrary and he noted that "several models could fit the data but .that it is conceptually convenient and logical to adopt a hyperbolic equation of the same style as the Michaelis Menten equation". In essence this treats the growth of a bacterial cell as analogous to an enzymic reaction; a master reaction that is the rate limiting reaction for growth (Roels, 1982). J 10 Monods equation often gives quite a satisfactory description of both batch and continuous culture growth. Many other expressions have been suggested due to descrepancies and short comings of different kinetic models: du/dCs = k(Umax-U)P Konak, (1973) (3) U = Umax{l-exp(-kCs)} Tessier, (1936) (4) • Cs = U*A +U*B/(Umax-U) Dabes et al (1973) (5) u = (Umax[S] / (Ks + [S])) * (l-[S]/[Sm])n Luong (1987) (6) However, both the waste water and the culture are complex hetrogenous mixtures. It has been suggested that a range of kinetic parameter values should be considered to characterize growth (Grady & Lim 1981). The values of the kinetic parameters for growth and substrate removal are strong functions of the type of organisms present and the quality of (i.e. substrates present in) the waste. Within limits (rarely, if ever, defined) the rate of growth increases with temperature. This also affects the rate of decay and death which seem to increase as growth does. Grady and Lim (1981) compiled typical values for the Monod parameters for mixed cultures growing on sewage (Table 1) and on defined media (Table 2). There is a fairly large range of values in the literature on continuous systems even under apparently similar conditions. More unsettling is a complete lack of confidence intervals or error estimation by which the reader could make assumptions about accuracy and precision of the presented values or if they even refer to similar responses. Similar value ranges are available for the specific decay coefficient and the Arrhenius temperature sensitivity coefficient . The Table 1 Some typical values for the Monod kinetic parameters Umax and Ks as determined by different researchers for sewage sludge growing on sewage. Umax Ks Reference 0.40 60 P e i l and Gaudy 1971 0.46 55 P e i l and Gaudy 1971 0.16 22 Benedek and Horvath 1967 0.55 120 Jordan et a l 1971 Table 2 Typical values of the Monod parameters for activated sludge growing on defined glucose based media as determined by different researchers. These are all laboratory studies under contolled conditions. Note the range of values for both of the kinetic parameters Umax and Ks. The table is taken from Grady and Lim (1981). Umax Ks Substrate Source 0.31 - 0.77 11-181 Glucose Gaudy and Gaudy 1971 0.69 26 Glucose Chiu et a l 1972 0.18 — Glucose Garrett and Sawyer 1952 133 Glucose Tench and Morton 1962 0.36 103 Glucose Muck and Grady 1974 0.38 - 0.49 11-29 Glucose P e i l and Gaudy 1971 0.35 8 Glucose Gates and Marlar 1968 conclusion can only be that the kinetic parameters can take on a wide range of values, especially as that is specifically reported by two of the reports. Substrate variability can be discounted as an effect as artificial feed was used in all of these studies. This leaves biotic differences in the sludge populations as the most likely cause of the discrepancies in results. In summary, there is a wide range of values reported in the literature for the Monod kinetic parameters Umax and Ks and the Streeter 12 Phelps temperature sensitivity coefficient. However, there is little to assist the reader at selecting a logical value or range of values. This problem is exacerbated for researchers investigating different reactor configurations for which the conventional theory might not apply. Despite widespread use and general acceptance, the Monod model does not accurately describe observations of ciliate populations in real situations (Sambanis et al, 1987;Bungay and Bungay, 1968; van den Ende,1973). It has already been noted that ciliates represent an important component of the sludge biotic community {e.g. Pillai and Subrahmanyan, 1942; Ardern and Lockett, 1914). Although ciliate reproduction seems to follow the basic Monod model (or something like it) mathematical simulations of reactor populations based on these models lead inevitably to extinction of one component of the system. Observations lead to stable equilibria. The concept of refuges was suggested as a possible explaination of the results. Wall attachment was postulated as a refuge for bacteria from ciliate predators. Williams (1967) model of cell growth is interesting because it is based on the idea that a cell has functional compartments: one for the synthesis of new cells and the other for the synthesis of structural and genetic material in the cell. The latter must grow to a critical point before reproduction occurs. He points out that surface area related feeding mechanism for bacteria leads one to conclude that spherical bacteria (e.g. cocci) should grow as the 2/3 power of their volume, while rods would grow exponentially. Williams (1967) model assumes that certain properties of cells change with time ie the cell size,the cell composition etc which does agree with observations. The model responds to perturbations as many 13 pure cultures do, however the fit is much better for continuous culture systems than for batch. The model does not allow for saturation phenomena such as Monods equation. Furthermore the model does not produce oscillations or damped oscillations observed in many situations. This is the price of simplicity, as the author states that oscillations can be produced if a cellular age distribution is included but that considerably increases the complexity of the model. Webster (1983) notes several problems with the use of and determination of the standard Monod parameters Umax and Ks. He is skeptical of the use of single batch experiments to estimate parameters for continuous systems (and presumably, vice versa). Micro organisms in batch culture experience constantly changing conditions and growth patterns are probably not representative of continuous culture systems. Further, the interpretation of single batch studies involves some extrapolation or integration, which is likely to compound errors in the data. Physiologically, the microbes are likely to be themselves changing in response to increasingly adverse conditions. Sheih and Mulcahy (1986) state that determining the intrinsic kinetic coefficients is the fundamental problem with all mathematical models. The often assumed steady state for reactor operation might not really be achieveable. Webster (1983) cites Herbert (1962) who noted that the saturation constant (Ks) seemed to decrease with the age of continuously cultivated lines of bacteria. This is a predictable result of selection for faster growth in the chemostat. The determination of growth is also prone to greater error at high and low levels of substrate, limiting the range over which accurate determinations can be made. Chase (1976) reports that the Monod model was developed to describe growth of a 14 system that is at steady state. Subjected to changing conditions, several researchers have noted that the growth of micro organisms lag behind the levels predicted by the Monod equation in response to changing feed concentrations. Perret (1960) called this "growth rate hysteresis". The proposed explanation has 4 hypotheses: 1) Substrate utilization depends on enzyme concentration and substrate concentration. 2) Enzymes decay at a given rate. 3) For any substrate concentration, there is an optimal enzyme concentration. 4) If the concentration of enzymes is greater than the optimal concentration, enzyme synthesis ceases, so there is a lag between synthesis and use. The Monod model was applied by Curds (1971) in a five component computer simulation model of the activated sludge process. The components were sludge bacteria, dispersed bacteria, flagellated protozoans, free swimming ciliates and attached ciliates. The growth response was expected in all cases to follow Monod kinetics for substrate limited growth. Kinetic constants are given for the model. It is assumed that the ciliates feed upon the bacteria. The results of the model conform well to observations of sewage plant operation and the resulting changes in the populations. The absolute levels of the various components do not correlate well with effluent quality. The protozoan populations in the model do oscillate as observed populations. The model predicts clear effluent in several situation were it does not occur. A number of other mathematical approaches have been attempted by various other disciplines for simulating interacting groups of 15 organisms. The Lotka Volterra equations used by many ecologists to model two (or more) interacting populations have the form: For population dXx/dt = ( a x + B n X 1 + B 1 2 X 2 ) X x (7) For population X 2 dX2/dt = ( a 2 + B 2 2 X 2 + B21X±) X 2 (8) This indicates that the change in population of a given species is related to some population density characteristic of the species, a population density characteristic of the second species and the present population density. The terms Bjj are functions that alter the growth the of the population in some way. For example if a^=r and B-Q = -r/k, then equation (7) becomes dXj/dt = ( r-(r/k)Xx)* (9) Where r is the intrinsic rate of increase of the population and k is a function of the environmental resources available to the population. This is the Verhulst Pearl equation relating population change to the intrinsic population growth rate, the carrying capacity of the environment and the present population density. The Monod equation has been incorporated into the Lotka Volterra predator-prey equations by several authors (e.g. Tsuchiya et al, 1972; Dent et al, 1976 ; Canale et al, 1973) and observed results agree qualitatively with those predicted by the model. The result predicts three potential responses of predator populations based on the loading rates of the systems. At very low loading rates, the rate of increase of the prey populations is low and the predator population oscillates depending on the availability of prey. As the loading rate increases, the oscillations become damped and approach an equilibrium. At high loading rates, the predator 16 population population approaches equilibrium smoothly and monotonically (Bazin,1980). A similar finding was reported by Rashit and Bazin (1987). Using a protocol similar to a sequencing batch reactor, they question if cyclical environmental perturbations lead to higher or lower biological diversity. The result depends on the scale of the perturbations relative to the growth of the species in the environment. Although not directly concerned with waste treatment the question has major implications for sequencing batch reactor operation. Goodman and Englande (1974) reveiwed the mathematical models of Eckenfelder (1954, and modified extensively since) and McKinney (1962 and also revised every other year or so) found them to be essentially identical. The author states that modifications of Monods' basic model have been made because attempts to apply it to activated sludge have had problems. Garrett and Sawyer (1952) are cited as examples. Drew (1981) states that balanced growth means that the cellular components (proteins, DNA, lipids, etc) increase at the same rate as cell mass or cell number. Unbalanced growth means that one or more of these components increases independently of the others, which is often indicative of a stress or change in the bacterial metabolism. Drew (1981) states "The techniques used to measure growth obscure the fact that all bacterial cultures are grossly heterogenous and really monitor average values that describe the growth of a population rather than of individual cells." He also notes the inherent variation in even pure cultures, and that most microbiological techniques are designed to minimize the heterogeneity and natural variability of these populations. Monod was cited as a description of balanced growth for simplistic bacterial systems when the substrate concentration is much greater than Ks. This obviously does not apply to all systems. Slater (1981) states "pure culture growth systems are highly unrepresentative of almost all the habitats which support the growth of micro organisms." It is his belief that there are probably important properties of microbial communities that have been overlooked due to the emphasis on pure culture work. Slater (1981) emphasizes ecological descriptions of microbial communities. They may be complex and intimate connections or simply interacting assemblages. Several terms from the literature are mentioned (consortia, syntrophic associations, synergistic assemblages, etc). He mentions that several reports of pure cultures have turned out to be tight associations of bacteria. 2.3 O n the measuring of sludge growth Ardern and Lockett (1936) stated that the most important test is microscopic examination of the sludge. They offer a characterization of sludge conditions based on the types of organisms present. Ideally there should be few flagellates and suctoria, amoebae rarely, and a preponderance of ciliates. Koch (1981) discusses the use of optical density, dry weights and conversion factors for the estimation of bacterial numbers from optical or dry weight data. Optical methods are quick and easy but can lead to errors, as they are indirect and involve assumptions about relationships that probably change over time (e.g. size to mass ratios, number of viable cells relative to the amount of cellular debris in solution). Calibration and estimation of error is strongly recommended. However, two equations are given for the relationship between bacterial numbers 18 and light absorbance at 2 wavelengths. The author goes on to a discussion of fundamental concepts in statistics, sample distributions and comparative tests which are absolutley required for quantitative work. The requirement for statistical approaches to quantifying bacterial growth should be emphasized. It is not clear if the response of sludge (with greater biological complexity) can be expected to be easier to quantify. Bacterial biomass is usually measured indirectly or is estimated by a number of techniques (Norland et al.,1987). The approach is easy but relies on assumptions that cannot be easily tested or calibrated. This paper details the direct measure of 337 bacterial volumes and subsequent measure of dry weight. The correlation coefficient was 0.946 and the standard error was 0.20. Small bacteria show a higher proportion of dry matter than larger bacterial cells. Peil and Gaudy (1971) claim to have verified the general applicability of the Monod model and the use of kinetic constants determined on glucose based substrates for estimation of sludge growth on sewage. They do take care to acclimatize the microorganisms in their study, although their techniques probably select for a small subset of the sludge community. Whether or not the culture resulting from their acclimation techniques is representative of sludge may be doubted. The color and texture of the sludge varied over time and no direct observations are reported. As they used a completely defined glucose-mineral salts medium it is possible that the nutritional needs of the entire sludge community would not be met. They state that "for most experiments d i f f e r e n t values of Umax and Ks were obtained i n experiments c a r r i e d out at d i f f e r e n t times". The experimental protocol reported did not include an estimation of error or variability, just 19 the statement that the Monod model provides a satisfactory description of the process. It is also noted that Umax and Ks cannot be considered as precise values as the system is subject to large fluctuations. The assessment that glucose estimates performance on sewage is not tested. It is apparently an "eyeball" judgement of the growth curves. Chiu et al. (1972a) modified the Monod model to include a decay term and used it to describe the growth of micro organisms in a continuous flow reactor. However, the saturation constant (Ks) changed with feed strength (higher at high feed strength, lower at lower feed strength). This was conjectured to be the result of a shift in microbial populations in the reactor vessel. The authors also note that extrapolation of batch culture estimations of kinetic parameters is prone to error, largely due to the heterogenity of the sludge community. Chiu et al. (1972b) reported that decay rates were very important at lower feeding rates. The kinetic models of Monod, Moser and Contois (all modified to include a decay term) were significantly better at describing growth responses of sludge than other models. The authors state reproducible results can be obtained in continuous flow systems. Konak (1974) claims to have derived general model describing growth as limited by any nutrient. The models of Monod and Tessier are shown to be special cases of this generalised form. The author also declares the saturation constant (Ks) to be the reciprocal of the maximum growth rate (Umax), a result that apparently surprised its author. The author does note that at low feed concentrations the relationship is adequately described by a straight line. At higher feed concentrations he suggests the use of Monod or (preferably) his own general model. 20 Ostojski (1987) attempted an ecological model on a sewage plant in France. Eckenfelder's and Naito's equations (both modifications of Monod's model) were used to estimate component functions. Estimations from actual operational data and the confidence interval on the functions were calculated. The results seem to indicate standard deviations as high as 46% of the value of the means. Table 3 presents a reduced summary of the data for the actual plant operation. Table 3 Summary data from Ostojski (1987). The data come from a sewage treatment plant in France. Of special interest is the level of variance on feed strength (influent concentration) and the sludge concentration in the treatment plant. Influent feed 152.3 mg BOD/1 +/- 43.7 Sludge con'c 3462.2 mg/1 +/- 613.7 Outflow 24.1 mg/1 +/- 10.7 Antheunisse and Koene (1987) observed that the bacteria recovered from a 2 stage treatment plant changed with conditions in the plant. The experiment was not replicated. Different genera seem to reach maxima on different dates. Roels (1982) is in favor of unstructured models for describing microbial processes as micro organisms adapt and adjust to their environment by a number of mechanisms. He includes mass action law functions (e.g. Monod's relationship) as they are compelling descriptions of the speed with which micro organisms respond to specific stimuli. Also important though are the various regulatory functions of cells ie enzyme activity adjustments, cellular composition changes and evolutionary change through selection processes. Present experimental 21 approaches do not allow for any estimation of these latter processes and there importance cannot be estimated. O'Neill and Gardner (1979) state that most ecological models are deterministic and fail to account for or assess three specific types of error and uncertainty: 1) model bias or error in model structure. 2) measurement error or parameter uncertainty. 3) natural variability of ecological systems. The analysis given in this paper is fairly brief, but emphasizes that response to a given variable is not really likely to be identical from year to year. Environmental variability, spatial heterogeneity in the environment and genetic variability are some specific causes for the differences. Model structure and measurement error are connected problems that may be resolved through application of error assessment procedures. 2.4 Ecology Hynes (1970), discussing the adaptation of benthic invertebrates, cautions that laboratory work often produces conclusions that are misleading in ecological terms. As an example he cites the impact of temperature on physiological rates, which can vary with the season even when temperture is completely controlled. Several investigators are cited who found the effect of temperature was different for organisms of different states of maturity and size. There are several observations about temperature effects that contradict the predictions of the Arrhenius temperature relationship: most fresh water environments are quite active biologically when the water temperature is close to 0°C. 22 Hynes (1970) also notes the importance of temperature in planktonic population upsets. Certain species of algae and diatoms undergo explosive population growth only when specific factors allow, the most notable of which is temperature. One of the most compelling points about planktonic communities is the seasonal succession of species that occurs. It has been found where ever it has been looked for. Additionally, research indicates a diurnal cycle of activity, with different species reaching daily maxima and minima at different times. Jackson and Berger (1984) found the survival of starved ciliates differed between species. The conditions under which starvation occured also seemed to affect the rates of survival studies. Crowded ciliates died off faster than uncrowded ciliates and the rate increased at higher temperatures. The rate of die off displayed a different temperature relationship than respiration or growth. Abrams and Mitchell (1978) investigated the survivability of nematodes at very low O 2 tensions. The conclusion is that nematodes can survive very low O 2 for short periods. They mention that nematode populations can be quite high in activated sludge and that populations peak and decline as sludge ages. The nematodes in this study became inactive if the P O 2 dropped below 5 mm Hg. It is often assumed that ecosystems tend to be more stable as species diversity increases. While this is a common idea in ecology, Bazin (1980) questions whether it must be so. Most of the basis for rejecting the notion comes from complex mathematical models. A definitive solution is not presented one way or the other in this paper. Rather he finishes by saying that stability is not an inherent property of complex systems but seems to be the result of specific interactions. The 23 question of species diversity as a function of environmental stability was tested somewhat by Rashit and Bazin (1987). They found that periodic disturbance could increase or decrease diversity depending on the duration and magnitude of the upset. This is directly relevant to the operation of an SBR as each cycle involves major physical and chemical changes within the reactor. Roszak and Colwell (1987) reviewed bacterial survival. Several investigators believe that any indirect method of enumerating bacterial biomass is inherently prone to error. The traditional exponential phase, stationary phase and population decline form of bacterial growth is modified to agree with direct observations: the die off is substantially slower than previous estimates. The variability of bacterial morphology, physiology and standard "life cycle" is also discussed. Many of the conventional ideas of the typical life cycle are now being criticized due largely to discoveries of high adaptation and biological flexibility. Unbalanced growth is one such concept where the bacteria grow in size but do not replicate, leading to filamentous forms. Vast differences in morphology apparently accompany transition from environments, that is culturing under one set of conditions and then transfer to another. The lab is emphasized as being a very unnatural environment for growth. Organisms interact in a number of ways, the definitions for which are not universally agreed upon. For the present discussion it is sufficient to say that organisms can have positive, neutral or negative effects on the populations of other organisms. Of course the effect one organism has on another may be quite independant of the effect the second has on the first (Slater, 1981) The trend of population change in a two species model leads to 3 posible scenarios. In the first, one species always 24 wins over another due to some inherent ability. The second possibility is that some stable balance will occur between the populations. This level may change due to environmental factors, and may even be completely altered by the environment (Birch, 1953). The final possible outcome between two interacting populations is that an unstable equilibria occur with several metastable areas being possible between the two populations. This type of relationship leads to either species winning out or the populations oscillating depending on the relative densities. Gause (1934) performed a number of predator prey studies. Although he did not observe the periodic predator-prey population oscillations predicted by the Lotka Volterra equations, Gause did explore some fundamental concepts such as refuges for prey from predators and immigration. The same type of work was extended by Huffaker {et oi.,1963) to include greater environmental complexity. In this situation, the predator and prey populations did oscillate as predictied by the mathematical models. Most of the theory concerning population interaction assumes consistency in the properties of the populations. This involves inherent (or genetic) characteristics as well as population parameters (such as yield or Monods saturation constant,etc). It has been shown (e.g. Pimentel et al., 1963) that significant evolutionary population change can occur in 15 to 20 generations. In bacterial systems, these levels of population turnover.can occur in less than a day. Another assumption commonly made in waste treatment theory is spatial homogeneity of the microbial populations. Although this may be a fair assumption for the mass of the liquid in the reactor vessel, it does not take into account the possible effect of the vessel walls or surface 25 scum layer as colonization sites or refuges for organisms. It is possible that a slight decrease in predator grazing efficiency would allow a prey species with a constant reproduction rate to escape from predator population control. Megee (et al.,1972) investigated a chemostat culture of Saccharomyces cerevisae and Lactobacillus casei. The Lactobacillus casei required riboflavin produced by the yeast in order to live. In such a case, the kinetic response of one species is totally dependant on the population of the other. In an system as biologically complex as a waste treatment reactor it seems likely that some species will inhibit or enhance other species and thus cause fluctuations in kinetic response. Wimpenny (1987) emphasizes the importance of heterogenous environments in modelling microbial populations. He differentiates between model system experiments and microcosms. Of course there is some continuity from the model to the microcosm and then to the real world. A number of experiments are recounted, many having qualitative agreement with field observations. As realism is a consistent concern with models of all kinds, this sort of veiw is necessary. Realism cannot be assumed, it must be verified. In this regard, most waste treatment experimentation is aimed at assessing the qualitative effects of the waste stream on the kinetics of the microbial populations. Not considered directly is the impact of the experimental equipment or the differences between the experimental conditions and those of the environment to which the results will be extrapolated. Hawkes (1963) emphasizes that waste treatment is acheived by a dynamic balance of a multitude of microbial, fungal, insect and protozoan populations. He discusses trophic levels and the ecological roles of different organisms. The diversity of habitats within a well mixed reactor are covered as well (ie on the floe particles, within the floe, dispersed in the aqueous phase, adhering to particles and to the walls of the vessel, etc). As in Hynes (1970) flow rate and food are dominant factors in affecting species occurence and distribution. Comparisons are made between communities found in. solid matrix treatment configurations and aqueous based systems {ie activated sludge). Essentially everything that occurs in a biological treatment process is the result of its ecology. This includes efficiency, sludge growth, system stability and system failure occurences such as bulking, effluent quality problems and washouts. The ability to withstand shock loads, flow variability and toxic compounds also relate to the fundamental characteristics of the biota involved with the process. Curds (1965) discussed the significance of protozoans in the activated sludge process. The implication is that specific roles must be filled for the treatment process to function properly. He discusses the maturation of sludges and the possible meaning of the successive types of protozoans that are found to occur. Specific groups are found to follow a sequence with sludge age. Sludge age is also consistently used by engineers as an operational parameter important in quality control in the treatment process. It is Curds' opinion that the successional stages reflect changes in food quality. He cites Hetherington (1933) who also observed successions of protozoans (in the laboratory) and postulated that the protozoan sequence followed a bacterial successional sequence. That different protozoans feed on different bacteria may have implications for batch reactors where the food quality changes throughout each cycle. The 27 concept of protozoans as indicators of treatment efficiency and of sludge health is also broached. The meaning of this succession is not clear for experiments on kinetics of activated sludge. If the sludge populations change significantly it is not logical to expect that the kinetic parameters must remain constant. It is logical to expect a given population to respond consistently to a given stimulus, but as sludge is generally biologically complex and undefined this assumption may not be applicable to all sludges. Woombs and Laybourn-Parry (1986) found that the effect of nematodes on the biofilm community appeared to be different at different locations in the biofloc. This emphasizes that the environment in narrow films is heterogenous enough to have a measureable effect on biological activity. Sladecek (1971) found a sequence within the genus Vorticella based on the food preferred by each species. Vorticella species have been described in this way by other authors (e.g. Curds, 1971; Bick, 1969) and the results agree that the types of sessile ciliates found in water streams can be indicative of the degree of pollution or purification. The bottom line is that even relatively slight differences in food quality favor detectable differences in the biota that survive in the reactor. McKinney and Gram (1956) comment that not understanding the functional role of protozoans has led to poor quality designs and operation of sewage treatment plants. The concept of sludge development is discussed and its relationship to treatment efficiency. A basic reveiw of the types of organisms associated with various levels of treatment is given. The importance of the operator knowing what 28 organisms are present is stressed. Quote: "The major problem (in sewage treatment systems) is that the operator and the engineer fail to realize that the biological indices of operation reflect the complete and instantaneous condition of the system." This agrees well with the work of Baines (et al., 1953). They found that the numbers of attached ciliates changed over time during the year and that treatment efficiency tracked the changes very closely. When ciliate numbers increased effluent BOD decreased and vice versa. Free swimming ciliate populations did not correlate as well as the attached species. Data is presented from 1947 and most of 1952. Of special relevance to the present study is that in both 1947 and 1952 large changes in the relative numbers of various organisms occurred in November. This may be expected to have some impact on the kinetics, of the process. Treatment efficiency changed radically at the same time. Curds and Vandyke (1966 ) fed five common ciliate species commonly found in sewage treatment plants 19 different species of bacteria. Triplicate estimations of the ciliate reproduction rates were made over a period of almost 20 days (15 sample dates). The results show that ciliates (important members of the sludge community) strongly vary their reproductive rates over time even on consistent diets. Six fold increases occurred and three or four fold increases happened often and rapidly. The standard deviation of the daily reproduction rate was often around 10% of the value. Standard binomial theorem indicates the data range is therefore +/- 30% of the value of the reproductive rate. Some of the bacteria were found to be extremely toxic to different ciliates. Only Opercularia species were found to multiply rapidly enough to remain in high flow rate sewage plants, the authors claim this has been confirmed by 29 years of observation but cite no report. This implies that variability is a major problem in quantifying sludge growth response. Orhon et al (1986) report that although performance of a standard CSTR is well understood, treatment efficiency can be impaired by fluctuations in flow and loading rates. This is not such a problem with SBR reactor configurations due to greater operational flexibility. Topiwala (1971) presents data from a number of experiments on the effects of temperature on the kinetic characteristics of Aerobacter aerogenes cultures. Although unreplicated and performed in a single reactor over a period of time, the results indicate that K g decreases with increasing temperature, U m a x increases and the lag phase of growth is affected by both the direction and magnitude of temperature change. The author says that an Arrhenius type model fit the temperature effect relationship for the kinetic parameters. Chohji and Sawada (1983) state shifts in nutritional status result in transient unbalanced growth in E.coli cultures. However, shifts in temperature without a change in medium quality result in instant adaptation to the new conditions. The effect of temperature is dependant on the nutrient availability. Transient growth appeared to be caused by a difference in the physiological state of the bacteria. Shifts also seem to affect the lag time before the exponential growth phase. Stephens and Lyberatos (1987) report a computer simulation of two hypothetical bacterial species. The cycling of feed strength as a method of altering the growth rates is examined. If one bacterium has a higher growth rate than the other, it will outperform and dominate the culture. However, if the kinetic constants are such that there are concentrations of feed strength in which either bacteria grows 30 faster, then feed strength cycles determine which bacteria will dominate the final population. This is relevant to SBR reactor operation: it implies different conditions favor different populations. LeChevallier et al. (1988) examined the resistance of bacteria to a chemical stress (chlorine). It was found to be strongly influenced by adherence to the walls of the container. This increased the lethal concentration 150 times. Growth history was also an important factor (ie previous conditions and temperature). The results are said to indicate that selective pressures can rapidly cause changes in characteristics of bacterial populations. 2.5 T E M P E R A T U R E EFFECTS In their own experiments, Ardern and Lockett (1914) found that the efficiency of the sludge at removing sewage solids increased with time as the process was repeated. The results of several of their experiments are presented. They present some work on the effect of temperature and it seems to indicate that 30°C treatment is often less complete than treatment at 20°C or lower temperatures. The process seemed to fall off rapidly around 10°C. Hissett (et al. 1982) investigated aerobic treatment of piggery wastes at different temperatures using one fermenter run once at each temperature. Care was taken to acclimate the biota by growing them for a period (apparently 4 weeks, which means the whole experiment would take 10 months to complete). It seems as though the respiration rate increases with temperature to a maximum at 45°C. The lag period is likewise affected by the temperature, with 40°C having the shortest lag. The authors do not indicate whether they encountered any variability 31 between samples or if the sludge in their experiments was comparable throughout. Novak (1974) questioned the validity of the Arrhenius equation due to the range of values of the Streeter-Phelps temperature sensitivity coefficient O. Novak (1974) cites Zanoni (1969) who found O depends on many factors, among them the temperature range, the substrate concentration, food : micro organism ratio, type of substrate and the procedure for the determination of the value of 0. These are all factors that might be expected to have some impact on the community composition. Biological growth effects of temperature are often summarized by a value known as the Q20. O is equal to (QlO)®'^®. Novak (1974) says that O has values of close to 1.000 at very low substrate levels, but up to 1.18 at high concentrations of certain substrates. Novak (1974) doesn't seem to really consider that the sludge at one temperature may be biologically different from the sludge at a different temperature. Benedict and Carlson (1974) say that use of the van't Hoff equation between.20°C and 35°C introduces less than 10% error into the estimation of reaction rates but cite no proof of the statement. They say O has been found to vary between 1.000 and 1.250, with most estimations being between 1.000 and 1.100. They cite Monod's observation that several curves fit his observed growth response curves but that it is logical and convenient to adopt the Michaelis-Menten equation. In rough terms then, a bacterium is considered as analogous to an enzyme interacting with a substrate. They go on to consider all the possible outcomes of the effects of substrate on the parameters Ks and Umax, (ie that one, both or neither 32 increase or decrease as a result of substrate cone). Their first conclusion is that the response of O is highly variable and should be assessed in terms of the basic parameters Ks and Umax. Even then, they note that acclimation effects and other sources of variability can still interfere. Kalantar (1987) investigated the effects of two types of error on the estimation of Arrhenius' temperature coefficient. The author found that simple linear regression gave the poorest estimates of the parameter if the error was constant at different temperatures and non-linear, weighted least squares and weighted non-linear regression gave the best estimates. For a constant relative error, non-linear regression gave the poorest estimates. Thus, the accuracy of a calculation technique is affected by the type of error to which the data are prone. 2.6 S B R systems Irvine et al. (1980) report that due to the vast complexity of biochemical reactions in a waste treatment reactor and the possibility of overlooking a crucial variable with standard kinetic approaches, a general scheme to handle the potential complexity of the kinetics problem is required. Initially their interest lay in resolving the apparent conflict in nitrogen removal results in continuous flow systems in a fashion that would allow shed some light on the problem of operating an SBR to remove nitrogen. As the kinetic model is an important predictive tool for anticipating responses to given stimuli, an overly simplistic model cannot provide realism. As biological waste treatment involves extreme complexity and any one reaction can critically upset the system, the model begins with all relevant reaction equations (that have been related to nitrogen removal). They state that it is of potentially more use to have a complex model with every term in it than it is to have a quantitative model 33 with no realistic terms. The former leads to better understanding while the latter is a black box that sheds no light on functional relationships. This is a multi dimensional view; that many factors are relevant and important so they had best all be considered. Irvine and Richter (1978) state that improved performance is often observed with unsteady state, periodic processes in chemical engineering. They suggest equations to approach the kinetics of the SBR. They say that SBRs could be on the order of 60% of the size of CSTR reactors. Less would be possible, except that SBR design is for 20mg BOD/1 effluent in all cases, whereas CSTR are designed for 20 mg/1 average effluent. It is required to know the variability of the waste flow and the reaction rate coefficient with relative accuracy in order to design and control the process. Barth (1983) states that what Ardern and Lockett (1914) demonstrated was the advantage of retaining substrate adapted organisms to perform treatment. Table 4 summarizes the differences between batch systems and continuous flow systems (after Barth (1983)). Barth (1983) recounts some experimental results. He found that an SBR bulked under certain conditions but recovered when switched back to a more fundamental regime. It was hypothesized that the SBR could favor certain organisms by operating under certain regimes. Dennis and Irvine (1979) used the fundamental lack of a steady state to explain the hypothetical flexibility at handling transient loads for SBR systems. The results of varying the fill and react portions of the cycle indicate that the development of suitable biota is paramount to the operation of the process. As to the effects of the fill and react ratio, it isn't really conclusive due to other factors obscuring a clear relationship (ie 34 Table 4 A comparison of the principle characteristics of batch reactors and continuous systems (from Barth,1983). Batch systems Continuous systems Concept Time sequence S p a t i a l sequence Inflow pe r i o d i c continuous Outflow p e r i o d i c continuous Loading c y c l i c a l even ( i d e a l l y ) conditions v a r i a b l e i n time v a r i a b l e i n space D.O. varies constant Aeration intermittant continuous f l e x i b i l i t y high l i m i t e d flow pattern perfect plug completely mixed No recy c l e c l a r i f i e r required monitoring end of cycle continuous changing O2 levels during very long fill periods and short react periods). Experimentally they alter more than one factor at a time by the regime they applied. Irvine et al. (1983) report results of the Culver city full scale implementation of the SBR. The results indicate that full scale batch systems can be extremely efficient and that the previous industry avoidance of batch systems is not justifiable now that technology relieves the operator of constant process manipulation. Two years of results are presented (monthly averages). The results seem to show a cyclic rise and fall in effluent BOD and reactor biomass in the two reactor basins. Ketchum (et al. 1979) emphasize the flexibility of the SBR system. The ability of the SBR to control filamentous organisms, handle 35 fluctuating flows and strengths and operate until a given level of treatment is achieved is discussed. Ammonia and nitrate levels in the effluent stream can also be controlled with relative accuracy through adjustment of the operating parameters. The authors state that before a suitable operation scheme can be designed the kinetics and potential byproducts of the process must be investigated. Ketchum (et al. 1987) report that with the proper balance of anoxic, anaerobic and aerobic conditions in the SBR cycle phosphorous removal can be achieved without the addition of chemicals. This has advantages in terms of environmental impacts of the plant, sludge disposal costs, chemical cost savings and potential uses of the excess sludge. The authors state that 4 main groups of organisms seem to be involved in the phosphorus removal process: denitrifiers, fermentive organisms that create certain metabolic byproducts, phosphorus accumulating organisms and aerobic autotrophs and heterotrophs. Ketchum (et al. 1987) observe that a fairly long time period is required to build up suitable populations in the reactor. They stress that the process depends on specific balances of anaerobic, anoxic and aerobic periods in the reactor. Manning and Irvine (1985) recount experiments on phosphorous removal in an SBR system. A completely defined medium was fed to reactor run in one of six control strategies. No replication was made. From five to eight weeks were required to develop phosphorus removal in the reactor populations. Loss of phosphorus removal ability was gradual over two weeks when operating conditions were changed but could be recovered rapidly (ie three days) by readjusting operating regimes. In all cases, the feed was quite weak, on the order of 330 ppm COD and 36 255 ppm BOD. Poor settling is slightly favored by the same conditions that favor phosphorous removal. Palis and Irvine (1985) discuss several graphs that emphasize the fluctuation of oxygen, nitrogen oxides and ammonia nitrogen in the reactor during SBR operation. Very rapid conversion of nitrate to insoluble forms occurs at low loading rates. Settling characteristics seem to relate to the level of O 2 in solution during the react phase of operation. Irvine and Richter (1976) report a simulation model showing that the sizing of SBR reactors changes with the variability of the mass-flow of the wastestream even if the average flow is constant. Knowledge of the variability of the influent concentration is therefore required before design can be adequately performed. Hoepker and Schroeder (1979) found no connection between the relative growth rate of micro organisms and the effluent quality from a reactor. Huge ranges of variability in the results are reported here graphically but not in the text of the report. The number of replicates or experiments used to calculate the variation is not given. Possible causes of the variability are not indicated. Braha and Hafner (1987) recognized that biological changes can occur in the sludge populations inside reactors. Different conditions lead consistently to different bacteria predominating. Moser (1974) states batch reactors have this inherent difference from continuous flow configurations: that easy to metabolize compounds are consumed first and as time proceeds poorly decomposable substances accumulate. This implies at least a sequence of metabolic reactions if not a succession of different organisms being dominant and active. Braha and Hafner (1987) 37 say that only in Moser (1974) and in Gaudy and Gaudy (1972) are kinetic constants regarded as being specific to experimental conditions. The experimental results of Braha and Hafner (1987) show an overall drift in kinetic values with a sudden shift in their last experiment (Table 5). To explain the sudden shift Braha and Hafner (1987) hypothesize a qualitative change in the inflowing feed to their reactor. While this is a possible cause, by their own reasoning the effect of such a change would be a, shift in the biotic components residing in the reactor. Among their conclusions is the statement that the consistency of results indicates that a single experimental run (presumably of a single reactor, for that is what they used) would be sufficient to estimate the kinetic parameters of the reactor. However, the numbers they report (see Table 5) as replicate determinations of the same response indicate that up to 20 samples would be required to achieve the precision implied in their report. Table 5 The results of Braha and Hafner (1987) of an activated sludge kinetic analysis conducted over several months. Note that theta (the Streeter-Phelps temperature sensitivity coefficient shows a trend as the experiments proceed and that a fairly wide range of values were determined for the yield of biomass and the kinetic parameter Umax. Also note that the last experiment shows a markedly different kinetic response than the other runs. RUN YIELD Ks Umax THETA 1 1.45 14 0.400 11.4 2 1.35 13 0.482 9.3 3 1.37 13 0.500 8.0 4 1.38 14 0.568 6.4 5 1.56 15 0.776 4.9 6 1.42 16 0.548 4.0 7 0.78 14.8 0.383 3.6 38 Yoo et al. (1986) address the problem of choosing a value for the kinetic parameters and proposes the use of an "information index" to estimate how useful or reliable the kinetic parameters really are. Although the mathematics of the "information index" is complex it definitely recommends repeated estimates to achieve some sort of reliabile estimates for the kinetic terms. Sheih and Mulcahy (1986) comment that determining the intrinsic kinetic coefficients is the major problem with kinetic models in general. Several of the major research efforts into determining the kinetic response of activated sludge systems do not verify that they are replicable and representative estimates of a meaningful parameter. The fact that many authors have found need to modify Monods kinetic model may be an indication that it does not always describe the growth of a biologically complex system such as activated sludge. Even for biologically simplified systems Monods model may not apply to the growth response in a changing environment. The same may be true for the response to temperature. The sludge community is complex enough to react differently than a simple enzyme system to changes in temperature. The basis for assuming continuity between enzyme response and cellular response may not support further extrapolation to ecological communities. With this in mind, the next step is to attempt to measure the kinetic reponse of an SBR system. MATERIALS AND METHODS 3.1 EXPERIMENTAL DESIGN The original purpose of the experiment was to investigate the kinetic response of the SBR system to three factors: temperature, feed strength and hydraulic retention time. The factors temperature and feed strength are known to have major effects on biochemical reactions and these are well explored in the literature. Hydraulic retention time is a major operational variable for waste treatment reactors and it was felt that defining kinetic respose in terms of such a factor may lead directly to practical conclusions for the present study. Of even greater interest, however, was the possible interaction of factors in reactor kinetics. Most of the literature on the subject of reactor kinetics approaches factors as discrete and simple in the modes by which they affect reactor performance. It is more likely from a biological point of view that factors will interact and modify the reactor's response to other factors. This type of experiment was not found in the literature. The original experiment was begun in June of 1987 using 9 reactors. Three successive runs would be required to complete one replicate of the experiment. Inoculum sludge was obtained from the pilot scale waste treatment plant operated by the Environmental Enginneering department of UBC. The total solids and the volatile solids were monitored as means of estimating the growth of the sludge community. This was done by withdrawing a sample from the reactor at the end of the aeration stage using the draw pumps. Approximately 20 ml of sample was dispensed to a dried, numbered and pre-weighed crucible using a Fisher adjustable pipette. The crucible with the sample was again weighed using a Mettler analytical balance and then placed into a 105°C drying oven overnight. The dried samples were weighed and placed in a muffle furnace at 600°C for 15 minutes. Complete volatilization of the organic material should have occurred in 40 that period. The crucibles were reweighed when cool, cleaned and dried in the drying oven for not less than 24 hours. Duplicate blanks were run with each sampling time. 3.2 PROBLEMS WITH HRT AS A FACTOR Designing an experiment for 3 factors with 3 levels of each factor requires 27 experimental points for each replicate of the experiment. This created logistical problems and forced a re-examination of the experimental plan. It was decided that hydraulic retention time should be dropped as a factor. The retention time is a complex factor and alters several other reactor parameters, such as mass balance and food to microorganism ratio. Additionally, to adequately manipulate the hydraulic retention time would require more pumps, feed reservoir controllers and reactors than were available. Completion of an experimental replicate over time was ruled out as it was likely to introduce more variation into the data. 3.3 REVISED EXPERIMENTAL DESIGN The revised experimental design called for just feed strength and temperature to be used as the factors of interest. Although the kinetic effects of both factors are well explored both in theory and experimentally, the interaction of these factors is not well documented. The response of the reactor sludge populations to the applied treatments is commonly quantified by the alteration of specific kinetic coefficients. For the Monod saturation coefficient "Ks" the effect of temperature has not been definitively explored. Likewise, the Streeter-Phelps temperature sensitivity coefficient "theta" has not been considered to be affected by feed strength. As aerobic sludges are ecological communities of relative complexity, finding these effects would be equivalent to finding that some components of the system respond differently to these factors than other components. 41 The total variance in the experimental procedure was also considered to be of interest as this relates directly to the confidence interval that can be placed on the resulting kinetic parameters. This also should be relevant in the selection of appropriate contingency factors and safety margins for reactor design and scaling. This type of information is not readily available in the literature. The design of the experiment was a 3 x 3 factorial experiment in a randomized block design. This required that nine sets of experimental conditions be determined for each replicate. Figure 1 illustrates the schematic equipment layout for the experiment and figure 2 displays the process control configuration. This format was felt to provide an efficient means of quantifying the effects of the factors themselves as well as the anticipated interaction terms. In addition, the error inherent in this type of experimentation was measurable over replicates of the experiment. 3.4 REPLICATION THROUGH TIME Replicates of the experiment were conducted over time. The revised (two factor) experimental design first being implemented on Aug 13, 1987. For the Aug 13 experiment a 3-hour cycle time was chosen. The experimental protocols were modified to address problems observed after each experiment. Subsequent experiments used a 4 hour (Sept 16 and Oct 5) or a 6 hour cycle (Oct 20, 28 and Nov 16). The majority of the results presented here are from the 6 hour cycle experiments. The six hour cycle of reactor operation consisted of a 15 minute fill period, 4.25 hours of aeration, a 1 hour settling period and the drawing off of the clarified effluent prior to the commencement of the next cycle. The reactors stood idle until the next fill stage. The feeding of the reactors occurred prior to the draw stage allowing a 30 minute period for the levels to equilibrate. Figure 3 illustrates the SBR cycle and its standard divisions. 42 FEED PUMP FEED EESEEVOII FEEDING CASSETTE FEED OVERFLOW FEEDING SOLENOID + * I WASTE T T RESERVOIR F i g u r e 1 The schematic equipment layout for the experiment. Nine reactors were run simultaneously under this configuration. Replication of the experiment was performed over time. 43 F i g u r e 2 The control schematic of the equipment used in the experiment. The system controller was a Chrontrol 4 channel programable timer. 44 1 F I L L R E A C T 7 S E T T L E 7 D R A W I D L E • o o o o o Raw waste added No aeration. Low dissolved G^. 10 min duration Aeration. Dissolved O2 rises. Turbulent mixing. 270 min duration Aeration off. Biomass settles Dissolved O2 declines. Clear supernatant forms. 60 min duration Supernatant decanted 10 min duration No operation. Sludge may become anoxic, NO3 and P 0 4 conversion possible 10 min duration Figure 3 A schematic layout of the sequencing batch reactor showing the operational stages. The steps cycle under the control of a programmable timer. Reactor conditions are noted on the right. 45 The total solids and volatile solids were measured in the nine reactors at intervals as a means of estimating growth. This was done by withdrawing a sample from the reactor at the end of the aeration stage using the draw pumps. 100 ml of sample was filtered through a dried, numbered and pre-weighed glass fiber filter (Whatman 94H, 30mm diameter). Duplicate measures were taken from each reactor. The filter paper was then placed into a 105°C drying oven overnight. The dried samples were weighed and placed in a muffle furnace at 600°C for 15 minutes. Complete volatilization of the organic material should have occurred in that period. The filter was reweighed when cool. Duplicate blanks were run with each sample time. These values were used to correct the calculation of the volatile solids present in the reactor vessels. The measurements made on the reactors were extended to include the optical density of the solution. This term should also be inversely proportional to the suspended bacterial density in the reactor vessel. The total solids and volatile solids were run as before. Three more runs of the experiment were performed (Oct 20, Oct 28 and Nov 16). An experiment was begun Dec 16 but was not completed due to bulking problems in the inoculum sludge. 3.5 D E T E C T A B L E SOURCES OF VARIATION The possible sources of variation that can be investigated in this experiment are due to the treatments temperature, feed strength and the interaction of the two terms. The homogeneity of variance over all experiments can be tested as can the magnitudes of the effects calculated from the data. Significant differences detected in these terms would indicate that the sludge had changed its kinetic responses to the applied treatments. Additional tests can be performed on subsections of the data set to determine the relative importance of different components of the measuring process. The variance of the blanks give an estimate of the precision of the 46 measurement process. The variance of the inoculum sludge concentration as measured in the first sample from all experiments should also give an estimate of the limit of the usefullness of the sludge volume index as an indication of the biomass of the inoculum. EQUIPMENT 3.6 OVERVIEW OF SYSTEM The requirements of the sequencing batch reactor system are that the reactor operates in a continuous cycle of stages without intervening interruption. The treatments should be as uniform as possible to each of the reactor vessels. Therefore all three reactors subjected to a given temperature treatment should recieve the same treatment. The system must be able to feed the reactors a specific quantity of feed of consistent quality at a precise time. The aeration applied to all reactors should be as consistent as possible both in duration and flow rate. It should cease completely in all reactors simultaneously. The draw cycle should reliably remove a set quantity of supernatant from each reactor vessel. Any source of variation that can not be controlled must not also be consistently correlated with a specific reactor or treatment allocation. For that reason, the treatments must be able to be randomly rearranged to the various components of the system. A "Chrontrol" (Lindburg Enterprises, California) four channel programmable timer was used to operate the feeder pump controller, the aeration solenoid, the feeder solenoids and the draw pumps controllers. The timer is adjustable to the second and capable of cycling without adjustment indefinetly. 3.7 REACTORS Nine seven liter (five liter working volume) acrylic reactor vessels were secured and throughly cleaned. All of them were disassembled, the seals and gaskets checked and repaired if necessary. Between runs the reactors were 47 disassembled, cleaned and reassembled. The assembling of a reactor involved the insertion of the cleaned air stone diffuser, attaching the reactor lid and installing the draw line. The draw line was adjusted to the four liter level (the volume of liquid remaining in the reactor when the one liter draw was removed) and the entire unit assigned to a temperature treatment. Plates 1 through 3 illustrate the equipment in place in the lab setting. 3.8 AIR STONES Nine six inch aquarium air stones were purchased to facillitate aeration to the reactors. Between runs the air stones were removed, cleaned in acid to remove any afixed sludge, rinsed in dilute NaOH and rinsed several times with tap water. No specific randomization scheme was used to assign air stones to reactors. 3.9 FEEDING SYSTEM As feed strength was a major factor of interest and of primary importance in reactor operation the maintenance of feed quality recieved some consideration. Feeding the reactors was facilitated from three main feed reservoirs capable of holding up to 24 hours worth of feed. The reservoirs were 16 liter food grade plastic pails with lids. The feed reservoirs were acid washed and rinsed thoroughly before use. To restrict the possibility of contamination in the feed, the pails were rewashed before each filling. After each experiment tap water was run through the entire feeding apparatus for several cycles. Tygon tubing lines out of the pails were weighted at the immersed end and connected to the feeding pump. The feeding pump was a Cole Parmer peristaltic pump drive with three separate drive heads (one for each feed reservoir) connected to a Masterflex pump speed controller. The speed controller was operated by the Chrontrol timer. Plate 1 View of the experimental equipment showing a reactor immersed in the warm water bath. Three reactors sit in each bath. The frame work visible on the left supports the feeding cassettes. At the bottom the feed reservoir is visible (with lid). 49 Plate 2 View of the experimental equipment showing the feeding cassettes on the steel framework. The overflow tubes protrude from each cassette triplet. Below the cassettes are the solenoids (not visible) and the nine reactors. Plate 3 View of the experimental equipment showing the three temperature baths in the lower left hand corner. Some of the feeding cassettes are visible in the upper right hand corner and the Julabo refrigerated bath can be seen on the on the right. In the rear is the air manifold used to control the air flow to each reactor. 50 The Tygon feed lines connected the main feed reservoirs with the metering feeding cassettes. These were fabricated from 4 inch PVC irrigation pipe so that each of the three chambers in each of the three cassetttes delivered one liter of feed. The original design of the experiment called for the feeding of one, two and three liter aliquots, but this was reduced in the final version of the experiment to a single liter of feed. The level was controlled by an overflow system that discarded excess feed into the waste reservoir. The feeding cassettes drained through holes in their bottoms connected with the reactors. Feeding from the cassettes to the reactors was controlled by 9 solenoid valves positioned between the feeding cassettes and the reactors. The valves were operated by one of the channels of the Chrontrol timer. It was not possible to secure nine identical solenoid valves for this experiment. Six of the solenoids were 1/4 inch brass and three were 1/8 inch teflon lined solenoids. All cassettes drained in from three to five minutes. 3.10 DRAW SYSTEM The draw system was responsible for drawing material out of the reactors and disposing of it in the waste reservoir. The beginning of the system is the 1/4 inch glass tubing inserted into each reactor. The depth of insertion could be adjusted using a clamp positioned at the lid of the reactor. This held the draw line intake at the 4 liter level of the reactor, thus allowing the lines to all withdraw only 1 liter even if the pumps ran at different rates. The glass tubes were connected the peristaltic draw pumps using lengths of 3/16 inch brown latex tubing. Two pumps and controllers were required to accomodate the nine peristaltic drive heads required for the nine reactors. From the pumps latex tubing directed the reactor effluent to the waste reservoir. The end of these lengths of tubing also proved to be the most convenient sampling point for the reactors. During the well mixed aeration phase, the draw pumps could be 51 manually turned on and a sample simultaneously withdrawn from all of the reactors at once. The nine 300 ml capacity sample bottles were placed in a wire frame holder and suspended above the waste reservoir. This allowed any overflow to spill directly into the waste bucket. Overflow always happened. It was never observed that pump heads, even on the same drive, consistently pumped at the same rate. 3.11 AIR SYSTEM Aeration to the reactors was achieved using a compressed air system. A half inch high pressure red butyl rubber hose connected the air valve with a half inch solenoid valve. The solenoid valve was connected to the Chrontrol controller. The air flow from the solenoid went to a 12 line manifold. The manifold allowed the control of the air flow to the 9 reactors and gave a rough estimation of whether the flows were the same. The flow rates in the various lines was controllable by the use of hose clamps on the specific air lines. The aeration rate was thought to affect not only the dissolved oxygen content in the reactor vessels, but also the turbulence, the shear forces, the floe size and the rate of collisions between the floe and the substrate material. In practice the manifold was not a completely reliable system, and the air flow as indicated by the height of air in the tubes tended to drift. This was probably due to changes the resistance to flow at the air stones. At the end of each experiment the stones often had noticeable growth on the surface and clogging was likely. 3.12 T E M P E R A T U R E CONTROL Temperature control to the reactor vessels was achieved using a warm and cool water bath and an ambient air temperature tank. The warm (30°C) and cool (10°C) tanks were aquaria large enough to hold 3 reactors. The 20°C treatment was a fabricated plywood box large enough to hold three reactors. 52 All three treatments sat side by side on the lab bench in room 80 of the MacMillan building at UBC. During the draw down period of reactor operation the reactors in the 10°C and 30°C water baths became bouyant as they were emptied. Several of the glass draw line tubes snapped when the reactors impinged on the feeding reservoir platform above. This required the placing of blocks of steel on top of the reactor lids to keep them weighted down. The water in the warm tank was heated by three Jaeger aquarium immersion heaters (two 250 watt heaters and one 100 watt). Temperature control in the water bath was good and the temperature in the reactors only varied by a couple of degrees around the desired point of 30°C. The cold treatment was applied using a Julabo circulating bath unit. Again, three reactor were placed in an aquarium immersed in fluid of much greater thermal mass. Ethylene glycol was used as the circulating fluid as freezing was initially a problem in the Julabo unit. Temperature control by the Julabo unit was good and the temperatures in the reactors did not vary much from the desired treatments of 10°C. A problem was soon detected with variance in the input and output rates of the Julabo unit, causing frequent over and under flows. This was reconciled by the use of a toilet float valve in the aquarium to control the level of the glycol input to the aquarium within acceptable limits. The 20° C treatment was about what the ambient temperature of the lab. A 3/4" plywood box was built to hold 3 reactors in an ambient temperature water bath, but problems with leakage made the water an unattractive method of temperature control. It was felt by the technical staff of the department that the lab temperature did not vary excessively and the water would therefore be unnecessary anyway. Monitoring of the temperatures in the 53 reactors in the plywood box bore out this opinion and the box was left empty. Temperatures in the reactors inside the box remained close to the desired treatment of 20°C. 3.13 Feed Composition The artificial waste used as substrate for the sludge to grow on in this experiment was modified from a recipe by Gaudy and Gaudy (1971). In addition to the defined mineral components plus glucose as a carbon source, a dilute solution of yeast extract was added. The basis for this change was the advice contained in the American Microbiological Society Handbook of Methods (1981). As sludge is a complex and undefined community it seems unlikely that a simple mineral medium will meet all of the nutritional requirements for all of the species that interact to give the sludge its properties. The supplement chosen was a yeast extract capable of supplying B vitamins, amino acids, proteins and various trace materials. The yeast extract used was Marmite, a yeast-extract food-product that does not contain preservatives (potential bactericides). The sludge growth on this medium was good and changes in gross morphological properties (color, floe size, bulking) did not occur as rapidly or consistently as on the unsupplemented medium. 3.14 Inoculum Sludge At the beginning of this experiment no thought was given to the nature of the inoculum sludge. The aerobic sludge resulting from dairy waste treatment was used initially as the inoculum for the SBR reactors in this experiment (preliminary experiments, May-June 1987). The results of these experiments were highly variable as were the gross morphological characteristics of the inoculum sludge itself (color, texture, settleability, etc). A more consistent source of sludge was sought and found in the pilot scale water treatment plant operated by the Civil Engineering department of UBC. The water treatment Table 6 The artificial medium used in all experiments. The desired concentration in the feed solution is given along with the concentration of the stock solutions that were prepared fresh for each experiment. All of the stocks were refrigerated. CONSTITUENTS COMPONENT K 2HP0 4 KH 2P0 4 (NH 4) 2S0 4 MgS04.7H20 MnS04 CaCl 2.2H 20 ZnN03 CoCl 2.6H 20 H 3 B O 3 NaMo04 CuS0 4.5H 20 F e C l 3 YEAST g/g glucose 0.05 0.10 0.05 0.01 0.0001 0.75 0.02 0.02 0.05 0.0125 0.006 0.0005 0.100 g/1 STOCK 50 100 50 10 1.0 7.5 0.2 0.2 0.5 0.125 0.06 0.005 100.0 b o t t l e 1 1 2 2 2 3 3 3 3 3 3 3 4 GLUCOSE 1.000 g/1 500.0 2.000 g/1 3.000 g/1 3 N NaOH to adjust pH to 7.0 - 7.5 55 reactors at this facility are fed on municipal sewage via a dosing tank. Qualitatively, the sludge from this source looked and behaved as though it was quite consistent in its morphological characteristics. Five to ten liters of sludge was collected in a 16 liter food grade plastic pail and returned to the lab where an air stone was used to provide aeration to the entire bucket. All experiments were inoculated with the sludge not older than 24 hours or younger than 16 hours. The sludge volume (as sludge blanket depth at 30 minutes settling) was adjusted with tap water to approximate 300 ml sludge per liter of stock. The inoculum of 250 ml sludge was delivered to the reactors as a 150 ml aliquot for each reactor followed by a 100 ml aliquot. A 150 ml syringe afixed with a length of tygon tubing was used for this purpose. All 9 reactors could be inoculated in less than 5 minutes by this method. The inoculation was performed into cleaned reactors filled to the 4 liter level with tap water. This was done consistently within 15 minutes of the feeding cycle that begins reactor operation, bringing the reactor volume up to the working level of 5 liters. Samples were taken within 5 minutes of the commencing of aeration to determine the initial inoculum density for each reactor. The preliminary experimental protocols were modified as the experiments progressed to attempt to find the best set of conditions to elucidate the . effects of the various factors. The main types of modifications imposed on the experimentation scheme were changing the inoculum density and changing the duration of the cycle. The inoculum was changed from about 500 ml of raw concentrated sludge in the premlinary experiments to 150 ml of sludge diluted to 300 ml settled sludge per liter of diluted solution after 30 minutes settling for the results reported here. The reactor cycle was changed from 4 hours to 6 hours. 56 3.15 Statistical Analysis The data were collected and manipulated on an electronic spreadsheet, graphed and printed. Statistical analysis was performed using Lotus 1-2-3 and Systat on IBM-PC style computers for the smaller calculations and on the UBC mainframe using the SAS statistical package for the backwards regressions on the combined datasets. The major factors feed strength and temperature were represented by "F" and "R" respectively in regression models. Elapsed time was coded as "E". Data sets were combined using dummy variable coding to compare growth on the different dates, the individual dates being coded as no code for October 20 t h, D2 for October 28 and D3 for November 16 t h. Data sets with major experimental differences (i.e. inoculum size and cycle length) were analyzed separately. Similarly, data displaying significant differences in the stepwise regression analysis were not combined in subsequent analyses. It was felt that the gains possible from the extra data were offset by the possible errors in non-comparable data being analyzed. All possible 2, 3 and 4 factor combinations of factors (i.e. F,R,E, FR, FE, FRE, D2F, D2E... etc.) were introduced into a backwards stepwise regression. This procedure was felt to be capable of finding the best fit to the data from a choice of 21 factors and first degree interaction terms. RESULTS 4.1 Sample variability Several authors comment on the variability of the activated sludge process (e.g. O'Shaughnessy, 1923; Jenkins, 1942). Rates of treatment varying by more than an order of magnitude between different plants have been reported (O'Shaughnessy, 1923). Not surprisingly, different biota have been found to be favored by different conditions in plant operation (Antheuse and Koene,1987)and it is entirely logical to assume that this is the basis for the vast range of activity in different sludges. Metcalf and Eddy (1979) comment that "the scatter in treatment rates and efficiency is remarkable". Variation within the treatment process is less well documented. To quantify growth, a precise measure of reactor contents must be made. Only then can a precise estimate of the change in reactor contents be calculated. The accurate measurement of growth in the reactors was hampered by a certain amount of variability in the samples. There may be several causes of variation between replicate measurements. These include measurement errors (e.g. weight of filters, blanks, dried samples; volumes of samples), heterogeneity in the reactors (which is credible in a flocculating culture) and sampling errors. Variations in the measured values between replicate samples indicates heterogeneitj' in the reactor vessels and sets the limit of numerical precision that can be calculated from a given data set. Quantification of the variation in the samples indicates the number of samples required to achieve a given level of resolution of the reactor contents and therefore in the parameter estimates. It is assumed that samples are taken randomly and are independent of each other, that is the contents of one sample has no effect on the contents of other samples. The number of samples 58 may be calculated for a given level of variation and limit of resolution by equation 9 (Snedecor and Cochrane, 1978). n = (t2) o 2/ L 2 (9) Where n is the number of samples t is the 2 tailed t test for a given level of confidence in the results o is the sample variance L is the limit to be resolved As an example, to reliably resolve between 20 and 21 mg/1 of material would require a resolution limit (L) of 0.5 mg/1. Using t = 1.96 (95 % o confidence in the results) and o = 2.0 mg/1 (10% coefficient of variation) would require 31 samples be taken at each experimental point. If the variance was half that (eg 1 mg/1 or 5% of the mean) it would still require 15 samples per experimental point to reliably resolve the second digit of the concentration of material in the reactor. Table 7 illustrates the variability of the samples taken from the freshly inoculated Nov 16 reactors. As can be seen the dry matter samples vary by from 0 to 10% of the mean. Note that the occaissional value is relatively high, indicating larger variation between samples. A higher coefficient of variation may indicate physical heterogeneity in the inoculum and the occurrence of lumps of biomass in the samples. This may relate directhy to some characteristic of the flocculant particles in the reactor, such as the size or rapidity of floe formation. It is assumed that each pair of samples reflects an independent estimate of the variation in the sampling of the reactor contents, inoculation variance and sampling error. This is valid if the reactors are all prone to the same error. Applying equation 9 indicates that to determine the reactor dry matter content to 0.1 mg/1 limits would require over 100 samples. However, a 0.5 mg/1 resolution 59 T a b l e 7 The variation in samples from a freshly inoculated experiment. The data is from the November 19th experiment and is similar to the variation data for all experiments intial inoculum. COEFFICIENT OF VARIATION (%) TEMP FEED °C mg/1 COD DRY MATTER VOLATILES 10 1000 9.9 13.7 10 2000 8.5 2.8 10 3000 9.9 14.8 20 1000 5.5 7.3 20 2000 0.0 0.9 20 3000 1.6 0.9 30 1000 2.1 6.2 30 2000 6.0 4.9 30 3000 4.6 9.2 Average 5.34 % 6.75 % (making the second digit significant) would require only 4 samples. As only duplicate samples were taken the dry matter should be regarded as not significant in the second digit. The volatile solids data are somewhat more variable, averaging 6.75%. Some samples had 15% variation between them. That level of variation implies that 34 independent samples would be required to reliably determine the second digit of the volatile solids content of the reactors. The volatile solids data could be corrected for changes in the blank filters whereas the dry matter samples could not be so corrected. The volatile solids data are therefore thought to be a more representative of the actual reactor contents. The variability of the samples is high enough that the second digit of the volatile solids data cannot be said to be significant with the number of replicates performed in this experiment. This level of variation was also present in the October 20th and 28th experiments. This may represent a fundamental range of variation for the measurement process that was applied in this study. 60 Table 8 illustrates the variability of the samples taken at the end of the same experiment. It will be seen subsequently from the growth curve for the 20°C 2000 and 3000 COD treatments (Figures 5 and 6) that something of a population explosion occurred relative to the other reactors. However, the variability of the samples (Table 8 20°C 2000 COD feed strength) is not greater than the other reactors. The variability of the response does seem to be at its lowest for both dry matter and volatile solids in the 20°C 2000 COD feed strength treatment. It increases with higher and lower temperatures and feed strengths. The most variable treatments are the most extreme: the combinations of high or low temperature and high or low feed strength. It may be that the sludge is most acclimatized to conditions resembling the treatment with the lowest variability, but this is purely speculative. Table 8 The variability of the samples at the end of an experiment. The data presented is from the November 19th experiment and the samples taken after 94 hours of growth. COEFFICIENT OF VARIATION (%) TEMP FEED ° c mg/1 COD DRY MATTER VOLATILES 10 1000 10.35 1.71 10 2000 2.67 1.44 10 3000 12.20 11.11 20 1000 2.93 3.85 20 2000 0.13 0.70 20 3000 2.17 1.60 30 1000 4.10 1.82 30 2000 1.86 2.42 30 3000 4.90 3.85 Average 4.59 % 3.92 % Table 9 illustrates the variability in the entire Nov 19 experiment. The variability of the data is still fairly high despite a larger data set from which to make the variance estimation. The overall variance is the average of the 61 variances of the individual samples throughout each treatment. As can be seen, despite high values (i.e. > 10%) at some individual sample points, the variation overall is less than 4%. This still emphasizes the absolute requirement for replicate samples. To reliably determine the second digit of the sludge biomass per liter would require 10 samples be taken per point. The lowest variation is still apparent in the 20°C 2000 COD feed. Table 9 The variability between samples for the entire November 19 experiment. As can be seen, the sampling process using 100 ml of reactor contents on a Whatman 94HC filter displays an inherent variability of from about 2% to 6%. COEFFICIENT OF VARIATION TEMP FEED °C mg/1 COD DRY MATTER VOLATILES 10 1000 5.58 3.74 10 2000 4.67 2.39 10 3000 5.12 6.33 20 1000 3.66 3.41 20 2000 3.10 2.99 20 3000 3.46 3.04 30 1000 3.46 3.70 30 2000 2.41 2.04 30 3000 3.13 2.87 Average 3.84 % 3.39 % Table 10 illustrates the variability of dry weight and volatile solids measurements taken at the end of the Oct 20 experiment. As can be seen, the coefficient of variation is relatively low and rarely exceeds 10%. Similar variation is apparent in this data set as in the November 16^ data. The lowest variation is observed in the moderate treatments, and the variation increases in the combinations of high and low temperatures and feed strengths. This may imply increasing heterogeneity in the reactor contents in these 62 Table 10 The average coefficient of variation for each treatment for the October 20 experiment. The table reports the variability between samples at the end of the experiment. COEFFICIENT OF VARIATION (%) TEMP FEED °C mg/1 COD DRY MATTER V0LATILES 10 1000 7.15 3.26 10 2000 3.22 3.94 10 3000 10.19 8.98 20 1000 3.34 3.26 20 2000 2.03 2.29 20 3000 1.39 1.26 30 1000 3.78 4.61 30 2000 2.18 1.50 30 3000 4.02 3.65 Av e r a g e 4.14 3.63 treatments. On a microbial level, this could be the result of growth of colonies or flocculant particles that are not forming in the moderately treated reactor. In general the variability of the volatile solids data is slightly lower than that of the dry matter data. This may reflect the correction for changes in the blank filters possible in these results. The variability is of the same general scale as that of the November 16th experiment. This is not to say that the reactor contents are similar, just that the limit on the accuracy with which reactor contents can be determined is approximately the same. Table 11 illustrates the variability observed in the Oct 28 samples after 52 hours of growth. Again, the coefficient of variation of the samples is between 2 % and 10% and that of the volatile solids data is less than the dry matter data. Again, the data is not inherently more variable than the previous experiments. The 20°C 2000 COD treatment again has low variation for both the dry matter and the volatile solids but the 3000 COD treatment is lowest. 63 Table 11 Data variability for the October 28th experiment. The coefficient of variation is on a par with the October 20th results (Table 9). COEFFICIENT OF VARIATION (%) TEMP FEED °C mg/1 COD DRY MATTER VOLATILES 10 1000 9.33 4.00 10 2000 3.07 3.01 10 3000 4.73 2.57 20 1000 3.92 2.58 20 2000 3.58 3.77 20 3000 4.20 3.03 30 1000 5.39 4.64 30 2000 5.65 4.12 30 3000 1.83 1.70 Av e r a g e 4.63 % 2.98 % The variation between samples defines the variation measured within a reactor. However, the experimental protocol calls for replication over time and therefore the data will also contain information about the differences between reactors. Table 12 illustrates the variation apparent in the data if the Oct 20 and 28 experimental results are pooled. It is much greater than either experiment alone. The coefficient of variation for some individual points reached 42%, and averaged 15% for both volatiles and dry matter. Although it will be shown subsequently that the growth response of these two experiments is not significantly different, this result illustrates that the response is certainly not identical. This means the variation between two different experiments is greater than within experiments. This implies that greater precision can be achieved by replicating an experiment at one date than by replicating experiments over time. The level of variation apparent in the pooled data set decreases confidence in the numerical results of any calculations. 64 Table 12 The variability of the data of the combined October 20th and 28th experiments. It is clear from this that pooling experiments increases the variability in the data. COEFFICIENT OF VARIATION (%) TEMP FEED OC mg/1 COD DRY MATTER VOLATILES 10 1000 14.17 14.52 10 2000 17.66 17.12 10 3000 10.51 10.20 20 1000 5.87 7.55 20 2000 14.08 15.99 20 3000 26.67 26.44 30 1000 12.60 11.03 30 2000 18.94 16.38 30 3000 13.65 16.32 Aver a g e 14.91 % 15.06 % The level of variation becomes even greater if the November data is pooled with the October data. This raises doubts about extrapolation of results calculated from one time of year to treatment plant operation at another time of year. The change in efficiency can be inferred from some treatment plant operational results (e.g. Baines et al.,1953) but the range of efficiency that this effect causes is not widely apparent in the literature. Variation within a reactor may convey information about how the micro organisms are growing (ie spatial homogeneity, floe size, the existance of colonies). Variation between dates conveys information about how the micro organisms change over time. From the results so far it is apparent that experiments performed at different times may reach different results. This agrees with reports in the literature but not with the conclusions drawn from such data (e.g. Peil and Gaudy, 1971; Braha and Hafner,1987). 65 4.2 Regression analysis Making inferences from a data set also involves the application of assumptions about characteristics of the data and the experimental process, such as the distribution of error and the independance of samples. Without delving into statistical theory, suffice it to say that statistical procedures provide complete confidence only when the assumptions are fully met. The results are approximate to the degree that the fundamental assumptions are violated. Data transformations can be applied to restore compliance with some assumptions of data characteristics. However, it may be more useful to have approximate information about a factor of interest than exact information about a transformed or abstract variable which is of little interest (G.W. Eaton, pers. comm. 1988). The accumulation of dry matter is illustrated by figures 4, 5 and 6 for the 1000, 2000 and 3000 COD feed solutions respectively. Note the increase of scale on the ordinate axis required to accomodate the increased growth at the higher feed strengths. Note the general form of the growth response has an early increase followed by a decline and then a more general dry matter increase. This form is not predicted by most kinetic models but is explainable in ecological terms as a succession of organisms growing or as an artifact of colonization of the reactor surfaces. This would imply that the first peak in reactor biomass corresponds to changes in the biotic composition of the maturing sludge community. Alternately, it could be only an apparent drop in reactor biomass. If most of the new growth is colonizing the clean reactor surfaces, it would not be available for pick up in the samples. Growth on the walls of the reactors was a common observation at the end of each experiment. It should be immediately seen from examination of all of the graphs that, although similar, the growth response occuring on different days is often different. This is especially clear for the 30°C treatments but is also Figure 4 The accumulation of dry matter in the 1000 COD feeding treatment at 10°C, 20°C and 30°C for the' October 20, October 28 and November 16th experiments. CO J , cr >-cr o 3 0 0 ^ 2 5 0 -2 0 0 -1 5 0 -100 D R Y M A T T E R A C C U M U L A T I O N 1 0 0 0 C O D F E E D 1 0 ' C O OCT 20 A OCT 28 • NOV 16 100 ELAPSED TIME [HOURS] 2 0 ' C en cr >-cr Q 3 0 0-i 2 5 0 -2 0 0 -1 5 0 -100 5 0 H 0 O OCT 20 A OCT 28 • NOV 16 20 4 0 60 8 0 ELAPSED TIME [HOURS] • 100 3 0 ' C en cr >• cr Q 3 0 0-i 2 5 0 -2 0 0 -1 5 0 -O OCT 20 A OCT 28 • NOV 16 100 OA-ELAPSED TIME [HOURS] 100 68 Figure 5 The accumulation of dry matter in the 2000 COD feeding treatment at 10°C, 20°C and 30°C for the October 20, October 28 and November 16th experiments. The ordinate axis scale is double that of the 1000 COD graph (Fig 6). 6 0 0 n r"'"i \ 5 0 0 -CT) £, 4 0 0 -LU r- 3 0 0 -!< 2 0 0 ->-cn 1 0 0 < Q L D R Y M A T T E R A C C U M U L A T I O N 2 0 0 0 C O D F E E D 1 0 ' C O OCT 20 A OCT 28 • NOV 16 cw5r 2 0 4 0 60 8 0 ELAPSED TIME [HOURS] 2 0 ' C 6 0 0-i I—I CT) 5 0 0 -4 0 0 -MATTER 3 0 0 -2 0 0 -DRY 100 £ O OCT 20 A OCT 28 • NOV 16 X T ' 2 0 4 0 60 8 0 ELAPSED TIME [HOURS] 100 • 100 CT> J , Q: UJ 6 0 0 n 5 0 0 -4 0 0 3 0 0 H 3 0 ' C O OCT 20 A OCT 28 • NOV 16 X T 20 4 0 60 8 0 ELAPSED TIME [HOURS] 100 70 Figure 6 The accumulation of dry matter in the 3000 COD feeding treatment at 10°C, 20°C and 30°C for the October 20, October 28 and November 16th experiments. The ordinate axis scale is increased. CD CT UJ >-Q 900 800 700 600 500 400 300 200 100 0 D R Y M A T T E R A C C U M U L A T I O N 3 0 0 0 C O D F E E D 1 0 ' C O A • OCT 20  OCT 28 NOV 16 20 40 60 80 ELAPSED TIME [HOURS] 100 2 0 ' C or UJ >-en 900 800 700 600 500 400 300 200 100 0 O OCT 20 A OCT 28 • NOV 16 &6~A^A-A D* 20 40 60 80 ELAPSED TIME [HOURS] 100 CT >-cr o 900-j 800-700-600-500 400H 300 2 0 0 -100 Q o4 0 o A • OCT 20  OCT 28 NOV 16 3 - A - o -20 3 0 ' C rr 40 60 80 100 ELAPSED TIME [HOURS] 72 manifest in the 2 0 ° C treatment. Hynes (1970) has amply i l lustrated the seasonal succession of biota occuring in running waters and Hawkes (1963) and Curds (1965) have demonstrated the biological complexity of the sludge community. A shift of kinetic response wi th respect to time is easily explainable in these terms. Note also that the growth of the 10°C treatment is poor at a l l feed strengths, whereas the 2 0 ° C treatment outgrew the 3 0 ° C treatment for the 2000 C O D feed on both October 28th and Nov 16th. This directly conflicts wi th the standard approach to the effect of temperature on activated sludge growth. The van ' t Hoff Streeter-Phelps modification of the Arrhen ius equation predicts an increase i n growth rates wi th temperature, but it is acknowledged that there is a l imit to the range of its applicability (Metcalf and Eddy, 1978; G r a d y and L i m , 1981). This m a y indicate that a 10 C° temperature change exceeds that l imit . It may also be that simple temperature change does not convey enough information about the degree of shock or perturbation to the system to adequately predict response of the sludge. The 10°C growth and the 1000 C O D treatment displayed a biomass reduction. This m a y indicate the energy input for 0 net growth of sludge changes wi th temperature. The capacity of a reactor to support a given amount of sludge depends on more than just the input of feed. It is apparent that, given a specific level of feed added to the reactor the biomass w i l l decline i f the temperature is low. Kinet ica l ly , this is the same as saying the endogenous respiration rate (spontaneous death rate of microbial cells) changes with temperature. Grady and L i m (1981) review literature on this topic and conclude the effect follows an Arrhenius- l ike effect, increasing wi th higher temperature. The present result indicates that the apparent death rate m a y also increase at lower temperatures. However , it maj ' be more in line with the ecological literature to speculate that the 73 predation rate remains high at lower temperatures and the production of biomass slows down (Hynes, 1970). The 3000 COD feed treatment (figure 6) shows an interesting response: the 20°C treatment vastly outgrew the 30°C treatment on both Oct 28 and Nov 16 but not on Oct 20. Similarly, the 2000 COD 20°C treatment also shows the Oct 20 experiment accumulating dry matter much slower than both the Oct 28 and Nov 16 experiments. Under the 30°C conditions the Oct 20 experiments consistently outgrew the others. This may indicate a population explosion of some component of the sludge community for the latter two dates. A possible explaination could be the presence of some organisms controlling the population in the Oct 20 sludge that are not present in the later replicates in the 20°C treatments. The 30°C response may indicate the presence of a fast growing organism better adapted to the warm conditions than those occuring in the later experiments. Figures 7, 8 and 9 illustrate the accumulation of volatile solids under the 1000, 2000 and 3000 COD treatments respectively. They are quite similar to the dry matter data. Again, the 10°C treatments did not grow appreciably at any feed strength. Again, the 20°C 2000 and 3000 COD feed accumulated biomass faster than the 30°C treatment contrarj' to what the standard kinetic models predict. The 1000 COD treatment accumulated material at a comparable rate. The standard kinetic approach suggested by Phelps (1944) equation suggests that the growth of sludge biomass at any temperature should be close to twice as fast as a treatment 10 C° cooler. This was never observed to occur. This is not surprising if one considers the sludge to be a community of organisms not equally suited to the changes in conditions. In this case, growth of sludge cannot be expected to be simulated by a simple enzyme Figure 7 The accumulation of volatile solids in the 10°C, 20°C and 30°C treatments and a 1000 COD feed strength. The data for these calculations could be corrected for the blanks. Note the same early peak and decline prior to the larger growth response in the reactors. JL to o _i o 00 o > 300 250 200 150 H 100 50 0 V O L A T I L E S O L I D S A C C U M U L A T I O N 1 0 0 0 C O D F E E D 1 0 ' C O OCT 20 A OCT 28 • NOV 16 j=r 20 4-0 60 8 0 ELAPSED TIME [HOURS] 2 0 ' C to 9 _i o O > 300 250-200-150-100 50 H 0 O OCT 20 A OCT 28 • NOV 16 i A ' 20 40 60 8 0 ELAPSED TIME [HOURS] 3 0 ' C 1 1 s. 300-, \ JL 250-.IDS 200-SOL 150-LATILE 100 p A 50-O OCT 20 A OCT 28 • NOV 16 n' o > 20 40 60 8 0 ELAPSED TIME [HOURS] 100 100 100 76 Figure 8 The accumulation of volatile solids in the 10°C, 20°C and 30°C treatments and a 2000 COD feed strength. 600 - j \ 500-IDS 400-SOLI 300^ Ul _ i 200-VOLATI 100^ OH V O L A T I L E S O L I D S A C C U M U L A T I O N 2 0 0 0 C O D F E E D 1 0 ' C O OCT 20 A OCT 28 • NOV 16 CO Q _ l O CO O > 600 - i 500 400 300 200-I 100 0 20 40 60 ELAPSED TIME [HOURS] 2 0 ' C 80 O OCT 20 A OCT 28 • NOV 16 O 20 40 60 80 ELAPSED TIME [HOURS] 100 100 — 600-\ , |> 500-] CO 400-I Q I O 300-CO __ 200-O > O OCT 20 A OCT 28 • NOV 16 3 0 ' C 20 40 60 80 ELAPSED TIME [HOURS] 100 78 Figure 9 The accumulation of volatile solids in the 10°C, 20°C and 30°C treatments and a 3000 COD feed strength. O) JL CO Q _) O CO UJ o > V O L A T I L E S O L I D S A C C U M U L A T I O N 3 0 0 0 C O D F E E D 1 0 ' C 1 — , 900-1 \ 800-700-CO 600-Q _ J 500-o CO 400-ILE 300-200-i 1 0 0 Q 0 + I0A OCT 20 O A OCT 28 • NOV 16 20 4 0 60 8 0 100 ELAPSED TIME [HOURS] 2 0 ' C 9001 800-700-600-500 400 300 200 O OCT 20 A OCT 28 • NOV 16 1OOQ . 0 - A A-A-' 20 4 0 60 ELAPSED TIME [HOURS] 80 100 1 1 900 - i \ 800-JL 700-CO 600-Q _ l 5 00-O CO 400-ILE 300-200-0L 1 0 0 Q > o4-o A • OCT 20  OCT 28 NOV 16 3 0 ' C 'Q-A* 20 4 0 60 8 0 ELAPSED TIME [HOURS] 100 80 kinetic system as it is much more complex. The ini t ia l problem m a y be traced to Monod (1942) who characterized bacterial growth as s imilar to an enzymic reaction. The idea of a continuity from the less complex parts conveying kinetic properties to the more complex whole m a y be a good approximation for simple systems but becomes strained when applied to ecological communities. Communit ies m a y not be wel l approximated by the response of individuals, let alone as some part of an individual . To simplify the response of activated sludge in experimental work several authors have adopted acclimatization procedures that would effectively select for some subsample of the sludge community (ie Pe i l and Gaudy, 1971). The assumption that the result ing calculated kinetic parameters is representative of the whole activated sludge system is never tested. Table 13 presents the backwards stepwise regressions on the combined data from the Oct 20, 28 and Nov 19 experiments. This compared the responses of growth in a l l reactors for a l l experiments simultaneously. The resulting models are the least squares fit of the combinations of factors required to describe the data with a factor significance level of 90%. A s can be seen, the standard error for each coefficient is fair ly large and the combined data set cannot be described without the inclusion of a date term (D3FE) , indicating significant differences between dates. This means there are two distinct growth response forms for the sludge dry matter and volatile solids, one that describes both the October 20 and 28th experiments and a different response for the Nov 16th experiment. In addition to this, most of the significant factors are interactions of simple factors. This indicates that the response to a given factor is significantly significantly modified by the levels of other factors. The implication of this is that the kinetic response of the sludge depends on the total set of experimental conditions. The effect of one factor can not be considered to be independent of the levels of other factors. 81 Table 13 Results for Nov 16, Oct 28 and Oct 20 combined data sets backwards stepwise regressions on the accumulation of dry matter in the reactor vessels. DRY MATTER R 2 = 0.7367 n=162 F=109.83 REGRESSION P < 0.0001 COEFFICIENT STD ERROR F P INTERCEPT 77.7732 FR -2.3153601 0.441261 27.5 0.0001 RE 0.0896085 0.022968 15.2 0.0001 FRE 0.00004747 0.00000842 31.8 0.0001 D3FE 0.0007426 0.00013844 28.7 0.0001 ( F = f e e d , R = temp, E = e l a p s e d t i m e , D = d a t e ) Note that time ("E" for elapsed time) is a significant modifier of the effects in 3 of 4 interaction terms. This implies the response to the factors changes as the experiment progresses. The implication of this is that the effect of feed strength and temperature on the sludge changes as the sludge matures. This is not predicted from any equation that models the influence of temperature on sludge growth, but is logical if one hypothesizes a succesion of biota in the reactor. If the kinetic response is indeed a very dynamic characteristic and succession and selection continually alter the sludge then the utilization of simple numerical kinetic constants may be somewhat limited. On the other hand, this implies that breeding sludge for improved performance is possible. An important characteristic of the regression results is the scale of the standard error of the estimate of the numerical coeffficients. Typically it is around 25% of the value of the coefficient. A confidence interval for the coefficients can be constructed from the standard error and the t value for the level of confidence desired in the results. Roughly, the 95% confidence (t= 1.96) range of the coefficients is about _+ 50% of the value in the table. This obviously hampers application of the results. 82 The results for the volatile solids (table 14) are quite similar to those for dry matter. Again, there are significant differences between the October experiments and the November experiment, giving rise to two response forms. The factors interact significantly, indicating that the response to one factor is significantly modified by other factors. The volatile solids data bare some relationship to the dry matter data as volatile solids are calculated from the dry matter results. It is not surprising that the results of these regressions are quite similar. Table 14 The accumulation of the volatile solids in the reactors as described by the predictor variables from the backwards stepwise regressions. VOLATILE SOLIDS R 2=0.7653 N=162 F=21.73 REGRESSION P < 0.0001 COEFFICIENT ERROR F P INTERCEPT 77.697813 E -1.944801 0.408076 22.71 0.0001 RE 0.085769 0.021241 16.31 0.0001 FRE 0.0000402 0.0000078 26.63 0.0001 D3FE 0.0007930 0.000128 38.36 0.0001 ( F = f e e d , R = temp, E = e l a p s e d t i m e , D = d a t e ) The models have a fairly high degree of unexplained variation. This is reflected by the relatively low rv value indicating that the model is far from a precise predictor of growth in the reactors. The observed accumulation of dry matter or volatile solids could not be completely related to the independent variables. A result of major importance here is that there are significant differences between the experimental dates. This is directly implied by the requirement of the D terms (which collect the data from the different dates for the regression anatysis) variable to model the growth response. This means there are two principle responses to the treatments, one for October 20th and 28th and a different response in the November 16 experiment. 83 The interaction between factors was of major importance in predicting the growth response. Note feed strength ("F"), elapsed time ("E") and the real temperature in the reactors ("R") all interact with each other. This indicates that the response of the sludge to a given temperature treatment changes with both feed strength and time. Again the scales of the error terms are high relative to the coefficients. This means that temperature altered the response to feed strength and vice versa. This particular interaction was not always the most significant predictor but an interaction term usually was. One variable modifies the response to the other. The major interactions are indicated on the table. Table 15 presents the results of the regressions on the % transmittance data. The three different dates are all significantly different, with the interactions between feed strength and temperature being different on each date. This result is credible in light of the indication of increasing variation from the coefficient of variation results reported previously (Tables 8 - 12). An increase in bacterial numbers will be represented as a decrease in transmission of light. The responses are quite variable but still show an initial decrease in bacterial numbers followed by more concerted growth. The % transmittance results are illustrated in figures 10, 11 and 12 and summarized in table 15. They indicate growth relationships that are fairly different from those of the preceding regressions. However this is not surprising as % transmittance reflects the growth of a small component of the activated sludge community. This component (primarily bacteria) grows and reponds differently from the community as a whole. This is similar to the findings of Zanoni (1969) who found that the value of some kinetic parameters (e.g. the temperature sensitivity coefficient) was affected by the procedure used to generate the data. 84 F i g u r e 10 The change of the reactor % transmittance for the October 20th October 28th and November 16th experiments for the 1000 COD feeding regime. to < y— to < or % T R A N S M I T T A N C E 1 0 0 0 C O D F E E D 1 0 * C 85 100-1 60-40-20-0-O OCT 20 A OCT 28 • NOV-1.6 20 I 40 60 80 ELAPSED TIME [HOURS] 2 0 ' C 100-1 40-20-0-O OCT 20 A OCT 28 • NOV J6 20 40 60 80 ELAPSED TIME [HOURS] 100 100 tO < (— 100 40-20-3 0 ' C O OCT 20 A OCT 28 • N.°.Y..1.6 20 40 60 80 ELAPSED TIME [HOURS] 100 Figure 11 The change of the reactor % transmittance for the October 20th October 28th and November 16th experiments for the 2000 COD feeding regime. Note that the 20°C response shows a more rapid change for the Nov 16th experiment than the 30°C response. This also occurred for the 3000 COD treatment response (Table 12). 87 O zz. GO < cn CO < cc I— 100 n 80 60 H 40 20 H 0 0 100-, 40-20-% T R A N S M I T T A N C E 2 0 0 0 C O D F E E D 1 0 ' C A" 3 O OCT 20 A OCT 28 • N0V_16 20 40 60 80 ELAPSED TIME [HOURS] 2 0 ' C O OCT 20 A OCT 28 • N 0 V J 6 20 40 60 80 ELAPSED TIME [HOURS] 100 100 100- , 3 0 * C 40 60 80 ELAPSED TIME [HOURS] 100 88 Figure 12 The change of the reactor % transmittance for the October 20th October 28th and November 16th experiments for the 3000 COD feeding regime. Note that the 20°C response is greater and more rapid than the 30°C response. 89 CO z : < i— 1 0 0 8 0 6 0 -4 0 -2 0 -0 0 1 0 0 - i 1 0 0 % T R A N S M I T T A N C E 3 0 0 0 C O D F E E D 1 0 ' C O OCT 20 A OCT 28 • NOV 16 2 0 4 0 6 0 8 0 ELAPSED TIME [HOURS] 2 0 ' C 4 0 6 0 8 0 ELAPSED TIME [HOURS] 3 0 ' C . 2 0 - O OCT 20 A OCT 28 • NOV 16 2 0 4 0 6 0 8 0 ELAPSED TIME [HOURS] 1 0 0 1 0 0 1 0 0 90 Again there is a significant difference between dates, however now all three dates are seen to be significantly different in their growth responses. In each case the best descriptor variables are complex interaction terms rather than simple variables such as feed strength or temperature. This indicates that the response to feed is significantly altered by the temperature or the feed strength or the elapsed time and vice versa. Stated another way, the description of growth as the result of of feed strength and temperature changes significantly with time, feed strength and temperature. The response measured on one date at one feed strength or temperature might not shed any light on the growth occuring on another date under different conditions. Table 15 The results of regressions performed on the complete % transmittance data set. Note that there are three responses: the October 20th response (the simple factors E, RE and FRE), the October 28th response (D2FRE factor) and the November response (D3E, D3FR and D3RE). % TRANSMITTANCE R2=0.7779 N=162 F=77.06 REGRESSION P < 0.0001 COEFFICIENT STD ERROR F P INTERCEPT 79.98998 E 0.438263 0 .075300 33.87 0.0001 RE -0.022032 0 .003626 36.92 0.0001 FRE -0.000006 0 .0000009 42.90 0.0001 D2FRE 0.000002 0 .0000010 6.46 0.0120 D3E -0.431822 0 .0854471 25.54 0.0001 D3FR -0.000083 0 .0000363 5.25 0.0233 D3RE 0.016969 0 .0039995 18.00 0.0001 ( F = f e e d , R = temp, E = e l a p s e d t i m e , D = d a t e ) The regression results were used to calculate predicted growth under the experimental conditions. This simplifies the response forms considerably. The aspects of the curves that are noteworthy are the slopes of the lines and the differences between the dates. To aid comparison between all of the 91 subsequent figures, the axis scales have been standardized. The increase of growth rate with temperature was apparent in all graphs from the regression equations. The predicted growth was always higher at a warmer temperature. This is not always what was observed in the actual growth situations, but is apparently the overall trend. Also, the difference between the different dates is not consistent. It seems to be greater at higher feed strengths and temperatures. Figure 13 illustrates the linear response of the dry matter accumulation to the 1000 COD feeding regime at 10°C, 20°C and 30°C. In all cases the November 16 experiment either accumulated biomass faster or lost biomass slower than the other dates. This trend was consistent throughout, all of the results. The 2000 COD results (Figure 14) and the 3000 COD results (Figure 15) share the same property but the disparity between the dates becomes larger at the higher feed strengths. The 10°C treatment lost biomass for all feed strengths during the October experiments but only the 10°C 1000 COD treatment lost biomass for the November experiment. The volatile solids growth response is similar to the dry matter growth response. Figures 16, 17 and 18 present the 1000 COD, 2000 COD and 3000 COD treatments respectively. Again the November 16 experiment accumulated biomass faster than the October experiments. However, volatile solids were consumed only in the October 1000 and 2000 COD treatments and not at all in the November experiment. The % transmittance regressions detected significant differences between the October 20 and 28 experiments. At the lowest feed strength, 1000 COD the two experiments produced almost identical predicted responses at all temperatures (Figure 19). However, as the feed strength becomes stronger (ie 2000 and 3000 COD) the predicted growth response becomes more disparate (Figures 20 and 21, respectively). Unfortunately, a time related trend in Figure 13 The growth of dry matter fed 1000 COD substrate as predicted from the regressions on the data set. The two lines represent the different response forms calculated from the data. D R Y M A T T E R A C C U M U L A T I O N 1 0 0 0 C O D F E E D 1 0 ' C 450- i 400-\ 350-300-111 250-i — 200-i — < 150-RY 100-Q 50-0-450 ^ 400 350-§j 300 250 200-150->-Q 100-J Q 50 - I O OCT 20 ond Oct 28 A NOV 16 10 20 30 40 ELAPSED TIME [HOURS] 50 2 0 ' C O OCT 20 ond OCT 28 A NOV 16 ^ y = = ^ ^ ^ o o 10 20 30 40 ELAPSED TIME [HOURS] 50 Figure 14 The growth of dry matter in response to 2000 COD feed as predicted from the stepwise regressions. The two lines represent the different response forms calculated from the data. 450 400 350 300 250 200 150 100 50 0 DRY MATTER ACCUMULATION 2000 COD FEED 10'C O OCT 20 ond Oct 28 A NOV 16 ELAPSED TIME [HOURS] 20' C 450 - i 0 10 20 30 40 ELAPSED TIME [HOURS] 30' C 96 Figure 15 The growth of dry matter in response to 3000 COD feed as predicted from the stepwise regressions. The two lines represent the different response forms calculated from the data. 450 -1 — 400-o> 350-_£ 300-cc 1.1 250-MATT! 200-150-DRY 100-50-0-DRY MATTER ACCUMULATION 3000 COD FEED 10'C O OCT 20 and Oct 28 A NOV 16 20 30 ELAPSED TIME [HOURS] 40 50 20' C 450 400 CT> 350 300-I 250 200 150 100 50-| 0 CC >-CC o O OCT 20 and OCT 28 A NOV 16 20 30 40 ELAPSED TIME [HOURS] 50 30' C cc >-cc a 450 400 350 300-250-200-150 100-! 50 H 0 O OCT 20 ond OCT 28 A NOV 16 10 20 30 40 ELAPSED TIME [HOURS] 50 Figure 16 The growth of volatile solids in response to 1000 COD substrate as predicted from the regression results. The two lines represent the different response forms calculated from the data. ,—I 4 5 0 - i \ 4 0 0 -3 5 0 -CO 3 0 0 -Q _J 2 5 0 -o 00 2 0 0 -Ld _ l 150-ATI 100-_ i 5 0 -o > 0 -V O L A T I L E S O L I D S A C C U M U L A T I O N 1 0 0 0 C O D F E E D 1 0 ' C O OCT 20 ond Oct 28 A NOV 16 0 = 6 = 10 20 3 0 40 ELAPSED TIME [HOURS] 50 2 0 ' C 1 1 450-, \ CJ) 4 0 0 -3 5 0 -CO 3 0 0 -o 1 2 5 0 -o CO 2 0 0 -Ld 1 150-ATI 100-i o 5 0 -> 0 -O OCT 20 ond OCT 28 A NOV 16 -A-i 10 20 3 0 ELAPSED TIME [HOURS] 40 -A O 50 3 0 ' C ,—I 450-, \ 4 0 0 -3 5 0 -CO 3 0 0 -Q 1 2 5 0 -O CO 2 0 0 -UJ _ l 150-ATI 1 0 0 -_ j 5 0 -o > 0 -0 O OCT 20 ond OCT 28 A NOV 16 10 20 3 0 40 ELAPSED TIME [HOURS] 50 Figure 17 The growth of volatile solids in response to 2000 COD substrate predicted from the regression results. The two lines represent the different response forms calculated from the data. 450-1 400-JL 350-CO 300-Q _ l 250-o CO 200-ILE 150-100-_ i 50-o > o -VOLATILE SOLIDS ACCUMULATION 2000 COD FEED 10'C O OCT 20 ond Oct 28 A NOV 16 101 10 20 3 0 4 0 ELAPSED TIME [HOURS] 5 0 E 450-i 400 350 H 20' C O OCT 20 ond OCT 28 A NOV 16 2 0 3 0 ELAPSED TIME [HOURS] 4 0 5 0 30' C 1 — , 450-j 400-JL 350-CO 300-Q 1 250-O CO 200-Lul _ l 150-ATI 100-_ _ i 50-o > 0-O OCT 20 and OCT 28 A NOV 16 - A 10 20 3 0 4 0 ELAPSED TIME [HOURS] 5 0 Figure 18 The growth of volatile solids in response to 3000 COD substrate predicted from the regression results. The two lines represent the different response forms calculated from the data. .—. 450-, \ u> 400-JL 350-CO 300-Q _l 250-O CO 200-ILE 150-• 100-o 50^ VOLATILE SOLIDS ACCUMULATION 3000 COD FEED 10'C O OCT 2 0 o n d Oc t 2 8 A N O V 16 103 -A • O A- •A A-O -o o A -o _~-A 10 20 3 0 4 0 ELAPSED TIME [HOURS] O 5 0 20' C I—I 450 - j \ 400-350-CO 300-o _ J 250-o CO 200-Ld _l 150-ATI 100-_i 50-o > o-O OCT 2 0 o n d OCT 2 8 A NOV 16 -A"' -A 10 20 3 0 4 0 ELAPSED TIME [HOURS] 5 0 30' C ,—. 450-j 400-JL 350-CO 300-Q _ J 250-O CO 200-LU _l 150-ATI 100-_i 50-o > 0-O OCT 2 0 o n d OCT 2 8 A NOV 16 ELAPSED TIME [HOURS] Figure 19 T h e growth of % transmittance in response to 1 0 0 0 C O D substrate as predicted from the regression results. T h e three lines represent the different response forms calculated from the data. 105 % TRANSMITTANCE 1000 COD FEED 10'C 100 CO < cn i— 40-20- O OCT 20 A OCT 28 • NO V J_6 10 20 30 40 ELAPSED TIME [HOURS] 50 Figure 20 The growth of % transmittance in response to 2000 COD substrate predicted from the regression results. The three lines represent the different response forms calculated from the data. % TRANSMITTANCE 2000 COD FEED 10'C 40-20-0 O OCT 2 0 A OCT 2 8 • N O V 16 0 10 ' 20 30 ELAPSED TIME [HOURS] 20* C 40 50 100 80 v Q 60-40-20 0 O OCT 2 0 A OCT 2 8 • N O V 16 10 20 30 40 ELAPSED TIME [HOURS] 30' C 50 100 n 80 u • • -•A A 60-40-20-• O -O • ~ l 10 O OCT 2 0 A OCT 2 8 20 30 40 ELAPSED TIME [HOURS] 50 Figure 2 1 The growth of % transmittance in response to 3000 COD substrate as predicted from the regression results. The three lines represent the different response forms calculated from the data. 109 CO -z. < cn t— 100 40-20-0 % TRANSMITTANCE 3000 COD FEED 10*C 10 O O C T 2 0 A O C T 2 8 • N O V J_6 20 30 40 ELAPSED TIME [HOURS] 50 100-, 20' C O 80-J CO zz. < cn (— 60-40 H 20 0 10 O O C T 2 0 A O C T 2 8 • N O V 1 6 •A-O 20 30 40 ELAPSED TIME [HOURS] •A -O 50 CO < cn t— 100 40-20-30' C •• O O C T 2 0 A O C T 2 8 • N O V 1 6 10 20 30 40 ELAPSED TIME [HOURS] 50 110 response did not occur between the predicted responses: October 20 and November 16 both accumulated opaque materials faster than the October 28 experiment. The results of Braha and Hafner (1987) are more appealing in this regard as kinetic parameters in their reported data show a consistent change with the sequence of experiments (see table 4). However, if some sort of biological succession is the cause of this effect, it does not necessarily imply a smooth sequence will result kinetically. Significant differences among dates are present in the earlier experiments as well. Tables 16 and 17 present regressions on the Oct 5 and September 16 dry matter and volatile solids data respectively. These experiments both had the same cycle length and inoculum size and should be comparable between themselves. Again, the dates are significantly different in their growth response to temperature and feed strength. The factors interact significantly and correlation between the models and the data is fairly high (R > 90% for both dry weights and volatiles). Table 16 The stepwise regressions on the dry matter from the combined results of the October 5 and September 16th experiments. DRY MATTER R 2 = 0.9210 N=90 F=135.40 REGRESSION P < 0.0001 COEFFICIENT ERROR F P INTERCEPT 34.4360 E -0.59971 0.119 25.35 0.000 RE 0.03385 0.004 48.65 0.000 DF -0.00971 0.001 44.73 0.000 DR -0.96766 0.08836 119.94 0.000 DFR 0.00044 0.00008 32.89 0.000 DFE 0.00021 0.00008 6.86 0.011 DFRE -0.00001 0.00000 9.43 0.003 I l l Table 17 The regression results on the combined October 5 and September 16 volatile solids results. V O L A T I L E SOLIDS RZ=0 .9053 N=90 F = l l l . 97 REGRESSION P < 0. 0001 COEFFICIENT ERROR F P INTERCEPT 27.87717 E -0.49839 0.10858 21.07 0.000 RE 0.03248 0.00442 53.91 0.000 DF -0.00766 0.00132 33.51 0.000 DR -0.79676 0.08054 97.87 0.000 DFR 0.00036 0.00007 26.34 0.000 DFE 0.00018 0.00007 6.48 0.013 DFRE -0.00001 0.00000 10.25 0.002 It must be stressed that the October 5 and September 16th experiments are not comparable to the October 20th, 28th and November 16th experiments due to differences in cycle time and inoculum density. The squared independent variables were regressed on the dry weights, volatile'solids data and % transmittance results. The resulting model is the least squares fit of the combinations of squared factors and should be more capable of fitting parabolic responses than simple linear regression models if such trends exist in the data. However, as pointed out by Williams (1967), the descrepancy between a linear model and a hyperbolic one would be at most 6%. The same regressions were run for the squared variables as for the untransformed variables. Tables 18, 19 and 20 illustrate October 20th, 28th and November 16th dry matter, volatile solids and % transmittance results respectively. As can be seen, the squared terms are not as good as the linear fits, indicating the higher order model is not a more accurate representation of growth. Table 18 Results for Nov 16, Oct 28 and Oct 20 squared variables from the combined dry matter data sets in the reactor vessels. D R Y M A T T E R R 2 = 0.7047 n=162 F= 73.14 REGRESSION P < 0.0001 COEFFICIENT STD ERROR F P INTERCEPT 90.2269 E -0.009653 0.00536 3.25 0.0735 RE 0.0000657 0.000015 20.24 0.0001 FRE 0.0000000 0.000000 5.99 0.0155 D3FE 0.0000000 0.000000 18.23 0.0001 D3RE -0.0000316 0.000011 7.83 0.0058 Table 19 Results for Nov 16, Oct 28 and Oct 20 squared variables from the combined volatile solids data sets in the reactor vessels. V O L A T I L E SOLIDS R 2 = 0.7091 n=162 F= 95.70 REGRESSION P < 0.0001 COEFFICIENT STD ERROR F P INTERCEPT 85.6564 RE 0.000028 0.0000057 24.60 0.0001 FRE 0.000000 0.0000000 17.62 0.0001 D3FE 0.000000 0.0000000 26.78 0.0001 D3FRE 0.000000 0.0000000 4.87 0.0288 Table 20 Results for Nov 16, Oct 28 and Oct 20 squared variables from the combined % transmittance data sets in the reactor vessels. % T R A N S M I T T A N C E R 2 = 0.6794 REGRESSION n=162 F= 35.80 P < 0.0001 COEFFICIENT STD ERROR INTERCEPT 81 .2186 R -0 .005446 0.003255 2.80 0.0964 E 0 .004055 0.001527 7.05 0.0087 FR -0 .000000 0.000000 3.20 0.0755 FE -0 .000000 0.0000000 19.71 0.0001 RE -0 .000012 0.0000024 24.18 0.0001 D2FR 0 .000000 0.0000000 4.52 0.0351 D3R -0 .007487 0.0034356 4.75 0.0308 D3E -0 .004714 0.0014441 10.66 0.0014 D3RE 0 .000009 0.0000024 15.34 0.0001 113 4.3 Kinetic parameter calculations The growth data obtained from the experiments were used to calculate standard Monod kinetic coefficients and Streeter-Phelps temperature sensitivity coefficients. Standard deviations, standard errors of the estimates and confidence intervals were also calculated. The results indicate that, despite control of temperature and feed quality and feed strength, the major kinetic parameters cannot be resolved with any real confidence. The reason may be the lack of steady states in the SBR, however it is possible that the same variability could be encountered in continuous flow systems. As the environment keeps fluctuating so a procession of organisms become active and alternately dominate the kinetic response. To calculate the Monod kinetic parameters Umax and Ks a least squares regression was calculated for the inverse growth rate versus the inverse feed strength (Lineweaver-Burke plot, 1/rate vs 1/[S]). The standard error of the estimate was included with each regression parameter. The standard error as a fraction of the mean is reported as an indication of the level of confidence that can be placed on the results. The result is that again, although the numerical results are often in line with the literature, they cannot be resolved with any confidence. The October data is from the final (52 hour) sampling point. The November results are from the cumulative results from 46 hours onwards. Again note that the variance within a single sample may be greater than across several samples. Hanes and Eadie-Hofstee form regressions were also calculated but the results did not provide any better resolution of the kinetic parameters. Indeed, the results seemed to be irrational more often than the Lineweaver-Burke form regressions. This is thought to result from the forms of the equations in which the kinetic parameters are calculated as the reciprocals of numbers close to zero. Table 21 The results of Line weaver-Burke form least squares regressions on the dry matter data from the pooled October 20th and 28th experiments and the November 19th experiment. Oct 20 28 Dry matter Average Umax .061 Average Ks -37.859 Nov 19th Dry matter UMAX 0.1358 KS 181.35 95% confidence i n t e r v a l as f r a c t i o n of mean _+ 26% + 20% Average error as f r a c t i o n of mean +, 26% + 29% Table 22 The results of Lineweaver-Burke form least squares regressions on the volatile solids data from the pooled October 20th and 28th experiments and the November 19th experiment. Oct 20 28 volatile solids 95% confidence i n t e r v a l as f r a c t i o n of mean Average umax .426 +_ 30 % Average Ks -92.213396 _+ 40 % November 16 volatile solids UMAX .2497 +_ 23 % Ks 342.5129 + 34 % 115 Table 23 The results of Lineweaver-Burke form least squares regressions on the % transmittance data from the pooled October 20th and 28th experiments and the November 16th experiment. Oct 20 28 % transmittance 95% confidence i n t e r v a l as f r a c t i o n of mean Average umax .392 _+ 85 % Average Ks 130.111 _+ 156 % Nov % transmittance UMAX .0103 _+ 33 % KS 206.0684 + 189 % Temperature effects on growth Table 24 illustrates the temperature response calculated from the October 20th and 28th pooled data. It was felt that, as these dates did not appear to be significantly different in growth from the regression analysis, the increased amount of data should aid resolution of the kinetic parameters. The Q ^ Q value was defined as either the increment in dry matter (or volatile solids) at 20°C divided by the increment in dry matter (or volatile solids) at 10°C or as the 30°C growth increment divided by the growth increment at 20°C. A Q 2 Q term was calculated by the growth increment of the 30°C treatments divided by the 10°C treatments. Theta was calculated as the 10th root of the Q ^ Q value. The confidence interval around each term would be +_(error)*(t value for the appropriate degrees of freedom and confidence level). For 95% confidence and 6 degrees of freedom t=2.447 (Snedecor and Cochrane, 1978). The October data are from the final (52 hour) sampling point. The November results are from the cumulative results from 46 hours onwards (to 94 hours growth, 8 degree of freedom, t= 2.306). Table 24 The effect of temperature on the growth of dry matter and volatile solids in the reactor vessels for the combined October 20th and 28th data sets. Dry matter Average Q 1 0 theta 2.6862 1.1039 95% Confidence i n t e r v a l + 61 % Volatile solids Average Q 1 0 theta 1.3328 1.0291 + 85 % % transmittance Average Q 1 0 1.715 +_ 58 % theta 1.055 Q20 3.730 +_126 % 1.068 The average temperature sensitivity coefficients are in line with those reported in the literature, but they are quite variable and cannot be resolved with any real confidence. Zanoni (1969) reported several factors that affected the value of the temperature sensitivity coefficients. Among them he lists feed strength, the specific temperature range and the procedure used to collect the growth data. These results are substantiated by the present experiment. The dry matter, volatile solids and % transmittance measurements all produce different Q10 and theta values, none of which are particularly meaningful due to high amounts of variation. This is a particularly important result in light of the experimental designs reported in the literature in which there is no assessment of experimental error or confidence interval. Novak (1974) questions the validity of the Phelps equation for application to sludge growth. experiment. Table 25 reports the effect of temperature on the growth rate of dry matter, volatile solids and % transmittance for the November 16th experiment. As before, the variance is large relative to the mean and the numbers are therefore Essentially the same result occurred in the November 19th 117 not extremely meaningful. The dry matter and volatile results were calculated only from reactors in which growth occurred. Of the 24 data points calculated from the 46 hour, 70 hour and 94 hour results nine data points were negative,indicating a net loss of biomass. Seven had more growth at the cooler temperature and therefore had a Q - ^ Q less than one and eight had a Q - ^ Q greater than one. Only the 8 with aQjQ greater than 1 were used here for the dry matter and volatile solids data in an attempt to reduce the variation in the results. As can be seen, the variance is still high despite this. Table 25 The temperature effects on growth of dry matter, volatile solids and % transmittance in the november 16th experiment. DRY MATTER 95% Confidence i n t e r v a l Q 1 0 2.4543 + 144 % THETA 1.0939 VOLATILES Q 1 0 3.1368 + 180 % THETA 1.1211 % TRANSMITTANCE Q20 Q 1 0 2.800 + 857% 1.289 + 851% THETA 1.108 1.026 The variation in the samples and the change in response over time create a situation where the calculation of kinetic parameters is hindered by large confidence intervals. The results are quite difficult to interpret in a meaningful way due to the potential range in the coefficients. It is not clear that this type of problem will be restricted to the SBR, as it is possibly due to a natural succession of organisms in the reactor vessel. 118 Overall, it is my impression that sludge growth is poorly estimated by simplistic equations and models. However, these models do agree with large scale trends and the growth of subsections of the biotic community. Within wide limits, simplistic models can offer guidance but numerical precision is difficult to obtain. Variation within reactor samples is moderate but the growth response of sludge changes with time. Sludge is a large dynamic ecological system probably having a constant succesion of biota dominating the activity of the community. This situation may be exacerbated in the sequencing batch reactor due fluctuating conditions. 119 Conclusions 1) An experimental design of this type is a valuable tool to find and quantify sources of variability in waste treatment research. Direct conclusions can be made about how to increase the accuracy and scope of further experimentation. 2) The experimental protocol and feeding treatment did not cause obvious morphological or other detectable changes in the sludge characteristics. It is felt that the sludge response was not obviously unrepresentative. 3) The sampling of activated sludge has an inherent variation, possibly due to heterogeneity in the reactor contents. This implies an absolute requirement for replicate samples to accurately determine the reactor contents. 4) The kinetic response of the activated sludge subjected to a sequencing batch reactor operational protocol changes over time. This may be due to a succession of species in the sludge community. The scale of the change of kinetic response may be quite large and has not been adequately documented in the literature. It is also not predicted in existing standard kinetic models, but may limit their utility. 5) The standard kinetic parameters for sludge growth cannot be determined with any practical level of confidence or accuracy. Nonetheless, the apparent kinetic response is quite close to values reported in the literature, which furthers the impression that the response is representative of what could be expected in a larger scale unit. 6) The values calculated for the standard kinetic coefficients are specific to the method of characterizing growth in the reactor. It is doubtful that one measure is better than another due to relatively large confidence intervals aroundff the calculated values. 121 Recommendations 1) Examination of the process biology should be made with an eye towards understanding if biological succession in the reactor can be controlled or manipulated. There may exist a stable communities that are more resistant to successional change. 2) Quantification of the range of process variation over a season is a poorly documented parameter with some relevance to process design. This is estimated by Arrhenhius van't Hoff temperature corrections but the effect is likely to involve biotic shifts in the process. 3) Further studies on the variation inherent in the process should be made. Of specific interest are factors like variation as a function of degree of perturbation or shocking, variation as a function of reactor biomass and general acclimation effects. 4) Studies should be undertaken to determine the range of variation present in other measures taken on waste treatment reactors. It seems unlikely that the level of variability reported here is restricted to the measuring techniques applied in this experiment. Similarly, the calculation of other kinetic parameters are therefore hampered by a general lack of certainty in the derived numbers. 122 References Abranis, B.I. and M.J. M i t c h e l l 1978 "Role of oxygen i n a f f e c t i n g s u r v i v a l and a c t i v i t y of Pelodera punctata (Rhabditae) from sewage sludge" Nematologica 24:456-462 American Society for Microbiology 1981 Manual of methods for general bacteriology American Society for Microbiology Antheunisse, J . and J.I.A. Koene 1987 " A l t e r a t i o n of th aerobic and f a c u l t a t i v e anaerobic b a c t e r i a l f l o r a of the A/B p u r i f i c a t i o n process caused by l i m i t e d oxygen supply" Water Res. 21(1): 129-131 Ardern, E. 1917 "A resume of the present p o s i t i o n of the a c t i v a t e d sludge process of sewage p u r i f i c a t i o n " J . Soc. Chem. Ind. 36(14):822-830 Ardern, E. and W.T. Lockett 1914 "Experiments on the oxidation of sewage without the a i d of f i l t e r s " J . Soc. Chem. Ind. 33(10):523-539 Ardern, E. and W.T. Lockett 1915 "The Oxidation of sewage without the ai d of f i l t e r s Part I I I " J . Soc. Chem. Ind. 34(18):937-943 Ardern, E. and W.T. Lockett 1936 "Laboratory tests for a s c e r t a i n i n g the condition of a c t i v a t e d sludge" J . Inst. Sewage P u r i f i c a t i o n 1(1)-.212-215 Arrhenius S. 1889 "Uber di e reaktionsgeschwindigkeit bei der i n v e r s i o n von rohrzucker durch sauren" Zeit Phys. Chem. 4:226-248 :— c i t e d i n Grady and Lim 1981 Baines, S., H.A.Hawkes, C.H.Hewitt and S.H.Jenkins 1953 "Protozoa as i n d i c a t o r s i n a c t i v a t e d sludge treatment" Sewage and I n d u s t r i a l Wastes 25 (9):1023-1033 Barth, E.F. 1983 "Implementation of sequencing batch reactors for municipal wastewater treatment" 6- symposium on wastewater treatment Ottawa, Canada Bazin M.J. 1980 Mixed Culture K i n e t i c s Chapter 2 of Mixed Culture Fermentations ME Bushell and JE Slater (eds) Benedek, P. and I. Horvath 1967 "A p r a c t i c a l approach to sludge k i n e t i c s " Water Res. 1:663-682 Benedict,A.H. and D.A. Carlson 1973 "Temperature acclimation i n aerobic systems" J . Water Poll.Con. Fed. 45(l):10-24 Benedict,A.H. and D.A. Carlson 1971 "Aerobic heterotrophic b a c t e r i a i n activated sludge" Water Res. 5:1023-1030 Benedict,A.H. and D.A. Carlson 1974 "Rational assessment of the Streeter Phelps temperature c o e f f i c i e n t " J . Water P o l l . Cont. Fed. 46(7): 1792-1799 123 Bick , H. 1969 "An i l l u s t r a t e d guide to c i l i a t e d protozoa used as b i o l o g i c a l i n d i c a t o r s i n freshwater ecology" W.H.O. V.B.C.69:136 — c i t e d i n V. Sladecek 1971 Birch L.C. 1953 "Experimental background to the study to the d i s t r i b u t i o n and abundance of i n s e c t s " E v o l u t i o n 7:136-144) Borrow, A.; S. Brown; E.G. J e f f e r y s and R.H.J. Kessel 1964 "The e f f e c t of temperature on the k i n e t i c s of metabolism of Gibberella fujikuroi i n s t i r r e d c u l t u r e " Can. J . Microb. 10: 445-466 Braha,A. and F. Hafner 1987 "Use of lab reactors to model b i o k i n e t i c s " Water research 21(1):73-81 Brown, A.D. 1957 "Some general properties of a psychrophilic pseudomonad: the e f f e c t s of temperature on some of these properties and the u t i l i z a t i o n of glucose by t h i s organism and Pseudomonas aeruginosa" J . General Microbio. 17: 640-648. Bungay, H.R. and M.L. Bungay 1968 "Microbial i n t e r a c t i o n s i n continuous c u l t u r e " Advances Appl. Microb. 10:269-290 Bushell, M.E. and J.H. Slater 1980 Mixed Culture Fermentations Canale, R.P.; T.D. Lustig, P.M. Kehrberger and J.E. Salo 1973 "Experimental and mathematical modelling studies of protozoan predation on b a c t e r i a " Biotech. Bioeng. 15:707-728) Chartrain, M.; L. Bhatnagar and J.G. Zeikus 1987 "Microbial ecophysiology of whey biomethanation: comparison of transformation parameters, species composition and st a r t e r culture performance i n continuous cu l t u r e " Appl. Enviro. M i c r o b i o l . 53(5):1147-1156 Chase, L.M. 1976 "A dynamic k i n e t i c model of the activated sludge process" Proc. 31st Indust. Waste Conf. Perdue pp 43-53 Chiu, S.Y.; L.T.Fan, I.C. Kao and L.E. Erickson 1972 " K i n e t i c behavior of mixed populations of activated sludge" Biotechnol. Bioeng. 14:179-199 Chiu, S.Y.;L.E. Erickson, L.T.Fan and I.C. Kao 1972 "Kinetic model i d e n t i f i c a t i o n i n mixed populations using continuous cu l t u r e data" Biotechnol. Bioeng. 14:207-231 Chohji, T. and T. Sawada 1983 "E f f e c t s of temperature s h i f t on growth rate of Escherichia coli BB at lower glucose concentration" Biotech Bioeng 25:2991-3003 Converti, A.; M. Del Borghi, M. Z i l l i and G. F e r r a i o l o 1987 "Kinetics of alcohol fermentations c a r r i e d out i n r o t a t i n g b i o l o g i c a l surface reactors" Biotechnol. Bioeng. 29:16-23 Curds, C.R. 1965 "An e c o l o g i c a l study of the c i l i a t e d protozoa i n a c t i v a t e d sludge" Oikos 15:282-289 124 Curds, CR. 1971 "Computer simulations of the microbial population dynamics i n the activated sludge process" Water Res 5:1049-1066 Curds, CR. 1973 "The r o l e of protozoa i n the a c t i v a t e d sludge process" American Zool. 13:161-169 Curds, CR. and J.M. Vandyke 1966 "The feeding habits and growth rates of some fr e s h water c i l i a t e s found i n activated sludge plants" Journal Applied Ecology 3:127-137 Curds, CR. and A. Cockburn 1968 "Studies on the growth and feeding of Tetrahymenapyriformis i n axenic c u l t u r e " J . Gen. Microbio. 54:343-358 Curds, CR. and G.J. Fey 1969 "The e f f e c t of c i l i a t e d protozoa on the f a t e of Escherichia coli i n a c t i v a t e d sludge process" Water Res 3:853-867 C u r t i s , E.J.C. and CR. Curds 1971 "Sewage fungus i n r i v e r s i n the United Kingdom: The slime community and i t s constituent organisms" Water Res. 5:1147-1159 Dabes, J.N.; R.K. Finn and C R . Wilkie 1973 "Equations for substrate l i m i t e d growth: The case for Blackman k i n e t i c s " Biotechnol. Bioeng 15:1159-1177 Dennis, R.W. and R.L.Irvine 1979 " E f f e c t of f i l l to react r a t i o on sequencing batch b i o l o g i c a l reactors" J . Water P o l l . Cont. Fed. 51(2):255-263 Dent,V.E, M.J. Bazin and P.T. Saunders 1976 "Behavior of Dictyostelium discoideum and Escherichia c o l i grown together i n chemostat cu l t u r e " Arch. M i c r o b i o l . 109:187-194 Drew, S.W. 1981 Liquid culture Chapter 10 of Manual of methods for general bacteriology American Society for Microbiology Eckenfelder, W.W. and E.L. Barnhart 1961 "The e f f e c t of organic substances on the transfer of oxygen from a i r bubbles i n water" Amer. Inst. Chem. Eng. J . 7(4):631-634 Eikelboom, D.H. and H.J.J, van Buijsen 1981 Microscopic Sludge Investigation Manual TNO Research I n s t i t u t e The Netherlands Friedman, B.A. and P.R. Dugan 1968 " I d e n t i f i c a t i o n of zooglea species and the r e l a t i o n s h i p to zoogleal matrix and f l o e formation" J . B a c t e r i o l . 95(5):1903-1909 Garrett, M.T. and C.N. Sawyer 1952 "Kinetics of removal of soluble BOD by a c t i v a t e d sludge" Proceedings 7 Ind. Waste Conf. Perdue U n i v e r s i t y Engineering Extension series #129: 537-552 Gates, W.E. and J.T. Marlar 1968 "Graphical analysis of batch culture data using the Monod expressions" J . Water P o l l . Cont. Fed. 40(11):R469-R476 125 Gaudy, A.F. and E.T. Gaudy 1971 " B i o l o g i c a l concepts for the design and operation of the acti v a t e d sludge process" E.P.A Water P o l l u t i o n Research series Report #17090 1972 "Mixed microbial populations" Advances Biochem. Eng. 2:97-143 Gause, G.F. 1934 The Struggle for Existence, Hafner books, New York Goodman, B.L. and A.J. Englande 1974 "A u n i f i e d model of the activated sludge process" J . Water P o l l . Cont. Fed. 46(2):312-332 Garrett, M.T. and C.N. Sawyer 1952 "Kinetics of removal of soluble BOD by acti v a t e d sludge" Proc. 7th Indust. Waste Conf. Perdue; Exten. Ser. 79 (36):3-51 Grady and Lim, B i o l o g i c a l Wastewater Treatment, Marcel Dekker 1980 Hawkes H.A. 1963 The ecology of waste water treatment Oxford, Pergamon Press Herbert, D. 1962 Continuous culture of microorganisms. Proceedings 2nd International Symposium Prague, I. Malek, K.Beran and J . Hospodka (eds) Academic Press publishing} — c i t e d i n Webster (1983) Hetherington, A. 1933 "The culture of some holotrichous c i l i a t e s " Arch. Protistenk 80:154-222 — c i t e d i n Curds (1965) Heukelekian, H. and H.B. Schulhoff 1938 "Studies on the c l a r i f i c a t i o n stage of the acti v a t e d sludge process IV: Preliminary notes on the c l a r i f y i n g organisms i n activated sludge" Sewage Works Journal 10:43-48 Heukelekian, H. and M.L. Littman 1939 "Carbon and nitrogen transformations i n the p u r i f i c a t i o n of sewage by the acti v a t e d sludge process II Morphological and biochemical studies of the zoogleal organisms" Sewage Works Journal 11 (5):752-763 His s e t t , R; E.A. Deans and M.R. Evans 1982 "Oxygen consumption during batch aeration of piggery s l u r r y at temperatures between % and 50°C" A g r i c u l t u r a l Wastes 4:477-487 Hoepker E.C. and E.D. Schroeder 1979 "The e f f e c t of loading rate on batch acti v a t e d sludge e f f l u e n t q u a l i t y " J . Water P o l l . Cont. Fed. 51 (2):264-273 Huffaker C B ; K.P. Shea and S.G. Herman. 1963"Experimental studies on predation: complex dis p e r s i o n and l e v e l s of food i n an acarine predator-prey i n t e r a c t i o n " H i l g a r d i a 34:305-330 Hynes H.B.N. 1970 The Biology of Running Waters Pergamon Press 126 Irvine, R.L.; L.H. Ketchum, R. Breyfogle and E.F. Barth 1983 "Municpal a p p l i c a t i o n of sequencing batch treatment" J . Water P o l l . Control Fed. 55(5):484-489 Irvine, R.L.; J.E. Alleman; G. M i l l e r and R.W. Dennis 1980 "Stoichiometry and k i n e t i c s of b i o l o g i c a l waste treatment" J . Water P o l l . Control Fed. 52 (7):1997-2006 Irvine, R.L. and R.O. Richter 1976 "Computer simulation and design of sequencing batch b i o l o g i c a l reactors" Proc. 31 s t Ind. Waste Conf. Perdue U n i v e r s i t y Jackson, K.M. and J . Berger 1984 "Survival of c i l i a t e d protozoans under starvation conditions and at low b a c t e r i a l l e v e l s " M i c r o b i a l Ecology 10:47-59 Jenkins, S.H. 1942 "Role of protozoa i n the activated sludge process" Nature 150 (3812):607 Jones, G.L. 1973 " B a c t e r i a l growth k i n e t i c s : measurement and s i g n i f i c a n c e i n the a c t i v a t e d sludge process" Water Res. 7:1475-1492 Jordan et a l 1971 "Evaluating t r e a t a b i l i t y of selected i n d u s t r i a l wastes" Proc. 26th Ind. Waste conf Perdue Uni v e r s i t y Engineering extension series c i t e d i n Gardy and Lim 1980 Kalantar A.H. 1987 "Estimation of the parameters of A r r h e n i u s - l i k e equations" Chem Eng J . 34:159-164 Ketchum, L.H.; R.L. Irvine, and R.W. Dennis 1979 "Sequencing batch reactors to meet compliance regulations" Amer. Inst. Chem. Eng. symposium seri e s 75(190):186-191 Ketchum, L.H.; R.L. Irvine, R.E. Breyfogle and J.F. Manning 1987 "A comparison of b i o l o g i c a l and chemical phosphorus removals i n continuous and sequencing batch reactors" J . Water P o l l . Control Fed. 59(1):13-18 Koch, A.L. 1981 Growth measurements Chapter 11 of American Society for Microbiology 1981 Manual of methods for general bacteriology American Society for Microbiology Konak, A.R. 1974 "Derivation of a generalised Monod equation and i t s a p p l i c a t i o n " J . Applied Chem. Biotechnol. 24:453-455 LeChevallier, M.W.; CD. Cawthon and R.G. Lee 1987 "Factors promoting s u r v i v a l of b a c t e r i a i n chlorinated water supplies" Appl. Environ. Microb. 54(3):649-654 Leduc, R.; T.E. Unny, and E.A. McBean 1986 "Stochastic model of f i r s t order BOD k i n e t i c s " Water Res. 20(5):625-632 Lloyd, L. 1944 "Sewage b a c t e r i a bed fauna i n i t s natural s e t t i n g " Nature 154 (3908):397 127 Luong, J.H.T. 1987 "Generalization of Monod k i n e t i c s f o r an a l y s i s of growth data with substrate i n h i b i t i o n " Biotechnol. Bioeng. 29:242-248 Manning, J.F. and R.L. Irvine 1985 "The b i o l o g i c a l removal of phosphorus i n a sequencing batch reactor" J . Water P o l l . Control Fed. 57(l):87-94 McKinney,R.E 1962 Microbiology for Sanitary Engineers McGraw H i l l New York McKinney,R.E and M.P. Horwood 1952 "Fundamental approach to the activat e d sludge process" Sewage and I n d u s t r i a l Wastes 224(2):117-123 McKinney,R.E and A.Gram 1956 "Protozoa and Activated Sludge" Sewage and In d u s t r i a l Wastes 28(10):1219-1231 Megee,R.D.,J.F.Drake, A.G.Fredrickson, and H.M. Tsuchiya; 1972 " Studies on i n t e r microbial symbiosis" Can. J . M i c r o b i o l . 18:1333-1742 Mines, R.O. and J.H. Sherrard 1986 "Evaluating and u t i l i z i n g b i o k i n e t i c constants i n activa t e d sludge" Water A i r and S o i l P o l l u t i o n 27:323-328 Monod J . 1949 "The growth of b a c t e r i a l c u l t u r e s " Annual Rev. Microbiol 3:371-394 Moser, A. 1974 "Die k i n e t i k der biologischen abwasserreinigung" Das Gas Wasserfach 115:411-417 — c i t e d i n Braha and Hafner, 1987 Moser,H. 1958 "The dynamics of b a c t e r i a l populations maintained i n a chemostat" Carnegie I n s t i t u t e p u b l i c a t i o n #614 — c i t e d i n Grady and Lim 1980 Muck, R.E. and C.P.L. Grady 1974 "Temperature e f f e c t s on microbial growth i n CSTRs" J . Environ. Eng. Div. A.S.C.E. Norland,S; M. Heidal and 0. Tumyr 1987 "On the r e l a t i o n between dry matter and volume of b a c t e r i a " M i c r o b i a l E c o l . 13:95-101 Novak J.T. 1974 "Temperature substrate i n t e r a c t i o n s i n b i o l o g i c a l treatment" J . Water P o l l . Control Fed. 46(8):1984-1994 O'N e i l l ,R.V. and R.H. Gardner 1979 Sources of uncertainty i n e c o l o g i c a l models i n Methodology i n systems modelling and simulation Z e i g l e r , B.P.; M.S. Elzas; G.J. K l i r and T.I. Oren (eds) North-Holland publishing Co. Orhon D; Y Cimsit and 0 Tunay 1986 "Substrate Removal Mechanism for Sequencing Batch Reactors" Water S c i Technol 18:21-33 O'Shaughnessy, F.R. 1923 "The phys i c a l aspects of sewage d i s p o s a l " J . Soc Chem Ind. 42(36):359T-370T Os t o j s k i , M.S. 1987 " Mathematical modelling ofsewage treatment plant operation" E c o l o g i c a l Modelling 39:67-83 128 P a l i s , J.C. and R.L. Irvine 1985 "Nitrogen removal i n a low loaded s i n g l e tank sequencing batch reactor" J . Water P o l l . Control Fed. 57(1):82-86 P e i l , K.M. and A.F. Gaudy 1971 "Kinetic constants for aerobic growth of microbial populations selected with various s i n g l e compounds and with municipal wastes as substrates" Applied M i c r o b i o l . 21 (2):253-256 Perret, C.J. 1960 "A new k i n e t i c model of a growing b a c t e r i a l population" J . General M i c r o b i o l . 22:589-617 P i l l a i , S.C. and V. Subrahmanyan 1942 "Role of protozoa i n the activat e d sludge process" Nature 150 (3809):525} P i l l a i , S.C. and V. Subrahmanyan 1944 "Role of protozoa i n the aerobic p u r i f i c a t i o n of sewage" Nature 154 August 5 1944 :179-180 Pimentel,D, W.P. Nagel and J.L.Madden; 1963 "Space-time structure of the environment and the s u r v i v a l of para s i t e host systems" American N a t u r a l i s t V 97:141-167 Prakasam, T.B.S. and N.C. Dondero 1967 "Aerobic heterotrophic b a c t e r i a l populations of sewage and sludge: Part III Adaptation to a synthetic waste" Appl. Microb. 15(5):1128-1137 Randall, C.W.; J.B. Richards, and P.H. King 1975 "Temperature e f f e c t s on aerobic d i g e s t i o n k i n e t i c s " J . Environ. Eng. Div. A.S.C.E. Rashit E. and M. Bazin 1987 "Environmental f l u c t u a t i o n s , p r o d u c t i v i t y and species d i v e r s i t y : An experimental study" Microbial Ecology 14:101-112 Reynoldson T.B.; 1942 Nature V.149:608 Richard, M.; O.Hao and D. Jenkins 1985 "Growth k i n e t i c s of Sphaerotilus natans species and th e i r s i g n i f i c a n c e i n activa t e d sludge bulking" J . Water P o l l . Cont. Fed. 57(1):68-81 Robinson, J.A. and W.G. Characklis 1984 " Simultaneous estimation of Vmax, Km and the rate of endogenous substrate production (R) from substrate depletion data" Microb. E c o l . 10:165-178 Roels, J.A. 1982 "Mathematical models and the design of biochemical reactors" J . Chem. Tech. Biotech. 32:59-72 Roels, J.A. 1983 Energetics and Ki n e t i c s i n Biotechnology E l s e v i e r Books. Rozak, D.B. and R.R. Colwell 1987 "Survival strategies of b a c t e r i a i n the natural environment" M i c r o b i o l o g i c a l Reviews 51(3):365-379 Rozich, A.F. and R.J. C o l v i n . 1986 "E f f e c t s of glucose on phenol degradation by heterogenous populations" Biotech. Bioeng. 28:965-971 129 Sambanis,A. S.Pavlou and A.G. Frerickson 1987 "Coexistence of b a c t e r i a and feeding c i l i a t e s : Growth of b a c t e r i a on autochthonous substrates as a s t a b i l i z i n g factor for coexistence" Biotech, bioeng. 29 :714-728 Sheih W.K. and L.T.Mulcahy 1986 "Experimental Determinations of I n t r i n s i c K i n e t i c C o e f f i c i e n t s for B i o l o g i c a l Waste water Treatment Systems" Water Sc i Technol V.18:1-10 Sheintuch,M.; 0 . Lev, P. Einav and E. Rubin 1986 "Role of exocellular polymer i n the design of activated sludge" Biotech. Bioeng. 28 :1564-1576 Sladecek,V 1971 "Saprobic sequence within the genus V o r t i c e l l a " Water Res. 5 :1135-1140 Sl a t e r , J.H. 1980 Mixed cultures and microbial communities Chapter 1 of Mixed culture fermentations M. E. Bushell and J.H. Slater (eds) Academic Press Snedecor, G.W. and W.G. Cochran 1978 S t a t i s t i c a l Methods ( 6 t h edition) Iowa State U n i v e r s i t y Press Somerville, H.J. 1980 Mixed cultures i n aerobic waste treatment Chapter 4 of Mixed culture fermentations Bushell, M.E. and J.H. Slater (eds) Academic Press Stephens, M.L.; and-G. Lyberatos 1987 " E f f e c t of c y c l i n g on f i n a l mixed cu l t u r e f a t e " Biotech. Bioeng. 29:672-678 Tench, H.B. and A.V Morton 1962 "The a p p l i c a t i o n of enzyme k i n e t i c s to activ a t e d sludge research" J . Inst. Sewage P u r i f . 5:478-486 Tessier, C 1936 Ann Physiol.Physiochim. B i o l . 12:527 Topiwala, H. and C.G. S i n c l a i r 1971 "Temperature r e l a t i o n s h i p i n continuous cu l t u r e " Biotech. Bioeng. 13:795-813 Tsuchiya,H.M. J.F.Drake, J.L.Jost and A.G. Fredrickson 1972 "Predator prey i n t e r a c t i o n s of Dictyostelium discoideum and Escherichia c o l i i n continuous c u l t u r e " J . of Bacteriology V 110:1147-1153 Van den Ende, P. 1973 "Predator prey i n t e r a c t i o n s i n continuous c u l t u r e " Science 181:562-564 Van Den Heuvel, J . C ; and H.H. Beeftink 1988 "Kinetic e f f e c t s of simultaneous i n h i b i t i o n by substrate and product" Biotech. Bioeng. 31:718-724. V a v i l i n , V.A. 1982 "The theory and design of aerobic b i o l o g i c a l treatment" Biotech. Bioeng. 24:1721-1747 130 Webster, I.A. 1983 "The continuously fed batch reactor f or measuring microbial growth rates" Biotech. Bioeng. 25:2981-2990 Williams F.M. 1967 "A model of c e l l grow dynamics" J . The o r e t i c a l Biology 15:190-207 Wimpenny, J.W.T. 1987 "New approaches to the study of microbial populations" J . Applied Bacteriology, Symposium Supplement 1987 :27S-41S Woombs, M. and J . Laybourn-Parry 1986 "The r o l e of nematodes i n low rate p e r c o l a t i n g f i l t e r sewage works" Water Res. 20 (6):781-787 Yoo Y.J.; M. Marino-Galarraga, J . Hong and R.T. Hatch 1986 "Experimental design for parameter estimation from batch c u l t u r e " Biotech. Bioeng. 28:836-841 Zanoni A.E. 1969 "Secondary e f f l u e n t deoxygenation at d i f f e r e n t temperatures" J . Water P o l l . Cont. Fed. 41:640-


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items