Sunteți pe pagina 1din 216

UNDERSTANDING BIOSOLIDS DYNAMICS IN A

MOVING BED BIOFILM REACTOR

by

Christopher Goode

A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy

Graduate Department of Chemical Engineering and Applied Chemistry, University of Toronto

Copyright by Christopher Goode (2010)


Understanding Biosolids Dynamics in a Moving Bed Biofilm Reactor

Christopher Goode

Doctor of Philosophy, 2010

Department of Chemical Engineering and Applied Chemistry

University of Toronto

Abstract

Biofilm systems such as the moving bed biofilm reactor (MBBR) are finding increased

application in wastewater treatment. One important process that governs MBBRs and yet is

poorly understood is the rate of biofilm detachment. The detachment of cells from biofilm

surfaces controls both the accumulation of biofilm and the quantity of biomass that is suspended

in the bulk liquid phase. This changing balance of attached and suspended cells, in this thesis

named the biosolids dynamics, can impact the efficacy of MBBRs. The goal of this research was

to investigate how the biosolids dynamics are influenced by process changes relevant to applied

wastewater treatment systems and suggest new routes to reactor design and optimization.

To achieve this goal, the work addresses three separate but interconnected lines of

inquiry. First, multivariate analysis (Principal Component Analysis, Partial Least Squares) was

used to examine 2 years of historical data from an MBBR operating at a Canadian pulp mill in

order to identify key process variables, perform process diagnostics, and act as a predictive tool.

Secondly, the effect of calcium concentration on biofilm structure, microbiology and reactor

performance was investigated in four laboratory-scale MBBRs operated at a range of calcium

concentrations (1 to 300 mg/L Ca2+). It was found that above a threshold calcium concentration

between 1-50 mg/L, MBBR biofilms were observed to be thicker with greater density, contain

larger anoxic regions adjacent to the carrier substratum, have more proteinaceous EPS, and have

ii
altered microbial community structure. The results suggest an important role for calcium that

should be considered in the design and operation of MBBRs. In the final line of inquiry, a

diffusion-reaction biofilm model was adapted to represent the key processes of the MBBR. The

model was found to simulate average trends observed in the lab-scale experiments allowing for

quantification of the detachment rate. Transient periods of reactor starvation were also simulated

by introducing a novel metabolic state function to account for down-regulation of metabolism as

a result of starvation. This approach was found to accurately simulate starvation response when

coupled with detachment expressions that were growth-dependant.

iii
Acknowledgments

I would like to acknowledge people who made this work possible. I first recognize my

research supervisor Prof. Grant Allen for the support and guidance of this thesis. His

commitment to training graduate students goes beyond providing direction for research. By

encouraging broad involvement in university life and supporting opportunities to gain leadership,

communications, and industrial experience, Grant enabled my doctorate training to reach far

beyond the work contained in these pages. I am deeply grateful for this rich experience.

I would like to further recognize my colleagues from the Allen lab old and new, as well

as those in other labs, such as the labs of Professors Elizabeth Edwards, Steven Liss, Ramin

Farnood, Emma Master, and Gideon Wolfaardt with whom I had the pleasure of collaborating.

What an outstanding collection of individuals to work with and share ideas.

The work was directly assisted by undergraduate students Teresa Chee, Manuel Popiol,

and Imtiaz Ahmed. I would also like to acknowledge our industrial partner in the work, Irving

Pulp & Paper, who provided financial assistance and tremendous support from mill personnel

such as Jon LeRoy, Andrew Booker, Cindy Milbury, and others. Alan Werker from

AnoxKaldnes also contributed to the work.

Family and friends have been patient with this long educational journey. Their love and

support has been vital to me. I must specifically mention my wife Laura: I couldnt have done it

without you.

This thesis is dedicated to my mom.

iv
1

Table of Contents

Chapter 1: Introduction ............................................................................................................... 1


1.1. The Moving Bed Biofilm Reactor (MBBR) and Biosolids Dynamics................................ 1
1.2. Initial Motivation: Variable Performance of an Industrial-Scale MBBR........................ 4
1.3. Lines of Inquiry for Investigation ........................................................................................ 6
1.3.1 Multivariate Statistical Analysis of an Industrial MBBR .................................................. 6
1.3.2. The Effect of Calcium on MBBR Biofilms ...................................................................... 7
1.3.3 A Theoretical Model of MBBR Biosolids Dynamics during Periods of Starvation........ 10
1.4. Structure of the Thesis......................................................................................................... 12
1.5. Scholarly Contributions ...................................................................................................... 13
Chapter 2: Literature Review.................................................................................................... 15
2.1. Microbial Biofilms ............................................................................................................... 15
2.1.1. Biofilm Microbiology ..................................................................................................... 16
2.1.1.1 Diffusion gradients cause community stratification ................................................. 16
2.1.1.2 Quorum sensing regulates many cell functions in biofilms ...................................... 19
2.1.2. Biofilm Structural Heterogeneity.................................................................................... 22
2.1.3. Extracellular Polymeric Substances................................................................................ 25
2.1.3.1 Composition of EPS .................................................................................................. 26
2.1.3.2 Chemical Interactions of EPS................................................................................... 26
2.1.4. The Role of Cations in EPS Cohesion ............................................................................ 28
2.1.5. Biofilm Detachment........................................................................................................ 32
2.1.5.1. External Forces Causing Detachment ..................................................................... 33
2.1.5.2 Internal Weakening Causing Detachment ................................................................ 34
2.1.5.3. The Influence of Growth Rate on Biofilm Detachment............................................ 40
2.1.6. Thermophilic Biofilms.................................................................................................... 42
2.2. The Moving Bed Biofilm Reactor (MBBR) ....................................................................... 44
2.2.1 Carrier Design Considerations......................................................................................... 47
2.3 Overview of Biofilm Modeling............................................................................................. 48
2.3.1 Modeling Biofilm Detachment ........................................................................................ 53
2.3.2. Modeling Moving Bed Biofilm Reactors ....................................................................... 56
Chapter 3: Multivariate Statistical Analysis of an Industrial MBBR ................................... 58
3.1. Introduction.......................................................................................................................... 58
3.1.1. The MBBR at Irving Pulp & Paper Ltd. ......................................................................... 58
3.1.2. Performance Fluctuations ............................................................................................... 60
3.1.4 Objectives ........................................................................................................................ 62
3.1.3. Overview of Multivariate Analysis................................................................................. 64
3.1. Methods................................................................................................................................. 65
3.1.1. Overview of Mill Treatment Data................................................................................... 65
3.1.2. Modeling Approach ........................................................................................................ 66
3.2. Results and Discussion......................................................................................................... 68
3.2.1. PCA Overview Model..................................................................................................... 68
3.2.2. Descriptive PLS Model................................................................................................... 70
3.2.3. Predictive PLS Model ..................................................................................................... 73
3.3. Conclusions........................................................................................................................... 75
Chapter 4:The Effect of Calcium on MBBR Biofilms............................................................. 76
2

4.1. Introduction.......................................................................................................................... 76
4.2. Methods................................................................................................................................. 78
4.2.1. Laboratory MBBR System ............................................................................................. 78
4.2.2. Experimental Conditions ................................................................................................ 80
4.2.3. Feed and Effluent Analysis............................................................................................. 80
4.2.4. Biofilm Areal Density and Organic Fraction.................................................................. 81
4.2.5. Dissolved Oxygen Microprofiles.................................................................................... 81
4.2.6. EPS Extraction and Characterization.............................................................................. 83
4.2.7. Confocal Microscopy...................................................................................................... 83
4.2.8. SEM ................................................................................................................................ 84
4.2.9. DNA extraction, amplification and characterization with DGGE .................................. 85
4.2.10. Enumeration of Protozoa and Metazoa......................................................................... 86
4.3. Results.................................................................................................................................... 86
4.3.1. Biofilm Density, Thickness and Oxygen Profiles........................................................... 87
4.3.2. Extracellular Polymers.................................................................................................... 89
4.3.3. Microscopy ..................................................................................................................... 89
4.3.4 Bacterial Community Analysis ........................................................................................ 91
4.3.5. Enumeration of Protozoa and Metazoa........................................................................... 92
4.3.6 Treatment Performance.................................................................................................... 93
4.4. Discussion .............................................................................................................................. 95
4.4.1. Biofilm Structure ............................................................................................................ 95
4.4.2. Biofilm Microbiology ................................................................................................... 100
4.4.3. Treatment Performance................................................................................................. 103
4.5. Conclusions.......................................................................................................................... 104
Chapter 5: A Theoretical MBBR Model of Biosolids Dynamics during Periods of
Starvation................................................................................................................................... 106
5.1 Introduction......................................................................................................................... 106
5.2 Model Formulation and Experimental Methods ............................................................. 109
5.2.1. Model Overview ........................................................................................................... 113
5.2.2. Model Nomenclature .................................................................................................... 114
5.2.3. MBBR Reactor Balances .............................................................................................. 114
5.2.3.2. Reactor balance assumptions ................................................................................ 115
5.2.4. Biofilm Component ...................................................................................................... 116
5.2.4.1. General equations.................................................................................................. 116
5.2.4.2. Challenges for solving general equations.............................................................. 118
5.2.4.3. Accounting for the diffusion boundary layer ......................................................... 120
5.2.4.4. Biofilm model components..................................................................................... 120
5.2.4.5. Biofilm Model Assumptions ................................................................................... 124
5.2.5. Correcting Effective Surface Area for Carrier Geometry............................................. 125
5.2.6. Rate of Biofilm Detachment ......................................................................................... 127
5.2.7. The Effect of Starvation................................................................................................ 130
5.2.7.1. Observed Responses to Starvation......................................................................... 130
5.2.7.2. Observed Starvation Response in Industrial MBBR.............................................. 131
5.2.7.3. Modeling Metabolic Downshift ............................................................................. 133
5.2.8. Experimental Data and Methods................................................................................... 139
5.2.8.1. Steady state data .................................................................................................... 139
3

5.2.8.2. Starvation experiments........................................................................................... 140


5.2.8.3. Activity measurements using iodonitrotetrazolium (INT) chloride ....................... 141
5.2.9. Model Parameter Selection and Fitting Methodology .................................................. 142
5.2.10 Model Solution Implementation Program ................................................................... 143
5.3. Results and Discussion....................................................................................................... 144
5.3.1. Modeling Acclimated MBBRs at Different Calcium Concentrations .......................... 144
5.3.2. The Impact of Surface Area Correction........................................................................ 151
5.3.3. Modeling the Response to Starvation ........................................................................... 153
5.4. Conclusions......................................................................................................................... 160
Chapter 6: Conclusions, Recommendations, and Engineering Significance....................... 163
6.1 Conclusions...................................................................................................................... 163
6.2 Recommendations ........................................................................................................... 169
6.3. Engineering Significance ............................................................................................... 171
Abbreviations ............................................................................................................................ 174
References.................................................................................................................................. 176
Appendix 1: Variable Identification for Multivariate Models ............................................. 196
Appendix 2: Higher Organism Counts by Mass .................................................................... 197
Appendix 3: Scilab code for theoretical MBBR model.......................................................... 198
4

List of Tables

Table 2.1. Summary of MBBR performance data ........................................................................ 46


Table 2.2. Principal rate expressions for modeling detachment ................................................... 53
Table 3.1. Summary of MBBR operating parameters .................................................................. 60
Table 3.2. Summary of generated models .................................................................................... 70
Table 4.1. Summary of reactor feed properties and acclimation period....................................... 79
Table 4.2. Protozoa and metazoa counts for biofilm and suspended biomass.............................. 92
Table 5.1. Nomenclature used in the derivation of the theoretical model .................................. 114
Table 5.2. Subscripts used for parameters .................................................................................. 114
Table 5.3. Model assumptions for reactor balances.................................................................... 115
Table 5.4. Process matrix for the MBBR heterotrophic biofilm................................................. 124
Table 5.5. Model assumptions for the biofilm component ......................................................... 124
Table 5.6. Common detachment rate expressions proposed in the literature ............................. 127
Table 5.7. Kinetic parameters used in the MBBR model ........................................................... 142
Table 5.8. Best-fit parameter values for each detachment term and each calcium condition..... 146
Table 5.9. Detachment rate and metabolic state constants used in the simulation of starvation
experiments ................................................................................................................................. 156
5

List of Figures

Figure 1.1. MBBR biofilm carriers................................................................................................. 1


Figure 1.2. Schematic of an aerobic MBBR ................................................................................... 2
Figure 1.3. Treatment performance variations at Irving Pulp & Paper Ltd.................................... 5
Figure 2.1. Diagram of biofilm structural heterogeneity .............................................................. 23
Figure 2.2. Schematic diagrams of MBBR carrier designs .......................................................... 47
Figure 2.3. Basic schematic of diffusion reaction models ............................................................ 50
Figure 2.4. Model-derived structures from Eberl et al. 2000 ....................................................... 51
Figure 3.1. Fluctuating treatment performance of the full-scale MBBR over 250 days of
operation ....................................................................................................................................... 61
Figure 3.2. Ratio of COD to BOD in the MBBR outlet ............................................................... 64
Figure 3.3.Score plot for model PC1 ............................................................................................ 69
Figure 3.4. PLSFull_1 model predictions, investigation of performance shift contributions, and
variable loadings ........................................................................................................................... 73
Figure 3.5. Predictions and observed BOD output values for PLSFull_2 model ......................... 74
Figure 4.1. Schematic diagram of the K1 carrier and methods of experimental interrogation..... 82
Figure 4.2. Combined biofilm areal density measurements for three experiments ...................... 88
Figure 4.3. Dissolved oxygen micoprofiles, thickness of anoxic region, and biofilm areal density
measurements................................................................................................................................ 89
Figure 4.4. Ratio of proteins to polysaccharides for biofilm EPS cultivated under different
calcium concentrations.................................................................................................................. 89
Figure 4.5. CLSM and SEM images of biofilm structure for different calcium concentrations .. 91
Figure 4.6. Cluster analysis for DGGE fingerprints for biofilm samples at different calcium
concentrations ............................................................................................................................... 92
Figure 4.7. COD removal efficiencies and effluent VSS for experiments II and III at different
calcium concentrations.................................................................................................................. 93
Figure 4.8. Turbidity of reactor effluents at incerasing calcium concentrations .......................... 94
Figure 5.1. Overview of MBBR model components and main governing equations................. 113
Figure 5.2. Schematic representation of the discretization of spatial derivatives through the
Method of Lines.......................................................................................................................... 120
Figure 5.3. Comparison of theoretical surface area correction to carrier specifications ............ 127
Figure 5.4. Four examples of effluent TSS and soluble BOD fluctuations following transient
periods of reactor shutdown........................................................................................................ 131
Figure 5.5. Conceptual diagram of batch growth following exposure to substrates................... 133
Figure 5.6. Theoretical metabolic downshift and recovery due to a hypothetical 10 hour
starvation period.......................................................................................................................... 138
Figure 5.7. 3-dimensional depiction of the modified specific maximum substrate utilization rate
for steady state and an 11 hour starvation period. ...................................................................... 139
Figure 5.8. Photo of carrier plugging in comparison to normal biofilm colonization ................ 141
Figure 5.9. Model predictions and experimental data for effluent VSS, COD, and biofilm
thickness...................................................................................................................................... 145
Figure 5.10. Example of model simulation of dissolved oxygen profile.................................... 146
Figure 5.11. Comparison of model predictions with and without the surface area correction ... 152
Figure 5.12. Performance data measured during starvation experiments and model prediction of
the starvation response................................................................................................................ 155
6

Figure 5.13. Comparison of model simulations with and without the metabolic state function and
growth-dependant detachment .................................................................................................... 157
Figuer 5.14. Changes in biofilm mass per carrier before and after an 11 hour starvation period
..................................................................................................................................................... 159
Figure 5.15. INT-formazan activity measure for three starvation experiments.......................... 159
7

List of Appendices

Appendix 1: Variable Identification for Multivariate Models.................................................... 196


Appendix 2: Higher Organism Counts by Mass ......................................................................... 197
Appendix 3: Scilab code for theoretical MBBR model .............................................................. 198
1

Chapter 1: Introduction
1.1. The Moving Bed Biofilm Reactor (MBBR) and Biosolids Dynamics

The moving bed biofilm reactor is a biofilm treatment system capable of degrading

polluting organic compounds and nutrients (N and P) in wastewater effluents. The key feature of

the MBBR is the use of small plastic biofilm support media to allow for a high concentration of

protected biofilm growth in a well mixed reactor vessel. A typical design for these carriers is

featured in Figure 1.1 and a schematic of the general configuration is shown in Figure 1.2. As

can be seen from the schematic, the retention of carriers is achieved by the installation of a

grating on the effluent pipe. Aerobic reactors such as that depicted are most common, and a high

rate of aeration is used to provide dissolved oxygen to the microorganisms and keep the reactor

well-mixed.

Developed in the late 1980s (US Patent

#548779, degaard, 1995) the MBBR is an emerging

technology that is finding increased application due to

two advantages inherent in the system design. First,

the reactors contain a high concentration of


1 cm
microorganisms enabling rapid degradation rates and
Figure 1.1. MBBR biofilm carriers (K1
carrier, AnoxKaldnes) thus small reactor sizes. Secondly, as a biofilm

system there is no need for returning settled sludge to

the reactor even at short hydraulic retention times, which simplifies the design and control of

effluent clarification (degaard et al., 2006). Without the highly concentrated suspended

bacterial population of activated sludge, the overall solids removal requirements are also

reduced, allowing for the use of alternative technologies such as dissolved air flotation (Lundt et
2

al., 2001). In general the reactors are straightforward to install and maintain, requiring only a

tank of adequate size and a bank of aerators. It has also been shown that the treatment

performance of MBBRs is proportional to the installed biofilm surface area (degaard et al.,

2000) so treatment upgrades can be performed by simply adding additional carriers to the same

tank.

MBBR

Biofilm Carriers Treated Effluent

Grating for
carrier retention

Influent Wastewater
Compressed Air

Figure 1.2. Schematic of an aerobic MBBR. Not drawn to scale.

In contrast to the macroscale simplicity of the MBBR, the microscale processes involved

in converting unwanted pollutants are complex. MBBR biofilms are heterogeneous three-

dimensional structures that can contain hundreds of bacterial, fungal and eukaryotic species in

the colonizing biofilm. Each of these species interacts through competition for substrates or by

predator-prey relationships that cause inherent dynamic changes in biofilm microbiology over

time (Briones and Raskin, 2003). The biofilm is held together through the microbial excretion of

various extracellular polymeric substances (EPS) that act as a gel-like matrix surrounding the

cells. This encapsulated structure results in the formation of concentration gradients for all

relevant dissolved compounds due to microbial activity and diffusion. In addition to this, the
3

biofilm-associated microorganisms detach from the surface of the biofilm to populate a

planktonic phase that can contribute significantly to pollutant removal (Yu et al. 2001). These

complex microscale processes make it difficult to develop accurate mathematical models,

identify routes to optimizing performance, or even establish which variables are the most

significant controlling factors for a given reactor system.

One of the interesting features of the MBBR is the importance of the detached suspended

phase, making the reactor not exclusively a biofilm or suspended phase system, but rather a

hybrid. The concentration of suspended organisms exists in a balance between the rates of

detachment from the biofilm, suspended phase growth and the continuous loss of mass due to

flow out of the reactor. Thus at short hydraulic retention times the reactor may have an

insignificant suspended population while at longer hydraulic retention times the detached phase

can grow, playing an increasing role in organic degradation. The direct exposure of the

suspended organisms to the bulk reactor liquor makes them better positioned to access organics

than the diffusion-limited biofilm, further increasing their significance to the system. Although a

more concentrated suspended population can cause increased rates of degradation, it also

increases the amount of suspended solids discharged from the reactor which can impact solids

settling equipment. It is for these reasons that changes in the relative amounts of biofilm and

suspended microorganisms, in this thesis referred to as the biosolids dynamics, are central to

MBBR treatment performance.

As mentioned, the biosolids dynamics are dependant on the rate of biofilm detachment.

The literature review that follows in Chapter 2 will show that detachment is dependant on

multiple factors and can be considered as a range of processes from steady erosion of particles

from the surface of the biofilm to discreet sloughing of large segments caused by instability in
4

the deeper regions of the biofilm (Stoodley et al., 2001, Wilson et al., 2004). There is also

considerable evidence to suggest that microorganisms are capable of inducing detachment when

they determine it is no longer favourable to remain in the biofilm phase through a range of cell

changes (Lazzazara, 2000; Rice et al., 2005; reviewed in greater detail in Chapter 2). Past

research has proposed various theories for how detachment should be modeled or explained but

there is little consensus outside of the conclusion that different systems are governed by different

mechanisms for detachment. There have been no studies specifically investigating how

detachment can be understood in the emerging MBBR reactor. Without a description of biofilm

detachment, a key feature of the biosolids dynamics remains unknown. This limits our ability to

predict how the reactor will respond to process change or devise ways to enhance the

performance of MBBRs.

1.2. Initial Motivation: Variable Performance of an Industrial-Scale MBBR

The initial motivation for this project was the observation of variable treatment

performance in a full-scale MBBR operated by Irving Pulp & Paper Ltd., an industrial partner in

the work. The MBBR (designed by AnoxKaldnes AB) at the mill is used to treat effluent from

the washing tanks of the pulp bleaching process and an example of the typical fluctuating

behaviour is shown in Figure 1.3. From this data it is clear that the MBBR removes Biochemical

Oxygen Demand (BOD) and produces Total Suspended Solids (TSS) at rates which change

dramatically from one day to the next. While the mill meets all regulations, this treatment

variability complicates environmental management. By determining the conditions that lead to

increased performance there may be opportunities to further reduce mill discharges, minimizing
5

the environmental impact of the effluent. Understanding, predicting and minimizing BOD and

TSS emitted from this operational MBBR were the initial objectives of this work.

Given the complexities of the microscale processes that govern degradation and

production of biosolids in MBBRs as described above, a wide range of variables are possible

candidates for explaining fluctuating treatment performance. In the industrial setting this is

further complicated by the fact that the

90% composition of wastewater is frequently


BOD Removal Efficiency

80%
70% changing due to upstream process changes in
60%
50%
40%
the pulp mill. With so many options for
30%
20% potential study, an initial scoping
10%
0% investigation was performed to better define
0 200 400 600 800

400 the project. This included site visits to talk to


350
Effluent TSS (kg/d)

300 engineers and operators, a review of pertinent


250
200
150
literature and a preliminary data mining
100
50 exercise using multivariate statistical analysis.
0
0 200 400 600 800 For brevity the results of this work are
Time (d)

Figure 1.3. Treatment performance variations over not presented in the thesis. However, through
aproximately 2 years of measurements at Irving
Pulp & Paper Ltd. this work several interesting avenues for

scholarly investigation were identified that form the basis for the chapters of this thesis. These

are captured by the three lines of inquiry described in the following section.
6

1.3. Lines of Inquiry for Investigation

1.3.1. Multivariate Statistical Analysis of an Industrial MBBR

Multivariate Statistical Analysis of an Industrial MBBR


Questions

(A) Is multivariate analysis a suitable modeling approach to understand and predict the observed
performance fluctuations at the Irving Pulp & Paper Ltd. mill?

(B) What are the significant variables controlling MBBR performance for this application?

Hypotheses

(A) Multivariate statistical analysis is a useful tool for understanding and predicting performance
behaviour of an industrially applied MBBR

(B) The model will indicate mill process variables that have a significant impact on the performance
of the MBBR.

Objectives

(A) Develop a methodology for multivariate statistical analysis to provide process insight.

(B) Develop an approach for generating accurate predictions of MBBR performance.

In order to make inferences about what is causing the treatment performance variations,

some type of system model is required. A model would allow for the use of historical data

collected at the mill to identify important variables. Given the complex and transient nature of

the industrially applied MBBR, building a theoretical model from first principles is challenging

and may require making assumptions that eliminate key variables. For instance, since the bleach

plant effluent is not chemically defined, there may be significant changes in the types of organic

compounds present in the effluent over time due to changes in pulping or bleaching conditions.

While an assumption could be made that the composite measures of Chemical Oxygen Demand

(COD) and Biochemical Oxygen Demand (BOD) define the important characteristics of

wastewater, they may not provide adequate information about inhibitory or kinetic effects

resulting from exposure to changing concentrations of different organic compounds. If these


7

factors were contributing to the observed fluctuations in performance, a theoretical model based

on BOD or COD would not be sufficient.

Multivariate analyses such as Principal Component Analysis (PCA) and Partial Least

Squares (PLS) are statistical techniques that reduce the dimensionality of large datasets to create

a correlation structure that can identify data trends and significant variables (Wold et al., 1987;

Geladi and Kowalski, 1986; Kourti and MacGregor, 1995). In contrast to theoretical models,

these methods allow for the composition of the wastewater to be chemically characterized

indirectly, through the inclusion of correlated upstream process variables. The only requirement

for conducting a multivariate analysis investigation is that it be based on an adequately sized

dataset of relevant measurements.

Considering the challenge of investigating the fluctuating performance at Irving, it was

determined that multivariate analysis could be a useful analytical tool. This thesis seeks to

address the applicability of state-of-the-art multivariate techniques to understand and predict

behaviour of an industrial MBBR. To date, no other study has examined a high-rate biofilm

treatment process like the MBBR using this technique. If successful, this line of inquiry would

indicate important process variables that influence treatment performance at the pulp mill. This

information would not only provide avenues for optimizing performance of the investigated

MBBR, but also suggest important considerations for the design and operation of MBBRs at

other facilities. Chapter 3 of this thesis outlines the approach, methodology, results and

conclusions in pursuit of this line of inquiry.

1.3.2. The Effect of Calcium on MBBR Biofilms


8

The Effect of Calcium on MBBR Biofilms


Questions

(A) How is MBBR biofilm structure, microbiology and function affected by the concentration of calcium
in wastewater?

Hypotheses

(A) Biofilm structure (thickness, oxygen diffusion profiles, morphology, EPS composition) is affected
by changes in the concentration of calcium.

(B) Biofilm microbiology is affected by changes in the concentration of calcium.

(C) MBBR performance is affected by changes in the concentration of calcium.

Objectives

(A) Construct a laboratory-scale MBBR apparatus

(B) Run long term experiments at different calcium concentrations to establish steady state conditions
that reflect

(C) Characterize the structure, microbiology, and performance for different calcium concentrations

The scoping review of literature relevant to biofilm systems indicated that divalent

cations may be an important variable for biofilm reactors like the MBBR. Divalent cations such

as calcium have been shown to influence biofilm structure and detachment (Turakhia and

Characklis, 1989; Applegate and Bryers, 1991; Huang and Pinder, 1995; Krstgens et al. 2001;

Ahimou et al., 2007a). This is potentially due to the role divalent cations play in bridging

negatively charged moieties of extracellular polymeric substances (EPS) through electrostatic

interactions (Flemming and Wingender, 2001a,b). In general, biofilms become thicker, denser

and more mechanically stable when exposed to increasing concentrations of divalent cations.

Further to this, Sobeck and Higgins (2002) propose that the EPS matrix acts as an ion exchange

resin, where divalent bridges can be formed or broken due to relative concentrations of

monovalent and divalent cations competing for negative sites. While this work highlights the

importance of divalent cations in cellular aggregation, all of the past studies are limited in their

applicability to the MBBR due to the use of single or binary species cultures, different reactors,
9

or different conditions. They also omit any investigation into how microbiology may be affected

by divalent cation concentrations.

By following this line of inquiry, a variable shown to be critical to biofilm structure in

several systems was assessed for importance in the increasingly-installed MBBR reactor.

Through understanding the significance of divalent cations to the function of MBBRs it is

possible to consider this variable in future MBBR designs. This would be particularly useful for

industrial wastewater treatment where there can be substantial site to site differences in

wastewater cation concentrations. For instance, in pulp and paper applications, a paper mill

effluent may have an elevated calcium concentration due to the use of calcium carbonate as a

paper component while a pulp mill effluent may have low calcium concentrations mainly

governed by the calcium in the mills local water supply. Improving reactor design to account for

these differences could lead to better functioning treatment systems. The enhanced understanding

of this variable would also improve the breadth of knowledge available for troubleshooting

current systems.

Beyond improved design, the role of divalent cations in increasing biofilm thickness,

density and strength suggests a strategy for enhancing performance through stratification of

multifunctional biomass. For example, it is desirable to perform simultaneous nitrification and

denitrification in a single continuous reactor, a process which requires both aerobic and anoxic

microbial communities. Increasing biofilm thickness or reducing diffusion of oxygen into the

deeper regions of biofilms through calcium addition could lead to better retention of both

nitrifiers and denitrifiers.

There may also be situations where avoiding thick biofilms is desirable, such as for any

biofilm system where reactor plugging is problematic. These could include submerged biofilters
10

but also MBBR systems where the loading conditions are such that the carriers become plugged

with biomass, thus reducing the biofilm surface area. By establishing the significance of divalent

cations in controlling biofilm accumulation it may be possible to account for different effluent

conditions through reactor and carrier selection, or by taking upstream measures to reduce

divalent cations.

Chapter 4 describes a study investigating these questions using a laboratory-scale MBBR

apparatus.

1.3.3. A Theoretical Model of MBBR Biosolids Dynamics during Periods of Starvation

A Theoretical MBBR Model of Biosolids Dynamics during Periods of Starvation


Questions

(A) How can theoretical biofilm models be adapted to describe the kinetics and biosolids dynamics
of MBBRs?

(B) Can the detachment process be modelled to simulate steady-state behaviour for different
calcium concentrations?

(C) Can an MBBR model be developed to describe the response to transient periods of starvation?

Hypotheses

(A) A theoretical diffusion-reaction biofilm model can be adapted to capture the key
features of MBBR behaviour

(B) Different divalent cation concentrations will affect the detachment rate

(C) The response to starvation can be simulated by the MBBR model

Objectives

(A) Derive a dynamic theoretical model of the MBBR

(B) Compare different detachment rate terms to steady state data from calcium experiments

(C) Compare model outputs to transient starvation experiments in the lab and industrial data

Biofilm theoretical modeling is a continually evolving field and biofilm models can be

divided into two general approaches: biofilm continuum models and cellular automaton models.
11

A discussion of the current advances in each type of modeling follow in the literature review and

it will be shown that a continuum model approach is the better approach for providing

information about biofilm reactor behaviour and average biofilm properties. Such models

involve coupling mathematical representation of diffusion, reaction, growth and detachment

processes in a one dimensional representation of the biofilm. By considering the biosolids

balance and other features specific to the MBBR, it is possible to adapt biofilm continuum

models to the MBBR.

Although a theoretical MBBR model is not the best tool for gaining information and

performing predictions for the highly dynamic reactor at Irving Pulp & Paper, a theoretical

model can be fit to laboratory-scale data to help directly quantify the biosolids dynamics. In

order to do this, several features of the MBBR must be integrated with a biofilm model. These

include creating reactor mass balances on substrates and biomass, considering the carrier

geometry in describing the surface area, and defining how detachment and flowrate changes

influence the biosolids dynamics.

As introduced earlier, the detachment expression has not been previously considered for

the MBBR. By using data derived from the lab-scale study of the effect of divalent cations,

different detachment expressions can be compared for their applicability. This also allows for

better quantification of the influence of calcium on the biosolids dynamics.

In addition to providing a detachment expression that predicts average behaviour, it is

also desirable to modify the MBBR model to so that it can account for the response to periods of

starvation. Starvation response is of particular interest to this work because, as will be shown in

the results of the multivariate study in Chapter 3, bleach plant shutdowns causing temporary

periods of MBBR starvation were found to cause many of the most significant performance
12

fluctuations. This type of transient operation may be common in a range of industrially applied

MBBRs where plant shutdowns for maintenance or other factors occur frequently. It is of interest

to extract how the detachment rate or other kinetic parameters may be affected by starvation so

that the response to starvation can be characterized. This will suggest if there are potential

strategies for minimizing the impact of poor performance caused by starvation periods.

Understanding the starvation response in MBBRs through modeling also provides

valuable information to the biofilm research community. Current reports of the microbial

response to starvation are conflicting, with some suggesting starvation promotes dispersion

(Hunt et al., 2004; Gjermansen et al., 2005), and others suggesting periods of starvation can

promotes aggregation (Li et al., 2006). A wide range of investigated biofilm systems may

experience starvation, including biofilms in natural environments, infectious films colonizing

medical equipment, and films fouling industrial equipment. A theoretical model could inform

research ongoing in all of these areas.

Chapter 5 describes the development of a theoretical MBBR model, its comparison to

lab-scale and industrial data, and how the model can be adapted to predict starvation behaviour.

1.4. Structure of the Thesis

As described above the thesis is structured to include a literature review followed by

three major chapters that cover all of the methodology, results and discussion related to the line

of inquiry described. This is followed by a chapter of comparative discussion (Chapter 6) that

draws general conclusions about the biosolids dynamics in an MBBR. The thesis concludes by

describing the significance of the major conclusions and recommendations drawn from all of the

lines of inquiry. The chapters are also related to papers published, submitted, and in preparation

as described in the following section.


13

1.5. Scholarly Contributions

The work has led to several contributions as it has been developed. These include:

Presentations

Goode, C. and Allen, D.G. (2008) Modeling starvation of wastewater treatment biofilms.
Canadian Chemical Engineering Conference. Ottawa, Ontario. October 22nd.

Goode, C. and Allen, D.G. (2008) The role of calcium in a moving bed biofilm reactor (MBBR).
IWA Biofilm Technologies Conference, Singapore. January 9th

Goode, C., Allen D.G. (2007) Calcium enrichment in a moving bed biofilm reactor (MBBR).
Ontario-Quebec Biotech Conference, University of Toronto, Toronto, Ontario. June 9th.

Goode, C. LeRoy, J. and Allen D.G. (2006) Multivariate analysis of a high-rate biofilm process
treating Kraft mill effluent. The 8th IWA Symposium on Forest Industry Wastewaters. Vitoria,
Brazil. April 12th.

Goode, C. (2005) Understanding biosolids dynamics in a moving-bed biofilm reactor. PAPTAC


Paperweek International. Montreal, Quebec. February 10th.

Goode, C. (2004) Understanding biosolids dynamics in a moving-bed biofilm reactor (MBBR)


Ontario-Quebec Biotech Conference, Universit Laval, Quebec City, Quebec. June 10th

Publications

Goode, C., LeRoy, J., and Allen, D.G. (2007) Multivariate analysis of a high-rate biofilm process
treating Kraft mill effluent. Water Science and Technology. 55 (6) 47-55.

Goode, C. and Allen, D.G. (2006) Multivariate Statistical Modeling of Effluent Treatment:
Understanding and Prediction. O Papel Magazine. Nov. 2006 Issue. (non-refereed magazine
article)

Additional publications in progress:

Goode, C. and Allen, D.G. (2008) The effect of calcium on moving bed biofilm reactor (MBBR)
biofilms. Submitted to Water Environment Research.

A comprehensive paper including all of the results presented in Chapter 4

Goode, C. and Allen, D.G. (2008) Theoretical modeling of the moving-bed biofilm reactor
system. In preparation
14

A shorter contribution, indicating how a diffusion/reaction model can be adapted to


represent the MBBR using an accessible software
Covering elements of chapter 5

Goode, C. and Allen, D.G. (2008) Modeling starvation of moving bed biofilm reactor (MBBR)
biofilms. In preparation.

A paper presenting the approach to modeling starvation shown in chapter 5


15

Chapter 2: Literature Review

This literature review presents a current understanding of the fundamental features and

principal governing variables of the MBBR, as well as provides a discussion of the modeling

research that informs the approaches used in this thesis. This is achieved through a description of

the key elements of biofilms, with a specific focus on biofilm microbiology, structural features,

and biofilm detachment. After presenting the review of biofilms, focus is shifted to literature

specifically relevant to the MBBR. The final part of the review concerns biofilm modeling

approaches, with more detailed discussion of the theoretical and statistical approaches used in

this thesis.

2.1. Microbial Biofilms

Biofilms are microbial aggregates attached to a solid surface. They are comprised of the

microbial cells themselves, their associated extracellular materials and other entrained materials,

such as particulate matter that comes into contact with the biofilm. Biofilms are found in diverse

environments including natural settings such as those on rocks in streams and on shorelines, in

water distribution piping and other industrial systems, or on the surfaces of other organisms such

as when infectious biofilms colonize plants and animals. From the thin and compact films that

grow on human teeth to the thick algal films that colonize marine environments, the features of

biofilms are tied to the conditions they exist in and are just as heterogeneous (Stoodley et al.,

2004). In fact it appears that the principal generalization one could draw about biofilms is that

they are diverse in nearly every respect, from microbiology to structure and function (Wimpenny

et al., 2000).
16

The diversity of biofilms and their relevance to the natural and engineered world has

sparked considerable academic interest. As noted by Wimpenny and Colasanti (1997), the 1990s

saw a rapid increase in publication on biofilms and the field has been thoroughly reviewed

through several books and comprehensive review articles (Lappin-Scott and Costerton, 1995;

Costerton et al., 1995; Davey and OToole, 2000; Wimpenny et al., 2000; Costerton, 2007;

Lewandowski and Bayenal, 2007; Kjelleberg and Givskov, 2007). Rather than present a

comprehensive analysis of the literature on biofilms, the following sections will highlight

important biofilm concepts that provide context for the hypotheses investigated in later chapters.

2.1.1. Biofilm Microbiology

Biofilms can be comprised of any type of microorganism, including algae, fungi,

bacteria, achaea, and protozoa/metazoa, and in most natural and engineered biofilms the

microbial community is complex with multiple species (Briones and Raskin, 2003). While many

general principles of microbial ecology apply to biofilms, biofilm communities have several

unique features that influence their microbiology and community dynamics (Wuertz et al.,

2004). These include the presence of substrate diffusion gradients causing stratification and

clustering of species and the cell density dependant signaling mechanisms (quorum sensing) that

govern fundamental genetic regulatory pathways.

2.1.1.1 Diffusion gradients cause community stratification

As dissolved substrates diffuse into the biofilm and are consumed, concentration

gradients manifest and cause differential exposure to organisms with depth. This influences

microbial selective pressure, resulting in stratification (in this case broadly defined as spatial
17

change) of the microbial community. Gradients of electron acceptors such as dissolved oxygen

or nitrate can even lead to stratification of microbial communities with drastically different

metabolisms, and from an environmental microbiology perspective, different degradative

functions (Okabe et al., 1996). For instance it is possible that oxygen-limited biofilms can

contain both aerobic and anaerobic microorganisms existing in stratified but close proximity.

There are several examples of stratified communities studied for their role in wastewater

treatment. Schramm et al. (1997) demonstrated diffusion gradients in ammonia, nitrite, nitrate,

and oxygen for a nitrifying fluidized bed biofilm reactor and the spatial stratification of ammonia

and nitrite oxidizing bacteria. These observations were made using microelectrode sensors for

dissolved compounds and florescence in-situ hybridization (FISH), an RNA labeling technique

for targeted species visualization. By manipulating the bulk concentrations of ammonia, and thus

the gradients of all nitrogenous species, it was found that the relative amounts and locations of

ammonia and nitrite oxidizers were affected. Specifically, where ammonia concentrations were

higher there were increased relative concentrations of ammonia oxidizers.

A similar microelectrode and FISH approach was used by Okabe et al. (1999) to study

wastewater treatment biofilms exposed to municipal wastewater. This work extended that of

Schramm et al. (1999) by using additional stains to show the stratification of a greater number of

nitrifying species as well indicating the location of heterotrophic bacteria. By coupling this with

the consumption and production rates of various substrates at different positions in the biofilm,

the authors were able to link gradients, species abundance, and degradation rates of different

substrates. The examination of a greater number of species and their location in the stratified

biofilm also allowed for inferences to be made regarding the specific metabolic needs of

different species. Other work by Satoh et al., (2000) characterized the relative abundance and
18

locations of nitrifiers and heterotrophs exposed to different carbon to nitrogen (C/N) ratios while

Gieseke et al. (2001) examined stratified populations of nitrifiers in batch reactors to consider

temporal as well as spatial changes in biofilm microbiology. These studies all confirm the

importance of biofilm stratification in nitrification, a microbial function of great current

importance to wastewater treatment.

Other studies have demonstrated the importance of stratification for systems beyond

those with aerobic nitrification. For instance, Santagoeds et al. (1998) observed a thriving

sulfate-reducing population in the anoxic inner regions of aerobic biofilms. Recent interest in

anaerobic ammonia oxidizing bacteria (anammox) has shown how anaerobic films without an

organic substrate can support heterotrophic organisms in deeper regions of the biofilm (Tsushima

et al. 2007). In this case the stratification was attributed to the accumulation of dead biomass in

the inner regions of the biofilm which through lysis was hypothesized to act as a carbon source.

This accumulation of inactive biomass is hypothesized to be caused by ammonia depletion at

greater depths causing starvation and cell death.

The observed relationship between substrate concentration gradients and biofilm

community stratification has also been predicted theoretically by a number of modeling efforts.

These include the widely cited multispecies biofilm model of Wanner and Gujer (1986), the

work of Horn and Hempel (1997), and recent work to model membrane aerated biofilm reactors

(Shanahan and Semmens, 2004; Matsumoto et al., 2007). All of this work uses kinetic

parameters of different species or species types to predict relative growth rates based on the

concentrations of substrates and electron acceptors. The simulations all predict stratified

communities, an observation confirmed by experimental analysis.


19

As will be detailed in section 2.1.2, biofilms are often structurally heterogeneous, an

observation that complicates the concept of stratification. The preceding discussion reviews

studies that generally depict gradients with respect to a measured biofilm thickness and the

presence of key organisms at different depths. Considering that several studies have observed

porous biofilm structures with protrusions and other heterogeneous forms, the observed gradients

and stratification do not only occur in a direction perpendicular to the substratum, but reflect the

contours of the heterogenous biofilm-water interface (de Beer and Stoodley, 1995; Picioreanu et

al., 2004).

2.1.1.2 Quorum sensing regulates many cell functions in biofilms

The study of biofilm systems has revealed what appears to be coordinated behaviour,

such as sudden detachment events (Dow et al., 2003; Thormann et al., 2005; Gjermansson et al.

2005), increased rates of the production of extracellular polymeric substances (Ramasamy and

Zhang, 2005; Sauer et al., 2002), and the induction of bacterial virulence factors (Zhu et al.,

2002; Hammer and Bassler, 2003). In many cases coordinated behaviour of biofilms has been

hypothesized to be controlled by cell-to-cell signaling through quorum sensing regulatory

pathways. Quorum sensing involves the release of signal molecules that trigger genetic

regulation in bacteria only when a threshold concentration is reached, thus requiring the close

proximity of many signal producers. The phenomenon was first described in detail by Nealson

and Hastings (1979) for bioluminescing bacteria Vibrio fisheri and since its proposal as a

signaling framework for bacteria, many other quorum sensing regulated processes have been

discovered (Waters and Bassler, 2005). In most pathways for gram-negative organisms, species

specific acyl-homoserine lactones (AHL) have been implicated as the signal molecule. For the
20

gram-positive bacteria, signal molecules have been found to be oligopeptide based, again with

species specific residues (Waters and Bassler, 2005).

Identification of quorum sensing signaling pathways has improved our understanding of

biofilm behavior and many processes of interest to researchers have been found to be controlled

by quorum sensing. For instance Davies et al. (1998) demonstrated through a gene knockout

study that Pseudomonas aeruginosa biofilm accumulation was governed by known quorum

sensing genes and signaling molecules. A similar study of Aeromonas hydrophila by Lynch et al.

(2002) revealed that mature biofilm formation was also inhibited in mutants without key

signaling genes. Smith and Iglewski (2003) review a body of work that suggests that quorum

sensing signaling pathways control the virulence of Pseudomonas aeruginosa biofilms. This has

also been found for virulence in Vibrio cholerae, the bacterium responsible for cholera infections

(Zhu et al., 2002).

In order to explore the specific signaling pathways involved in this fascinating

phenomenon, much of the work investigates pure culture biofilms or suspended cultures. Critical

reviews by Kjelleberg and Molin (2002) and Parsek and Greenberg (2005) question the

relevance of quorum sensing in multi-species biofilms where structural and functional processes

are complex and may be controlled by multiple environmental and genetic factors. Nevertheless,

work by McLean et al., (1997) has shown that AHL signaling activity can be measured in mixed

species natural biofilms growing on rocks in a stream environment. Valle et al. (2004) detected

AHL activity in an activated sludge community treating phenolic wastewater suggesting the

importance of quorum sensing in such systems. In fact these authors augmented the wastewater

with a dose of a relevant AHL and observed some evidence of improved phenol degradation.

However in light of the complexity of the microbial community involved, a causal link between
21

AHL addition and performance cant be irrefutably concluded from this work. Additional

investigation of activated sludge biomass by Morgan-Sagastume et al. (2005) further

demonstrates that quorum sensing signals exist in wastewater treatment systems.

In many of the examples studied so far, quorum sensing regulates behavior to enhance

the survival or proliferation of the biofilm (Waters and Bassler, 2005). This type of genetic

control allows cells to be aware of the concentration of neighbouring cells, an important

consideration for timing of infection or dispersion. An example of how this enhances survival is

given by Rice et al. (2005) who present data indicating that the decision of biofilm cells to

actively detach and disperse from the biofilm is controlled by both substrate concentrations and

quorum sensing signals. The authors hypothesize that biofilm organisms may regulate dispersion

to account for a calculation of how much substrate is present and how many other microbes are

present. This multi-factor control would ensure cells depart from situations where they are likely

to face starvation, either through excessive competition or inadequate food supply.

The role of quorum sensing regulation in bacterial survival can also be considered in a

multi-species context. Keller and Surette (2006) review a body of literature to argue that quorum

sensing signaling is an important component of bacterial competition, influencing the evolution

of microbial community structure. They substantiate this by reviewing evidence of organisms

that secrete compounds that interfere with or modify signal chemistry of other species as well as

cooperative interactions between species through signal eavesdropping. West et al. (2006)

present a more comprehensive theoretical discussion of microbial social cooperation and assess

the costs and benefits for organisms to secrete signaling molecules that may be useful to other

species. These perspectives are of particular relevance to multi-species biofilms where

coordinated behaviors may be required for biofilm-wide responses to occur.


22

In reviewing the quorum sensing literature one can conclude that such processes control a

number of important genes that regulate biofilm structure and function. This conclusion is not

surprising since the close proximity of cells in biofilms allows for threshold concentrations to be

reached. The prevalence of quorum sensing suggests that it should always be considered as a

possible mechanism for biofilm response to changing environmental conditions.

2.1.2. Biofilm Structural Heterogeneity

Although early concepts of biofilms depicted homogenous films of uniform thickness

(Rittmann and McCarty, 1980), a number of studies in the 1990s began to report observations of

heterogeneous structures found in biofilms of laboratory or natural origin. These observations

coincided with the development of confocal laser scanning microscopy (CLSM) and

microelectrode investigations (Wimpenny et al., 2000). For instance de Beer et al., (1994),

Stoodley et al., (1994), and de Beer and Stoodley (1995) used CLSM, fluorescent staining, and

the temporal visualization of dye and microsphere injection to observe a porous biofilm structure

with open channels that allowed for convection of fluid to deeper biofilm regions. These

observations were further validated by demonstrating spatial heterogeneity in the dissolved

oxygen profiles measured with depth using microsensors. The observation of a porous structure

was found to allow greater diffusion of dissolved substrates compared to a homogenous film, a

result that caused the research community to re-examine biofilm modeling approaches which

had, until this time, assumed a uniform thickness (Wimpenny and Colasanti, 1997).

Some of the work to characterize biofilm heterogeneity led to a generalized concept of

biofilm structure as a collection of mushroomlike microcolonies with open channels between

the stalks, promoting fluid flow to deep regions of the biofilm (Figure 2.1. Diagram is credit of
23

the Center for Biofilm

Engineering at Montana

State University

Bozeman. Used with

permission). The

observations that support

such a generalized structure

Figure 2.1. Diagram of biofilm structural heterogeneity. Mushroom-like


and the adaptive
structure has been observed by various biofilm researchers. Diagram
used with permission. mechanisms hypothesized

to cause its formation are

discussed in the review by Costerton et al. (1995). Cell signaling through quorum sensing

pathways was specifically identified as the likely control mechanism governing development of a

mushroom microcolony structure. Following the presentation of this generalized biofilm

structure, a number of biofilm modelers published efforts to determine conditions that predict

mushroom morphology (Picioreanu et al., 1998; Eberl et al., 2000; Chambless and Stewart,

2007) using three-dimensional discretized approaches. This work indicated that reasonable

assumptions for biological parameters could predict mushroom or other porous forms simply

based on growth patterns caused by availability of substrate.

Although some experimental and modeling results support the generalized mushroom

structure proposed by Costerton et al. (1995), other work has shown that this conceptual model

does not reflect the wide range of observed biofilm structures (Wimpenny and Colasanti; 1997,

van Loosdrecht et al., 1997). In fact, through these comprehensive reviews it was argued that

different biofilm morphologies fit some unifying structural theory. The general factors that have
24

been implicated in defining the heterogeneity of biofilms are the substrate concentration gradient

and the forces acting on the biofilm to cause detachment. Essentially, these authors argue that

two competing processes define structure: (1) that steeper concentration gradients increase

porosity and (2) that increased detachment forces at the surface of the biofilm shear extended

structures, causing consolidation to more homogenous films. Beyond these general competing

factors the authors recognize that microbial diversity and the expression of extracellular

polymeric substances add additional complexity to biofilm structure.

A number of experimental observations support the contention that heterogeneity can

take a variety of forms. For instance Masol-Deya et al. (1995) observed large channels forming

in the thick biofilms that colonized the surface of granulated activated carbon in fluidized bed

reactors treating toluene. Channels were observed to penetrate at least half way through the depth

of the biofilm, a morphological feature observed both in the lab and at a number of industrial

systems treating contaminated groundwater. Neu and Lawrence (1997) studied biofilms

cultivated from a culture in a stream and observed long ridges formed parallel to the direction of

bulk water flow. These authors specifically noted that these riverine biofilms did not feature the

inverted biofilm density profile predicted by the mushroom concept and suggested that this

may be due to inert material being incorporated into the channel structure. Gjaltema et al. (1994)

examined biofilms cultivated in an experimental rotating drum reactor observing various

heterogeneous structures including patchy growth, the formation of colonies, and streaming

filamentous structures. The authors attributed the range of structures to variability of the flow

regime experienced by the biofilms even though the rotating drum was designed to provide a

constant shear environment. Further complexity in structure was observed in rotating biological

contactor (RBC) reactors treating municipal wastewater observed by Martn-Cereceda et al.


25

(2002). Using confocal microscopy, the mixed community biofilms were observed to be

comprised of a consolidated base structure of densely packed bacteria with outer layers

consisting of clumps of bacterial colonies in a porous extracellular polymeric matrix with a

substantial population of ciliated protozoa. When one considers that higher organisms and

filamentous bacteria and fungi can act as structural elements with dimensions on the order of

hundreds of microns, generalized concepts of structural features such as those discussed by

Costerton et al. (1995) reflect only a component of biofilm heterogeneity.

2.1.3. Extracellular Polymeric Substances

Extracellular polymeric substances (EPS) are polysaccharides, proteins, lipids and

nucleic acids that have been observed to accumulate on the surfaces of bacterial cells (Flemming

and Wingender, 2001a,b). These biopolymers are produced by most types of microorganisms

(including bacteria, archaea, fungi, and algae), in a variety of environments. Since EPS are the

primary compounds found on the cell surface, interactions between the EPS of neighbouring

cells govern cell-to-cell aggregation (Wingender et al., 1999). In the case of biofilm formation,

the adhesion of cells to a solid substratum is governed by interactions between EPS and the

substratum material (Brading et al. 1995). While EPS are found in both suspended and

aggregated bacterial cultures, the production of EPS has been shown to increase during biofilm

formation (Davies et al. 1993; Vandevivere and Kirchman, 1993), leading to the development of

an EPS matrix in which microorganisms are embedded. This EPS matrix acts as the main

structural component of biofilms.

The characterization of EPS compounds and their interactions has been the focus of a

significant body of research. This work has been thoroughly reviewed elsewhere (Cooksey,
26

1993; Neilsen et al., 1997; Flemming et al., 1999; Wingender et al., 1999; Flemming and

Wingender, 2001a; 2001b; Sutherland, 2001; Liu et al. 2004; and Starkey et al., 2004). The

following presents the most pertinent literature related to the research presented in this thesis.

2.1.3.1 Composition of EPS

The composition of the EPS surrounding bacterial cells has been found to contain

polysaccharides, proteins, lipids, humic substances and nucleic acids, the relative amounts of

which depend on the cellular environment and chosen extraction technique (Flemming and

Wingender, 2001a). While earlier research with pure cultures suggested that most EPS was

various polysaccharides (Cooksey, 1992), the predominance of protein (Frlund et al., 1996;

Dignac et al., 1998; Bura et al. 1998) and significant amounts of DNA (Palmgren and Nielsen,

1996; Steinberger and Holden, 2005) and humic substances (Liu and Fang, 2002) have been

found in activated sludge.

Some of the environmental factors shown to affect EPS production and composition

include substrate limitation (electron donor or acceptor) and nutrient (N & P) limitation (Nielsen

et al., 1997). This has an important implication for biofilms, since gradients exist for these

factors in most biofilms. It is possible that biomass in deeper regions of the biofilm could

experience limitations in substrate or nutrients, leading to different amounts and compositions of

EPS.

2.1.3.2 Chemical Interactions of EPS

The EPS matrix in biofilms is formed due to adhesive chemical interactions between the

EPS. Several interactions have been proposed to be important in EPS adhesion including London

dispersion forces, hydrogen bonding and electrostatic interactions (Flemming et al., 1999).
27

London dispersion forces are attractive forces generated by the establishment of a dipole

due to a temporary shift in an atoms electron density. These weak dipoles are the primary

attractive force between non-polar compounds. Though weak individually, high molecular

weight biopolymers can exhibit a significant attractive force due to the summation of thousands

of London dispersion forces. The importance of this attractive force in cellular aggregation has

been shown by Liao et al. (2001) and Urbain et al. (1993), where increased hydrophobicity in

activated sludge was found to correlate with improved flocculation.

Hydrogen bonding is the attractive force caused by the dipole between hydrogen and an

electronegative atom, usually oxygen in hydroxyl groups. Hydrogen bonding is responsible for

protein tertiary structure and interactions between the hydroxyl groups found on polysaccharides.

Electrostatic interactions are the interactions caused by charged surfaces. Many EPS

carry negative surface charge due to negatively charged functionality, such as deprotonated

organic acids. Examples of organic acids commonly isolated from EPS include uronic acids from

polysaccharides (Frlund et al., 1996) and glutamic and aspartic acids from proteins (Dignac et

al., 1998). While the interaction of two negatively charged species is repulsive, it has been

proposed that divalent cations (such as Ca2+ and Mg2+) can form ionic bridges between negative

functional groups on neighbouring EPS (Eriksson and Alm, 1991; Sobeck and Higgins, 2002).

Other researchers suggest that the electrostatic interactions are not beneficial to cohesive

strength, and thus the addition of either monovalent or divalent cations simply reduces the

repulsion of EPS (Zita and Hermansson, 1994; Cousin and Ganczarcyck, 1998). Although there

is debate about the importance of either mechanism, there is substantial evidence to suggest that

the concentration of cations plays a critical roll in EPS cohesion. A survey of the literature

investigating the roll of cations in EPS cohesion is presented in the following section.
28

2.1.4. The Role of Cations in EPS Cohesion

Several mechanisms for the cation-induced cohesion of EPS have been proposed. These

include Derjaugin-Landau-Verway-Overbeek (DLVO) theory, cation bridging, and alginate

formation. While the literature contains different opinions on the importance of each of the

proposed mechanisms, it is likely that each mechanism plays some role in EPS adhesion.

DLVO theory predicts interaction energies for charged surfaces with respect to separation

distance in a liquid phase with electrolytes (Derjaugin and Landau, 1941; Vervey and Overbeek,

1948). According to DLVO, increasing the ionic strength of the intercellular solution will reduce

the repulsion between EPS compounds. This would allow for biopolymers to move within a

range where the London dispersion (hydrophobic) interactions could lead to increased cohesion.

This DLVO mechanism has been used by Zita and Hermansson (1994) to explain the observation

that increasing ionic strength led to larger activated sludge flocs and improved sludge settling

characteristics. Their study was performed using a batch addition of electrolyte and the improved

flocculation was observed to occur similarly for potassium and calcium ions. Cousin and

Ganczarczyk (1998) found similar results, again in a batch study, using high concentrations (10-

45 g/L) of NaCl as an electrolyte. The DLVO theory also compares well with observations that

decreased sludge surface charge improves floc settling (Liao et al. 2002).

The cation bridging mechanism involves divalent cations (usually Ca2+, Mg2+ and Fe2+)

ion-pairing with the negative functional groups of two EPS molecules. Evidence of the cation

bridging mechanism was presented by Eriksson and Alm (1991) through their observation of

activated sludge deflocculation after treatment with a chelating agent (EDTA). They concluded

that the chelating agent preferentially bound the divalent ions in the EPS matrix of the flocs,
29

causing a removal of the cation bridges and deflocculation. Biggs et al. (2001) used particle size

analysis to show an increase in activated sludge floc size following the addition of calcium.

While this work measured the effect of cation addition directly (as opposed to using indirect

settling parameters) it did not offer insight into the relative importance of the DLVO or cation

bridging mechanisms. Higgins and Novak (1997a;b) found that by increasing the ratio of

divalent to monovalent cations (DM ratio) in the feed to a lab-scale activated sludge reactor, the

settling and dewatering properties were improved. They further determined that below a DM

ratio of 2, the settleability of the sludge flocs began to become problematic. This suggested that

for their experiment, the cation bridging mechanism was of greater importance than the DLVO

mechanism. The authors also found that simply adding monovalent sodium to their reactor led to

poorer settling and dewatering, an observation in direct contradiction to DLVO theory. It has

since been concluded that discrepancies between the DLVO and cation bridging proponents are

likely due to the use of batch versus continuous experiments (Novak et al. 1999; Sobeck and

Higgins, 2002). When applying cations to a batch study, the immediate effects most likely

manifest themselves as a DLVO response, leading to improved settleability. However the

continuous application of cations to a growing system leads to incorporation of ions into the EPS

matrix. If the ions are divalent, then increased bridging is observed. However, if the added ions

are monovalent, an ion exchange can occur, reducing the amount of cation bridging and

aggregate strength. In fact, Sobeck and Higgins (2002) found that the observed sludge property

improvements due to addition of divalent cations were only fully realized after 20 days of

continuous operation. This speaks to the time scale of each mechanism and their ultimate

applicability to real systems.


30

The formation of an alginate gel with the specific binding of calcium ions has also been

proposed as a mechanism for cell adhesion. This mechanism is really a type of cation bridging

mechanism that involves only alginate and calcium. Other divalent cations such as magnesium

are not of the right size to properly form a gel with alginate, leading to the polysaccharides

selectivity. Work involving the pure strain of Pseudomonas aeruginosa, an organism commonly

found in wastewater treatment systems, led to the isolation of alginate and its implication in cell

binding (Bruus et al., 1992). While it is likely that many wastewater treatment systems contain

organisms capable of producing alginate, it has been shown that both Mg2+ and Ca2+ can improve

the strength of cellular aggregates, which suggests that cation bridges with other compounds are

being formed (Sobeck and Higgins, 2002).

By examining the different proposed mechanisms of EPS cohesion and the supporting

evidence, it seems likely that cation bridging is the most prominent force at work in typical

wastewater treatment systems. It has been found that some activated sludge plants can show

significant improvement in sludge settling (due to improved flocculation) by continuously

adding calcium and thus increasing the DM ratio (Higgins and Novak, 1997b).

Much of the work discussed so far involves aggregation in activated sludge, but there

have also been investigations in biofilm systems that indicate the importance of divalent cations.

For instance Turakhia et al. (1983) used the chelating agent ethylene glycol-bis(-aminoethyl

ether) N,N-tetraacetic acid (EGTA) to selectively remove calcium ions from a mixed culture

biofilm. It was found that by adding EGTA, increased rates of biofilm detachment could be

induced. Turakhia and Characklis (1988) investigated Pseudomonas aeruginosa biofilms

growing under different calcium concentrations at constant shear and organic loading. They

found that increasing the calcium ion concentration led to the establishment of a thicker biofilm
31

due to greater biofilm strength and consequently a lower detachment rate. It was further

determined that calcium addition did not affect the specific growth rate, suggesting a growth-

independent process. Applegate and Bryers (1990) also observed increased biofilm accumulation

for Pseudomonas putida biofilms at higher calcium concentrations. This was observed for both

oxygen and carbon substrate limited biofilms, which further supports the finding that the

enhanced growth was not a metabolically-driven response.

Since divalent cations are hypothesized to enhance bonding, some investigations have tried

to quantify the effect of divalent cations on the cohesive properties of biofilms. Krstgens et al.

(2001) measured the compressive strength of biofilms grown on agar plates with different

concentrations of calcium ion using a film rheometer. It was found that increasing the calcium

ion concentration increased the films Youngs modulus and yield stress, suggesting a stronger

biofilm. Further evidence for the role of cation bridging in enhancing biofilm strength was

presented by Stoodley et al. (2001), who observed and quantified deformation of colonies

through digital image analysis of time lapse microscopy. Deformation was measured before and

after exposure to a 1g/L dose of AlCl 3 for 3 hours. This batch exposure to elevated cations (in

this case trivalent) was found to increase the apparent elastic modulus of biofilms grown in flow

cells. Using a more elegant technique to directly measure cohesive strength with atomic force

microscopy, Ahimou et al. (2007) also found that biofilms grown under elevated calcium levels

were significantly more cohesive. These studies all suggest that cation bridging is an important

mechanism for cellular cohesion in biofilms.


32

2.1.5. Biofilm Detachment

Biofilm detachment is the process by which material from the biofilm breaks free and

enters the suspended phase. Detachment is a complex phenomenon based on physical-chemical

interactions of EPS and microbes at the cellular level as well as biologically mediated processes

such as those discussed in the section on quorum sensing. Though complex, detachment is a

highly important process that governs the accumulation of biomass, the production of suspended

solids, and biological survival and proliferation strategies.

To begin to understand detachment, it is important to consider that this process represents

a range of mechanisms. Bryers (1987) characterized four general modes of detachment that

remain the basis for our consideration of the subject:

a) Erosion of small particles that detach from the biofilm surface in a continuous

manner;

b) Sloughing, where large segments of the biofilm dislodge in discreet events,

sometimes dislodging material right down to the substratum;

c) Abrasion, where biomass is continuously removed through collisions with other

solid objects; and

d) Grazing, where higher order organisms such as protozoa cause detachment due to

film weakening resulting from predation on bacteria.

While all detachment mechanisms are observable in various biofilm systems, the importance of

each mechanism for a given system depends primarily on the shear forces present

(hydrodynamic or through contact with external solid entities), substrate loading and the

presence of higher organisms. The modes are also not exclusive, with several reports of various

types of detachment occurring from the same biofilm. For instance several studies examine
33

simultaneous erosion and sloughing phenomena (Stoodley et al. 2001; Telgmann et al., 2004;

Wilson et al., 2004) while others have investigated simultaneous erosion and abrasion (Nicolella

et al., 1997; Kwok et al., 1998).

In addition to different modes of biofilm detachment, there are many factors that can

influence the rate of detachment. Since detachment is a balance between the shear forces applied

to the biofilm surface and the cohesive properties of the biofilm itself, factors that influence

biofilm detachment rates either increase the applied forces (hydrodynamic shear, collisions with

particles) or change the cohesive strength of the biofilm (biologically mediated weakening,

biofilm chemistry changes, formation of internal gas bubbles). In many cases, detachment has

been studied to develop a mathematical expression for the rate of detachment so that this can be

incorporated in biofilm modeling.

2.1.5.1. External Forces Causing Detachment

First considering changes in external forces applied to the biofilm surface, hydrodynamic

shear is an important factor present in all but the most quiescent biofilm systems and its effect

has been characterized by several investigations. Rittmann (1982) and Trulear and Characklis

(1982) performed initial work using rotating annular reactors to demonstrate that the detachment

rate can be a function of the shear force experienced by biofilms. Greater measured shear forces

at higher rotational speeds were related to a non-linear increase in the specific rate of

detachment. Peyton and Characklis (1993) examined this relationship in more detail to

distinguish between short term shear changes and steady-state conditions suggesting that at

steady-state, the magnitude of constant shear was not considered to be a significant predictor of

detachment rate. All of these authors recognize that different biofilm systems may have different
34

dependencies on shear and this may explain the conflicting findings. More recently, Choi and

Morgenroth (2003) used online particle size analysis to monitor changes in detachment

phenomena as a result of dynamic changes in hydrodynamic shear, linking shear change to

biofilm detachment rates and the size distribution of detached particles.

Shear can also be applied through abrasive contact of solid objects against the surface of

the biofilm, as is the case for fluidized bed type reactors. Chang et al. 1991 observed that steady-

state detachment rates were higher at increasing levels of abrasive shear achieved by adding

more granulated activated carbon (GAC) carriers in a GAC airlift reactor. Gjaltema et al. (1995)

published results depicting similar trends in a biofilm airlift suspension (BAS) reactor with basalt

particles as a biofilm carrier. In this experiment the airlift reactor was not fed any substrate

during detachment trials with the aim of calculating a detachment rate that was specifically

related to shear under non-growing conditions. This work demonstrated that increased abrasive

shear caused increased detachment in non-growing biofilms.

2.1.5.2 Internal Weakening Causing Detachment

There are also a number of factors that can cause weakened cohesive properties of

biofilms, leading to increased detachment. These can generally be separated into external

changes in the aqueous chemistry and biologically mediated mechanisms.

Biofilms can be subjected to external chemical changes that influence detachment. An

example of this is where detachment of biofilms in industrial piping is promoted by application

of oxidative antimicrobial chemicals such as chlorine, peroxo acids, iodophores and others that

break chemical bonds and kill microorganisms (Meyer, 2003). However detachment is not only

induced by oxidative chemicals and work by Chen and Stewart (2000, 2002) demonstrated that
35

exposure of biofilms to various ionic species and other chemicals could modify the viscosity of

homogenized biofilms and cause detachment in intact biofilms. The two measures were used to

infer that the ionic species affected the cohesive properties of biofilms which was the

mechanistic explanation of detachment. Some of the most significant chemical factors

investigated included NaCl (0.3M), NH 2 Cl (25-100mg/L), SDS (1g/L), and Urea (2M) where the

applied doses all resulted in greater than 50% biofilm detachment as measured through protein

quantification. Their work demonstrated the importance of electrostatic interactions in biofilm

strength, suggesting that chemicals targeting these interactions play the most significant role in

enhancing detachment. The importance of electrostatic interactions to biofilm cohesion and the

rate of biofilm detachment is a key motivation for the work in this thesis and the literature that

supports this connection has been also been described in section 2.1.4. In general the ionic

chemistry of the EPS matrix plays an important structural role and the removal of divalent

cations through chelation (Turakhia et al. 1983) or through ion exchange with monovalent

cations (Sobeck and Higgins, 2002) can weaken the biofilm, making it more susceptible to

detachment.

Biologically mediated detachment mechanisms are also of significance to wastewater

treatment biofilms. One such mechanism is the coordinated secretion of EPS-specific lyases.

Boyd and Chackrabaty (1994) demonstrated how Pseuomonas aeruginosa biofilms secreted an

alginate-specific lyase that allowed biofilm microorganisms to break free from their

characteristic alginate-based EPS matrix. A similar observation was made by Allison et al.

(1998) where detachment was enhanced when Pseudomonas fluorescens biofilms were exposed

to a solution containing an exopolysaccharide lyase secreted by the organism. Kaplan et al.

(2004) found that a polysaccharide lyase produced by Actinobacillus actinomycetemcomitans


36

could induce detachment not only in Actinobacillus actinomycetemcomitans biofilms, but also

Staphylococcus epidermidis biofilms exposed to the lyase. This suggests that certain secreted

lyases may act on EPS components that are not specific to their own species, an important

finding that supports the hypothesis that organisms can effectively invoke detachment in mixed

community systems.

For a biofilm detachment event to be triggered by secretion of lyases, coordinated

behaviour is likely required through cell signaling. The role of quorum sensing is often

implicated in coordinated detachment events and the review of quorum sensing in section

2.1.1.2. discusses some of the evidence supporting this link (i.e. Davies et al. 1998; Dow et al.,

2003; Thormann et al., 2005; Gjermansen et al. 2005). However it should be noted that

detachment, being a multifactor process, is not necessarily regulated by quorum sensing

pathways in all systems. For instance Wilson et al. (2004) studied a Pseudomonas aeruginosa

system where cultures with and without the lasI gene responsible for quorum sensing signal

production were found to have similar detachment rates and detached particle size distributions.

For situations where quorum sensing is involved in controlling detachment, these regulatory

pathways may also be linked to other environmental cues beyond cell density such as nutrient

cues (Lazzazera, 2000; Rice et al. 2005).

Nutrients, and in particular the lack of nutrients causing starvation, have been shown to

be an important factor governing detachment. Sawyer and Hermanowicz (1998) measured

detachment and nutrient depletion in flow cell system using microscopic image analysis of

changing microcolony sizes. They concluded that detachment increased when nutrients were

depleted to a greater extent. Hunt et al. 2004 used a combination of experimental observations

and biofilm modeling to evaluate the hypothesis that nutrient limitation in inner regions of
37

biofilms causes sloughing-type detachment. The authors used confocal laser scanning

microscopy to observe Pseudomonas aeruginosa biofilms with mushroom-like microcolony

structure (depicted in Figure 2.1) to show that as biofilms mature, voids develop in the inner

regions of microcolonies that the authors hypothesized result from starvation-induced cell death.

The voids cause structural instability that were linked to sloughing events, an observation that

corroborated other studies (Kaplan et al. 2003; Sauer et al. 2004). The modeling work of the

study by Hunt et al. (2004), which relied on a stochastic cellular automaton calculation technique

depicting an evolving grid of cells over a theoretical sample area, sought to reproduce the

observed voids by stipulating that cells detach and are removed from the model if substrate falls

below a critical concentration for longer than 24 hours. The resultant simulations did predict

biofilm microcolonies with void central regions and sloughing type detachment, however this is

not surprising since the voiding of inner regions was explicitly connected to the substrate

concentration which is expected to be low in deep regions of the biofilm. In fact, since many

species have adaptive mechanisms to survive during periods of starvation (e.g. spore formation)

disappearance of cells from the center of biofilm microcolonies may not be a plausible

explanation for the observed phenomena even if the simulated structure matches experimental

images. Chambless and Stewart (2007) also noted inconsistency in this conceptualization of

starvation-induced detachment.

Perhaps a better explanation for void-related sloughing events is the development of

oxidative stress through the production of reactive nitrogen intermediates such as NO and

ONOO- (Romeo, 2006; Barraud et al. 2006). Evolution of reactive nitrogen intermediates has

been shown to occur during anaerobic metabolic and endogenous processes and these authors

hypothesized that that Pseudomonas aeruginosa biofilms used NO-dependant cell signaling
38

pathways to regulate detachment. Detachment was found to be induced by exposing biofilms (for

P. aeruginaosa strains both with and without the NO-dependant signal pathway) to non-toxic

concentrations of NO. Van Alst et al. (2007) also demonstrated the importance of NO signaling

in dispersal processes of Pseudomonas aeruginosa biofilms. However these authors describe

how NO signaling is part of a complex response to nitrate concentrations that governs biofilm

virulence through increased motility mediated by the expression of extracellular rhamnolipids.

The specific role of the expressed rhamnolipids in detachment of P. aeruginosa biofilms has also

been more thoroughly investigated by Boles et al. 2005. Since these findings relate to behaviour

that may be specific to the well-studied opportunistic pathogen, it is uncertain how big a role NO

could play in a strategy to induce detachment of a mixed community. So far only limited

engineering work has been conducted to make use of NO as a dispersing agent. For instance

Charville et al. (2008) created NO-releasing xerogels for use as biomedical implant materials

showing that these materials prevented attachment of E. coli, S. aureus, and S. epidermidis.

Detachment can also be induced when microorganisms inadvertently weaken the biofilm

when gas bubbles are evolved under certain metabolic conditions. For instance Harramoes et al.,

(1980) found that detachment in pilot scale biofilm reactors used for tertiary denitrification was

caused by the evolution of nitrogen gas bubbles. Ohashi and Herada (1994) took microscopic

videos of gas bubble formation in biofilms cultivated in flow cells, demonstrating the resultant

detachment. Some common gas-inducing scenarios for wastewater treatment are for the

evolution of methane, hydrogen sulfide, or nitrogen and the likelihood of this mechanism being a

significant contributor to a biofilm detachment rate is dependant on the concentration profiles of

substrates such as oxygen, sulfate, and/or nitrogenous substances such as ammonia/nitrate/nitrite

in the biofilm.
39

A final biologically mediated detachment mechanism is detachment caused by grazing by

eukaryotes such as ciliates, amoeba, flagellates, rotifers, and nematodes. Grazing results in the

metabolism of ingested bacteria and the potential structural weakening of the biofilm causing

additional shear-induced detachment. The loss of biofilm mass due to grazing can be significant

as shown by the measurements of Huws et al. (2005), where average steady-state biofilm

thickness was observed to be reduced from 500 m to 200 m when mixed community biofilms

were exposed to the ciliate Colpoda maupasi. It can also significantly impact bioreactor

performance as shown by Lee and Welander (1994), who found that grazing limited nitrification

performance of pilot moving bed biofilm reactors which they hypothesized to be a result of a

reduced mean cell residence time (MCRT) in the biofilms. These authors suggested that the

effect of grazing applied a detachment pressure that led to more rapid turnover of cells in the

biofilm and this caused the slow-growing nitrifiers to be selectively washed out of the biofilm.

Given that grazing depends on a complex predator-prey relationship, it can be

challenging to predict or describe. Even a controlled study by Canale (1973) to examine and

mathematically describe one ciliate (Tetrahymena pyriformis) grazing on a pure suspended

culture of Aerobacter aerogenes led to unpredictable changes in populations following

adjustments to the dilution rate of the continuous system. If one adds to this the observation that

certain protozoa selectively feed on particular bacterial species (Rnn et al., 2002), the

possibility that different higher organisms have different affinities for grazing biofilms or

suspended bacteria (Caron, 1987), and the observation that aggregated bacteria can employ

protective strategies to avoid consumption by higher organisms (Matz and Kjelleberg, 2005),

mathematical description of this complicated phenomenon is challenging. This may explain the
40

fact that biofilm modeling approaches do not often explicitly account for grazing detachment

even though this process appears to be significant for systems with higher organisms.

2.1.5.3. The Influence of Growth Rate on Biofilm Detachment

The importance of starvation on biofilm detachment discussed in the previous section

could also be considered as the extreme extent of a broader link between detachment and growth

rate. There is evidence that the rate of detachment can be predicted by a function of the growth

rate of a biofilm. For instance Peyton and Characklis (1993) compared the rate of detachment of

measurements from a rotating drum reactor to a function dependant on the growth rate to give a

significant correlation. This finding was further supported by comparing data from earlier

investigations in similar reactor systems. Gjaltema et al. (1995) also observed a dependence of

the detachment rate on growth in biological airlift suspension (BAS) reactors.

This prompts questions about the mechanism that underlies this correlative connection.

Kwok et al. (1998) posited a hypothesis to explain the significance of the growth rate on

detachment as a result of their measurements that further confirmed this relationship for BAS

systems. They postulated that as biofilm grows, weakly bound protrusions form, which are more

easily detached than the well-established film underneath. This also fits with the description of

biofilm formation described in Bryers and Characklis (1982) where they explain a link between

early biofilm development and high rates of detachment. Bryers and Characklis note that early

colonization of a surface is characterized by rapid growth of cell clusters that they suggest are

easily detached, leading to a rapid initial specific detachment rate. They further note that as

biofilms mature and form a substantial extracellular polymeric matrix the specific detachment

rate decreases due to increased biofilm stability. Other work by Nakhla and Suidan (2002)
41

demonstrates this relationship in anaerobic fluidized bed biofilms. However, while the

hypothesis that colonies or protrusions are more easily detached seems plausible, the majority of

the evidence that supports such a hypothesis is either based on correlation (for instance, Kwok et

al. 1998; Nakhla and Suidan, 2002) or theoretical approaches such as the modeling work

presented in van Loosdrecht et al., 1997.

The importance of genetic control of biofilm phenomena highlighted by recent work to

describe quorum sensing and morphological changes that occur as biofilms mature (Sauer et al.,

2002) suggests that there may also be a microbiological explanation for the dependence of

detachment on growth rate. Bester et al. (2005) reported observations of high growth-dependent

rates of cell detachment from Pseudomonas aeruginosa biofilms cultivated in flow cells. These

authors hypothesized that biofilms act as a cell factory producing a high rate of detached

suspended cells throughout the development and maturation of the biofilm. Their measurements

suggested that cell yield from biofilms was higher than for a suspended Pseudomonas

aeruginosa grown under similar conditions in a chemostat. It could be speculated that there may

be evolutionary advantages for biofilms to enhance detachment under high substrate flux

conditions since this would represent a favourable situation to actively colonize neighbouring

surfaces. Given the previous discussion of genetic regulatory pathways involved in detachment

processes, for instance through excretion of lyases, it is plausible that growth rate-dependant

signals could be linked into such pathways.

Whether the mechanism underlying the dependence of detachment on growth rate is

based on genetic or physicochemical factors it appears to be an important relationship for various

systems. Several modeling studies have explicitly incorporated the growth rate into a

mathematical representation of the detachment rate (Speitel and DiGiano, 1987; Peyton and
42

Characklis, 1993; Horn et al. 2003) finding that such a description is both predictive of biofilm

behaviour and simple to implement mathematically. It should be noted, however, that separating

the influence of biofilm growth and nutrient concentration can be difficult due to their direct

dependence.

2.1.6. Thermophilic Biofilms

Temperature is an environmental parameter that controls microbial community structure

and function. Since the industrial MBBR studied in this thesis operates under thermophilic

conditions a brief review of the differences observed between mesophilic and thermophilic

biofilms is presented for context.

An important functional effect of temperature is how it influences the kinetics of

microbial communities. It has been generally observed that thermophilic organisms have a

higher maximum specific growth rate than mesofiles (Rittmann and McCarty, 2001), an

observation that has also been confirmed for whole-community growth rates for suspended

cultures degrading wastewater acclimated at a range of temperatures (LaPara et al., 2000). More

rapid degradation of COD has also been measured for thermophilic biofilm treatment systems.

For instance, Liao and Liss (2007) observed that a membrane aerated biofilm reactor (MABR)

could achieve 90% COD removal when operated at 55oC while in a similar system at mesophilic

temperatures (18-28oC) COD removal was only 67%.

Community diversity has also been found to be influenced by higher temperatures.

Studies investigating the diversity of bacterial species present in communities cultivated under

different temperatures indicate that the community diversity decreases at higher temperatures.

For instance Norris et al., (2002) measured decreasing numbers of bacterial species in soil
43

samples across an increasing temperature gradient associated with geothermal heating.

Seckiguchi et al., (1998) compared granular sludge from upflow anaerobic sludge blanket

(UASB) reactors operated at 25oC and 55oC to find that the higher temperature granules had

fewer distinct species. These observations may be explained by the evolution of relatively fewer

thermophilic bacterial species due to the reduced prevalence of such conditions in natural

environments.

A final consideration of importance to biofilm systems such as the MBBR is the

influence of higher temperatures on structural characteristics of biofilms such as thickness, EPS

composition, and detachment rate. To date there are limited studies systematically comparing

thermophilic and mesophilic biofilms with respect to structural properties, however, some

evidence suggests temperature may be important. Liao and Liss (2007) reported that steady-state

biofilm thickness was significantly lower and biofilms were more hydrophobic when grown

under thermophilic conditions in MABR reactors in comparison to a mesophilic control.

Vogelaar et al. (2005) found that activated sludge systems operated at thermophilic temperatures

resulted in sludge that settled poorly in comparison to a mesophilic control. Since hydrophobicity

and zeta potentials of sludge flocs were found to be similar in both cases, the authors concluded

that temperature likely affected polymer bridging behaviour. However, in contrast to these

observations, other researchers have found that a thermophilic temperature of 45oC was optimal

for producing sludge with a low sludge volume index (SVI) when running sequencing batch

wastewater treatment reactors at a range of temperatures from 35-60oC (Tripathi and Allen,

1999). If temperature does decrease the cohesive properties of biological aggregates, this may be

important for biofilm detachment. The thinner biofilms observed by Liao and Liss (2007) under

thermophilic operation of an MABR indicate that for this biofilm system higher temperatures
44

caused more rapid detachment rates. However the lack of comprehensive investigation on the

effect of thermophilic conditions on biofilm structural properties constrains the potential for

detailed conclusions.

This brief review indicates that there may be differences between the structural and

functional properties of biofilms grown at mesophilic and thermophilic temperatures. A more

fulsome review of thermophilic biological treatment from a performance perspective is presented

by Suvilampi and Rintala (2003).

This section concludes the review of literature of the salient features of biofilms relevant

to this thesis. The literature clearly depicts the complexity of biofilms and mixed community

biofilm treatment technologies such as the moving bed biofilm reactor. This highlights the

importance of investigation that provides information about the key process variables important

in applied systems to aid in system control and optimization.

2.2. The Moving Bed Biofilm Reactor (MBBR)

The MBBR was invented as an alternative wastewater treatment process that could

provide advantages over activated sludge or other biofilm technologies. Since its invention, there

has been considerable work to demonstrate the reactors effectiveness in different treatment

situations. This research has explored the functionality of the reactor in different industries and

applications as well as the removal of different compounds of interest.

For examples of treatment in different applications, a report by Rusten et al., (1997)

provides a good summary of performance data from 13 different Swedish municipal plants using

MBBRs. These selected plants were all small-scale, with the largest designed for <2200 m3/day

capacity, including examples of both newly built MBBRs as well as those converted from
45

existing activated sludge or other treatment systems. In all cases studied, the MBBRs achieved

greater than 90% Biochemical Oxygen Demand (BOD 7 ) removal. This work demonstrated the

broad applicability of the MBBR to small treatment scenarios, a situation seen as an initial niche

market for the reactor since it requires limited operator control. Other studies by degaard et al.

(1993 and 1994) have demonstrated the effectiveness of MBBRs in full-scale municipal

treatment with similar reductions in BOD.

The MBBR has also been applied to a range of industrial wastewaters. Rusten et al.

(1999) investigated an MBBR used as a pretreatment step for a chemical plant effluent. As a

pretreatment reactor, the studied system was exposed to elevated organic loads (>50 g/m2d) but

was still able to consistently reduce BOD by 60-80%. Other work on MBBRs treating pulp and

paper effluents (Rusten et al. 1994; Broche-Due, 1994; Jahren and degaard, 1999) has found

the MBBR to be a suitable technology for the sector. Nitrification of aquaculture waste has also

been demonstrated and Rusten et al., (2006) reviewed design considerations specific to this

industry.

MBBRs have been used to treat all common wastewater pollutants including dissolved

organics, nitrogenous compounds, and phosphorous. A summary of performance data collected

from lab, pilot and full scale is shown in Table 2.1. As can be seen from this table, a large body

of pilot experience has been published to confirm the general applicability of the MBBR to a

range of treatment situations.

With a large and increasing number of MBBR installations around the world, research

continues to explore new operational regimes and design optimizations. Recent work by Ahl et

al., (2006) investigated the potential for installing membrane filters in lieu of the typical MBBR

grating for retention of carriers and suspended solids. Their work examined the potential for
46

membrane fouling in such reactors which may begin to find increased application due to their

potential for compact and highly effective solids control retention. Gaul et al., (2005) also

examined the potential for simultaneous nitrification/denitrification by operating MBBRs to

promote the growth of anaerobic ammonia oxidizing (Annamox) bacteria. Their work describes

specific operational strategies that establish a community of such bacteria in MBBRs,

specifically a low hydraulic retention time and a finely tuned dissolved oxygen concentration.

There has yet to be any study examining how divalent cations may influence treatment

performance or promote conditions for multiple degredative processes.

Table 2.1. Summary of MBBR performance data from the literature

Reference Application Experimental Details Treatment Performance


degaard et al. , 1994 Municipal wastewater, nutrient Pilot >85% TKN removal
removal
degaard et al. , 1993 Small municipal treatment plant Small full scale plant, serviced 96% BOD7 removal
~250 people. 94.5% COD removal
97.1% P removal
41.5% TKN removal
Rusten et al., 1992 Dairy farm effluent Pilot and full scale data 40-95% COD removal for pilot
85-90% COD removal for full scale
Broche-Due et al., 1994 Pulp mill effluent, sulphite semi- Pilot 75% soluble COD removal
chemical pulping.
Rusten et al. , 1997 Summary of various small scale Small full scale plants, 3 newly >92% BOD7 removal
municiple treatment plants in installed and 2 converted from
existing activated sludge plants.

Rusten et al. , 1999 Chemical plant effluent. Pilot. Two reactors at low (18 80% (low) and 60% (high) soluble BOD
gBOD/m2d) and high (40 removal
gBOD/m2d) loadings
Jahren and degaard 1999 Pulp mill effluent, Pilot 60-65% soluble COD removal.
thermomechanical pulping, high
Embley, 2001 Pulp mill effluent, kraft pulping, Pilot 75% BOD5 removal at 38C
both mesophilic and 63% BOD5 removal at 58C
thermophilic effluent (38C and ~30% COD removal at both temperatures
Helness and degaard, 2001 Lab-scale removal of both Lab-Scale, SBR operation with >95% P removal
Nitrogen and Phosphorous. both aerobic and anaerobic 70-90% TKN removal
Synthetic wastewater with phases.
acetate as a carbon source.
Tal et al. , 2003 Lab-scale investigation of 2000L reactor run with 0.59-0.75 mgNH3/m2/day
nitrification and denitrification different loadings to evaluate Note: lab scale rate rate determined in batch
for an aquaculture treatment nitrogen removal tests
Wang et al., 2006 Lab-scale MBBR degrading 13 L Lab-scale reactor run at 77% BOD removal at D.O. = 2mg/L
BOD and removing Nitrogen different D.O levels. Loading 71% COD removal at D.O. = 2mg/L
was 0.45 kg BOD/m3 d, 1.17 89.9% TKN removal at D.O. = 2 mg/L
degaard, 2006 Review of various municiple Full-scale municipal treatment 91-94% COD removal
treatment plants plants with various treatment 73-85% TKN removal
objectives. 3700-5700 m3 94-98% P removal
MBBR volume
47

2.2.1 Carrier Design Considerations

The fundamental characteristic of the MBBR is the specially designed biofilm carriers,

for which the geometry, sizing and materials of construction have been considered carefully to

maximize performance. Since the original patents for the MBBR were assigned to a company

now embodied by AnoxKaldnes AB, much of the research literature investigating carrier

properties has focused on the AnoxKaldnes carriers such as the K1 carrier investigated in this

thesis.

One of the most important features of MBBR carriers is that they contain a large

protected surface area for biofilm colonization. The colonized surface area in an MBBR appears

to be one of the most important design characteristics governing the rate of substrate conversion

as shown by degaard et al., (2000). This is a key difference from the activated sludge process

where treatment performance is more directly tied to reactor volume. In the MBBR, surface area

can be increased by designing carriers with a higher specific surface area or by adding a greater

quantity of carriers to a reactor volume degaard et al., (2000). This offers flexibility for future

treatment capacity upgrades without requiring the construction of additional reactors. However,

as noted in degaard (2006), the

volumetric fill fraction for carriers can only A B

be increased to roughly 70% before mixing

problems begin to occur so there is a limit

to the potential for upgrading the surface

area for a given reactor volume.

Beyond adding additional carriers,


Figure 2.2. Schematic diagrams depicting
surface area can be manipulated through structures of the original K1 carrier (A) and the
recently patented Biofilm Chip design (B).
48

design of the carrier geometry. For instance a recent patent by Lfqvist et al. (2007) describes a

new carrier design marketed under the name Biofilm Chip with a high surface area of 1000

m2/m3 (compared with the 500 m2/m3 for the commonly used K1 carrier also investigated in this

thesis). This is achieved with the wafer-like design depicted in Figure 2.2b where the small wafer

holes (1mm 1mm 2mm deep) offer protected surfaces for biofilm growth. In moving to a

higher specific surface area, the geometry necessarily results in smaller inter-wall spaces. This

presents the possibility of biofilm growth bridging the gaps between adjacent walls, causing

plugging of the carrier material. Since many studies have measured mature biofilm thicknesses >

1mm (Murga et al. 1994; Okabe et al., 1999, Horn et al., 2003, as examples), this plugging of

carrier designs such as for the Biofilm Chip appears to be a likely scenario for wastewater

biofilms. Plugging dramatically reduces the effective surface area and could lead to significant

amounts of biofilm in diffusion-limited regions. This changes the nature of the biomass in

MBBR reactors underscoring the importance of the current work to better understand how

biofilm structure is impacted by divalent cations or periods of starvation, particularly if these

impact the potential for carrier plugging. So far the potential for biofilm plugging of carriers has

not been addressed in the literature.

2.3 Overview of Biofilm Modeling

The complex features of biofilms make them more challenging to describe

mathematically than suspended cultures. Specifically the diffusion gradients of substrates,

structural heterogeneity, and detachment phenomena add complexity to traditional approaches to

predict the kinetics and growth of microbial communities. To confront the unique features of

biofilms, a number of models have been proposed in the literature, encompassing different levels
49

of this complexity. Several review articles discuss the merits of various proposed models (Arvin

and Harremos, 1990; Wimpenny and Colasanti, 1997; Chaudry and Beg, 1998; van Loosdrecht

et al. 1997; 2002; Wik, 2003; Wanner et al., 2006). Most of these reviews share the view that

selecting a biofilm modeling approach should be guided by the golden rule of modeling: a

model should be as simple as possible, and only as complex as needed (Wanner et al., 2006).

With this in mind, biofilm models can be divided into three general categories that may be most

suitable to the objectives of a given study. These include:

1. Empirical or semi-empirical models

2. Theoretical diffusion-reaction models. These are typically described by a system of

differential or analytical equations based on average biofilm properties.

3. Discrete cellular automaton models. These are evolving simulations in a theoretical grid

of space with biofilm cells that propagate into neighboring gridpoints according to

algorithms tied to kinetic parameters.

Multivariate statistical analysis is an example of empirical approach used in this thesis

and the multivariate techniques applied are reviewed in section 3.1.3. Other empirical or semi-

empirical approaches have been employed to describe biofilm systems, such as Plattes et al.

(2006) description of moving bed biofilm reactor biofilms. These may be highly effective

predictors for process control but not necessarily informative with respect to a broader

understanding of biofilms. For this reason empirical modeling papers are not reviewed except for

multivariate approaches and those that are directly relevant to the MBBR.
50

Of the theory-based modeling approaches both diffusion-reaction and discreet cellular

automaton models have been useful in furthering our understanding of biofilms and helping to

provide a framework for process design and control. Diffusion-reaction models are useful for

predicting the flux of substrates into the biofilm using measured average parameters and thus can

be the best choice for modeling and validating a reactor system where macroscopic properties are

of interest. On the other hand, if an investigation seeks to test hypotheses related to the

microscale structure and/or determine how test variables affect biofilm morphology, a cellular

automaton approach is required.

Examples of diffusion-reaction models include those proposed by Jennings et al. (1976),

Rittmann and McCarty (1980a,b), and Wanner and Gujer (1986). The foundation for these

models is a coupling of the substrate flux into the biofilm with the growth, decay and detachment

Substratum Biofilm Boundary Bulk


Layer Liquid Nomenclature Used

High S f,b,o Limiting substrate


Concentration concentration in film, bulk, of
Substrate in the reactor inlet (f, b, or o)
Substrate Concentration

Rs Rate of microbial substrate


consumption
D f,bl Diffusion coefficient for the
biofilm and the boundary
layer (f, bl)
Low J Flux of substrate at biofilm-
Concentration water interface
Substrate a Specific biofilm surface area
HRT Hydraulic retention time

Note: Bulk formula for CSTR reactor

dS f d 2S f dSb S o Sb
Df Rs Ja Rs
dt dx 2 dt HRT
dS
D bl ( S b S f )
dz
Figure 2.3. Basic schematic of diffusion-reaction models. Equations based on Wanner and Gujer (1986)
and are arranged for a CSTR biofilm system.
51

of cells. A schematic depicting the general trends of these models can be seen in Figure 2.3.

Depending on the substrate loading and biofilm detachment rate, a biofilm may either be shallow

(with a substrate concentration >0 at the biofilm-substratum boundary) or deep (complete

substrate depletion at some depth).

The Rittmann-McCarty model is a simplified, one-dimensional model, depicting a

homogenous single-species biofilm with one limiting substrate at steady state. These

simplifications result in a steady state biofilm thickness that is dependant on the rate of biofilm

growth and detachment. To further simplify the use of this model at a time when computation

was considerably slower Saez and Rittmann (1992) developed a pseudoanalytical solution that

avoids the repetitive solution of non-linear differential equations. This model was validated by

Rittmann and McCarty (1980b) using a column reactor with controlled hydrodynamic conditions

and substrate loading. While the experimental results fit the model predictions, the highly

controlled system used for validation is not representative of most industrially relevant biofilms.

The biofilm model proposed by Wanner and Gujer (1986) extended the diffusion-reaction

model to include dynamic conditions, multiple microbial species and substrates, and customized

detachment rate terms. Solutions for several

case studies were presented to explain

observed autotrophic/heterotrophic

competition dynamics, as well as the effects

of erosion-type detachment and sloughing.

One of the notable features of this model

was how it allowed for dynamic changes in


Figure 2.4. Model-derived structures from
biofilm thickness by representing all spatial Eberl et al. 2000
52

coordinates in a dimensionless grid. Wanner and Reichert (1996) extended the model further by

adding representations of water flow within the biofilm, the movement of particulates, EPS, and

both attachment and detachment processes. These terms were all incorporated in the general

differential equation calculation framework of the original Wanner and Gujer (1986) model.

Reichert (1994) developed a customizable software platform for performing model calculations

called AQUASIM which remains a well-cited tool for aquatic system modeling (Wanner et al.

2006).

There are a number of examples of cellular automaton models (Hermanowicz 1998; Kreft

et al., 1998; Picioreanu et al., 1998; Eberl et al., 2000; Pizarro et al. 2001; Lapsidou and

Rittmann, 2002). These models rely on discreet cellular automaton calculations, whereby the

system is defined as a collection of model cells in 2 or 3 dimensions with given rules of

interaction. Such rules depict the amount of substrate consumed by each cell, the rate of cell

growth and the diffusion of important chemical species occurring in a given time step.

Calculations are performed to generate new system coordinates at each time step allowing the

biofilm structure to develop. Structures generated from these cellular automaton models have

shown that relatively simple cellular rules yield complex biofilm structures, such as the

protrusions featured in Figure 2.2 (Eberl et al. 2000). These derived structures compare well

with the structural heterogeneity observed in real biofilms using confocal microscopy and

microelectrodes (van Loosdrecht et al. 2002). One of the limitations of the cellular automaton

approach is that it requires definition of a large number of cell properties which may be difficult

to validate besides examining the resultant biofilm structures they produce. However, these

models have become an important tool for testing structural hypotheses (Wanner et al., 2006) in

conjunction with confocal microscopy of biofilms.


53

2.3.1 Modeling Biofilm Detachment

The earlier review of biofilm detachment presented the complexity of this process and the

many factors that can be important for different biofilm systems. This has also resulted in a range

of mathematical expressions describing detachment. Most of the diffusion-reaction models

involve a term for detachment based on a constant rate of erosion/abrasion in units of detached

mass over time. In conjunction with the work to define variables that influence detachment, there

are models that express the rate based on shear stress, the rate of growth, and biofilm mass or

thickness. Table 2.2 contains some of the principal studies where biofilm detachment rate

expressions were postulated.

Table 2.2. Principal rate expressions for modeling detachment

Rate Expression Variables Reference


kd f L f 0.58
Volumetric density, thickness, shear stress Rittmann (1982)
Concentration of carrier particles, Reynolds
k b1 k b2 C p k b3 Re k b4 number, Shear stress
Chang et al. (1991)


L f k d' k d" Thickness and growth rate Speitel and Di Giano (1987)

kd X f Areal density Rittmann and McCarty (1980)

k d f L2f Volumetric density and thickness Wanner and Gujer (1986)

Operation, 0

Backwash, k d L f Lbase Thickness and operational mode Morganroth and Wilderer (2000)

Studies by Rittmann (1982) and Chang et al. (1991) depicted the detachment rate with

respect to a linear dependence on shear either explicitly or through variables such as the

Reynolds number and the concentration of glass beads (for abrasion). Shear was found to be

predictive of detachment for the systems in question, though even using dimensionless

parameters such as the Reynolds number may result in system-specific characteristics. For
54

instance the biofilm surface structure may strongly influence the hydrodynamic forces applied to

the biofilm due to the formation of heterogeneous ridges and channels. This could lead to

different detachment dependencies in comparison to a smooth biofilm.

The physiological state of biofilms has also been found to influence detachment and

Speitel and Di Giano (1987) expressed detachment as a term dependant on the growth rate, an

expression that fits well with experimental observations (Tijhuis et al. 1995; Nakhla and Suidan,

2002; Garney et al. 2008). While modeling detachment as being dependent on growth helps in

predicting detachment phenomena, theoretical detachment expressions with a simple direct

dependence on growth do not yield a steady state solution unless detachment always equals net

growth (accounting for cell decay processes). For instance, if the rate was greater than growth it

would result in complete detachment while a rate less than growth would cause indefinite

expansion. For this reason, Speitel and Di Gianos (1987) expression also includes dependence

on biofilm thickness (L f in Table 2.2.) which avoids unrestricted biofilm buildup. It could also be

argued that a thickness or biofilm mass-based detachment rate is justifiable as a representation of

biofilm stability, where thicker biofilms are more likely to detach than thinner biofilms. Other

significant modeling efforts have solely based detachment on thickness or mass either directly

(Rittmann and McCarty, 1980) or as a second order function (Wanner and Gujer, 1986). This

specific detachment rate may not be independent of other factors such as shear or growth rate,

but may be useful modeling framework in which a specific detachment rate constant (i.e. k b in

Table 2.2.) is considered a variable.

Evidence of sloughing events has led some modelers to model these discreet events.

Early work by Howell and Atkinson (1976) depicted sloughing in trickling filters by creating a

critical limiting substrate concentration criterion that triggered complete removal of the biofilm.
55

Although the general biofilm modeling approach used in the study was not sophisticated, the

simulation of sloughing events led to output that reflected observed reactor performance data.

Morganroth and Wilderer (2000) proposed biofilm models with several detachment mechanisms,

including sloughing, and projected their impact on microbial community structure. The authors

adapted the model of Wanner and Gujer (1986) as the foundation for their study. It was found

that for systems characterized by sloughing, a significant fraction of the biofilm was comprised

of heterotrophic bacteria at all biofilm depths. This differed from simulations using erosion-type

detachment, where a significant autotrophic population was predicted, with a much shallower

penetration of heterotrophs. This was explained by the fact that sloughing events exposed deeper

sections of the biofilm to oxygen, and thus favoured heterotrophic growth at deeper levels. While

these results are interesting, particularly for their application to biofilm

nitrification/denitrification, they have not been experimentally validated.

The detachment models discussed so far still do not connect clearly with a mechanistic

understanding of the process, but rather tie it to measurable process variables that may not be

widely applicable to various systems. Stewart (1993) sought to improve our mechanistic

understanding by describing a framework for deriving detachment expressions. In this approach,

the detachment rate was calculated as the integral of all detachments (of various sizes and at

various depths) as a function of their probability of occurrence. Thus the probability function can

be selected under a variety of assumptions (such as a direct dependence on growth, biofilm mass,

shear etc.) to derive detachment expressions. Making use of recently developed cellular

automaton models, the same group investigated how different detachment mechanisms could

impact the evolution of biofilm structure (Chambless and Stewart, 2007). By using a cellular

automaton model, the detachment rate could also be expressed as a probabilistic function
56

impacting simulated cells. Mechanisms investigated included detachment based on starvation,

shear, and the number of neighbouring cells (to reflect surface erosion). This work led to the

commonly observed mushroom structures only when all three detachment mechanisms were

combined in the simulation.

Through review of the different postulated detachment model expressions it can be

concluded that there are a number of options for modeling the MBBR and the most suitable

approach may be one that includes more than one dependency. Furthermore, there has been no

modeling exploration of the connection between cellular cohesion and the rate of biofilm

detachment. Factors such as the concentration of divalent cations may influence biofilm strength

and thus affect biofilm detachment rates. There has yet to be a study investigating this

relationship in either diffusion-reaction or cellular automaton modeling.

2.3.2. Modeling Moving Bed Biofilm Reactors

Even with their increasing application, moving bed biofilm reactors have been the subject

of limited modeling research. Although a significant body of performance data has been

published to guide reactor sizing and design, modeling has been used only to a limited extent to

explore the fundamental processes that govern reactor behaviour. Given the preceding discussion

of evolving approaches to model biofilms, there is potential to gain insight into this hybrid

reactor system.

Some research has led to the generation of MBBR models. Havla et al., (2002) took an

engineering practitioners approach to the simulation of the upgrade of an activated sludge plant

to an MBBR. Their work involved using the GPS-X process simulation software platform

(Hydromantis Inc., Ontario, Canada) to represent the system. This included a biofilm modeling
57

component included in the GPS-X code. The biofilm component simulated the biofilm by

dividing it into layered compartments reflecting changing kinetic parameters with depth. By

integrating the biofilm component into the treatment train, it was possible for the authors to

predict changes in performances due to transient loading. Fouad and Bhargava, (2005) used a

steady state analytical approximation of flux based on the biofilm model of Suidan and

Rittmann, (1989) to allow for solution of the biofilm component of the MBBR. This was coupled

with suspended phase equations to predict performance, although only under steady state

conditions. Plattes et al. (2006) also developed an MBBR model that simply addresses the

biofilm component by adjusting the Monod half-saturation constant (K s ) to account for diffusion

limitation. The authors added to this work by creating a dynamic MBBR model using a zero-

order assumption for the biofilm rate equations to simplify the solution. This led to good

predictions of transient nitrification behaviour.

All of the MBBR modeling investigations were conducted with the objective of creating

a predictive model that could be used by an engineering professional to optimize or control the

system. The emphasis has been on pollutant degradation and not biomass properties and

dynamics. To date no research has been conducted using more advanced approaches to

investigate biofilm activity, stratification, and detachment processes in MBBRs.


58

Chapter 3: Multivariate Statistical Analysis of an Industrial MBBR


3.1. Introduction

The findings and discussion contained in this chapter are substantively based on a published

account in:

Goode, C., LeRoy, J., and Allen, D.G. (2007) Multivariate analysis of a high-rate biofilm process
treating Kraft mill effluent. Water Science and Technology. 55 (6) 47-55.

3.1.1. The MBBR at Irving Pulp & Paper Ltd.

Irving Pulp & Paper Ltd., a Kraft pulp mill located in St. John, New Brunswick, Canada,

uses a moving bed biofilm reactor to reduce BOD in effluent from its pulp bleaching reactors

before discharge to the St. John River. Brought online in 2001, the reactor has been an important

component of the mills environmental management since its installation. Conversations with

engineers at Irving (personal communication, 2005) described several unique features of the

MBBR design, including the reactors thermophilic operating temperature (~58oC) and the fact

that the MBBR is run at a high rate (hydraulic retention time <2 hours is common) without a

solids separation device. These unique features were selected as design considerations to create a

treatment system that could be as compact as possible while still providing treatment

performance that went beyond regulatory compliance. Compactness was an important design

objective for the historical mill site which operates in the urban area of Saint John with limited

physical space to install new process units.

Perhaps the most interesting feature of the Irving MBBR is its operation at thermophilic

temperatures. The operating temperature of 58oC was chosen because unlike many treatment

scenarios, the bleach plant effluent was too hot to support biological treatment and required

cooling. This cooling step was designed to be achieved through direct injection of cooling water
59

to eliminate the need for a heat exchange unit. For direct injection cooling a higher MBBR

temperature (for instance, thermophilic instead of mesophilic) would require less cooling water,

which in turn leads to less effluent and a smaller reactor. While working at high temperatures

was desirable for these reasons, thermophilic wastewater treatment carries some uncertainty due

to there being limited operational experience for biofilm systems at this temperature range. To

overcome this uncertainty, the installation of the full scale reactor was preceded by a pilot study

that compared efficacy for treating the mill effluent at both mesophilic and thermophilic

temperatures (Embley, 2001). It was found that adequate BOD removal could be achieved at

both mesophilic and thermophilic temperatures following a period of acclimation. This data

supported the decision to install the thermophilic reactor and the subsequent full scale operation

at 58oC has further demonstrated effective operation at this temperatures. However, as a

thermophilic reactor with thermophilic microbiology, the Irving MBBR remains a unique

system. For the purpose of developing a model or understanding how process changes affect

performance, the Irving MBBR may not be comparable to theoretical relationships generated for

mesophilic temperatures.

A second route to minimizing the reactor size was through operation with a short

hydraulic retention time and by omitting a solids separation device. High rate operation is

possible at Irving because the reactor was constrained by one principal treatment objective: BOD

reduction. Since the pulp mill effluent is deficient in nitrogen, there was no need for the longer

retention times often required for nitrification. The retention of a high concentration of biofilm

microorganisms made possible by the MBBR design allowed for BOD reduction to be achieved

at a high volumetric rate. The decision to omit a solids separation device was also made in order

to simplify and reduce the size of the treatment system. The mills installation of the MBBR
60

coincided with several other environmental projects that together reduced the production of

treatable effluents through increased recycling of materials. This meant that at the reduced

effluent production rates, the suspended solids generated by the MBBR remained below the

mills regulated limits even without a separation step. The modeling challenge addressed by this

work, as described in the following sections of the chapter, is motivated in part by the fact that

minimizing suspended solids discharges in this unique situation will lead directly to water

quality benefits and continuous improvement in environmental management.

For background, a summary of typical operating specifications and performance is given

in Table 3.1.

Table 3.1. Summary of MBBR operating parameters. Temperature and pH are controlled

Treatment Variable Average Value Standard Deviation (n = 1117)


BOD in 10612 kg/d 1921 kg/d (18%)
BOD out 4291 kg/d 1004 kg/d (23%)
COD in 28683 kg/d 5482 kg/d (19%)
COD out 21677 kg/d 4833 kg/d (22%)
3 3
Total Flow 26556 m /d 3437 m /d (13%)
Hydraulic Retention Time (HRT) 1.61 hr 0.23 hr (14%)
Total Suspended Solids (TSS) in 1146 kg/d 744 kg/d (65%)
Total Suspended Solids (TSS) out 4951 kg/d 1240 kg/d (25%)
o o
Reactor Temperature 57.2 C 3.3 C (5.7%)
Ph in 6.8 0.3 (4.7%)

3.1.2. Performance Fluctuations

MBBR treatment performance in this work is judged based on the discharge of BOD and

TSS since these are the principal environmental measures that the MBBR impacts. As introduced

previously, the MBBR at Irving experiences considerable day-to-day variation in these

performance measures. Examples of this variable behaviour were introduced in Chapter 1 and

are depicted over a shorter time period in Figure 3.1 to better demonstrate the nature of the

fluctuations.

From examining Figure 3.1 several observations can be made. First it is important to note
61

10000

TSS (kg/d)
8000

6000

4000

2000

B
10000
Total BOD (kg/d)

8000

6000

4000

2000

8000
Soluble BOD (kg/d)

6000

4000

2000

0
0 50 100 150 200 250

Time (days)
Figure 3.1. Fluctuating treatment performance of the full-scale MBBR over 250
days of operation. (A) indicates 1-3 week shifts in the average behaviour while (B)
depicts short but highly abnormal decreases in performance.

that the discharge of suspended solids from the MBBR plays a significant role in the emission of

total BOD. This can be best observed by noting many of the days that feature high TSS and high

total BOD without a corresponding increase in soluble BOD. Thus for a reactor without solids

separation, the biosolids dynamics become directly tied to BOD removal efficiency because the

concentration of the suspended population will be included in the assessment of BOD. Looking

at the nature of the fluctuations, it is also possible to observe that there are multi-day trends in
62

performance. This is most clear for the TSS data which show distinct shifts to higher or lower

solids for periods of 1-3 weeks.

In addition to this, there are a number of shorter incidents where performance deviates

significantly from the average trend leading to poor treatment in all parameters. In some cases it

appears that this is due to poor removal of soluble BOD (e.g. days 28, 95, and 245) and others

where an unexplained discharge of solids causes poor performance (e.g. days 83, 133, 209). On

top of the general trends in performance and distinct abnormal events, there also exist

considerable day to day process changes. These daily fluctuations appear to be primarily caused

by the inconsistent discharge of TSS which in turn contributes to the total BOD.

Considering the different types of performance variability discussed, the MBBR at Irving

Pulp & Paper Ltd. exhibits complex behaviour that is likely influenced by a range of processes

that act on different timescales. The magnitude of the variability, particularly for the highly

abnormal periods, suggests that better understanding of the process may present significant

opportunities for improving performance.

3.1.4 Objectives

This chapter describes a multivariate statistical analysis of the MBBR system at Irving

Pulp & Paper Ltd. The central hypothesis of this work is that multivariate analysis is a suitable

tool for this purpose. If this hypothesis is accepted, the model should be able to provide

considerable insight into the causes of the observed fluctuating behaviour. The approach to this

question involved achieving two objectives.


63

Objective (A) was to create multivariate models to provide explanations for poor

performance and isolate possible control variables for improving treatment. These descriptive

models were designed so that they could be adequately predictive of BOD removal and TSS

discharge while remaining simple enough to allow the user to easily determine which process

factors were most important.

Questions

(A) Is multivariate analysis a suitable modeling approach to understand and predict the observed
performance fluctuations at the Irving Pulp & Paper Ltd. mill?

(B) What are the significant variables controlling MBBR performance for this application?

Hypotheses

(A) Multivariate statistical analysis is a useful tool for understanding and predicting performance
behaviour of an industrially applied MBBR

(B) The model will indicate mill process variables that have a significant impact on the performance
of the MBBR.

Objectives

(A) Develop a methodology for multivariate statistical analysis to provide process insight.

(B) Develop an approach for generating accurate predictions of MBBR performance.

Objective (B) was to further demonstrate the usefulness of a multivariate method by

providing a forecasting tool to predict BOD discharges without having to wait the required 5

days for the results of a BOD test. Such a tool would be useful in the case studied because as can

be seen from the plot of COD to BOD ratio in Figure 3.2., the effluent COD is not a sufficient

predictor of BOD. While the performance of many treatment plants can be monitored on short

time scales by COD, such an approach requires that the ratio of COD to BOD be relatively

constant. To provide a method for short term monitoring a more complex model was created to

use any data necessary to provide the most accurate predictor.


64

12

Effluent COD:BOD ratio


10

0
0 50 100 150 200 250

Time (days)
Figure 3.2. Ratio of COD to BOD in the MBBR outlet. Ratio is not constant.

3.1.3. Overview of Multivariate Analysis

In order to explain the variability in treatment performance (BOD, TSS) and predict

future performance, a model of the system was required. As introduced previously, for an

industrial situation with complex dynamic behaviour and a poorly defined wastewater

composition, theoretical modeling approaches may be limited. As an alternative to theoretical

modeling, multivariate statistical tools such as Principal Component Analysis (PCA) and Partial

Least Squares (PLS) have found application in a variety of industrial situations (Kresta et al.,

1991). The approach involves the generation of an empirical model based on measured historical

process data. Detailed explanations of the modeling concept and the algorithms used for

calculation have been previously presented by Wold et al. (1987), Geladi and Kowalski (1986),

and Kourti and MacGregor, (1995). In general it requires the calculation of a small number of

latent variables (principal components) from a larger number of original process variables

according to the following matrix form:


65

X = TP + E

Where X is the modeled data set, T is the matrix of the principal component scores such that the

columns represent each principal component, P is the transposed matrix of variable loading

factors and E is the matrix of model residuals. The vectors of each principal component are

orthogonal to each other and selected to maximize their description of the X variance (for PCA)

or to maximize their description of the Y output variance (for PLS). The principal components

summarize the correlation structure of a dataset and can indicate the importance of process

variables. The generated model can be used to detect abnormal operation, predict output

variables (in the case of PLS) and provide useful visualizations of multidimensional process

behaviour. The main requirement for successful multivariate modeling is a historical dataset that

is large enough to describe the significant operating trends of the process. Since multivariate

statistical analysis can be conducted without a priori variable relationships, it is useful for

situations where theoretical modeling is insufficient.

3.1. Methods

3.1.1. Overview of Mill Treatment Data

For this study, multivariate analysis was performed on data from January 3rd 2002 to

December 31st 2004, representing 3 years of operation. The dataset contains measurements taken

from the MBBR, the bleach plant and digestion for a total of 87 process parameters. Variables

were originally included if they were directly involved in MBBR control or had the potential for

impacting the chemical characteristics of the wastewater entering the MBBR. For all of the

predictive models, a segment of the data (June 9th 2004 to Dec 31st 2004, 20% of total dataset)

was reserved to independently validate the model.


66

3.1.2. Modeling Approach

Before creating a model to predict BOD from the MBBR, an overview of the process was

conducted by building a PCA model. A PCA overview model is useful for indicating if the

modeled data form clusters of similar operation. If data clustering is observed, separate

predictive models for each data cluster may improve predictability.

Following PCA analysis of the dataset, predictive models were developed to achieve

objective (A), the explanation of changes in treatment performance. Starting with the full list of

87 potential variables, the final modeled variables were selected by removing unnecessary

variables. The first step in doing this was to eliminate variables that were collinear and appeared

to be physically connected in the process. For instance the measurement of total flow could be

represented by including the flowrates of the different components that contribute to this value.

Although one of the strengths of multivariate statistical analysis is that it is not compromised by

collinearity (Wold et al., 1987), removing unnecessary model variables simplifies interpretation.

After removing collinear variables, the model was further refined by removing variables which

limited predictability. This assessment was based on examining how removing the variable

impacted the Q2 statistic of the model. This statistic is a measure of how the model would have

predicted the Y values (TSS and BOD) used for training the model. If removing a variable led to

a new model with at higher Q2, the variable was considered unnecessary. After performing these

refinements a list of model variables was generated that was small enough to allow for easy

interpretation.

As has been highlighted in several previous papers for multivariate analysis of biological

treatment, the time scale of process responses to changes in variables is important (Van Dongen
67

and Geuens, 1998; Eriksson et al., 2001; Barampouti et al., 2005). While some input variables

may cause immediate process change, others could impact the system on longer time scales. For

instance, the exposure of microorganisms to a toxic substance may cause process inhibition that

continues for a period of time after the exposure event. Several strategies to account for process

dynamics have been employed to improve multivariate models. These include variable lagging

and time scale decompression using wavelets and multiresolution. In this study, the variable

lagging approach was used due to its simplicity and demonstrated effectiveness (Van Dongen

and Geuens, 1998; Barampouti et al., 2005). The technique involves supplementing

observational data at time t with selected variables from day t - i for i = 1 to n where n is selected

according to the significance of the lagged variable. The predictive model was found to be

enhanced by adding several lagged variables.

The second modeling objective was to generate a predictor of MBBR effluent BOD that

could be used in lieu of simply assuming the COD:BOD ratio was constant. This allowed for the

inclusions of any available measurement without worrying about the interpretability of the

model. Variables were selected only to maximize predictive power of the model.

All model generation was performed with SIMCA-P v10.5 software licensed from

Umetrics AB (Ume, Sweden, www.umetrics.com). The software computes principal

components and performs cross-validation using a row and column-wise algorithm presented in

Eastman and Krzanowski (1982). Other computational software such as MatLab (Applied

Mathematics), SAS (SAS, North Carolina, USA, www.sas.com), or R (open source, available

online at www.r-project.org) can be used to perform multivariate analysis. SIMCA-P was

selected for the fact that it is specifically tailored to PCA and PLS analysis with built in

algorithms that do not require coding and easily manipulated graphical interface.
68

Variables are presented in this chapter with variable IDs in numeric form. The text

highlights the identity of important variables for discussion but a table of descriptions for

variable IDs can be found in Appendix 1.

3.2. Results and Discussion

3.2.1. PCA Overview Model

An initial PCA model was created to gain information about the overall system and

isolate any abnormal behaviour. The model specifications are listed for model PCAFull in Table

3.2. As can be seen from the table, the model reduces information from 57 variables into 9

principal components while explaining 69.7% of the variability in the X matrix. Of these, the
Principal Component 2 Score

-10

-20
-9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9
Principal Component 1 Score
Figure 3.3. Score plot for model PC1 with observations indicated by woodtype pulped and for days
with significant bleach plant shutdowns. ( ) denotes maple, ( ) denotes birch, ( ) denotes softwood,
and ( ) denotes a shutdown longer than 6 hours.
69

first 2 components are the most significant and can be used to visualize the system in 2-

dimensional space. A plot of PC1 and PC2 scores with an elliptical Hotelling T2 boundary ( =

0.05) is shown in Figure 3.3.

It can be seen from this plot that the data cluster and contain some outlier data points. The

Irving mill pulps wood in batches from three classifications of wood type; maple, birch and a

softwood mix. By labeling the data points according to these wood types, it is obvious that the

clustering observed in Figure 3.3 correlates with wood type pulped. This is not surprising since

the parameters for pulping and bleaching are wood type dependant, likely causing different

chemical compounds to enter the MBBR. To investigate the cause of the abnormal points in the

bottom left quadrant of the score plot, data were further labeled for days in which the MBBR is

cut off from feed due to a bleach plant shutdown longer than 6 hours. These shutdowns occur for

planned maintenance and for unexpected process upsets. The 6 hour cutoff time was chosen due

to an analysis of the effect of shutdown time on treatment performance. As can be seen from the

score plot, all the abnormal points occur on days affected by a >6 hour shutdown.

The results of the PCA model helped to guide the strategy for modeling treatment

performance. Since bleach plant shutdowns resulted in highly abnormal measurements, it was

decided that the predictive models of treatment performance should exclude these points. This

led to the development of normal operation models. The wood type clustering effect observed

raises the question: would the creation of three wood type specific models improve

predictability? In segregating data into wood type classes, there is a tradeoff between

advantageous noise reduction for a given wood type model and the negative impact of reducing

the available data pool. In the case of softwood, the most commonly pulped wood type, there

was enough data to create a separate model for improved prediction. For maple and birch data
70

were inadequate to create reliable models, and thus these wood types were modeled by a PLS

model of the whole data set. This approach was applied to both objectives for predictive

modeling.

3.2.2. Descriptive PLS Model

Two models were created to describe the MBBR process with respect to effluent BOD

and provide explanations for shifts in performance. These were a model of the full dataset

(PLSFull_1) and a softwood-specific model (PLSSoftwood_1). A summary of the key statistics

for each model is contained in Table 3.2.

Table 3.2. Summary of generated models.


Model Name Model Description Total Vars. Lagged Vars. PC's R2X R2 Y Q2 RMSEP (kg/d)

PCAFull Data overview 57 0 9 0.7 - 0.47 -


PLSFull_1 Descriptive, full dataset 53 12 4 0.52 0.33 0.29 985
PLSSoftwood_1 Descriptive, softwood data 35 3 3 0.48 0.32 0.28 713
PLSFull_2 Predictive, full dataset 58 12 5 0.54 0.52 0.45 941
PLSSoftwood_2 Predictive, softwood data 43 3 2 0.38 0.53 0.48 606

To assess the predictability of a given model, both the Q2 statistic and the Root Mean Squared

Error of Prediction (RMSEP) can be considered. While the Q2 represents predictability of the Y

values used in training the model, RMSEP is calculated from the deviation of the predicted

values for a previously unused set of validation data. The RMSEP values from Table 3.2 suggest

that the softwood model has enhanced predictability over the full dataset model. While the Q2

values are not high, there is an adequate fit for making inferences about the importance of

variables.

Figure 3.4 illustrates the descriptive ability of the PLSFull_1 model. Plot (a) features the

observed vs. predicted BOD over the validation set, while Figure 3.5 features the same plot for

TSS. Overall the model describes the major trends in the effluent BOD. However some regions

of the data are not well described by the model, such as the maple run from day 130-140. These
71

regions may be due to process behaviour that is unique to what has been measured in the training

dataset. For instance, in the maple data mentioned, the mill had begun a new mode of bleaching

operation that only involved 3 bleaching stages. Since there had been no previous operation with

3 stages, the model was unable to provide accurate predictions. In comparing the fit of the model

for solids, it appears that again while general trends are predicted, there some periods of

operation that are not well described. It is clear from the variable loading plot in Figure 3.4c that

TSS and BOD are correlated to each other, a finding that is not unexpected when considering

that the solids are primarily organics which would contribute to effluent BOD.

To demonstrate the descriptive ability of the PLS model, two regions of data are

indicated in Figure 3.4a (A and B) for comparison. Both regions exist within a period of maple

operation but depict a significant increase in BOD output from A to B. By creating a contribution

plot (Figure 3.4b), the relative change in model-weighted process measurements between A and

B can be observed. This identifies variables responsible for the observed increase in BOD output.

The contribution plot indicates that variables related to pulping and bleaching were the most

significant contributors to the observed change. These included variables 75 (methanol flow to

bleach plant), 72 (bleach plant production rate), and 70 (stock concentration) which were found

to increase, while 80, 81, and 82 (Kappa numbers at different stages of pulping) and 67 (black

liquor flow from digesters) were found to decrease. This suggests that the pulp mill had shifted

into an operation mode that altered the chemical composition of wastewater. It is possible that

this shift led to a wastewater that was either less degradable or inhibitory to MBBR

microorganisms.
72

Treated Effluent BOD (kg/d)


a)

6000 B

4000

2000 A

0 20 40 60 80 100 120 140 160 180


Time (days)
Variable Contributions (B - A)

5 b)
4
3
2
1
0
-1
-2
-3
2
3
4
5
6
8
9
11
17
22
25
26
45
57
58
59
60
61
62
63
64
65
67
68
69
70
71
72
74
75
76
77
78
79
80
81
82
83
84
Variable ID
5 9
c)
0.30 2
68 71
Variable Loading in PC2

0.20 325 BOD 8


426TSS57
62 83 11
76 6
58
78 6563
0.10 80
81 6779
61 60
8274 64
69
1(M)
1(S) 59
17 70
77 72
0.00 84 1(B)

75
-0.10

-0.20
22
-0.30 45

-0.20 -0.10 0.00 0.10 0.20


Variable Loading in PC1
Figure 3.4. PLSFull_1 model predictions, investigation of performance shift contributions, and
variable loadings a) PLSFull_1 model predictions ( ) and actual BOD values ( ) over the
validation dataset. Point A and B are highlighted for contribution analysis. b) Relative change in
selected variable contributions from region A to region B. Variables of note identified in text c)
Loading plot for all variables used in PLSFull 1. BOD and TSS Y variable denoted ( ).
73

As can be seen from the loading plot in Figure 4c, many of the variables implicated in

this BOD upshift are heavily weighted in the model, suggesting that the trend observed between

A and B is representative of other process events. The most important variables for predicting

BOD and TSS output from the MBBR were found to be flow parameters (D 0 stage effluent flow,

HRT, the importance of flow stoppage during process shutdowns), the wood type pulped, reactor

temperature, and pH. For the case of reactor temperature and pH the correlation is of note since

these variables were held constant through process control. The significance of these variables

arises from identified periods of inadequate control that coincided with decreased performance.

Residual effluent nitrogen that was not assimilated by the biomass after its addition (the mill

effluent is deficient in nutrients and N and P are added) was also found to be correlated with

treatment performance. This variable may act as an indirect measure of biofilm activity since a

decrease in growth rate would result in a reduction in the demand for nitrogen. As a comparison

with activated sludge, the MBBR is a very rapid treatment process that is much more susceptible

to flow changes. Without the dilution factor typical of an activated sludge process, the MBBR

may also be more susceptible to toxic inhibition. The effectiveness of this reactor system is

likely only made possible through the inherent resiliency of the biofilm itself.

3.2.3. Predictive PLS Model

A second set of PLS models was created for enhanced predictability of the output BOD.

The principal difference between the predictive and descriptive models was that the predictive

models included daily measurements of lab-based variables such as COD and TSS.

As in the previous case, both full-dataset (PLSFull_2) and softwood (PLSSoftwood_2)

models were created. The summary of model statistics is listed in Table 2. Again,
74

PLSSoftwood_2 was found to have improved predictability over PLSFull_2 both with respect to

Q2 and RMSEP. Both models were significantly more predictive upon the addition of the new

variables.

In the best case, the PLS technique was found to predict softwood values with an RMSEP

of 600 kg/d. This represents a relative error of 14.5%. Through analysis of the environmental test

methods at Irving Pulp & Paper Ltd., it has been found that the BOD measurement carries a 9%

relative standard deviation. With this in mind, the predictive power of the PLS technique is quite

good.
Treated Effluent BOD (kg/d)

7000
6000
5000
4000
3000
2000

0 20 40 60 80 100 120 140 160 180


Time (days)
Figure 3.5. Predictions ( ) and observed BOD output values ( ) for model PLSFull_2
over the validation set.

A plot of the model predictions and the observed BOD output values in the validation set

for PLSFull_2 is presented in Figure 3.5. The prediction trends are very similar to those of the

descriptive model even though the Q2 is much higher (0.483 compared to 0.291). With respect to

the current modeling problem, the findings suggest that the daily monitoring of COD does not

substantially improve the prediction of BOD beyond that achieved through online measurements.
75

3.3. Conclusions

Several conclusions can be drawn from this work. From PCA modeling it was found that bleach

plant shutdowns lead to abnormal treatment performance and the influence of wood type pulped

was significant. Descriptive PLS modeling was able to predict major trends in the validation set

to identify variables responsible for shifts in performance. Both BOD and TSS are linked due to

the contribution of organic solids to the BOD measurement. The most important variables

governing MBBR performance were found to be flow parameters (D 0 stage effluent flow, HRT),

the wood type pulped, faults in the temperature or pH control of the reactor, and some potential

indirect indicators of biomass activity (residual nitrogen and pH out). Predictive modeling using

additional variables showed some improvement in fitting the validation set. The best predictor

for modeling softwood data had an RMSEP of 606 kg/d representing a 14.5% margin of error.

This represents a good fit given the measurement error of the BOD test (9%). Overall,

multivariate statistical modeling was effective in providing information and predicting

performance for the MBBR process.


76

Chapter 4: The Effect of Calcium on MBBR Biofilms


The findings and discussion contained in this chapter are substantively based on an article

submitted to Water Environment Research.

4.1. Introduction

Biofilm technologies such as the Moving-Bed Biofilm Reactor (MBBR) are increasingly

being implemented in wastewater treatment due to their advantages with respect to smaller

reactor sizes, ease of operation, less demanding solids separation requirements, and the increased

specialization of attached biomass (degaard, 2006). As the application of MBBRs becomes

more widespread, the knowledge of how wastewater characteristics affect biofilm structure and

reactor performance gains importance for reactor design and optimization. In particular there is

limited understanding of how wastewater composition influences biofilm detachment, a complex

process that is critical to the stability and function of biofilms in MBBRs.

Divalent cations such as calcium are a component of wastewater that has been shown to

influence biofilm structure and detachment (Turakhia and Characklis, 1989; Applegate and

Bryers, 1991; Huang and Pinder, 1995; Krstgens et al. 2001; Ahimou et al., 2007a). This is due

to the role divalent cations play in bridging negatively charged moieties of extracellular

polymeric substances (EPS) through electrostatic interactions (Flemming and Wingender,

2001a,b). In general, biofilms become thicker, denser and more mechanically stable when

exposed to increasing concentrations of divalent cations. Further to this, Sobeck and Higgins

(2002) propose that the EPS matrix acts as an ion exchange resin, where divalent bridges can be

formed or broken due to relative concentrations of monovalent and divalent cations competing

for negative sites.


77

These findings suggest that the ionic composition of wastewater is an important

consideration in the design and operation of MBBRs, and that there may be opportunities for

increasing the performance of existing reactors through modification of the divalent to

monovalent cation ratio. Such an approach has already been successfully applied at the industrial

scale for improving density and settling properties of activated sludge bio-flocs by switching

bases used in pH control to Ca(OH) 2 or Mg(OH) 2 (Higgins et al. 2004a,b). However in biofilm

systems like the MBBR, manipulation of the cohesive properties of biofilms may lead to other

significant performance enhancements. For instance, systems operated to perform simultaneous

nitrification/denitrification may be more effective if the biofilms are thicker or denser, due to the

creation of a larger anoxic zone within the biofilm. Conversely, in MBBRs where the primary

function is BOD removal, thin biofilms may be optimal in order to increase available surface

area (Lazarova and Manem, 1995) and reduce the variability of solids discharge due to

sloughing.

While the divalent to monovalent cation ratio appears to be a potentially cost effective

biofilm optimization tool, it is unclear whether the results of past researchers can be translated to

multi-species MBBR biofilms. The majority of the relevant biofilm work has been done in pure-

culture flow cell or annular reactor systems that have different hydrodynamic conditions than

those experienced by biofilms within the MBBR carrier material. Also by working with a pure

culture, the results may overly magnify species-specific responses such as alginate gel formation

by Pseudomonas aeruginosa in the presence of calcium (Sarkisova et al., 2005). Of the studies

conducted with mixed microbial communities, the majority of the work has investigated flocs

and granules (for instance Keiding and Neilsen, 1997; Sobeck and Higgins, 2002; Chang and

Lin, 2006) or anaerobic biofilm systems (Huang and Pinder, 1995). While it is interesting that
78

there seems to be some consistency in response between flocculating, granulated, and different

biofilm systems, there has been no direct assessment of the effect of divalent cation

concentration on aerobic MBBR biofilms. Furthermore, throughout this work the microbiology

has not been characterized to determine if some of the results are explained by a community

shift.

The goal of this study was to investigate how different concentrations of divalent cations

influence biofilm structure, microbiology and reactor performance of lab-scale MBBRs. The

results should provide some direct evidence for the significance of divalent cation concentration

in design and optimization of this emerging treatment technology.

Questions

(A) How is MBBR biofilm structure, microbiology and function affected by the concentration of calcium
in wastewater?

Hypotheses

(A) Biofilm structure (thickness, oxygen diffusion profiles, morphology, EPS composition) is affected
by changes in the concentration of calcium.

(B) Biofilm microbiology is affected by changes in the concentration of calcium.

(C) MBBR performance is affected by changes in the concentration of calcium.

Objectives

(A) Construct a laboratory-scale MBBR apparatus

(B) Run long term experiments at different calcium concentrations to establish steady state conditions
that reflect

(C) Characterize the structure, microbiology, and performance for different calcium concentrations

4.2. Methods

4.2.1. Laboratory MBBR System


79

Four replicate 1.85 L continuous flow reactors were filled 60% by volume with 1000

plastic MBBR carriers (K1, AnoxKaldnes) each and aerated at the base with a stone diffuser

connected to a pressure and flow controlled air stream. The reactors were constructed from glass

with outflow spouts that were small enough to retain the carriers and a recirculating water jacket

to maintain the temperature at 24oC. All reactors were inoculated with a fresh homogenized 500

mL sample of return activated sludge collected from the North Toronto Treatment Plant,

Toronto, Canada, and the remaining volume of the reactors filled with a synthetic media as

defined in Table 4.1. After allowing the inoculum to colonize the biofilm carrier surfaces for 6

hours, flow of the defined media was commenced at a rate of 12 ml/min to all reactors to give a

hydraulic retention time of 2.6 0.2 h. Synthetic feed was prepared fresh daily and contained in

storage tanks in a refrigerator while being pumped to the reactor inlets. The pH of the influent

was adjusted to 5.9 to yield a steady reactor pH of 7.4 0.2. Feed adjustment was used instead of

feedback pH controllers to avoid changes in the ionic composition of the reactors over time.

Table 4.1 - Summary of reactor feed properties and acclimation period

Parameter Experiment I Experiment II Experiment III

COD 500 mg/L


Glucose 290 mg/L
Acetate 174 mg/L
NH4-N 50 mgN/L
KH2PO4-P 4 mgP/L
Trace Minerals Recipe according to Liao et al. (2001)
CaCl2-Ca 1, 1, 300, and 300 mgCa2+/L 1, 1, 100, and 100 mgCa2+/L 1, 50, 100, and 200 mgCa2+/L
Divalent to
Monovalent Ratio 0.012, 0.012, 3.72, 3.72 0.012, 0.012, 1.24, 1.24 0.012, 0.619, 1.24, 2.48
(me-eq/me-eq)
Acclimation Period 55 days* 30 days 100 days
*Experiment 1 involved acclimating all four reactors at 1 mgCa2+/L for 35 days after which two were switched to 300 mgCa2+/L and allowed to
acclimate for an additional 19 days.
80

4.2.2. Experimental Conditions

Three experiments were performed to assess the influence of the divalent cation calcium.

Calcium was selected as the divalent cation of interest to provide better comparison to past

studies which have primarily focused on calcium. The calcium concentrations and acclimation

periods used in the three experiments (I, II, and III) are shown in Table 1. The first experiment

was a preliminary study to investigate how the MBBR biofilms change when calcium in the

reactor feed is increased from 1 to 300 mg/L. The second experiment was conducted to compare

steady state biofilms grown at two calcium concentrations (1, 100 mg/L) by quantifying structure

and performance differences within the inherent variability of these complex systems. The final

experiment was performed to identify important trends over a wider range of calcium

concentrations (1-200mg/L). Measurements reported here were primarily gathered in

experiments II and III and were taken when the biofilms had achieved a pseudo-steady state

condition following the acclimation period.

4.2.3. Feed and Effluent Analysis

Standard analyses of wastewater characteristics were performed on influent and effluent

streams. Chemical Oxygen Demand (COD), Total and Volatile Suspended Solids, NH 4 -N and

Turbidity were measured through strict adherence to protocols (5220D, 2540D, 4500-NH 3 F, and

2130 respectively) from APHA (1998). The ionic composition was confirmed by ICP-AES

(PerkinElmer). All tests were performed immediately following collection of a liquid sample

except for some COD and ammonia tests where samples were stored frozen at -20oC and then

thawed for analysis. Storage was found to have an insignificant affect on concentration values.

CODs reported were measured on the soluble fraction after filtration using the filter paper from
81

the protocol for suspended solids analysis. Turbidity measurements were performed on effluent

samples collected from the unsettled fraction after performing a 1 hour settling test according to

the APHA standard method 2710C.

4.2.4. Biofilm Areal Density and Organic Fraction

Biofilm areal density was determined by randomly sampling ten carriers from the reactor

of interest and gently rinsing them with two 150 mL volumes of deionized water to remove

suspended biomass. The carriers were dried at 102-105oC for 24 hours and weighed. The carriers

were then cleaned of all biomass by extraction in 150 mL of 0.1 M NaOH at 90-95oC for 30

minutes under magnetic stirring. The clean carriers were thoroughly rinsed with deionized water

and dried again at 102-105 oC. The difference in dry weight before and after extraction was used

to determine the areal density given the average carrier dimensions. The volatile and ash

fractions were determined by scraping biofilms from 3-5 carriers into a weighing dish using the

dull edge of a sterile scalpel to avoid abrading plastic from the carrier surface. Collected mass

was characterized for total and volatile solids as described above. Following analysis, clean

carriers were returned to the reactors after being marked by a small incision to avoid re-

sampling. In all experiments < 30% of the total carriers were sampled.

4.2.5. Dissolved Oxygen Microprofiles

In experiment III, MBBR carriers were collected and carefully cut to expose the interior

biofilm surfaces so that a dissolved oxygen microelectrode could be lowered into the biofilm as

depicted in Figure 1B. Profiles were collected with a glass Clark-type oxygen microsensor with
82

an 8-12 m tip diameter and fast (<2 s) response time (Unisense OX-10, Unisense AB)

connected to a picoammeter (PA2000, Unisense AB).

(A) (B) (C)


Biofilm O2 Microelectrode
Section
Removed

Central
Biofilm Isolated

Figure 4.1. (A) Schematic diagram of the cross-section of the


cylindrical the Kaldnes K1 carrier. (B) Same with a section cut out
to demonstrate how dissolved oxygen microprofiles were collected.
(C) Diagram of internal sections used for microscopy.

The microelectrode was positioned using a manual micromanipulator with depth resolution of 10

m and observed using a wide-angle objective. Prior to measurement, the samples were mounted

in a clamp and the microelectrode tip was positioned along the biofilm-free edges of the sample

to ensure the alignment of the carrier within the clamp and to determine the depth of the biofilm

substratum. The clamp apparatus was then submerged in an aerated bath of synthetic media of

composition similar to the reactor feed except that the COD was adjusted to 100 mg/L to be

closer to the mixed reactor concentration. The positions of the aerator and sampled carriers were

kept constant between tests to minimize variability in hydrodynamic conditions. The bulk

dissolved oxygen concentration was measured to be close to saturation (~8.4 mg/L) which

represented the steady state reactor concentration due to the heavy aeration used for mixing.

Three or more profiles were collected for each sample by moving the electrode to new

locations on the probed surface. Measurements were taken from a randomly selected biofilm

carrier from each reactor at 3 timepoints during the steadystate phase of experiment III.
83

4.2.6. EPS Extraction and Characterization

In experiment II, extracellular polymers were extracted from biomass collected by

dissecting and scraping 30 carriers from each reactor with a sterile scalpel. The collected

biomass was suspended in 50 mL of phosphate buffer at pH 7.5, homogenized by vortex mixing

and cooled to 4oC in an ice bath. Following sampling the homogenized suspension for dry solids,

extractions were performed using a cation exchange resin method (Frlund et al. 1996) modified

such that the impeller speed was at 900 rpm with a duration of mixing of 1 hour.

In experiment III, an alternate extraction protocol was followed to reduce the number of

carriers sacrificed for analysis. This time the biomass from 1 carrier (performed in triplicate) was

scraped into a 1.5 L eppendorf tube and heated for 1 hour in an 80oC water bath (Brown and

Lester, 1980; Zhang et al., 1999). The dry solids basis weight for comparing polymer amounts

was calculated from the average areal density for the sampling day and corrected for

inefficiencies in scraping by conducting a biofilm mass measurement on the scraped carrier

pieces. This correction was less than 10% of the average biofilm mass per carrier.

Extracellular extracts were characterized for proteins using the method of Lowry et al.

(1951) with adjustment for humic acids (Frlund et al., 1995) and for polysaccharides using the

method of Dubois et al., (1956). For the small volume extracts of experiment III, the same

characterization methods were used through adaptation to microplates according to the work of

Fryer et al., (1986) and Masuko et al., (2005) for the Lowry and Dubois methods respectively.

4.2.7. Confocal Microscopy

Samples were prepared for confocal laser scanning microscopy (CLSM) by cutting a flat

section of the carrier material with a sterile scalpel taking care to avoid manipulating the biofilm
84

coated surface as shown in Figure 1C. Carrier pieces were set in 2% agarose by filling a

planktonic slide with the hot liquid and then submerging the biofilm just prior to gel formation.

The fixed samples were then stained with both Sypro Orange (MolecularProbes) and a wheat

germ agglutinin (WGA) lectin stain. These were selected to bind representative moieties on

proteins (Sypro Orange) and polysaccharides (WGA) to show the hydrated biofilm structure.

Staining was done by adding 1 mL of a mixed dye solution to the surface of the agarose-bound

biofilm and allowing the samples to incubate for 30 minutes in the dark. Following incubation

the samples were rinsed with 5 aliquots of phosphate buffered saline. A Zeiss Axioplan LSM 510

scanning confocal laser microscope was used to capture images at various optical depths for each

sample. Prior to capturing an image, a consistent optimization procedure was performed to

determine the laser intensity and detector gain/offset. Confocal images were collected at various

biofilm depths, analyzed and rendered into 3D reconstructions using ImageJ image analysis

software (Abramoff et al. 2004) with the Volume Viewer plug-in (Barthel, 2005). Three replicate

samples were prepared for each reactor at one steadystate timepoint for experiment II and at

three steadystate timepoints for experiment III. Presented images reflect representative

observations

4.2.8. SEM

Samples were collected for scanning electron microscopy (SEM) using the same method

as for CLSM (Figure 1C). These samples were placed in a fixative 2% gluteraldehyde solution

(in a 0.1M sodium cacodylate buffer) for 20 minutes and then dehydrated by sequential exposure

to 50%, 70%, and 100% ethanol (v/v) solutions with each step equilibrated for 1 hour and the

final 100% step repeated three times. Dehydrated samples were then transferred to a fresh
85

volume of anhydrous ethanol for storage until processing (<1d). To prepare for SEM imaging,

samples were dried in a critical point dryer (Bal-Tec CPD 030) and sputter coated with gold

(Denton, Vacuum Desk II). Images were collected with an FEI XL30 (Phillips) microscope.

4.2.9. DNA extraction, amplification and characterization with DGGE

Biofilm samples were collected for DNA fingerprinting analysis on several days

following the acclimation period. Biomass samples were collected by sampling three carriers at

random from each reactor and cutting the carriers into small biofilm coated pieces with a sterile

scalpel. The cut carrier materials from all three sampled carriers were transferred to a 2 mL

sterile centrifuge tube and frozen by immersion in liquid nitrogen. Frozen samples were stored at

-20oC until DNA extraction.

DNA extraction was performed using a commercially available extraction kit (UltraClean

Soil DNA Isolation Kit, MO BIO Laboratories) and the extractions followed all specifications in

the kit method. Final extracts were dried in a vacuum drier (Savant DNA 120, Thermoelectron

Corp.) and resuspended in 50 L DNase-free water (UltraPURE, GIBCO).

16S rDNA of the extracted samples was amplified by polymerase chain reaction using

bacterial primers as defined by Muyzer et al.(1993) under a similar temperature profile in an

PTC-200 peltier thermal cycler PCR block heater (MJ Research). PCR products were confirmed

by running an agarose gel against a known ladder (NEB). A second bacterial ladder with a range

of 16S guanine-cytosine content was also amplified to provide a standard for comparing band

distances on the DGGE gels.

DGGE gels were prepared in a Biorad Model 475 gradient delivery system by mixing a

gradient from 30-60% urea and formamide in an 8% acrylamide/bis-acrylamide gel. Gels were
86

suspended in TAE buffer at 60 oC and run for 5 minutes at 25V and 5hours, 25 minutes at 160V

in a Biorad DCode electrophoresis system. Following electrophoresis, the gels were stained in

ethidium bromide, illuminated and imaged in a UV transilumination box (G:Box, Syngene).

Band data were compared for similarity using the UPGMA algorithm to compare Jaccard

distances of bands through image analysis using GelComparII software (Applied Mathematics).

4.2.10. Enumeration of Protozoa and Metazoa

Protozoa and metazoa populations were monitored in the biofilms and the suspended

phase. To enumerate the biofilm, a carrier sample was collected with tweezers and gently rinsed

with phosphate buffered saline (pH 7.5). The biofilm was then carefully scraped into a 2 mL

centrifuge tube with a scalpel and suspended in 1.5 mL of PBS. All biomass collection was done

to avoid introduction of external microorganism through the use of sterile equipment and buffer.

The suspension was homogenized on a vortex mixer for 2 minutes and the resulting mixture was

pipetted on to an Improved Neubauer counting chamber (Hausser Scientific). Organisms were

identified using a light microscope in phase contrast following a protocol similar to that of Fried

et al., (2000), except that organisms were classified into general categories as free swimming

ciliates, stalked (attached) ciliates, rotifers, rotifer cysts, or nematodes. Flagellates and amoeba

were not enumerated. Eight replicate counts were performed on each sample to generate an

average count and the volume of the counting chamber was used to calculate numbers of

microorganisms per carrier (approximately 1 cm3). Suspended samples were analyzed in a

similar fashion except that a volume of the suspended solids were directly applied to the

improved Neubauer counting chamber for enumeration.

4.3. Results
87

The MBBRs were run to conduct the three experiments as defined in Table 4.1. The

majority of the results were collected for experiments II and III which provided an assessment of

the reproducibility (II) and the response to a range (III) of calcium concentrations.

4.3.1. Biofilm Density, Thickness and Oxygen Profiles

Biofilms cultivated in MBBRs with higher calcium concentrations led to increased

biofilm areal density (mgVS/cm2) as depicted in Figure 4.2. In all experiments there were

significant ( = 0.05) increases in biofilm accumulation in comparing 1 mg/L with any of the

higher calcium concentrations. However between the areal densities for 50, 100, 200 and 300

mgCa2+/L, there were no significant continued increases with concentration. This is best shown

by the results of experiment 3 where the 50, 100 and 200 mgCa2+/L trials were not found to be

statistically different.

1.4 300
200 mg/L
100 mg/L
1.2
Biofilm Areal Density (mg/cm )

50
2

mg/L
mg/L
1

0.8
1 mg/L
0.6

0.4

0.2

0
I I II II III III II II III III I I
Experiment
Figure 4.2. Combined biofilm areal density measurements for three experiments listed in order of
increasing concentrations of calcium as noted above. Shaded region represents the fraction of
organic matter. Error bars represent standard deviations for >5 measurements at timepoints after
acclimation.
88

Figure 4.2 also shows the organic and inorganic fractions of the biofilm, which for all conditions

except the 300 mgCa2+/L trials were greater than 90% volatile. For biofilms grown at 300

mgCa2+/L, the biofilm mass was composed of a large fraction of inorganic solids, indicating the

precipitation of calcium carbonates or phosphates. The presence of precipitates was supported by

bright field light microscopy of the biofilm which showed significant inorganic crystal formation

(images not shown).

The oxygen microprofiles of the 1 mgCa2+/L and 200 mgCa2+/L reactors of experiment

III (Figure 4.3A) suggest that the increase in calcium led to a thicker biofilm with a larger anoxic

region (anoxic = oxygen not detected by sensor). These trends are found to continue over time,

as shown in Figure 4.3B. It was also found that biofilms are heterogeneous with respect to

thickness at different locations on the biofilm surface as indicated by the standard deviations of

profile measurements.

(A) (B)
9

Biofilm density (kg/m 3)


1600 25
8
Dissolved oxygen (mg/L)

Thickness of biofilm and

1400 20
7 Biofilm -w ater
anoxic region ( m)

interface 1200 15
6
1000 10
5
800 5
4
600 0
3
2 400

1 200
Biofilm -w ater interface
0 0
6 13 21
0 500 1000 1500
Distance from substratum ( m) Days after acclimation

Figure 4.3. (A) Average dissolved oxygen profiles for 1 mgCa2+/L (white) and 200 mgCa2+/L (grey)
with average biofilm thickness noted. Error bars on all measurements represent standard deviations
for 3 microprofiles taken in different positions on each sampled biofilm. (B) Comparison of average
biofilm thickness (columns) anoxic zone (hatched portion, no detection of oxygen), and densities
(points, on inset axis) for reactors fed 1 mgCa2+/L (white) and 200 mgCa2+/L (grey) over three
measurement times when reactors were in pseudosteady state in experiment III. Error bars
represent standard deviations for at least three profiles in different locations for a given sample.
89

Thickness also was found to vary over time to a much greater extent than observed in the areal

density measurements.

4.3.2. Extracellular Polymers

The extracted extracellular polymer compositions depicted a clear trend: as calcium

concentration was increased, the EPS became more proteinaceous (Figure 4.4). This trend was

statistically significant (R2 of 0.83, P < 0.005 for slope = 0) even when grouping data from two

different experiments (II and III) using two different extraction techniques.

7
Ratio of Proteins to Polysaccharides

4
R2 = 0.8335

3 Slope = 0 (P<0.005)

0
0 50 100 150 200

Influent Calcium Concentration (mg/l)


Figure 4.4. Increasing calcium concentration leads to enrichment of
protein in EPS. Measurements from experiments II ( ) and III ( )
are plotted togther, yielding a significant correlation.

4.3.3. Microscopy

Biofilms grown at 1 mgCa2+/L and 100 mgCa2+/L in experiment II were imaged with

confocal and scanning electron microscopy. Images depicting typical biofilm morphology are
90

shown in Figures 4.5A and B using CLSM. The images are a 3-D reconstruction of the upper

regions of the biofilm and it can be seen that in the 100 mgCa2+/L reactors the biofilm structure

included intertwined filamentous organisms in a less dense configuration compared to the 1

mgCa2+/L case. While biofilms are heterogeneous, the images presented represent the general

structure observed in all replicate samples.

(A) 1 mg/L Ca2+ (B) 100 mg/L Ca 2+

14 72 m
68 m

6 m
m
146

(C) 1 mg/L Ca2+ (D) 100 mg/L Ca 2+

Figure 4.5. CLSM images of biofilms stained with with Sypro Orange (Proteins) and
WGA lectin (Polysaccharides) for 1 mgCa2+/L (A) and 100 mgCa2+/L (B). (C) and (D)
depict SEM images of the biofilm surface for biofilms grown under similar conditions (1
and 100 mgCa2+/L, respectively).
91

It should be noted that these reconstructions do not represent the complete biofilm since laser

penetration is limited to 200-300 m. The scale dimensions depicted also reflect a further

reduction in penetration depth due to the mounting procedure which fixes the biofilms in a

hydrated condition under a thin layer of agarose which further reduced the viewable depth.

SEM images shown in Figure 4.5C and D further confirm the structures observed with

CLSM. Although the sample preparation technique for SEM leads to structural alteration due to

dehydration of the biofilms, it is evident that the 100 mgCa2+/L biofilm surfaces are composed of

an abundance of intertwined filamentous bacteria.

4.3.4. Bacterial Community Analysis

To evaluate if calcium concentration had an effect on the bacterial community

composition, DGGE fingerprints of the 16S rDNA for bacteria amplified through PCR were

compared for similarity in a dendrogram. The DGGE fingerprints for duplicate reactors (noted A

or B for 1 mgCa2+/L and C or D for 100 mgCa2+/L) at three time points (15, 25 and 31 days after

% Similarity of Fingerprint
30 40 50 60 70 80 90 100 Ca2+ Rep Time
100
30

40

50

60

70

80

90

R43
1 1Amg/L25
R44
1 1Amg/L31
R42
1 1Amg/L15
R13
1 1Bmg/L25
R14
1 1Bmg/L31
R22 100
100 C mg/L
15
R12
1 1Bmg/L15
R33 100
100 D mg/L
25
R34 100
100 D mg/L
31
R32 100
100 D mg/L
15
R23 100
100 C mg/L
25
R24 100
100 C mg/L
31
Figure 4.6. DGGE fingerprints at 15, 25 and 31 days after acclimation for duplicate reactors at 1 and
100 mgCa2+/L in experiment II. Bands are organized in a dendogram according to similarity of
banding pattern using the UPGMA similarity algorithm comparing the Jaccard distance of fingerprints.
Except for one replicate, samples from different replicate reactors and times cluster according to
calcium condition.
92

acclimation) and the resulting dendrogram are shown in Figure 4.6. With one exception, samples

at all times and in all replicate reactors clustered by similarity according to the concentration of

calcium. This suggests that while all reactors were inoculated in exactly the same manner,

differences in calcium concentration caused the selection of a different community of

microorganisms either directly or indirectly.

4.3.5. Enumeration of Protozoa and Metazoa

As in real effluent treatment systems, the MBBRs contain a complex microbial

community that includes a diversity of protoza and metazoa. It is of interest to determine whether

calcium influences the number or community structure of the biocenosis.

Table 4.2 - Protazoa and metazoa counts for biofilm and suspended biomass

[Calcium] 1 mg/L 50mg/L 100mg/L 200mg/L


1 2 1 2 1 2
Biofilm Suspended Biofilm Suspended Biofilm Suspended Biofilm1 Suspended2
Ciliates
Stalked 1.1 (3.1) 2.6 (1.7) 126 (84) 19.3 (13.8) 55.5 (31.5) 3.2 (1.8) 0.4 (1.0) 2.0 (2.1)
Free Swimming N.D. 1.9 (1.5) N.D. 5.5 (3.7) N.D 0.8 (1.1) N.D. 0.5 (0.9)
Rotifers
Active Rotifers 5.3 (3.1) 0.8 (1.5) 13.5 (7.1) 1.5 (2.1) 27.7 (6.2) 1.6 (1.7) 46.1 (21.3) 13.0 (10.3)
Rotifer Cysts 12.3 (6.5) 2.6 (2.3) 113 (60) 3.3 (5.3) 63.7 (38.4) 0.4 (0.9) 40.9 (18.8) 4.5 (4.5)
Nematodes 9.7 (10.2) N.D. 9.0 (2.2) 1.0 (1.5) 13.5 (8.0) 0.8 (1.1) 9.3 (8.1) 9.8 (8.2)
Total Organisms 28.5 (18.9) 5.5 (5.5) 262 (145) 30.5 (20.8) 161 (74.4) 6.8 (2.3) 96.7 (39.2) 29.8 (18.7)

1 Biofilm counts are presented as counts (x1000) per MBBR carrier (standard deviations for 8 measurements)
2 Suspended counts are presented as counts (x1000) per mL of reactor mixed liquor (standard deviations for 8 measurements)
N.D. None detected

Numbers of higher organisms were counted for biofilm and suspended phase samples in

experiment III. The results of this counting are depicted in Table 4.2. It is clear from these data

that there are substantially more protozoa and metazoa for reactors at 50, 100 and 200 mgCa2+/L

than at 1 mgCa2+/L, particularly in the biofilm samples. The high variability of counting data

makes it hard to determine differences between the three reactors with elevated calcium

concentrations, however there does appear to be a positive correlation between calcium


93

concentration and active rotifer population which are highest in the 200 mgCa2+/L trial. It should

be noted that the data are presented from a functional engineering perspective reflecting the

number of organisms per carrier to allow for comparison to other MBBR systems. If higher

organism counts are to be considered from a community structure perspective, it may be more

appropriate to consider counts/mg biomass. Appendix 2 contains the data of Table 4.2 in this

form. Though the differences between the low and high calcium conditions are reduced due to

the changes in biomass associated with these conditions, the trend toward higher protozoa and

metazoa counts is still present.

4.3.6 Treatment Performance

The MBBRs were monitored for their


100
degradation of organics (as COD),
COD Removal Efficiency (%)

95
ammonia and for the production of
90
volatile suspended solids (VSS) to
85
determine if calcium concentration
80
influenced performance.
75
400

350
Removal efficiencies for soluble COD
Reactor VSS (mg/L)

300

and volatile suspended solids for 250

200
reactors following acclimation are 150

100 50
1 mg/L 100 mg/L 200 mg/L
shown in Figure 4.7. It was found that mg/L
50

0
the three trials conducted at 1 Figure 4.7. Pseudosteady-state COD removal efficiencies and
reactor VSS for experiments II ( ) and III ( ) at different
mgCa2+/L had lower average removal calcium concentraitons. Error bars represent standard
deviations.
94

efficiencies than all other trials at higher calcium concentrations. For the elevated calcium

concentrations there were no statistically significant differences in removal efficiency and when

comparing the average reactor VSS data a similar trend is observed. In fact, it was observed that

the soluble COD removal efficiency and the average VSS for each experimental trial were

correlated (R2 = 0.54, P<0.02 for slope = 0).

The turbidity data shown in Figure 4.8 demonstrate the influence of calcium on the

properties of the suspended biomass in MBBRs. The 1 mgCa2+/L condition led to visibly turbid

effluent after settling for 1 hour, while the 50, 100 and 200 mgCa2+/L trials were much clearer.

Microscopic examination of the turbid effluents indicated that the cause of the turbidity was a

high concentration of non-flocculated bacteria that would not settle by gravity. This finding was

consistent across the entire measurement period.

Analysis of ammonia in the reactor effluents was also conducted in each experiment. It

was found that removal efficiency for all


80
70 calcium conditions was similar at around 50-
60
Turbidity (NTU)

50 60%, which could be attributed primarily to the


40
30 amount of ammonia nitrogen required for
20
heterotrophic growth. The high carbon to
10
0
nitrogen ratio likely led to an environment
1 mg/L 50 mg/L 100 mg/L 200 mg/L
Calcium Concentration where nitrifiers were out-competed for oxygen
Figure 4.8. Turbidity of reactor effluents at increasing
calcium concentration for experiment III by heterotrophs.
95

4.4. Discussion

The goal of the reported experiments was to investigate how the concentration of calcium

affects the biofilms and overall reactor performance in MBBRs to determine if this variable is an

important consideration in the design and operation of these emerging reactors. To help focus

the discussion, the analysis of the results is divided into three sections on biofilm structure,

biofilm microbiology, and MBBR treatment performance.

4.4.1. Biofilm Structure

Several aspects of biofilm structure were characterized to gain insights into the role of

calcium including the biofilm thickness and density, the dissolved oxygen profile, the ratio of

organic to inorganic matter and the composition of the extracellular polymeric substances. In all

structural measures there were significant changes observed at different calcium concentrations

indicating the general importance of this wastewater component.

Biofilm areal densities and the organic/inorganic ratios (Figure 4.2) can be used with the

oxygen microprofile data (Figure 4.3) to draw conclusions about how calcium influences cellular

cohesion in biofilms. It can be seen from these figures that there is a significant increase in

biofilm thickness and density between the 1 mgCa2+/L experiment and the trials at higher

calcium concentrations. The greater biofilm thicknesses achieved at elevated calcium levels also

provided for a larger anoxic region within the biofilms which suggests that calcium may be used

as a variable for improving the function of MBBRs for removal of nitrogen through

nitrification/denitrification where a protected anoxic region is required for conversion of nitrate.

These data extend the results collected so far on the role of calcium in enhancing biofilm

accumulation. For instance, Turakhia and Characklis (1989) found that bulk calcium
96

concentrations from 0.4-50 mg/L led to areal density changes from 0.09 to 0.25 mg/cm2 and

while the biofilm age (140 hours), microbiology (pure culture Pseudomonas aeruginosa), and

hydrodynamics (RotoTorque) were different, a similar trend between those calcium

concentrations was found in this study. It was also found that beyond 50 mgCa2+/L there was no

further biofilm mass increase with increasing calcium concentration, a result also reported by

Huang and Pinder (1995) for an anaerobic system. Huang and Pinder (1995) used biofilm Total

Organic Carbon (TOC) as a measure of biofilm areal density and with this measure found a

maximum organic density at 120 mgCa2+/L, beyond which the density decreased. The authors

speculated that this could be caused by metabolic inhibition at elevated calcium levels, which

may be a more significant concern in anaerobic systems. However the data they present also

suggest that inorganic salt accumulation occurred at higher calcium levels, reducing the organic

fraction of the biofilm and thus decreasing biofilm areal density on a TOC basis. In examining

the inorganic fractions in Figure 4.2, it can be observed that the 300 mgCa2+/L biofilms became

primarily composed of inorganic material. By considering only the organic fraction, a trend

comparable to that observed by Huang and Pinder (1995) emerges: increasing biofilm mass with

calcium concentration to 50 mg/L where some optimum concentration exists in the range of 50-

200 mg/l followed by a decrease in accumulation at a concentration above 200 mg/L due to salt

precipitation.

There is considerable evidence for the mechanistic explanation of enhanced aggregation

at higher calcium levels due to greater electrostatic interactions between EPS molecules.

Increases in biofilm strength as a result of calcium concentration have been quantified by a

variety of measures including viscometry (Chen and Stewart 2000, 2002), uniaxial compression

(Krstgens et al. 2001), and most recently through direct determination of the cohesive force
97

using atomic force microscopy (Ahimou et al., 2007a). It has also been shown that the

application of the calcium-specific chelating agent EGTA caused biofilm detachment (Turakhia

et al., 1983). It is plausible that more strongly adhered cells and EPS at the biofilm surface

reduce the specific rate of detachment, leading to more massive biofilms.

It is interesting to note that beyond 50 mgCa2+/L there was no significant difference in

biofilm accumulation. This suggests that while calcium may be a necessary building block of the

EPS matrix, other factors begin to mediate biofilm accumulation once calcium has been provided

above some threshold concentration. For instance, several genetic pathways governing biofilm

dispersal have been shown to be controlled by a quorum sensing response which may prevent

biofilms from exceeding a critical concentration (Hall-Stoodley and Stoodley, 2002; Waters and

Bassler, 2005). This result may also be specific to the MBBR carrier material which has an

internal space for biofilm colonization that becomes constricted as biofilms increase in thickness.

This constriction can alter the hydrodynamics of the carriers as they circulate in the MBBR

causing reduced mass transfer and lowering the growth rate.

The observed inorganic salt formation at 300 mgCa2+/L highlights the importance of

considering precipitation processes when investigating the role of divalent cations. Formation of

calcium phosphates such as hydroxyapatite or calcium carbonates can occur under favorable

conditions that are difficult to predict theoretically due to the concentration gradients that

develop in biofilms. Precipitation can be problematic in wastewater treatment because inorganic

scale formation can foul equipment and increase maintenance costs. If the precipitated salts

contain a significant amount of phosphate a second problem arises in that the concentration of

phosphorous available for cellular metabolism is reduced. This can cause operational problems

in operating treatment systems and confound laboratory results examining the specific role of
98

calcium as a structural component. To evaluate this concern for calcium concentrations below

300 mg/L a solubility calculation using an equilibrium speciation model (MINTEQA2 v.2.40,

KTH, Sweden, data not shown) was performed to determine the likelihood of hydroxyapatite

precipitation for the composition and pH of the feed. This work indicated that at calcium

concentrations above 2 mg/L, significant amounts of phosphorous (>10% by mass) would be

converted to hydroxyapatite. However these calculations only provide the equilibrium state and

do not indicate if the rate of hydroxyapatite formation would be fast enough to occur in

competition with the rapid assimilation of phosphorous by cells. Fortunately this issue has

received considerable attention from researchers investigating biological phosphorous removal

processes. The work of Carlsson et al., (1997) and Maurer et al., (1999) demonstrate that for the

phosphorous and calcium concentrations used in this study, the rate of hydroxyapatite formation

would be sufficiently slow to cause minimal precipitation. With respect to calcium carbonate, the

equilibrium speciation prediction suggested that with the feed composition at reactor pH and

temperature calcium carbonate is not formed even at 300 mgCa2+/L. However it is difficult to

evaluate the carbonate balance for biofilms producing carbon dioxide as a metabolite and it is

possible internal biofilm conditions were more basic and therefore favorable for precipitation.

It should be noted that the measure of biofilm mass reflects the average mass of a sample

of 10 MBBR carriers and that the data presented are average values for several samples over

time in the pseudo-steady state phase of growth. Thus while biofilm heterogeneity and discreet

sloughing events may cause dynamic changes in biofilm thickness on different time scales for

different carriers, the reported areal densities describe average reactor conditions. The use of the

term pseudo-steady state speaks to this average state which masks aspects of the reactors that

are undergoing constant dynamic change as described in previous work (Horn et al., 2003;
99

Lewandowski et al., 2004). This informs the interpretation of Figure 4.3 where the average

oxygen microprofiles vary considerably within one calcium condition and that over time there

are changes in biofilm thickness that are greater in magnitude than the differences between

reactors. Thus an average measure such as the areal density can be a more reliable tool for

comparing reactors at different calcium concentrations.

Biofilm structure was further characterized by measuring the composition of the biofilm

EPS. It was observed that as calcium concentration increased the biofilms became enriched in

protein (Figure 4.4). This result was obtained through comparison of the data from two

experiments using two extraction procedures (CER and heating) which suggests that the trend

with calcium is more significant than differences in extraction procedure. The ratio of proteins to

polysaccharides is an important structural feature because the recent measurements by Ahimou et

al., (2007b) implicate polysaccharides as having stronger cohesive properties, though these

results are correlative in nature. Enrichment in proteins may also lead to functional changes in

the EPS matrix due to the increased retention of extracellular enzymes. This finding suggests that

concentration of calcium could be used as a control variable for manipulating EPS composition

for potential applications such as extracting useful chemicals from waste sludges.

There are several possible explanations for the observed change in EPS composition. The

first is that calcium binds preferentially to negative moieties on nucleic acids as compared to

negative sugar residues on polysaccharides. The complexity of protein-protein interactions and

the prevalence of glycoproteins makes such a conclusion speculative, though it is interesting that

work by Park and Novak (2007) using a variety of extraction procedures to target EPS associated

with different cations concluded that calcium and magnesium were bound more to the

polysaccharide fraction than other cations investigated. However the measured protein to
100

polysaccharide ratios for calcium associated EPS in their study were still as high as 2.8, which

fits within the range of results in Figure 4.4. While other divalent cations may bind even more

preferentially to proteins, the enhancement of protein due to selective calcium bridging is a

plausible hypothesis. An alternate explanation is that the calcium concentration led to the

selection of different microbiology (see Figure 4.6) which in turn produced a different

composition of EPS that was more proteinaceous. It has been shown that the composition of EPS

produced by bacteria is highly species dependent (Sutherland, 2001) and since the experiments

were conducted as a mixed community for relevance to wastewater treatment, this explanation of

the observed changes cannot be ruled out.

4.4.2. Biofilm Microbiology

In addition to the structural role calcium plays in biofilms, there is evidence that calcium

plays a physiological or regulatory role in bacteria. Aside from the known requirement for

calcium as a cofactor for certain proteins it has been found that calcium acts in cell signaling

(Dominquez, 2004), biofilm virulence (Kierek and Watnik, 2003), alginate regulation

(Sarkisova, 2005), cellular and extracellular product formation (Patrauchan et al., 2005) and the

dechaining process of a filamentous organism (Wright and Klaenhammer, 1983). For these

reasons the current study evaluated microbiology of the biocenosis through microscopy, DGGE,

and counting of higher organisms.

Microscopy of MBBR biofilms depicted differences in the type and arrangement of

microorganisms in the upper regions and surfaces of the biofilm as a result of calcium

concentration. From the confocal and SEM images shown in Figure 4.5 the loose filamentous

network at 100 mgCa2+/L calcium can be compared to the more tightly packed structure at 1
101

mgCa2+/L. The two techniques were used to provide both the hydrated 3-D structure and a high-

resolution view of the surface microbiology on samples collected at the same time.

The observation of decreased biofilm density at higher calcium concentrations is contrary

to the measured values in Figure 4.2. This can be explained by the fact that the images collected

for CLSM analysis were taken at 1000x magnification and that at this magnification, the

fluorescent stains added to the biofilm were concentrated in the glycocalyx and cell-associated

proteins giving the appearance of cells floating in space. However the void spaces observed in

Figure 4.5 A and B were likely filled with an EPS gel that was at a low enough concentration as

to be considered void after optimization of the laser parameters of the confocal instrument. The

images also depict only the upper 70 m of the biofilm and these upper regions are typically

found to have a lower density than the internal regions of biofilms (Zhang and Bishop, 1994).

Thus the microscopic images give a good view into the morphological changes that occurred due

to increased calcium but they do not reflect the average film density.

Filamentous organisms may have a competitive advantage on the surface of biofilms in

reactors like the MBBR when operated at dilution rates near washout because they can extend

from the surface of the biofilm into the nutrient rich bulk without being detached from the

biofilm. Perhaps by increasing the concentration of calcium, this selective advantage is increased

due to greater aggregation of EPS on the biofilm surface, providing additional diffusion

resistance to the non-filamentous cells. However an alternate explanation for the observation of

increased filaments is simply that they require a higher calcium concentration for growth. For

instance Kmpfer et al., (1995) found that below 20 mgCa2+/L only a few filamentous strains of

68 tested grew significantly. It is possible that the filamentous organisms contained in the

inoculum were unable to grow at the low calcium condition due to this growth requirement.
102

Although the filamentous organisms were the most apparent differences in biofilm

microbiology, analysis of DGGE fingerprints (Figure 4.6) suggested that the bacterial

community as a whole had changed due to calcium. The statistical analysis of banding pattern

similarity indicated that all samples had considerable diversity but that between replicates over

time and in different reactors, the DNA fingerprints clustered consistently according to calcium

condition.

A final measure of the MBBR microbial community was the enumeration of protozoa

and metazoa as shown in Table 4.2. The populations of ciliates, nematodes, and rotifers can

change rapidly over time in response to changes in reactor conditions (Kinner and Curds, 1987;

Freid et al., 2000). Since the higher organism population contains species that are fixed in the

dynamically changing biofilm (stalked ciliates, rotifer cysts) and that they exist in a predator-

prey relationship with bacteria, the population numbers are variable with respect to sample

location and over time. Thus when looking at the standard deviations presented in Table 4.2,

limited conclusions can be drawn about average population behaviour. However, it is clear that

the total numbers of higher organisms at 1 mgCa2+/L are much lower than for any of the other

conditions. This is true even if the data are considered with respect to the biofilm areal density

differences. It can also be observed that the average active population of rotifers appears to

increase with increasing calcium concentration. While this observation was consistent over a 1

week sampling period, the large number of rotifer cysts embedded in the 50 mgCa2+/L biofilms

had the potential to hatch and drastically change this active population. These dynamic changes

represent challenges to understanding reactor behaviour in real wastewater reactors. Although

the basis for modeling the response of biofilms to predation exists (Canale, 1973) such

considerations have not as yet been incorporated in biofilm models.


103

Through looking at the results depicting biofilm microbiology, it can be concluded that

calcium concentration causes changes to important microbial populations. Such an analysis has

never been performed previously on biofilm systems exposed to different calcium concetrations

and the result further highlights the potential differences between results obtained in pure culture

and those from mixed-community systems.

4.4.3. Treatment Performance

Soluble COD removal efficiencies during the pseudo-steady state phase of experiments II

and III indicated that the removal efficiency increased for calcium concentrations 50 mg/L and

above (Figure 4.7). This increased performance could be a meaningful process improvement for

systems that have naturally low concentrations of calcium in wastewater. The cause of the

improved performance could be due to the increased biofilm areal density. While diffusion

resistance may mean that much of the biofilms are not actively exposed to organic substrate, if

the carriers were oxygen limited there could be a performance advantage in having a stable

anoxic region where facultative organisms would provide additional degradation of soluble

COD. The calcium-induced changes in microbiology may have also led to the development of a

more active consortium of microorganisms.

Coupled to soluble COD removal is the production of suspended solids, which as noted

above forms a significant correlation. This is not surprising since the detachment of biomass has

been hypothesized to be a function of the biofilm growth rate (Peyton and Characklis, 1993). In

hybrid reactors such as the MBBR, the increased suspended phase also contributes to additional

substrate removal causing additive effects.


104

The generation of higher concentrations of suspended particles in the calcium-enriched

reactors could impact the quality of MBBR effluent. However, in most practical situations the

effluent solids are attenuated by a settling or floatation device. The turbidity data presented in

Figure 4.8 demonstrate that it is actually the low calcium condition that is likely to lead to

suspended solids challenges. The large population of planktonic cells in the 1 mg/L case would

not settle after 1 hour in a standard settling test causing dramatically higher turbidity. This

observation indicates that detachment of cells from the 1 mgCa2+/L is occurring mostly in the

erosion mode as defined by Stewart (1993), and that after detachment there is no significant

flocculation in the suspended phase. It is most likely that this observation could be explained by

reduced cohesive force between cells and the EPS matrix when exposed to low calcium

concentrations. However the impact of the lower numbers of ciliates, nematodes and rotifers

must also be considered since these organisms have been found to be important for clarification

of effluents in biofilm systems (Lee and Welander, 1996).

The amount of ammonia added in the reactor feed was reduced by an extent that did not

show significant nitrification in all calcium conditions. This was likely due to the carbon to

nitrogen ratio (and loading rate) which was high enough make competition for dissolved oxygen

by nitrifiers unlikely (degaard, 2006). The structural results achieved for this primarily

heterotrophic biofilm suggest that for systems with lower carbon to nitrogen ratios, the

manipulation of calcium may prove to be a very important optimization variable for

nitrification/denitrification.

4.5. Conclusions

Through analysis of the results presented the following conclusions can be drawn:
105

Calcium concentration is an important variable to consider in the design and optimization

of MBBRs

Above a threshold calcium concentration (in this system 50 mg/L) the biofilms became

thicker and denser with larger anoxic zones. This indicates that divalent cation

manipulation could be a possible strategy for improving functionality of reactors

requiring anoxic activity.

Above a higher threshold value (in this system 300 mgCa2+/L) salt precipitation was

observed to occur suggesting that reactors operated at such a high concentration may

have scale formation issues, including the possibility of phosphorous reduction due to

hydroxyapatite precipitation.

EPS becomes enriched in proteins at higher calcium concentrations suggesting that

calcium may selectively form electrostatic bridges between nucleic acids

Bacterial populations became differentiated according to calcium concentration, and the

proliferation of filamentous organisms was observed in the upper region of biofilms

grown at higher calcium concentrations

Calcium concentrations at 50 mg/L and above led to higher numbers of protozoa and

metazoa, which may influence the performance of MBBRs, particularly with respect to

reduction of poorly-settling planktonic bacteria

MBBR treatment performance with respect to COD removal was higher at elevated

calcium concentrations which could be attributed to the observed increases in biofilm

accumulation or changes in microbiology

Settling properties of suspended biomass were dramatically reduced at the lowest calcium

concentration due to the abundance of non-flocculated bacteria


106

Chapter 5: A Theoretical MBBR Model of Biosolids Dynamics


during Periods of Starvation

5.1 Introduction

Moving bed biofilm reactors have been widely applied to treat industrial effluents such as

in the pulp and paper, chemical, and food manufacturing sectors (degaard, 2006). Industrial

effluents can be highly variable with respect to flow and composition due to manufacturing

process changes. Batch operation, scheduled plant maintenance, process changes to meet product

specifications, and unexpected shutdowns caused by equipment failure are just some of the

situations commonly experienced in industry that can cause episodes of transient wastewater

inputs to a treatment plant. Although equalization tanks are an option for controlling this

dynamic behaviour, they add treatment volume and thus limit the principal advantage of compact

treatment systems such as the MBBR.

There is evidence to suggest that transient operation can influence the biofilm detachment

rate and reactor biosolids dynamics in MBBRs. Operational changes have been shown to affect

other biological treatment systems, particularly with respect to the cohesion and stability of

cellular aggregates. For instance, deflocculation of activated sludge systems has been triggered

by transient effluent temperatures (Morgan-Sagastume and Allen, 2003, 2005), periods of low

dissolved oxygen (Zhang and Allen, 2007), shock organic loads (Galil et al., 1998), and through

exposure to pulse doses of electrophilic toxic chemicals (Bott and Love, 2002). Microbial

biofilms have been found to detach and disperse when subjected periods of starvation (Sawyer

and Hermanowicz, 1998; Hunt et al., 2004; Gjermansen et al., 2005). In contrast, granule-based

systems have been purposefully exposed to a transient feast/famine feeding regime during start-

up to promote stable granulation with better settling properties (Li et al., 2006; Liu and Tay,
107

2007). These examples suggest that transient operation could also impact MBBR biosolids

dynamics; however they reflect a range of systems and transient stresses that likely involve

situation-specific mechanisms.

The variable nature of an MBBR treating industrial effluent was the initial motivation for

this thesis and chapter 3 describes a multivariate statistical approach to modeling treatment

variability of a reactor treating pulp mill effluent. This work helped to identify important process

variables that influence treatment performance such as changes in the wood type pulped, changes

in reactor flow rate, and the occurrence of complete flow stoppages due to plant maintenance

activities. This provided considerable insight into one MBBR application and some indication of

variables that may be important in other industrial situations. However, the empirical and

correlative nature of model parameters limits the multivariate approach in its ability to broadly

describe biosolids dynamics from a theoretical or mechanistic perspective.

In order to perform a more quantitative investigation of MBBR biosolids dynamics, a

dynamic theoretical model of the MBBR was derived. The derivation and analysis of lab and

industrial scale data are the focus of this chapter.

As discussed in the literature review in section 2.3.2., while biofilm modeling has been an

actively developing research area there have only been a few studies that implement theoretical

biofilm models to simulate MBBRs (i.e. Havla et al., 2002; Fouad and Bhargava, 2005; Plattes et

al. 2006). None of these have specifically characterized factors that influence the biofilm

detachment rate or described biosolids dynamics resulting from transient operation.

This chapter seeks to address this gap through the development of a dynamic theoretical

model of the MBBR. The model is then used to investigate the questions and hypotheses posed

for this line of inquiry:


108

Questions

(A) How can theoretical biofilm models be adapted to describe the MBBR?

(B) Can the detachment process be defined to simulate steady-state behaviour for different calcium
concentrations?

(C) Can the MBBR model be defined to describe the response to transient periods of starvation?

Hypotheses

(A) A theoretical diffusion-reaction biofilm model can be adapted to capture the key
features of MBBR behaviour

(B) Different divalent cation concentrations will affect the detachment rate

(C) The response to starvation can be simulated by the MBBR model

Objectives

(A) Derive a dynamic theoretical model of the MBBR

(B) Compare different detachment rate terms to steady state data from calcium experiments

(C) Compare model outputs to transient starvation experiments in the lab and industrial data

Periods of starvation caused by reactor flow stoppage were selected as the transient

process change for focus in this work as reflected in Hypothesis (C). Multivariate analysis of

pulp mill data presented in Chapter 3 indicated that treatment plant shutdowns were one of the

principal contributors to process variability. Historical data indicated that flow stoppage due

process upsets and maintenance occurred frequently and it is likely that similar shutdowns are

common to a range of industrial treatment systems. Microbial starvation caused by nutrient

scarcity during process shutdowns is hypothesized to be the cause of the MBBR performance

fluctuations. However, it is unclear how the biosolids dynamics are influenced by starvation or if

changes in the biofilm detachment rate contribute to the observed performance reduction. To test

hypothesis C, this chapter reviews starvation modeling approaches in the literature and proposes
109

a novel adaptation to Monod kinetics that can be easily integrated into dynamic biofilm models.

The approach is validated through comparison to controlled lab-scale starvation experiments and

the industrial data. By describing the transient response with a theoretical model, the mechanism

of the response can be evaluated, providing insights into potential mitigation strategies.

5.2 Model Formulation and Experimental Methods

As reviewed in section 2.3, there are a number of approaches to modeling biofilm

systems. To guide engineers and researchers in selecting amongst these approaches the

International Water Association (IWA) convened the Task Group on Biofilm Modeling to assess

the applicability of different models to benchmark case study scenarios. Their work was reported

in Wanner et al. (2006) and the guidance provided in this review was used to select a theoretical

modeling approach for the current work. The IWA Task Group states that:

the selection of an appropriate mathematical model must consider the objective of modeling,

the data available, the simplifications imposed, and the consequences of such simplifications.

(Wanner et al. 2006)

In this chapter of the thesis, the modeling objectives were to describe biosolids dynamics at the

reactor scale and to simulate the time-dependent response to flow stoppage. In considering these

objectives, two conclusions can be drawn: (1) the model must be able to simulate dynamic

behaviour and (2) a decision must be made regarding model complexity.

The need for dynamic simulation rules out some of the analytical and pseudo-analytical

models that implement a steady-state simplification (i.e. the steady state approach of Rittmann

and McCarty, 1980a,b and the MBBR model of Fouad and Bhargava, 2005). As a result,
110

dynamic models often employ numerical computation techniques to resolve partial differential

equations (in space and time) or are based on Individual Based Modeling (IbM, also called

cellular automaton, CA) simulations. Selecting among these dynamic models becomes a

judgment of how much mechanistic complexity to include.

One major factor that influences model complexity is the number of spatial dimensions

used for model parameters. While reactor-scale trends in substrate consumption can be generated

from 1-, 2-, and 3-dimensional models, the higher dimension models are needed if micro-scale

biofilm structural heterogeneity is to be simulated. By working in 2 and 3 dimensions, biofilms

can be characterized on the micro-scale, answering questions about how structure influences

biofilm function (Beyenal et al. 2004). However this level of detail does not necessarily add

accuracy to reactor-scale trends and the IWA Task group found that 1-, 2-, and 3-dimensional

models provided similar reactor-scale results for the benchmark problems studied (Wanner et al.,

2006). Higher dimension models also introduce a larger set of model parameters that require

estimation and increased computation burden.

When considering the MBBR system investigated in this thesis, two features limit the

potential of working in 2 and 3 dimensions: the MBBR hydrodynamics are not well defined and

the biofilm microbiology is a mixed community including filamentous organisms and protozoa

that make structural representations challenging.

MBBR hydrodynamics at the biofilm interface are difficult to estimate because they

result from the carriers circulating through the reactor volume due to vigorous bubbled aeration.

The circulating nature likely results in fluctuating flows through the internal spaces of the

carriers that are dependant on the orientation of the carriers and position within the reactor as

they circulate. Direct in situ measurement of hydrodynamics in the carrier system is not feasible,
111

leaving simulation as the only path to this value. Representing the system with computational

fluid dynamics (CFD) calculations is possible, but has yet to be reported on for MBBR systems

in the literature. This complex task was considered beyond the scope of this work but would be

required to validate parameter assumptions in a higher dimension model.

The microbiology present in MBBR systems is also challenging to simulate because IbM

modeling approaches have not yet addressed the growth of filamentous organisms or protozoa.

IbM simulations are based on a grid of theoretical cells that can be considered to be comprised

of bacteria, extracellular polymeric substances, or void spaces that evolve over simulated time

steps according to specified rules. Attempting to describe structural changes caused by

proliferation of filamentous bacteria and the growth of biofilm-grazing protozoa is an interesting

modeling challenge that has yet to be addressed in the literature. It may not be impossible to

represent these processes theoretically for instance by using rules for cell division that connect

newly divided cells in a filamentous chain but validation through controlled experiments poses

significant challenges. As with modeling the hydrodynamics, it was considered beyond the scope

of the thesis to perform this work. These as yet unaddressed complexities limit the usefulness

and expected predictability of 2- and 3-dimensional modeling for MBBRs with mixed microbial

communities. For this reason, a dynamic 1-dimensional diffusion-reaction model was selected as

the chosen approach.

Through reviewing the literature, and particularly the recommendations of the IWA task

group on biofilm modeling, the multi-species model of Wanner and Gujer (1986) was selected.

Although published in the late 1980s, this approach is still considered the theoretical standard for

1-dimensional modeling (Wanner et al. 2006) and it continues to be referenced in current

modeling studies (Xavier et al., 2004; Shannahan and Semmens, 2004; Matsumoto et al, 2007).
112

The model involves coupled differential equations representing temporal and spatial changes in

substrate concentrations and particulate materials such as microbial species. While only

representing gradients with biofilm depth and surface-averaged properties, the Wanner and Gujer

model can include any number of substrates or species. Work by Wanner and Reichert (1996)

demonstrated how this framework could be used to build a complex representation of a biofilm,

accounting for the transport of all solid and dissolved substances using limited model

assumptions.

The following sections describe the MBBR model derived in this work, beginning with a

diagrammatic overview of model and definition of nomenclature. Subsequent sections detail

MBBR reactor mass balances, the biofilm component, carrier geometry-based surface area

correction, the biofilm detachment rate terms used, and the response to transient periods of

starvation.
113

5.2.1. Model Overview

Moving Bed Biofilm Reactor


S effluent , X effluent
Effluent Substrate Concentration
S effluent , X effluent S b , X b (Assuming CSTR)

dS i ,b V
S i,o S i,b J max,i a rS ,i X i,b
dt Q
Substrate Biofilm Suspended
So Feed - Effluent Flux Phase
D d i
Effluent Suspended Solids Concentration

dX i ,b V
X i,o X i,b rdet oi X i,b
dt Q
Biomass Biofilm
Suspended
Feed - Effluent Detachment
Grow th

Biofilm Surface Area


a 8r 2 L f 2 r 2 L f h r 2 r 2 L f
2

Biofilm Component
S

J (Maximum Biofilm Flux)

Lf L
z

S i t , z 2 S i t , z
t
Di
z 2
ri t , z J i ,dbl
Di , w
L
S i ,b S i,L f
Biofilm Detachment
rdet k dm L f X f
2
Substratum Biofilm Diffusion Bulk rdet k dm 2 L f X f
Boundary
Layer
Reactor
rdet k dm L f X f k dg
Liquor

Figure 5.1. Overview of MBBR model concepts and main governing equations.

The diagram in Figure 5.1 gives an overview of the principal features and governing equations of

the derived MBBR model. A more detailed description follows in subsequent sections.
114

5.2.2. Model Nomenclature

Table 5.1. Nomenclature used in the derivation of the theoretical model.

Term Description Units Term Description Units

General Terms Biofilm Properties

3 3
S Concentration of substrate M/L Xf Biofilm volumetric density M/L
3 2
X Concentration of suspended bacteria M/L D Diffusion Coefficient L /t
2
J Flux of dissolved substrate M/L t
Kinetic Terms Lf Biofilm thickness L
L Diffusion Boundary Layer L
3
rS Rate of substrate consumption M/L t
3
Observed rate of growth M/L t Biofilm Detachment Terms
-1
max Maximum specific growth rate constant t
-1 3
q Specific rate constant of substrate consumption t rdet Rate of biofilm detachment M/L t
3 -1
K Monod half saturation constant M/L kdm Rate constant for biofilm detachment due to thickness t
2 -1 -1
Y Cell yield M/M, unitless kdm2 Rate constant for biofilm detachment due to thickness M t
-1
kd Specific rate of biomass decay t kdg Rate constant for biofilm detachment due to biofilm growth unitless
kI Specific rate of inactivation to inert material
Starvation Response Terms
Reactor Variables
-1
qmax Maximum specific rate constant for degradation of substrate t
3 -1
Q Flow rate through reactor L /t qmin Minimum specific rate constant for degradation of substrate t
3 -1
V Volume L Metabolic state function representing qmax(s,t) t
-1
HRT Hydraulic retension time = V/Q t qlim Absolute lowest value for qmin t
2 3
a Specific biofilm surface area L /L ksd Rate constant for down regulation of q unitless
r Radius of biofilm carrier element L ksu Rate constant for up regulation of q unitless

Table 5.2. Subscripts used for parameters

Subscript Description

b Bulk reactor liquor property, ie. suspended phase


f Biofilm property
O Oxygen substrate
C Organic substrate
o Inlet property, i.e. inlet substrate concentration
i Species or component index
h Property of heterotrophic biomass
in Property of inert biomass
dbl Property of diffusive boundary layer
crit Critical concentration of substrate causing starvation

5.2.3. MBBR Reactor Balances

The MBBR is conceptualized in this modeling effort to be a well-mixed reactor with a

fixed number of biofilm carriers and a population of suspended organisms. This concept leads to

straightforward mathematical representation as given in the reactor mass balances below for i

substrates and species:


115

dS i ,b V
S i,o S i ,b J max,i a rS ,i X i,b (1)
dt Q

dX i ,b V
X i,o X i,b rdet oi X i,b (2)
dt Q

As can be seen from eq. (1), substrates enter the reactor and are degraded by both the

biofilm (J max , maximum substrate flux into biofilm) and the suspended phase organisms before

being discharged. Rates of degradation, rS ,i , and growth, oi (note: oi Y rS ,i ), are based on

Monod kinetics with the exception of adjustments to model starvation response as described in

section 5.2.7.3.

Suspended solids are governed by a similar balance in equation (2) where solids enter and

exit the reactor and are generated through the rate of biofilm detachment and the growth of

suspended organisms. For the lab scale data simulated in this chapter, it is assumed that no

particulate matter enters the reactor since the reactor feed was a synthetic medium. This

simplifies equation (2) to:

dX i ,b V
X i ,b rdet Y rS ,i X i ,b (3)
dt Q

5.2.3.2. Reactor balance assumptions

Additional assumptions made to arrive at the reactor balances given above are described

in Table 5.3
116

Table 5.3. Model assumptions for reactor balances

Assumption Rationale

Reactor is well mixed Vigorous mixing caused by aeration was


observed by carrier motion in lab scale
reactors.
Kinetic parameters reflect average Kinetic parameters fit to reactor
community characteristics performance and within ranges reported in
the literature
There are sufficient carriers to render Lab-scale reactors contain 1000 carriers.
biofilm variability between carriers Sampling and measurement of biofilm
negligible properties done to minimize and account
for this variability
Biofilm and suspended biomass are This has been found to be a fair assumption
governed by similar kinetic parameters (Bakke et al., 1984; Rittmann and
McCarty, 2001)
Biofilm detaches at a much greater rate The net balance of solids indicates that
than suspended solids attach to the biofilm detachment is a more significant process. It
surface. The two processes can be modeled is challenging to directly measure the rate
as one relationship. of attachment in a reactor system to
develop terms for modeling this process.

5.2.4. Biofilm Component

The biofilm component is based on the work Wanner and Gujer (1986) who have

presented the derivation in detail (See also Wanner and Reichert, 1996).

5.2.4.1. General equations

The model defines mass balances on substrates and growth of microorganisms over a

finite element of the biofilm and these balances are then integrated across the biofilm thickness

and coupled to bulk reactor values. The mass balance for substrate i at a position in the biofilm is

given by:

S i t , z 2 S i t , z
Di ri t , z (4)
t z 2
117

The first term of equation (4) represents the Fickian diffusion of substrate into the biofilm

and the second term is the rate of degradation at that point. Solving this partial differential

equation can yield profiles in substrate through the depth of the biofilm over time. In doing so it

is possible to calculate the maximum flux,

Lf
S i
J max ,i (5)
z

at the biofilm surface. It is this value that governs the reactor mass balance on substrate as given

in equation (1)

In the Wanner and Gujer (1986) approach, the growth of the biofilm is derived as the

velocity of expansion of the biofilm a given depth. Framing the derivation in this way allows for

easy calculation of the biofilm thickness with time and for the composition of biofilm particulate

components. The velocity of expansion u(t,z) (units in L/t) is defined mathematically as

z
u t , z ot , z dz (6)
0

where o is the average observed specific growth rate at position z. Expansion is caused by the

integral of all growth at positions deeper than z. The net change in biofilm thickness is given by

this relationship specifically at the biofilm surface, L f , while subtracting the velocity of biofilm

detachment, u det as follows:

dL f z
ot , z dz u det (7)
dt 0

The velocity of detachment can be related to the detachment rate by multiplying by the

specific biofilm surface area, a, and the biofilm density, X f :


118

rdet u det a X f , (M/L3t) (8)

Mathematical description of the biofilm detachment rate for an MBBR is one of the

objectives of this chapter and greater detail on the detachment rate terms considered is given in

Section 5.2.5. When contemplating more than one particulate component, such as multiple

species, inert material, or EPS, the fraction of component i is given by:

f i t , z f t , z
oi t , z ot , z f i t , z u t , z i (9)
t z

Equation (9) can be integrated to yield the composition profile of different species with depth, a

modeling tool that has been used to answer questions related to the stratification of species in the

biofilm (Morganroth and Wilderer, 2000; Shanahan and Semmens, 2004).

Equations (4), (7), and (9) provide a complete one-dimensional description of the biofilm

in time and in space. Depending on the number of substrates and particulate components used,

the model can simulate a range of complexity to suit a variety of experimental situations.

5.2.4.2. Challenges for solving general equations

Three general features of the biofilm modeling equations make them challenging to solve:

1. The use of Monod kinetics leads to non-linear differential equations which are not readily

linearized.

2. The system includes partial differential equations that must be manipulated to use

ordinary differential equation solving techniques.

3. The evolution of biofilm thickness leads to a moving boundary problem which requires

normalized spatial coordinates to reduce degrees of freedom.


119

These challenges are not insurmountable, and can be solved as follows:

1. Non-linear differentials can be numerically solved through iterative computational

techniques

2. Rather than directly solving the partial differential in space, the spatial dimension is

converted to a system of linked ordinary differential equations through the Method of

Lines technique (Schiesser, 1994). This requires approximation of the spatial derivatives

using finite differences and an adequate number of grid points to reduce deviations from

the true solution.

3. As described by Wanner and Gujer (1986), the biofilm thickness can be normalized to 1,

with discretized differentials referenced to some fraction of the biofilm thickness. On

S
S t 1

S t 0.9
S t 0.8

S t 0.7
S t 0.6
S t 0.5
S
S t 0.4
S t 0.3
S t
t 0.1
0.2

t 0

S
z
z
Figure 5.2. Schematic representation of the discretization of spatial derivatives through the
Method of Lines
120

each time step the biofilm thickness is solved as a separate relationship to give the real

spatial coordinates.

Figure 5.2 provides a conceptual graph of the manner in which the solution is carried out.

5.2.4.3. Accounting for the diffusion boundary layer

The biofilm component is linked to the reactor mass balances through the concentrations

of substrates at the surface of the biofilm, the flux of substrates into the biofilm, and the

detachment rate. Unlike the flux and detachment rate, it is not possible to directly transfer the

substrate values in the bulk to those at the biofilm surface due to the presence of a diffusion

boundary layer common to solid-liquid interfaces. This laminar flow layer of thickness L can be

modeled by a first order diffusion term:

J i ,dbl
Di , w
L
S i ,b S i,L f (10)

Since the biofilm equations require the substrate concentration at the biofilm surface as a

boundary condition, equation (10) can to substitute in the bulk substrate concentration and relate

biofilm to reactor balances. In doing so it is necessary to set the boundary layer thickness L as a

model parameter. In this work, the parameter L is estimated from dissolved oxygen profiles

collected when probing biofilm samples.

5.2.4.4. Biofilm model components

So far the biofilm model has been described generally without defining the substrates and

particulate components modeled in this work. In order to select the appropriate model terms, the
121

lab-scale experiments will be used for validation and comparison require consideration. Chapter

4 provides a description of the experimental conditions and these details are not recounted in full

here.

The selection of species types represented in the model system is an important decision

point for the derivation. For mixed community systems this is often done by considering

common metabolic types such as heterotrophs, ammonium oxidizing bacteria, nitrite oxidizing

bacteria, dentrifiers, and/or anaerobic ammonia oxidizing (ANAMOX) bacteria. The species

types selected should reflect the operational conditions of the modeled system and the observed

conversion of substrates.

The lab-scale studies investigating the effect of calcium were conducted in a manner that

favoured the growth of heterotrophs and were not likely supportive of nitrifiers, anamox bacteria,

or denitrifying functionality. This is due to the experimental design which sought to capture

some of the general properties of the industrial MBBR treating pulp mill effluent where BOD

removal is the principal treatment objective. Parameters such as the hydraulic retention time,

organic loading, reactor fill ratio, and level of aeration were set to simulate similar operational

conditions. The organic loading to the lab-scale MBBRs was 22 gCOD/m2d (note COD BOD

for the lab system since carbon substrates, acetate and glucose, were readily degradable) and this

is indicative of a heterotroph-favouring state. For instance degaard (2006) investigated the

potential for MBBRs to carry out nitrification in a review of MBBR operational experience. This

experience indicates that BOD loadings greater than 7 gBOD/m2d require dissolved oxygen

concentrations greater than 10 mg/L (i.e. exceeding saturation with air) to achieve simultaneous

BOD removal and nitrification. Thus the lab conditions far exceeding the 7 gBOD/m2d

nitrification inhibition limit reported for MBBRs by degaard (2006).


122

Denitrifiers and ANAMOX bacteria are other possibilities for consideration since the

MBBR biofilms were sufficiently thick to have significant anoxic regions. However the low

levels of nitrogen supplied would be consumed preferentially by heterotrophs for growth

requirements and would not be adequate to support significant denitrification or anaerobic

nitrification.

Beyond the type of species present, other solid components of the biofilm can be

considered for inclusion in the model such as inert biological material generated through cell

lysis, extracellular polymeric substances (EPS), and other inert particulate matter from external

sources. The modeling approach considered inert material generated from cell lysis as the only

additional solid component.

The inclusion of lysed inert material was thought to be important because the lab-scale

MBBRs involve mature biofilms as defined according to terminology common in the literature

(Sauer et al., 2002). All lab-scale data were collected after the reactors had acclimated for at least

30 days. After such a period of growth, the death of cells can result in buildup of non-degradable

material, particularly in the inner regions of the biofilm (Mason et al., 1986; Horn et al., 2003).

Inert material was included in the model and evaluated for its significance.

The other commonly modeled element of cell death is the endogenous decay, which

represents mass loss due to the metabolism of dead cell components. This does not require the

definition of a new solids component but is reflected in the rates that govern evolution of

heterotrophic biomass.

Extracellular polymeric substances were evaluated for inclusion in the MBBR model.

Experimental measurements collected in this thesis indicated the presence of various EPS

compounds and a body of evidence suggests the important role EPS plays in biofilm structure
123

and function (Flemming et al., 2007). Nielsen et al., (1997) have developed a theoretical

framework for modeling the production of different EPS components and Horn et al., (2001)

implemented this framework in a simulation study. This work concluded that separately defining

EPS in the model led to growth and substrate conversion outputs that were similar to those

produced by models that treated the biomass as singular component. Adding EPS components

and associated rates also increases model complexity and requires validation with measurements

of EPS amounts collected through extraction. Currently EPS extraction and quantification is an

active research area and it is clear from evaluations of extraction methods (i.e. Brown and Lester,

1980; Zhang et al., 1999; Liu and Fang, 2002; Comte et al., 2006;) the extracted EPS amounts

and composition are highly dependant on extraction procedure. Thus in this work it was

determined that inclusion of separate EPS components would add complexity that may not be

verifiable experimentally and would not improving the models potential to answer the questions

investigated. It is likely for similar reasons that many other diffusion-reaction modeling studies

do not include separate EPS rate terms (i.e. Horn et al. 2003).

Inert particulate material present in the wastewater entering the reactor was also omitted

based on the assumption that the synthetic effluent fed to the reactors did not contain particulate

material.

Following the definition of solid components, it was necessary to define substrates

included in the model. These were selected to be dissolved organics as biochemical oxygen

demand and dissolved oxygen. Growth and decay rates for heterotrophic biomass rely on these

substrates and with no other species type defined, ammonia or other nitrogenous species were

omitted.
124

With this discussion in mind, Table 5.4 shows a matrix of the kinetic rates and

stoichiometry for all substrates and solid components modeled in the MBBR model.

Table 5.4. Process matrix for the MBBR heterotrophic biofilm

Dissolved Components Solid Components


Process SC SO Xh Xin Process Rate
SC SO
Growth -1 -(1-Y) Y qh Xh
KC SC KO SO
SO
Endogenous Decay -1 -1 kd Xh
KO SO

Lysis -1 1 k in X h

The selection of model components and process rates is intentionally a relatively simple

representation based on the reasons presented. This reduces the number of adjustable parameters

that are not related to the detachment rate and thus the biosolids dynamics. A similar approach

was taken by Horn et al. (2003) when investigating detachment phenomena under different

hydrodynamic conditions.

5.2.4.5. Biofilm Model Assumptions

The biofilm model used in this study relies on the assumptions discussed for the reactor

balances and some additional assumptions:

Table 5.5. Model assumptions for the biofilm component

Assumption Rationale
The effect of heterogeneous biofilm This was the conclusion of the IWA
structure on reactor function can be Biofilm Task Group as reported in Wanner
adequately represented by averaged et al., (2006).
parameters fit to experimental data
Biofilm density is considered to be Some experimental work has shown that
constant with biofilm depth biofilm density increases with the depth of
the biofilm (i.e. Zhang and Bishop, 1994;
Bishop et al., 1995; ). However, including
density profiles is time consuming
125

experimental work that was omitted when


characterizing biofilms in the lab. Instead it
was assumed that by defining an average
constant biofilm density, the general
properties of growth and detachment could
be simulated.
Numerical solution techniques and number Solution errors are minor with respect to
of spatial and time divisions used for variability of other model parameters.
estimating continuous trends are good Model outputs such as spatial gradients are
representations of actual differential resolved according to the number of space
equations and time divisions used and were always
selected to be smaller than intervals
between measurements.

5.2.5. Correcting Effective Surface Area for Carrier Geometry

The suspended carriers in MBBRs are typically compared on the basis of specific surface

area or specific protected surface area, referring to the surface area colonizable by biofilms under

turbulent reactor conditions. This parameter has been found to be a key design variable for

MBBR treatment performance, even when comparing carriers of different geometries, sizes, and

fill ratios (degaard et al., 2000). From a theoretical perspective, the dependence of degradation

kinetics on carrier surface area will be true for systems where reactor biomass exists primarily in

the biofilm phase (e.g. limited suspended phase due to low hydraulic retention times), and where

biofilms are sufficiently deep as to have substrate flux that is independent of thickness. For many

MBBR applications these conditions are generally met and commercial carrier designs range

from 200 to 1200 m2/m3 protected surface area (degaard, 2006).

While the biofilm surface area is an important parameter for predicting reactor

performance, the design surface areas for MBBR carriers are based on the geometry of the bare

carrier element, an assumption that has implications for modeling. For instance, the Kaldnes K1

carrier used in this work is specified as having a 490mm2/carrier protected surface area

(corresponding to 500 m2/m3 for packed carriers or 300 m2/m3 for a typical reactor fill ratio of
126

60%). Using the bare dimensions assumes that the surface area does not change as the biofilm

grows. However, as biofilm thickness increases the internal spaces of the carriers become more

constricted, leading to a lower functional surface area. For mature wastewater treatment biofilms,

biofilm thickness increases may present a significant reduction in surface area, particularly for

carriers with small internal spaces.

To demonstrate the relationship between biofilm thickness, L f , and surface area, a, the

following equation was derived for the K1 carrier geometry:

a 8r 2 L f 2 r 2 L f h r 2 r 2 L f
2
(11)

[A] [B] [C]

where r is the carrier radius and h is the carrier height. As depicted in Figure 5.1, term [A]

represents the area of the radial surfaces, term [B] represents area of the circumference, and term

[C]
Specific Surface Area (m /m )
3

600
2

S
500

400

300
t
200

100

0
0 500 1000 1500 2000 2500
Biofilm Thickness (m)
Figure 5.3. Comparison of theoretical surface area correction to carrier
specifications
127

represents the surface area created on the open sides of the carrier as thickness increases. This

equation is plotted in Figure 5.3 in comparison to the specified surface area of 500 m2/m3.

As can be seen from this relationship, a biofilm thickness of 300m leads to a reduction

in surface area of 9% from the design specification while a thicker biofilm of 700 m would

result in a surface area decrease of 22%. Thickness greater than 700 m have been reported in

other long-term biofilm studies (Zhang and Bishop, 1994; Murga et al. 1994; Okabe et al., 1999;

Horn et al., 2003), and even complete carrier plugging has been observed in MBBR experiments

(Gieseke et al., 2001). This suggests that correcting for the thickness-dependant surface area is

important, and the model developed in this work uses the relationship in equation (11) to do this.

The results section presents a comparison of model calculations to evaluate the significance of

this correction for describing reactor behaviour.

5.2.6. Rate of Biofilm Detachment

Biofilm detachment is perhaps the most challenging process to represent in mechanistic

terms since the physical and biological factors that contribute to detachment phenomena are

complex. A discussion of research investigating the detachment process and mathematical

representations used in modeling has been presented in section 2.1.5. and 2.3.1. of the literature

review, respectively. This review collected some of the common rate terms used in diffusion-

reaction modeling studies and these are shown again in Table 5.6. below:
128

Table 5.6. Common detachment rate expressions proposed in the literature

Rate Expression Variables Reference


kd f L f 0.58
Volumetric density, thickness, shear stress Rittmann (1982)
Concentration of carrier particles, Reynolds
k b1 k b2 C p k b3 Re k b4 number, Shear stress
Chang et al. (1991)


Lf k k
'
d
"
d Thickness and growth rate Speitel and Di Giano (1987)

kd X f Areal density Rittmann and McCarty (1980)

k d f L2f Volumetric density and thickness Wanner and Gujer (1986)

Operation, 0

Backwash, k d L f Lbase Thickness and operational mode Morganroth and Wilderer (2000)

One objective of this work was to evaluate detachment expressions to determine which

forms could be used to describe biosolids dynamics under different calcium conditions and in

response to transient starvation. Of the expressions in Table 5.6, the first two are based on shear

due to fluid flow acting on the surface of the biofilm. These expressions were not considered in

this work since the hydrodynamics were not known due to the complex flow experienced by the

carriers as they moved in the reactors. Since reactor variables that might influence

hydrodynamics such as the rate of aeration and number of carriers were held constant during

experiments, it was assumed that changes in shear did not affect the detachment rate.

The rates proposed by Speitel and DiGiano (1987), Rittmann and McCarty (1980), and

Wanner and Gujer (1986) are all based on either biofilm accumulation and/or the rate of biofilm

growth. Accumulation-based terms suggest a mechanistic link to a process of weakening as the

biofilm grows thicker. This may be explained by internal weakness caused by endogenous

respiration of decaying biomass or greater applied shear force on thicker biofilms for some
129

reactor configurations. Computationally, the accumulation-based terms are useful in that they

assure a steady state for most conditions and this may be the main reason they are so commonly

used (Rittmann and McCarty, 2001). On the other hand, growth-based terms suggest that rapidly

developing biofilm is more susceptible to detachment. Kwok et al, (1998) hypothesized that this

was due to the structure of growing biofilms which the authors characterized as extending from

the biofilm in protrusions that are easily sheared off the surface. Although it has been shown that

detachment is correlated to growth rate in a range of biofilm systems (Peyton and Characklis,

1993; Tijhuis et al. 1995; Nakhla and Suidan, 2002; Garney et al. 2008) growth-based

detachment terms alone do not yield a steady state film thickness. In order to overcome this

computational challenge, the detachment rate must also include other terms such as the growth

and thickness dependence proposed by Speitel and DiGiano (1987). It should also be noted that

the accumulation-based detachment rate proposed by Wanner and Gujer (1986) is based on Lf2.

Although this was selected without a mechanistic justification of the second order relationship,

the formulation may indirectly affect the time dependency of detachment on thickness changes

allowing the biofilm to reach a steady state more rapidly.

In the current work, all three forms of detachment rate are considered; dependency on

thickness, thickness2, and growth + thickness. These were formulated as follows:

rdet k dm L f X f (12)

2
rdet k dm 2 L f X f (13)

rdet k dm L f X f k dg (14)

The equations used are not taken in the exact form of cited studies, but provide a simple

and consistent basis to evaluate the mechanistic links implied by each modeling approach. Each
130

rate was fit to experimental data to evaluate rate constants k dm , k dm2 , and k dg . Data fitting

methodology is described in section 5.3.9.

5.2.7. The Effect of Starvation

5.2.7.1. Observed Responses to Starvation

Periods of starvation are an inescapable reality for microorganisms in natural settings and

survival has demanded the evolution of adaptation strategies. As cells pass into a condition

without nutrients they have been observed to undergo a range of physiological changes. Some

observed starvation-induced changes include down-regulation of metabolism (Augustin et al.,

2000a; Konopka et al. 2002; Moreno-Andrade et al. 2006), synthesis of enzymes that allow

metabolism of alternative substrates (Bernhardt et al., 2003), consumption of internal energy

storage molecules (Malmcrona-Friberg et al., 1986), decreases in cell volume (Amy et al. 1983;

Sanin et al. 2003), changes in cell surface hydrophobicity (Sanin et al. 2003, Haznedaroglu et

al., 2008), growth of flagella to increase motility (Malmcrona-Friberg et al., 1990), and spore

formation for organisms that can differentiate in this manner (Roszak and Colwell, 1987). The

response is species-specific and it represents a continuum of cell changes as the starvation

duration increases. For instance, the synthesis of starvation response proteins, metabolic down

regulation, and consumption of energy storage molecules may happen within the first hours of

starvation while cell size reduction and transformation to spores may only happen after

prolonged starvation (Konopka, 2000).

The observed responses clearly have the potential to affect biofilm structure and function.

Cell shrinking, surface property changes, and increased motility could cause reduced cohesion

between cells, resulting in weaker biofilms and increased biofilm detachment. Extracellular
131

polymers that form the biofilm matrix may also be actively degraded as a dispersal strategy

(Boyd and Chackrabarty, 1994) or consumed as an energy source, a process that would further

weaken the biofilm. As introduced previously, biofilm detachment has been observed as a result

of starvation (Sawyer and Hermanowicz, 1998; Hunt et al., 2004; Gjermansen et al., 2006) and

these may be some of the underlying causes. A shift into a stationary or low-metabolic mode of

growth as a result of starvation can also impact biofilm function even after nutrients are

reintroduced. Depending on the duration of starvation, biofilm cells that have shifted into low

metabolic states may take time to re-acclimate to maximum growth rates when they are exposed

to nutrients. For biofilm wastewater treatment reactors, a lag phase would cause pollutants to

pass through the system without being degraded.

These responses suggest two modeling approaches: (1) modification of the detachment

rate and (2) reduction in maximum degradation and growth rates as a result of starvation. To

select an appropriate strategy, historical data from known transient periods experienced by the

industrial MBBR were examined.

5.2.7.2. Observed Starvation Response in Industrial MBBR

Over a 3 year period of historical data, 51 reactor stoppages greater than 5 hours were

observed to occur. Shutdowns ranged from 5.5-16 hours in duration with a mean period of 9.3

hours. Four characteristic examples of the BOD and TSS trends before and after shutdowns are

presented in Figure 5.4.


132

Concentration of Effluent Total Suspended Solids or Soluble BOD 300


300

250 250

200 200

150 150

100 100

50 50

0 0
-5 -4 -3 -2 -1 0 1 2 3 4 5 -5 -4 -3 -2 -1 0 1 2 3 4 5
(mg/L)

300 300

250 250

200 200

150 150

100 100

50 50

0 0
-5 -4 -3 -2 -1 0 1 2 3 4 5 -5 -4 -3 -2 -1 0 1 2 3 4 5
Time (d) Time (d)

Figure 5.4. Four examples of effluent TSS ( ) and soluble BOD ( ) fluctuations following
transient periods of reactor shutdown (grey bars) in the operating data from a full-scale MBBR

These examples show how the recovery from shutdown causes a period of poor BOD removal. It

is also interesting to note that for each example it appears that the suspended solids are reduced

over the day following shutdown. This is not indicative of a widespread detachment event, but

rather that the detachment rate decreases in the day following the shutdown period.

By assessing this response it is proposed that the primary effect of starvation is the

reduction in metabolic rate of the biofilm and suspended organisms. A temporary period of low

growth rate accounts for the poor treatment performance in the days following the shutdown.

This also provides the basis for the observed reduction in suspended solids if one considers a

detachment rate that is dependant on growth. For growth-dependant detachment, the metabolic

decrease would also reduce the rate of biofilm detachment which could explain the observed
133

trends in suspended solids production. This was considered to be preferable to arbitrarily

adjusting the detachment rate term since this would introduce new model parameters that may

not be mechanistically descriptive. The discussion section provides greater rationale for this

selection by assessing other approaches in comparison to the lab-scale and industrial data.

5.2.7.3. Modeling Metabolic Downshift

The shift from a substrate limited state to one where substrates are present in abundance

has been observed to result in growth that cannot be explained strictly with Monod kinetics

(Grady et al. 1996; Wirtz, 2002). Acclimation of a community of microorganisms to new

substrate conditions results in some period of adjustment where growth rates are sub-optimal.

The most common way this is addressed mathematically is through the definition of a lag phase

of growth.

Lag Phase Exponential Stationary There exists a significant body of


Phase Phase
Nmax research on modeling microbial lag as it

is of great importance to food


Ln N

engineering and predicting microbial

contamination risk. Critical reviews by


No Baranyi and Roberts (1995), Swinnen et
Lag al., (2004), and Prats et al., (2008)
Time
Figure 5.5. Conceptual diagram of batch
survey this research and provide general
growth following exposure to substrates.
Adapted from Swinnen et al., (2004). N is the
number of cells. mathematical frameworks for lag

modeling of an inoculum of microorganisms transferred to a substrate-rich environment. A

conceptual diagram of the growth of bacteria over time is shown in Figure 5.5. with the lag
134

defined as the time between exposure to substrates and the transition to exponential growth at

max .

The duration of the lag is a measurable system property but challenging to predict since it

is dependant on a number of factors such as temperature (Augustin et al., 2000a), microbial

species (Tappe et al., 1999), and pre-inoculation growth conditions (Augustin et al., 2000b). For

instance, bacteria that have been subjected to longer periods of starvation have been found to

have a longer lag phase (Konopka et al., 2002). Much of the work to characterize how these

variables affect lag has been conducted in a way that simulates microbial contamination and is

not directly transferable to a wastewater treatment scenario.

Modeling lag can simply involve fitting a lag time parameter to experimental data such as

in the work of Mtris et al, (2001) and Augustin et al., (2000b) but this has some disadvantages

for modeling the recovery from starvation and specifically for diffusion-reaction modeling of

biofilms. Lag times must be defined explicitly and referenced to explicit time events such as the

re-exposure to nutrients. This requires the timing of events to be known a priori for a simulation

and applies an arbitrary response time instead of basing the response on some internal microbial

state variable that may be more mechanistically descriptive. By using explicit time lags, the

approach is not easily compatible with differential equation solution techniques used in

diffusion-reaction modeling. These are best solved by describing the system using differential

terms that can be integrated to define system evolution.

An alternative approach that addresses these challenges is the adaptation of kinetic

parameters based on environmental conditions. As reviewed by Grady et al. (1996) and Ferenci

(1999), there have been a number of ways in which constant kinetic parameters such as the

maximum growth rate, max , and affinity constant, K, can be made into functions of intrinsic cell
135

properties that reflect culture growth history. For adaptation of max due to starvation, Powell

(1967) introduced a mathematical framework that continues to be relevant (Grady et al., 1996).

Powell suggested that max is a state variable that represents an organisms intrinsic metabolic

state. Thus as organisms enter conditions that reduce the metabolic state, such as a period of

starvation, max decreases. When metabolically limited organisms are returned to conditions that

promote unrestricted growth, max will begin to return to the constant value. Thus instead of

explicitly delaying growth by a lag period, the lag can be reproduced by defining a function of

metabolic state change.

Wood et al., (1995) took such an approach to predict breakthrough curves in a simulated

aquifer system. Their work defines a critical substrate concentration that acts as a switch for

down-regulation of the metabolic state. When the substrate concentration falls below the critical

value, max remains constant for some lag duration and then decreases linearly until it reaches a

minimum value. When substrate concentration rises above the critical value a similar lag and

linear increase in max results, returning the organism to full metabolic capacity. This metabolic

state function is more mechanistically linked, but it still relies on piecewise functions that need to

be calculated based on a system memory function (which coordinates the lag duration).

The current MBBR model proposes a novel approach that builds on the work of Powell

(1967) and Wood et al., (1995) by adapting the metabolic state function to not require a memory

function and to be more amenable to biofilm modeling. This is done by starting with the

definition of degradation kinetics, where degradation rate is based on a metabolic state function.

S
rs ( S , t ) X (15)
Ks S
136

The metabolic state, (S,t), a function of time and substrate concentration, replaces q max

in the Monod equation. Note that growth kinetics would be directly related to the degradation

rate according to o = Yr s and from this point forward equations will only present derivations for

rates of substrate utilization since this is of greater relevance to treatment performance.

As in Wood et al. (1995), the MBBR model specifies a critical substrate concentration,

S crit , below which starvation occurs. By setting this value sufficiently low at K S /10, the model

ensures that for most situations strict Monod kinetics apply, reflecting the general applicability of

the historical growth equation. The change in (S,t) over time is then given by the following set

of logistic functions:

d if S S crit , k sd q max q min


(16)
dt if S S crit , k su q max q min

where k sd is a negative constant describing the rate of decrease in while k su is a positive

constant describing the rate of increase in . This set of functions results in a trend toward q max

when S > S crit and a trend toward q min , a minimum defined growth rate, when S < S crit . The

logistic nature of the functions was selected because it achieves a constant state at either q max or

q min over time and it provides a function that can be rationalized theoretically. Figure 5.6

presents a diagram for a theoretical starvation period of 10 hours to demonstrate the form of this

function.

The logistic response can be manipulated by adjusting the rate constants k sd and k su to

match observed behaviour. Although this introduces two new model parameters, it is

hypothesized that microbial communities will be sufficiently different in their response to

starvation and recovery to require unique rate constants.

The logistic function itself was selected as an appropriate form for the MBBR because it

reflects the starvation response of a mixed community. The tailed ends of each downshift and
137

upshift are indicative of a normal distribution of responses. For instance, as the community

enters starvation some of the species will be rapidly susceptible to substrate limitation and begin

to respond while the majority of species will become down-regulated around the median time.

The tail at the end represents the few resilient species that will take longer to become susceptible

to starvation. It is proposed that the logistic function is more representative of mixed community

behaviour than a strict linear relationship.

Aside from the rate constants governing the metabolic state, the model introduces a

minimum degradation rate q min that requires definition. It is proposed that for the durations of

60
Substrate Concentration (mg/L)

50

40

30

20

10
Scrit
0
MetabolicStateFunction, (d )

qmax
1

8
7
6
5
4
3
2
1 qmin
0
0 10 20 30 40 50
Time(hr)
Figure 5.6. Theoretical metabolic downshift and recovery due to a 10 hour
starvation period caused by reduction in substrate concentration (dashed line)
below Scrit . Curve depicted for k downshif t and k upshif t of -3.5 and 3.5, respectively.
138

starvation experienced during industrial shutdowns (<1 day), the MBBR microbial community

would not reach a state of complete inactivity. This is assumed because of evidence showing

some immediate metabolic activity even for longer starvation periods. Konopka et al., (2002)

observed an immediate but low growth rate after an 8 day starvation of an activated sludge

community and Bollmann et al., (2005) found that nitrifiers were able to oxidize ammonia at a

reduced rate following a 7 day starvation.

One must also consider the validity of setting the limiting q min value as a biofilm and

community-wide constant. An important consideration when modeling transient starvation in

biofilm reactors is that mature biofilms have significant regions that undergo starvation even

when the system as a whole is substrate-rich. When applying equations (15) and (16) to the

spatial profile of a biofilm, it is plausible that the inner regions will fall below the critical

substrate concentration and trend towards the low metabolic state of q min . If q min were defined as

a constant, steady state biofilms would be arbitrarily divided into an active region governed by

Monod kinetics and a discrete inactive region at the lower q min value with the division point at

the position where S < S crit . This discrete stratification would be difficult to justify, particularly

considering that biofilms are heterogeneous structures and the diffusion-reaction model provides

an averaged picture. To account for this, the q min value is proposed to scale with the

concentration of substrate so that as the substrate concentration falls further and further below

the Scrit value, the q min boundary also falls. This scaling function is given below:

S (17)
q min q max
S crit
S

q min q max qlim qlim (18)
S crit
139

where equation (17) is for an absolute limiting metabolic state, q lim , of zero, and (18) is scaled to

some non-zero absolute limiting value (q lim > 0). This results in a gradient of starvation response

over the depth of the biofilm as depicted in Figure 5.7a. Figure 5.7b shows how the metabolic

state function changes in both time and space for a simulated starvation period.
Specific Maximum Substrate Utilization Rate, q, (d-1)

Specific Maximum Substrate Utilization Rate, q, (d-1)


(A) (B)
8 8

6 6

4 4
2 2
0 0
0
24 1
60
36 0.5 48
48 1.0 36
Time (h) 0.5 0 24
60 0.0 0 Time (h)

Normalized Position
(1 = biofilm surface)

Figure 5.7. 3-dimensional depiction of the modified specific maximum substrate utilization rate for (A)
steady state and (B) an 11 hour starvation period.

5.2.8. Experimental Data and Methods

5.2.8.1. Steady state data

The data used for comparison and validation of the MBBR model were collected during

the lab-scale experiments investigating the effect of calcium, specifically experiment III which

tested 1, 50, 100, and 200 mg/L Ca2+ doses. This allowed for the model to be fit to four data sets

to show general validity and to compare biofilm detachment terms for each calcium

concentration. Filtered chemical oxygen demand (COD), suspended solids (VSS), and biofilm

properties such as the biofilm areal density and oxygen microprofiles were collected during a
140

steady state period after acclimation. The methodology used for collecting these measurements is

given in Chapter 4.

5.2.8.2. Starvation experiments

The industrial data depicting the response to starvation presented in Figure 5.4 are

somewhat limitated in their potential to provide process insight. One limitation is that BOD and

TSS measurements are based on composite daily samples that reflect average values over a 24

hour period. This sampling protocol dampens variability and does not reflect change over small

enough time intervals to make conclusions about what is happening to the biosolids dynamics

and degradation rate over the transient period. The historical industrial data also do not represent

controlled experimentation, and each shutdown may be affected by other changing process

variables. For these reasons, lab-scale starvation experiments were conducted to simulate

shutdown conditions.

The lab-scale starvation studies were undertaken using the acclimated MBBRs in

experiment III, described in Chapter 4. The experiments involved stopping flow to the four

reactors for a period of 11 hours while maintaining a constant rate of aeration. Flow was then

restarted and the reactors were monitored until they returned to steady state. A measurement

campaign involving COD removal, suspended solids, biofilm areal density, and biofilm and

suspended phase activity was undertaken throughout the starvation and recovery period. Except

for the activity measurement technique which is described in the next section, all methods have

been previously detailed in Chapter 4.

The results depicted for the starvation trials omit data from the 1 mgCa2+/L reactor. This

experiment was not included because it was observed that just prior to the starvation trial, fungal
141

species had rapidly grown in the reactor, causing the carriers to become completely plugged.

Figure 5.8 shows a photo of this unexpected occurrence.

This resulted in abnormal reactor


1 cm
Typical biofilm thickness behaviour that could not be modeled in the same

context as the rest of this study. Though this

situation prevented comparison of starvation

response in the 1 mgCa2+/L case with the other

Fungal plugging of conditions, it is an interesting observation because


1 mgCa 2+/L condition
it suggests that complete plugging is a possible and

stable growth situation for MBBRs. Such events


Figure 5.8. Photo of carrier plugging in
comparison to normal biofilm colonization
may not be desirable since it results in reduced

biofilm surface area and it was found that during the period of plugging, reactor degradation

rates had diminished considerably.

5.2.8.3. Activity measurements using iodonitrotetrazolium (INT) chloride

Biofilm and suspended phase microbial activity was quantified by measuring electron

transport chain (ETS) activity through reaction with iodonitrotetrazolium chloride using a

method described by Blenkinsopp and Lock, (1990).

Biofilm samples were collected by scraping the inner surfaces three MBBR carriers with

a sterile scalpel. Biomass was place in a 15mL centrifuge tube, suspended in 1.5 mL sterile

phosphate buffer, and gently vortexed on the lowest setting to homogenize the sample. For

specific activity rates, average biofilm mass per carrier was measured using 10 separate carriers
142

at the same time as described in section 4.2.4. To account for losses due to scraping biomass,

clean carriers were also assessed for biofilm mass and subtracted from the average value.

Suspended phase samples were collected by centrifuging (2100g, 10 minutes) 50 mL of

reactor mixed liquor. After decanting the supernatant, the biomass pellet was re-suspended in 1.5

mL of PBS. To quantify biomass, suspended solids measurements were collected at the same

time.

Activity was measured in the collected samples by adding 2 mL of a 0.2% (w/v) INT

chloride solution and incubating them in the dark at 37oC under shaking at 200 rpm. After a 2

hour incubation, the reaction was stopped by adding 1 mL of 37% formaldehyde solution.

Samples were then centrifuged at 2100g for 10 minutes to collect the biomass and evolved INT-

formazan crystals. Supernatant was decanted and the pellet re-suspended in 10 mL of methanol

with a vortex on high to allow for INT-formazan extraction. Following a 15 minute extraction

period in the absence of light, the samples were centrifuged at 3000g for 10 minutes and the

supernatant collected for spectrophotometric analysis. The extracted red INT-formazan was

examined at 464nm in comparison with a standard curve prepared in methanol.

5.2.9. Model Parameter Selection and Fitting Methodology

Model parameters were defined by using constants from the literature and through fitting

to experimental results. As a starting point, kinetic parameters for activated sludge communities

were taken from Rittmann and McCarty (2001), which are also similar to parameters used in

Horn and Hempel (1997) and Horn et al. (2000) for heterotrophic biofilms. These values are

shown in Table 5.7


143

Table 5.7. Kinetic parameters used in the The remaining model parameters describing
MBBR model
Parameter Value Unit biofilm properties, biofilm detachment, and the
-1
max 4.8 d
-1 response to starvation were fit using experimental
qmax 8d
KS 10 mg/L data within the defined kinetic framework. This
KO 0.1 mg/L
Y 0.6 mgVSS/mgCOD was done through the following steps:
-1
kd 0.08 d
-1
kI 0.4 d

1. Biofilm properties such as X f and L were set so that biofilm areal densities and oxygen

microprofiles (revealing biofilm thickness and mass transfer) were best fit by the model

through a minimization of errors technique.

2. Detachment rate constants for each detachment rate equation were selected to fit steady

state data for the four calcium conditions. Selection was done to minimize errors in the

biofilm thickness and suspended solids measure.

3. Parameters governing starvation were fit to starvation experiments for each calcium

concentration. Different detachment expressions were also fit to the experiments to

evaluate the validity of the proposed mechanism.

5.2.10 Model Solution Implementation Program

The MBBR model was solved using Scilab programming platform. Scilab is an open-

source numerical computing language that has a number of similarities with the commercial

MATlab platform. Scilab was developed and is managed as an initiative of the French National

Institute for Research in Computer Science and Control (INRIA) who host a website at

http://www.scilab.org with information about the language and open source downloads. The
144

model made use of Scilabs ordinary differential equation solver which uses a Backward

Differential Formula method for solving stiff problems.

As noted by the IWA Task Group on Biofilm Modeling (Wanner et al., 2006), the

common platform for evaluating diffusion-reaction biofilm models is AQUASIM, a program

developed to solve biofilm and other aquatic system problems. The model terms used in the

MBBR model could likely be incorporated into AQUASIM as the solution technique (method of

lines, stiff ODE solver) is similar. However, encoding the model in scilab allows complete

control of the calculation technique and has advantages in accessibility as a freely available open

source program.

The Scilab code for the central method in the model is presented in Appendix 3.

5.3. Results and Discussion

5.3.1. Modeling Acclimated MBBRs at Different Calcium Concentrations

To test the general applicability of the MBBR model and to compare the effect of

different calcium concentrations on model parameters, the model was used to predict

performance data and biofilm properties measured in the lab-scale experiments. Figure 5.9

demonstrates measurements and model predictions for the experiments at 4 influent calcium

concentrations after a 100 day acclimation period.


145

0 .7 rdet 0.15Lf X f 0.65


2+ 2+
1 mgCa /L rdet 0.34 L f X f 50 mgCa /L
400 400

Effluent COD and VSS (mg/L)

Effluent COD and VSS (mg/L)


350 350
300 300
250 250
200 200
150 150
100 100
50 50
0 0
0 10 20 0 10 20
Time (d) Time (d)

1000 1000

Biofilm Thickness ( m)
Biofilm Thickness ( m)

800 800
Xf = 30 mg/cm3
Xf = 10 mg/cm3
600 600

400 400

200 200

0 0
0 10 20 0 10 20

Time (d) Time (d)

2
r det 6 L
2+
100 mgCa /L
2+
rdet 0.4 L f X f
200 mgCa /L f X f
400
Effluent COD and VSS (mg/L)

Effluent COD and VSS (mg/L)

400
350
350
300
300
250
250
200
200
150
150
100
100
50
50
0
0
0 10 20
0 10 20
Time (d)
Time (d)

Xf = 15 mg/cm3
1000 1000
Xf = 30 mg/cm3
Biofilm Thickness ( m)

Biofilm Thickness ( m)

800 800

600 600

400 400

200 200

0 0
0 10 20 0 10 20
Time (d) Time (d)

Figure 5.9. Model predictions (lines) and experimental data for effluent VSS ( ), effluent
COD ( ), and biofilm thickness ( ). Simulations used best-fit detachment expressions
and density, X f , listed.
146

Table 5.6. Best-fit parameter values for each detachment term and each calcium case.
2+ 2+ 2+ 2+
1 mgCa /L 50 mgCa /L 100 mgCa /L 200 mgCa /L
Detachment RMSEP* RMSEP* RMSEP* RMSEP*
Parameter Value Value Value Value
Expression X Lf X Lf X Lf X Lf
rdet k dm L f X f kdm 1.1 29.6% 23.6% 0.35 19.3% 20.2% 0.4 36.6% 9.4% 0.5 29.6% 28.0%
2
rdet kdm2 L f X f kdm2 30 29.1% 36.3% 7.5 20.3% 15.6% 9 40.9% 5.4% 6 29.0% 24.55%
rdet k dm L f X f kdm 0.34 0.15 0.13 0.25
28.0% 18.1% 12.9% 10.5% 38.6% 11.9% 27.8% 35.3%
k dg kdg 0.7 0.65 0.7 0.5
* RMSEP denotes the root mean squared error of predictions

Dissolved Oxygen (mg/L) 8


7
6
5
4
3
2
1 Biofilm Surface
0
0 500 1000 1500 2000
Distance from Substratum ( m )

Figure 5.10. Example of model simulation (line) of dissolved


oxygen profile. Measured data ( ) represent average
measurements for 3 or more different positions in the biofilm.
147

The best-fit detachment expressions used in Figure 5.9 are compared along with best-fit

cases derived for each of the investigated terms in Table 5.8. This table also includes a root mean

squared error of prediction (RMSEP) statistic for the suspended solids and biofilm thickness,

respectively. Figure 5.10 demonstrates how the model can simulate dissolved oxygen profiles

through the depth of the biofilm in comparison to microelectrode measurements.

These results indicate that the derived MBBR model is capable of simulating average

values for steady state behaviour including substrate profiles with biofilm depth.

For each calcium concentration an alternate detachment rate term was required in order to

explain experimental biosolids dynamics. However, as can be seen from the RMSEP values, for

most cases there is only a small difference in the predictability of the model which suggests that

any of the proposed detachment rates could be used to simulate steady state. In comparing the

detachment rate constants for each calcium concentration it was found that for a given rate term,

the rate constants were lower for 50, 100 and 200 mgCa2+/L than for the 1 mgCa2+/L. This is

consistent with biofilm areal density measurements taken during the experiment and suggests

that elevated calcium concentrations enhance the cohesive strength of biofilms.

Although run as steady state systems, the MBBRs were subjected to small fluctuations in

the influent flow and substrate concentration. This was due to imperfect control of the peristaltic

feed pump and in formulating the synthetic wastewater. These minor process changes were

accounted for as inputs to the model resulting in the observed fluctuations in predicted steady

state values. It can be seen from Figure 5.8 that the predicted fluctuations in both COD and VSS

are relatively small, representing a well controlled state. In comparison to fluctuations in the

measured data, the predictions were not found to correlate with either the trend or the magnitude
148

of experimental variability. RMSEP calculations for the steady state data suggest that

unexplainable variations range from 12.9-36.6% as deviations from best-fit model predictions.

The inherent variability in the biosolids generated in the MBBR operated under

controlled conditions is an interesting finding, though not unexpected. Other studies have found

that biofilms grown to a mature state experience unpredictable changes in biofilm detachment

caused by sloughing events. For instance Horn et al. (2003) and Telgmann et al. (2004) observed

unpredictable changes in biofilm thickness and suspended solids when biofilms grown in tubular

reactors had been allowed to acclimate for more than 15 days. Morganroth and Wilderer (2000)

proposed the use of a random number generator to modify the specific detachment rate to

generally describe the bounds of this dynamic change. This approach was used by Horn et al.

(2003) to simulate their observations of unpredictable detachment events. The unpredictable

nature of the mature steady state has also been highlighted by Lewandowski et al. (2004), who

through highly controlled replicate flow cell experiments determined that biofilms cease to be

reproducible structural entities after sloughing events begin.

One of the interesting aspects of the current observations in the lab-scale MBBRs is that

suspended solids fluctuations are occurring on a reactor-wide basis. Sloughing events have been

measured to involve large particles (i.e. >1000 cells, Wilson et al. 2004) but the phenomenon is

essentially local, directly impacting only the region of the biofilm where the sloughed particle

detached. Individual sloughing events may cause variability in thickness measured at one

location in the biofilm or for small enough systems where the detached piece impacts the system

in general, such as for flow cells. For larger reactor systems such as the lab-scale MBBRs, it is

expected that sufficient biofilm surface area would be present to average out individual

sloughing events. Thus, the observed inherent variability of biofilm detachment in MBBRs
149

must result from coordinated phenomena occurring across the majority of the biofilms in the

reactor.

Coordinated sloughing on a reactor-wide basis was also described by Horn et al. (2003)

and Telgmann et al. (2004) where fluctuations in biofilm detachment were observed across the

entire length of the tubular reactors studied. They hypothesized that initial sloughing events

create biofilm roughness that weakens the biofilm and causes subsequent detachment events.

This explanation is reasonable for tubular systems since directional flow over the biofilm surface

allows upstream sloughing to modify the hydrodynamics on adjacent sections. The authors

propose that better characterization of biofilm structural properties will provide the mechanistic

link to predicting reactor-wide detachment events. In the case of MBBRs, coordinated

detachment events are not likely to be a result of hydrodynamic changes. A sloughing event on

one of the 1000 carriers is not likely to significantly affect the shear experienced by the rest of

the biofilms. Two more plausible explanations for the inherent variability in MBBR biosolids

dynamics at steady state are the effect of predation by higher organisms and detachment

coordination based on quorom-sensing signals.

Higher organisms were observed in MBBR biofilms for all calcium conditions to varying

extents. The impact of higher organisms such as ciliates, rotifers, and nematodes on biofilm

detachment has not been thoroughly quantified for a range of species, though it has been shown

that these organisms play a significant role in consuming suspended and aggregated bacteria

(Caron, 1987; Eisenmann et al., 1998). Fried et al. (2000) investigated protozoa and metazoa in

MBBR microbial communities and found that even minor process changes could cause dramatic

shifts in composition of these higher organisms. Since the higher organisms exist in a predator-

prey relationship with bacteria, their abundance and impact on MBBR biomass may be
150

inherently cyclical in nature. Furthermore, the reproductive processes for higher organisms may

result in non-linear responses over time. In several of the biofilms examined in this thesis, rotifer

cysts were observed in high abundance. Cysts may remain dormant for some period of time until

environmental and other cues activate a hatching event, a property that could influence

coordinated biofilm detachment. If higher organism population changes are responsible for the

observed fluctuating detachment rate this may prove to be a difficult process to account for

mathematically. Higher organism abundance counts are often highly variable in time and space

(Fried et al. 2000, this work) and the mathematical description of how different species affect

biofilm detachment has yet to be validated.

The variability in biofilm detachment rates could also be caused by biologically induced

detachment coordinated through quorum sensing. Some experimental evidence suggests that

sloughing events can be coordinated by quorum sensing signals following a cell density or

nutrient availability cue (Dow et al., 2003; Rice et al. 2005). In these studies the quorum-sensing

mechanism has been contemplated for a communication between regional cells within one

biofilm. However it is possible that communication could extend to other biofilms within a

reactor system. This would depend on production of signal molecules to a sufficient

concentration in the bulk reactor liquor to coordinate responses between carriers and cause

reactor-wide change. The validity of this explanation also depends on the potential for the

microbial community to generate non species-specific signals that can induce detachment. So far

research describing the effect of quorum-sensing mediated detachment has not been extended to

mixed community wastewater treatment systems.

The importance of either higher organism community dynamics or quorum-sensing

regulation to govern the observed variability in biofilm detachment requires further study. This
151

could be achieved by a measurement campaign to enumerate protozoa and metazoa and testing

for the presence of quorum-sensing signal molecules over a frequent timescale that captures the

observed dynamic changes. Such an approach would be challenging to conduct due to the

abovementioned variability in higher organism enumeration and for the complexities in

characterizing quorum-sensing signal pathways in a mixed community environment (Morgan-

Sagastume et al., 2005).

It is possible that from a practical perspective modeling biofilm detachment based on

such measurements will be too onerous to be useful in applied wastewater treatment systems.

Though future work in this area may reveal the mechanistic relationships, models for MBBR

design and operation may never be able to move beyond a concept of inherent variability in

describing these complex microbiological processes. Even so, the fact that MBBRs are

inherently variable, as shown by the results presented in this work, is an important consideration

for designing solids separation equipment. The variability in detached solids could influence the

type or size of settling equipment selected. This finding is also significant for systems that

operate with solids separating equipment, such as in the pulp mill application studied in this

thesis. Without a settling step, the variations in biofilm detachment cause uncertainty in

predicting BOD removal performance.

5.3.2. The Impact of Surface Area Correction

The rationale for correcting the biofilm surface area to account for carrier geometry and

biofilm growth was provided earlier and it was proposed that such a correction would be

increasingly significant for biofilms of greater thickness. To show how such a correction is
152

significant for the simulations performed, Figure 5.11 compares reactor predictions for the 1

mgCa2+/L case with and without the surface area correction.


Effluent COD and VSS (mg/L)

450
400
350
300
250
200
150
100
50
0
0 5 10 15 20 25
Time (d)

900
800
Biofilm Thickness ( m)

700
600
500
400
300
200
100
0
0 5 10 15 20 25
Time (d)

Figure 5.11. Comparison of model predictions with (dashed lines) and without (solid
lines) the surface area correction for the prediction the 1 mg/L experiment.
Experimental data for effluent VSS ( ), effluent COD ( ), and biofilm thickness ( ).
Simulations used identical parameters except for the surface area correction.
153

As can be seen from this comparison the surface correction improves model fit,

particularly with respect to the biofilm thickness. However the importance of the surface area

correction for modeling carrier-supported reactors must be considered with a heterogeneous

biofilm concept in mind. Biofilm structural heterogeneity can also influence the effective surface

area through the formation of protrusions and pores. Since biofilm structure has been found to be

affected by substrate concentrations it is possible that thicker biofilms undergo adaptive

structural rearrangement to increase mass transfer. Such processes could reduce the effect of

carrier constriction and limit the significance of the surface area correction. However without an

accurate description of the MBBR biofilm structure, the proposed surface area correction is a

good first step to improve the accuracy of MBBR models. Without this explicit relationship it is

likely that the effects of carrier constriction will be embodied in other model parameters such as

kinetic constants. This could cause inconsistencies in comparing between biofilms of different

thicknesses and on different carrier geometries.

5.3.3. Modeling the Response to Starvation

Following model validation using the steady state performance data, 11 hour starvation

experiments were simulated using the best-fit, growth-based detachment terms. These

simulations are shown in comparison to experimental data in Figure 5.12. The figure only

includes starvation experiments for the 50, 100, and 200 mgCa2+/L trials due to the fungal

plugging event in the 1 mgCa2+/L experiment, discussed previously.

The lab-scale data are consistent with observations at the industrial scale and clarifies

how MBBRs respond over the course of starvation and recovery. During the shutdown period,

substrate is rapidly depleted and starvation ensues causing the biomass to become metabolically
154

down-regulated. The model accounts for this through the metabolic state function where the

substrate utilization rate, q, is shifted towards minimum values. Suspended solids continue to be

generated during the shutdown as evidenced by the spike in VSS following flow restart. This is

not unexpected since hydrodynamic shear will continue to be applied to the biofilms and this

observation confirms the need for a thickness-based detachment rate term. If detachment was to

be strictly based on the growth rate, this spike would not be predicted. Following the starvation

period, substrate concentrations (as measured by COD) are found to increase due to incomplete

conversion caused by reduced biomass activity. It takes between 7 and 40 hours for the reactors

to return to maximum substrate utilization rates and the COD spike to diminish to pre-starvation

effluent concentrations. At the same time, VSS values decrease substantially, an observation

confirming that in addition to a dependency on thickness, biofilm detachment is tied to the

metabolic state which is reduced for the starvation recovery period.

By comparing the model predictions to the experimental data it can be concluded that

although certain experimental data points deviate from the model predictions, generally the

trends are well described. Metabolic state function kinetic constants used to fit the starvation

response are shown in Table 5.7 along with the best-fit detachment constants.
155

50 mgCa 2+/L 100 mgCa 2+/L 200 mgCa 2+/L


120 120 160
Effluent Filtered COD (mg/L)

(mg/L)
Effluent Filtered COD (mg/L)

COD (mg/L)
140
100 100
120

Filtered COD
80 80
100
60 80

Effluent Filtered
60 80
60
60
40 40
40
40

Effluent
20 20 20
20
0 0 00
-10 0 10 20 30 40 -10 0 10 20 30 40 -10
-10 0 10
10 20 30
30 40
Time (h) Time (h) Time (h)
Time (h)

500 450 600


600
450 400
500
500

(mg/L)
Effluent VSS (mg/L)

400
Effluent VSS (mg/L)

VSS(mg/L)
350
350 400
300 400
300
250

EffluentVSS
250 300
300
200
200

Effluent
150 200
200
150
100 100
100
50 100
50
0 0 0
0
-10 0 10 20 30 40 -10 0 10 20 30 40 -10 10 30
-10 0 10 20 30 40
Time (h) Time (h) Time (h)
Time (h)

Figure 5.12. Performance data ( ) measured during starvation experiments and model prediction (dotted lines) of the starvation response.
Shaded regions represent period of flow stoppage
156

Table 5.9. Detachment rate and metabolic state constants used in the simulation of
starvation experiments.
Calcium Concentration (mg/L)
50 100 200
Detachment Constants
Thickness k dm 0.15 0.13 0.25
Growth k dg 0.65 0.7 0.5
Starvation Constants
Downshift k sd 3.5 4 4
Upshift k su 1.25 1 1

The simulation of the starvation response suggests both the importance of modeling a

metabolic downshift and coupling the detachment rate to the metabolic state. To clearly

demonstrate the importance of these model specifications, Figure 5.13 presents three alternative

simulations of the 100 mgCa2+/L starvation experiment. One simulation uses the proposed

approach with a metabolic state function and the detachment term rdet kdm L f X f kdg to

achieve a good fit with the data as presented previously. A second simulation employs the

metabolic state function but models detachment as being a thickness-based term rdet k dm L f X f

using a constant derived from the steady state data (k dm = 0.4). The final simulation does not

employ the metabolic state function, keeping q constant at the maximum value. In this last case,

detachment is modeled as both a function of growth rate and thickness using the same

parameters as in Table 5.9.

The alternate simulations demonstrate why constant kinetic parameters and thickness-

based detachment rates are insufficient to capture the response to starvation. Both the metabolic

state function and the detachment term rdet kdm L f X f kdg were required to generate the

observed transient response. For the case without a growth-dependent detachment term, the

shutdown period was predicted to result in a spike in suspended solids much larger than those
157

observed. This was predicted as a result of the higher rate constant of 0.4 which was necessary to

fit steady state data. For the case with no metabolic state function, there was a very limited

response to starvation with small changes in detachment and COD removal. Thus by

implementing an un-adjusted Monod equation the model is not able to simulate trends observed

as a result of transient flow stoppages.

120
Effluent Filtered COD (mg/L)

100

80

60

40

20

0
-10 0 10 20 30 40
Time (h)

900
800
Effluent VSS (mg/L)

700
600
500
400
300
200
100
0
-10 0 10 20 30 40
Time (h)

Figure 5.13. Comparison of model simulations with no metabolic state function


(------), with metabolic state function and a thickness dependant detachment
rate (- -), and with both a metabolic state function and a growth-dependent
detachment rate ( ).
158

Some additional measurements were collected to provide a picture of what is happening

during the starvation period. Biofilm mass before and after starvation was compared to assess the

extent of biofilm detachment. These measurements are shown in Figure 5.14. The changes in

biofilm mass appear small, but they account for significant discharges of suspended solids. For

instance the 0.28 mg/carrier reduction in biofilm mass for the 100 mgCa2+/L case would

theoretically result in an increase in reactor VSS by 126mg/L under a stop flow condition. This

indicates an important feature of MBBRs operated at short hydraulic retention times: that more

than 80% of the biomass may reside in the biofilm and that only minor shifts in biofilm thickness

can create larger fluctuations in suspended solids. These results also demonstrate that the

transient starvation periods investigated do not lead to widespread detachment events that

remove the majority of the biofilm as found in other studies (Gjermansen et al. 2005; Thormann

et al. 2005). In fact, the detachment rates suggested by these biomass changes over the 11 hour

period are similar to those predicted by the detachment constants derived to fit the steady state

data. The discrepancy between this finding and observations made by Gjermansen et al. (2005)

and Thormann et al. (2005) is most likely due to the fact that their experiments were conducted

with pure culture developmental biofilms grown for 16 hours and 4 days respectively. These

biofilms may have quite different properties than the mature biofilms cultivated in the MBBR

and the different findings suggest caution in extrapolating results from simplified experimental

systems to mixed-community wastewater treatment situations.


159

8
7
Biofilm Mass (mg/carrier) 6
5
4
3
2
1
0
50 100 200
Calcium

Figure 5.14. Changes in biofilm mass per carrier before (shaded) and after (white) a
11 hour starvation period.

16
Specific Cellular Activity (gINT/mg hr)

14

12

10

0
50 100 200 Negative Control

Calcium Concentration (mg/L)

Figure 5.15. INT-formazan activity measure for three starvation experiments. Activity
measured immediately prior and post a 11 hour starvation period. Error bars reflect standard
deviation.
160

Since the starvation response was derived based on changes to the metabolic state, a

measure of biofilm activity was tested before and after starvation. Figure 5.16 depicts the

change in biofilm activity as measured through conversion of INT-formazan.

This figure shows small reductions in activity for all reactors following the starvation

period; however variability in the measurements does not make the reductions statistically

significant. If one assumes that the average values reflect the reduction in activity, this finding

does not represent the scale of activity reduction required to simulate the starvation response. In

the model simulations, an 11 hour starvation results in a q value of nearly zero throughout the

biofilm and suspended phase. The smaller activity reduction measured through the INT method

may be due to the metabolic target of the assay. The INT-formazan reaction is catalyzed by the

cells electron transport chain (Blenkinsopp and Lock, 1990), which is only one component of

the metabolic process. It is possible that metabolic down-regulation initiated by starvation

primarily affects other components of cell metabolism without impeding function of the electron

transport chain. In this sense, the conversion of INT by the electron transport chain could be

considered more a measure of viability. Comparing all the measured activities to the negative

control (cells deactivated by formaldehyde) in Figure 5.16, it appears that starvation for the

durations studied in this work does not significantly affect the viability of the MBBR

community.

5.4. Conclusions

The theoretical modeling of MBBR biosolids dynamics supports the following conclusions:
161

Diffusion-reaction biofilm models can be adapted to simulate MBBRs under steady state and

transient conditions. This provides greater mechanistic insight than current models published

in the literature.

Common biofilm detachment expressions proposed in the literature were found to fit average

data for experiments investigating the effect of different calcium concentrations. Calcium

concentrations 50 mg/L and greater were found to have lower specific rates of biofilm

detachment.

Laboratory MBBRs fed a constant synthetic wastewater were found to exhibit biofilm

detachment fluctuations at steady state that signified coordinated reactor-wide phenomena

that could not be explained by influent changes. Dynamic changes in protozoa and metazoa

populations and quorum-sensing regulated detachment were implicated at potential

explanations that require further study.

MBBRs and other biofilm reactors that exploit surface geometries with small internal spaces

should account for surface area reductions caused by growth of biofilms. For the K1 carrier a

correction equation was found to improve model fit.

Transient periods of starvation common in an industrial setting were found to cause poor

treatment performance and fluctuations in suspended solids due to metabolic down-

regulation and growth-dependent biofilm detachment.

A novel approach to modeling metabolic down-regulation due to starvation was proposed for

biofilm systems and implementation allowed for simulation of the transient starvation

response in MBBRs.

Of the biofilm detachment expressions investigated, only expressions that are dependent on

the metabolic state could simulate the response to periods of starvation. This suggests that
162

detachment expressions of the form rdet kdm L f X f kdg are the best approach for

modeling biosolids dynamics in MBBRs.


163

Chapter 6: Conclusions, Recommendations, and Engineering


Significance

6.1 Conclusions

This thesis explores biosolids dynamics in Moving Bed Biofilm Reactors using a range of

mathematical and experimental approaches according to three lines of inquiry. While each

chapter contains detailed conclusions, it is of interest to evaluate how the principal conclusions

drawn test the hypotheses postulated for each line of inquiry. For the first line of inquiry the

questions and hypotheses are given below followed by the responses and supporting conclusions

derived from the research:

Multivariate Statistical Analysis of an Industrial MBBR

Related Questions/Hypotheses

Question

(A) Is multivariate analysis a suitable modeling approach to understand and predict the observed
performance fluctuations at the Irving Pulp & Paper Ltd. mill?

Hypothesis

(A) Multivariate statistical analysis is a useful tool for understanding and predicting performance
behaviour of an industrially applied MBBR

Multivariate analysis was found to be well-suited to the challenge of modeling the

dynamic performance fluctuations of the MBBR at Irving Pulp & Paper Ltd. Models were used

to identify the most significant process variables that influence the MBBR and to predict

treatment performance. Conclusions that support the hypothesis:

The descriptive PLS MBBR model indicated process variables with the highest average

correlation to effluent BOD and TSS.


164

The model was used for diagnostic investigations of specific incidents of decreased

performance to identify process variables that were likely causative.

A predictive PLS model was found to simulate reactor performance with a Root Mean

Squared Error of Prediction (RMSEP) of 14.5% which is an effective predictor in the

context of the average measurement error of the BOD test (9%).

Question

(B) What are the significant variables controlling MBBR performance for this application?

Hypothesis

(B) The model will indicate mill process variables that have a significant impact on the performance
of the MBBR.

The multivariate analysis of the MBBR at Irving Pulp & Paper Ltd. indicated a number of

process variables that influenced treatment performance. These were:

Flow related parameters such as the D 0 stage effluent flow and the hydraulic retention

time. The most significant effect of flow variation was when the reactor was subjected to

flow stoppages for longer than 6 hours due to process shutdowns.

The wood type pulped. The different types (maple, birch, or a softwood blend) impacts

the chemical composition of wastewater and potentially its degradability.

Reactor temperature and pH. Although these variables were held constant through

process control, identified drift (<10%) in temperature/pH due to control errors coincided

with decreased performance.

Residual nitrogen. This variable was measured to assist in control of the nutrient feeding

system. Since nitrogen consumption is dependent on cell growth, high residual nitrogen is

an indication of reactor activity.


165

The Effect of Calcium on MBBR Biofilms

Related Quesitons/Hypotheses

Question

(A) How is MBBR biofilm structure, microbiology and function affected by the concentration of calcium
in wastewater?

Hypothesis

(A) Biofilm structure (thickness, oxygen diffusion profiles, morphology, EPS composition) is affected
by changes in the concentration of calcium.

Biofilm structure was found to be affected by the concentration of calcium in wastewater.

The conclusions that support this hypothesis are:

Above a threshold calcium concentration (in this system some point between 1-50 mg/L)

MBBR biofilms were observed to be thicker and more dense.

Above this threshold, the thicker biofilms were observed to have anoxic zones.

Above a higher threshold value (in this system somewhere between 200 and 300

mgCa2+/L), salt precipitation was observed to accumulate in MBBR biofilms.

Biofilm EPS was found to be enriched in protein at higher calcium concentrations

suggesting that calcium may selectively form electrostatic bridges between nucleic acids.

Question

(A) How is MBBR biofilm structure, microbiology and function affected by the concentration of calcium
in wastewater?

Hypothesis

(B) Biofilm microbiology is affected by changes in the concentration of calcium.


166

Biofilm microbiology was found to be affected by the concentration of calcium. The

conclusions that support this hypothesis are:

Bacterial populations became differentiated according to calcium concentration, and the

proliferation of filamentous organisms was observed in the upper region of biofilms

grown at higher calcium concentrations. This also contributed to observed changes in

biofilm morphology.

Calcium concentrations at 50 mg/L and above led to higher numbers of protozoa and

metazoa, which may influence the performance of MBBRs, particularly with respect to

reduction of poorly-settling planktonic bacteria

Question

(A) How is MBBR biofilm structure, microbiology and function affected by the concentration of calcium
in wastewater?

Hypothesis

(C) MBBR performance is affected by changes in the concentration of calcium.

MBBR treatment performance was found to be enhanced at higher calcium

concentrations. The conclusions that support this hypothesis are:

MBBR treatment performance with respect to COD removal was higher at elevated

calcium concentrations which could be attributed to the observed increases in biofilm

accumulation or changes in microbiology.

Below 50mg/L calcium, suspended biomass was found to settle poorly due to the

abundance of non-flocculated bacteria.


167

A Theoretical MBBR Model of Biosolids Dynamics during Periods of Starvation

Related Questions/Hypotheses

Question

(A) How can theoretical biofilm models be adapted to describe the kinetics and biosolids dynamics
of MBBRs?

Hypothesis

(A) A theoretical diffusion-reaction biofilm model can be adapted to capture the key features of
MBBR behaviour

A biofilm diffusion-reaction model was adapted to describe the MBBR system. The

following conclusions support the tested hypothesis:

The MBBR model was able to simulate the substrate profiles, biosolids dynamics, and

treatment performance of lab-scale MBBRs under both steady state and transient conditions.

Adapting the model to define biofilm surface area as a function of biofilm thickness was

found to improve the model fit.

While the MBBR theoretical model predicted average reactor performance and biosolids

dynamics, in evaluating steady-state data from the lab-scale reactors another important

conclusion was drawn:

Biosolids dynamics and the rate of MBBR biofilm detachment is inherently variable beyond

that which can be predicted by environmental conditions (such as wastewater composition,

temperature, shear) alone. This inherent variability, which has been estimated for the lab-

scale system studied to be between 13 and 37% error compared to theoretical model

predictions, may be explained by processes such as fluctuating protozoa and metazoa

populations and quorum-sensing mediated detachment.


168

Question

(B) Can the detachment process be modelled to simulate steady-state behaviour for different
calcium concentrations?

Hypothesis

(B) Different divalent cation concentrations will affect the detachment rate

The concentration of calcium was found to influence the rate of detachment modeled

with three general detachment expressions proposed in the literature. Conclusions that support

the hypothesis are:

Calcium concentrations 50 mg/L and greater were found to have lower specific rates of

biofilm detachment.

All detachment expressions could be fit to average strady state data for experiments

investigating the effect of different calcium concentrations.

To account for the observed influence of calcium on the detachment rate, each calcium

concentration was best fit by a unique detachment expression.

Question

(C) Can an MBBR model be developed to describe the response to transient periods of starvation?

Hypothesis

(C) The response to starvation can be simulated by the MBBR model

The response to starvation observed at pulp mill application was reproduced in the lab

and sucessfully simulated by the MBBR model. The following conclusions support this

hypothesis:
169

Transient periods of starvation common in an industrial setting were found to cause poor

treatment performance and fluctuations in suspended solids due to metabolic down-

regulation and growth-dependent biofilm detachment.

A novel approach to modeling metabolic down-regulation due to starvation was proposed for

biofilm systems and implementation allowed for simulation of the transient starvation

response in MBBRs.

Of the biofilm detachment expressions investigated, only expressions that were dependent on

the metabolic state could simulate the response to periods of starvation. This suggests that

detachment expressions of the form rdet kdm L f X f kdg are the best approach for

modeling biosolids dynamics in MBBRs.

6.2 Recommendations

The findings of this work further our understanding of biofilms, biofilm wastewater

treatment processes, and the Moving Bed Biofilm Reactor specifically. A number of

recommendations to scientists and engineering practitioners can be made in light of this research:

Transient conditions impact treatment performance of MBBRs and add complexity to

modeling and process control. It is recommended that transient operation be minimized

wherever possible through enhanced process control or equalization tanks in order to

maximize MBBR performance.

Flow stoppage causing microbial starvation is a specific example of transient operation that

has been shown to significantly reduce MBBR BOD removal following recommencement of

flow. It is recommended that flow stoppages be avoided or mitigated wherever possible.

Potential options for mitigating the impact of reactor shutdowns include providing an
170

alternate carbon source to the reactor during shutdown and planning a graduated

recommencement of flow to avoid significant BOD breakthrough while the reactor recovers.

The concentration of calcium in wastewater affects the structure and function of MBBR

biofilms. It is recommended that for wastewaters with calcium concentrations below 50

mg/L, augmentation be considered as a means of improving treatment performance and

enhancing the settleability of detached biomass. It is also recommended that calcium

concentrations greater than 200 mg/L be avoided if possible, due to the potential for

precipitation to occur within the biofilm where carbonate equilibrium may be different than

in the bulk wastewater. Precipitation can lead to inorganic accumulation that reduces the

effective surface area of the MBBR carriers and thus treatment performance.

The enhance biofilm accumulation and larger anoxic regions observed at higher calcium

concentrations suggest a potential role for using calcium to customize biofilms for different

treatment objectives. It is recommended that calcium be considered as a means of increasing

the efficacy of systems that require both aerobic and anoxic conditions such as for

simultaneous nitrification and denitrification.

The enrichment of protein in biofilm EPS when exposed to higher calcium concentrations is

an interesting finding that has significance for future biofilm research. It is recommended that

calcium be considered as a potential control factor for future efforts to extract useable

products from waste biosolids. It is also recommended that divalent cation concentration be

considered when comparing EPS extraction data between different biomass sources.

Theoretical modeling of the MBBR was enhanced by a mathematical description of carrier

geometer to address pore restriction due to growth. Surface area correction of this type has

not been proposed in current MBBR models and it is recommended that this be accounted
171

for. This recommendation will have particular significance for new carrier designs that seek

to maximize specific surface area by reducing the internal voids of the carrier material.

The inherent variability of the MBBR limits the ability of models to predict reactor

behaviour. It is recommended that future work be directed towards mathematical description

of the growth of higher organisms and quorum-sensing mediated detachment. Although

future research may uncover approaches to account for and control these fluctuations, until

this becomes feasible it is recommended that MBBRs be designed and operated with an

expectation of such variability.

It is recommended that biofilm detachment expressions used in theoretical modeling of

MBBRs carry a dependency on the metabolic state or growth rate of the microorganisms.

This work presents a novel approach to modeling starvation of biofilms that allowed for

simulation of lab-scale and industrial data. It is recommended that the metabolic state

function be applied to other situations where starvation is important, such as in fouling

biofilms and sequencing batch reactors. The specific logistic function form of the metabolic

state function has a conceptual basis in the response of a mixed community, however it is

recommended that further work be conducted to tie the metabolic state function to theoretical

quantities that could be independently validated (such as ATP production). This would

provide greater insight into the starvation process.

6.3. Engineering Significance

The findings of this thesis carry engineering significance. Some of the ways in which the

work advances understanding of biofilm wastewater treatment are described below:


172

The use of multivariate statistical analysis for understanding and predicting performance for

MBBR reactors in a dynamic industrial setting has been demonstrated for the first time. The

work completed and published stands as a guide to engineers and researchers who wish to

gain process understanding in complex environments where a priori modeling not feasible.

This analysis has also enhanced the capacity of engineers at Irving Pulp & Paper Ltd. to

understand key variables that affect MBBR performance, allowing them to avoid situations

that cause poor performance. Co-author and mill engineer Jon LeRoy has indicated that

insights gained through the work have reduced the environmental impact of mill operations.

This work has confirmed the importance of divalent cations such as calcium in biofilm

wastewater treatment systems. While there is a small body of research suggesting the

importance of calcium to biofilm structure and function, there has been no evaluation of how

calcium concentration affects MBBRs or other carrier-type treatment systems. The impact of

calcium on biofilm microbiology has also received little attention. The findings from this

work have been submitted for publication and will fill a knowledge gap in the literature.

The work has provided a theoretical modeling approach to MBBRs based on diffusion-

reaction biofilm models. This modeling approach can be more mechanistically descriptive

than current published models and is disseminated in this thesis in a coding platform (Scilab)

that is widely accessible. This work will also be submitted for publication so that other

modeling practitioners can use the unique features of the model (such as the surface area

correction) to investigate MBBR systems.

A novel approach to modeling the response of biofilms to short term starvation has been

proposed that is easily integrated into biofilm models with minimal adjustable parameters.

There is potential for these parameters to be tied directly to internal cell processes to reveal
173

greater insight into starvation. The approach could be used to predict the impact of process

shutdowns on MBBR treatment performance to aid decision making of process engineers.

Since starvation is a critical process to a range of natural and technical biofilm systems the

dynamic simulation tool presented in this thesis could benefit a range of modeling

investigations. Biofouling in industrial/medical systems and sequencing batch reactors are

two scenarios that appear to be particularly relevant.

This work investigated biofilm structure and function in systems that were either applied or

in lab reactors designed with the goal of closely replicating an applied scenario. In doing so,

the findings represent a realistic picture of the significance of the effect of calcium and

periods of starvation in wastewater treatment. For this reason, the thesis extends previous

observations made in more simplified experimental systems and highlights important

considerations that arise in applied systems. For instance, the work adds to evidence that

protozoa/metazoa play an important role in applied biofilm treatment systems. Such

considerations deserve greater attention if advances in biofilm modeling and conceptual

understanding are to be applied to wastewater treatment.


174

Abbreviations

BAS Biofilm Airlift Suspension reactor. A bioreactor with a fluidized bed of biofilm
support material such as sand or granulated activated carbon

BOD Biochemical Oxygen Demand. The amount of dissolved oxygen consumed as a


result of microbial degradation of the organic content of a wastewater.

CLSM Confocal Laser Scanning Microscopy

COD Chemical Oxygen Demand. The amount of dissolved oxygen consumed through
chemical oxidation of the organic content of a wastewater.

DGGE Denaturing Grade Gel Electrophoresis

DLVO Derjaguin, Landau, Verwey, and Overbeek. Authors of a theory of colloidal


interaction in electrolyte solutions.

EPS Extracellular Polymeric Substances

GAC Granulated Activated Carbon

HRT Hydraulic Retention Time. The residence time for a control volume to pass
through the reactor

ICP-AES Inductively Coupled Plasma Atomic Emission Spectroscopy

INT Iodonitrotetrazolium. Reagent used in activity analysis

MBBR Moving Bed Biofilm Reactor

MCRT Mean Cell Residence Time. The theoretical average time for a microbial cell to
move through a bioreactor

PCA Principal Component Analysis. Multivariate statistical technique

PLS Partial Least Squares. Multivariate statistical technique

RBC Rotating Biological Contactor. A biofilm reactor with biofilm on rotating disks
partially submerged in a liquid phase.

SEM Scanning Electron Microscopy

TSS Total Suspended Solids. Total dry mass of filterable solids in a wastewater
175

VSS Volatile Suspended Solids. Dry mass of filterable solids that does not remain after
heating to 550oC
176

References

Ahimou, F., Semmens, M.J., Novak, P.J., and Haugstad, G. (2007) Biofilm cohesiveness
measurement using a novel atomic force microscopy methodology. Applied and Environmental
Microbiology. 73 (9) 2897-2904.

Ahl, R.M., Lelknes, T., and degaard, H. (2006) Tracking particle size distributions in a moving
bed biofilm membrane reactor for treatment of municipal wastewater. Water Science and
Technology. 53 (7) 33-42.

Allison, D. G., B. Ruiz, C. San Jose, A. Jaspe, and P. Gilbert. (1998) Extracellular
products as mediators of the formation and detachment of Pseudomonas
fluorescens biofilms. FEMS Microbiology Letters. 167:179184.

Amy, P.S., Pauling, C., and Morita, R.Y. (1983) Starvation-survival processes of a marine
Vibrio. Applied and Environmental Microbiology. 45 (3) 1041-1048.

Applegate, D.H. and Bryers, J.D. (1991) Effects of carbon and oxygen limitations and calcium
concentrations on biofilm removal processes. Biotechnology and Bioengineering 37, 17-25.

Arvin, E. and P. Harremos. (1990) Concepts and models for biofilm reactor performance. Water
Science and Technology. 22 (1-2) 171-192.

Augustin, J.-C., Rosso L., and Carlier, V. (2000a) A model describing the effect of temperature
history on lag time for Listeria monocytogenes. International Journal of Food Microbiology. 57
(3) 169-181.

Augustin, J.-C., Brouillaud-Delattre, A., Rosso L., and Carlier, V. (2000b) Significance of
Inoculum Size in the Lag Time of Listeria monocytogenes. Applied and Environmental
Microbiology. 66 (4) 1706-1710.

Barampouti, E.M.P., Mai, S.T., and Vlyssides, A.G. (2005) Chemical Engineering Journal. 106,
53-58.

Baranyi, J. and Roberts, T.A. (1995) Mathematics of predictive food microbiology. International
journal of food microbiology. 26 (2) 199-218.

Barraud, N., Hassett, D.J., Hwang, S.-H., Rice, S.A., Kjelleberg, S., and Webb, J.S. (2005)
Involvement of nitric oxide in biofilm dispersal of Pseudomonas aeruginosa. Journal of
Bacteriology. 188 (21) 73447353.

Barthel, K.U. (2005) Volume Viewer. http://rsb.info.nih.gov/ij/plugins/volume-viewer.html.

Bernhardt, J., Weibezahn, J., Scharf, C., and Hecker, M. (2003) Bacillus subtilis During Feast
and Famine: Visualization of the Overall Regulation of Protein Synthesis During Glucose
Starvation by Proteome Analysis. Genome Research. 13: 224-227.
177

Bester, E., Wolfaardt, G., Joubert, L., Garny, K., and Saftic, S. (2005) Planktonic-cell yield of a
pseudomonad biofilm. Applied and Environmental Microbiology. 71 (12) 7792-7798.

Beyenal, H., Lewandowski, Z., and Harkin, G. (2004) Quantifying biofilm structure: facts and
fiction. Biofouling. 20 (1) 1-23.

Biggs, C.A., Ford, A.M., and P.A. Lant. (2001) Activated sludge flocculation: direct
determination of the effect of calcium ions. Water Science and Technology. 43 (11) 75-80.

Bishop, P.L., Zhang, T.C., and Fu, Y.-C. (1995) Effects of biofilm structure, microbial
distributions and mass transport on biodegradation processes. Water Science and Technology. 31
(1) 143-152.

Blenkinsopp, S.A. and Lock, M.A. (1990) The measurement of electron transport activity in
river biofilms. Water Research. 24 (4) 441-445.

Blom, H.A. (1996) Indirect measurement of key water quality parameters in sewage treatment
plants. Journal of Chemometrics. 10, 697-706.

Boles, B.R., Thoendel, M., and Singh, P.K. (2005) Rhamnolipids mediate detachment of
Pseudomonas aeruginosa from biofilms. Molecular Microbiology. 57 (5) 12-10-1223.

Bollmann, A., Schmidt, I., Saunders, A.M., Nicolaisen, M.H. (2005) Influence of starvation on
potential ammonia-oxidizing activity and amoA mRNA levels of Nitrosospira briensis. Applied
and Environmental Microbiology. 71 (3) 1276-1282.

Bott, C. and Love, N.G. (2002) Investigating a Mechanistic Cause for Activated-Sludge
Deflocculation in Response to Shock Loads of Toxic Electrophilic Chemicals. Water
Environment Research. 74 (3) 306-315.

Boyd, A. and Chackrabarty, A.M. (1994) Role of alginate lyase in cell detachment of
Pseudomonas aeruginosa. 60 (7) 2355-2359.

Brading, M.G., Jass, J., and Lappin-Scott, H.M. (1995) Dynamics of bacterial biofilm formation.
In Microbial Biofilms. pp. 46-63. Edited by Lappin-Scott, H.M., Costerton, J.M. Cambridge:
Cambridge University Press.

Briones, A., and Raskin, L. (2003) Diversity and dynamics of microbial communities in
engineered environments and their implications for process stability. Current Opinion in
Biotechnology. 14. 270-276.

Broch-Due, A., Andersen, R., and O. Kristoffersen. (1994). Pilot plant experiences with an
aerobic moving bed biofilm reactor for the treatment of NSSC wastewater. Wat. Sci. Tech. 29 (5-
6) 283-294
178

Brown, M.J. and Lester, J.N. (1980) Comparison of Bacterial Extracellular Polymer Extraction
Methods. Applied and Environmental Microbiology. 40 (2) 179-185.

Bruus, J.H., Nielsen, P.H., and K Keiding. (1992) On the stability of activated sludge flocs with
implications to dewatering. Water Research. 26 (12) 1597-1604.

Bryers, J. D. (1987) Biologically active surfaces: Processes governing the formation and
persistence of biofilms. Biotechnology Progress 3 (2) 57-68.

Bryers, J.D., and W.G. Characklis. (1982) Processes governing primary biofilm formation.
Biotechnol. Bioeng. 24 (11) 2451-2476.

Bura, R., Cheung, M., Liao, B., Finlayson, J., Lee, B.C., Droppo, I.G., Leppard, G.G., and Liss,
S.N. (1998) Composition of extracellular polymeric substances in the activated sludge floc
matrix. Water Science and Technology. 37 (4-5) 325-333.

Canale, R.P. (1973) Experimental and Mathematical Modeling Studies of Protozoan Predation
on Bacteria. Biotechnology and Bioengineering. 15, 707-728.

Caron, D.A. (1987) Grazing of attached bacteria by heterotrophic microflagellates. Microbial


Ecology. 13 (3) 203-213.

Carlsson, H., Aspegren, H., Lee, N., and Hilmer, A. (1997) Calcium phosphate precipitation in
biological phosphorous removal systems. Water Research. 31 (5) 1047-1055.

Chambless, J.D. and Stewart, P.S. (2007) A three-dimensional computer model analysis of three
hypothetical biofilm detachment mechanisms. Biotechnology and Bioengineering. 97 (6) 1573-
1584.

Chang, F.-Y., and Lin, C.-Y. (2006) Calcium effect on fermentative hydrogen production in an
anaerobic up-flow sludge blanket system. Water Science and Technology. 54 (9) 105-112.

Chang, H.T., Rittmann, B.E., Amar, D., Heim, R., Ehlinger, O., and Y. Lesty. (1991) Biofilm
detachment mechanisms in a liquid-fluidized bed. Biotecnology and Bioengineering., 38 (5) 499-
506.

Chaudry, M.A. S., and Beg, S.A.(1998) A review on the mathematical modeling of biofilm
processes: advances in fundamentals of biofilm modeling. Chemical Engineering Technology 21
(9) 701-710.

Charville, G.W., Hetrick, E.M., Geer, C.B., and Schoenfisch, M.H. (2008) Reduced bacterial
adhesion to fibrinogen-coated substrates via nitric oxide release. Biomaterials. 29: 4039-4044.

Chen, X., and Stewart, P.S., (2000) Biofilm removal caused by chemical treatments. Water
Research. 34 (17) 4229-4233.
179

Chen, X., and Stewart, P.S., (2002) Role of electrostatic interactions in cohesion of bacterial
biofilms. Applied Microbiology and Biotechnology 59 718-720.

Choi Y.C., and E. Morgenroth (2003) Monitoring biofilm detachment under dynamic changes in
shear stress using laser-based particle size analysis and mass fractionation. Water Science and
Technology 47 (5) 69-76.

Cooksey, K. E. (1993) Extracellular polymers in biofilms. In: Biofilms-Science and Technology.


Melo, L.F., Bott, T.R., Fletcher, M., and B. Capdeville (eds). Kluwer Academic Publishers. pp
137-47.

Comte, S., Guibaud, G., and Baudu, M. (2006) Relations between extraction protocols for
activated sludge extracellular polymeric substances (EPS) and EPS complexation properties
Part I. Comparison of the efficiency of eight EPS extraction methods. Enzyme and Microbial
Technology 38: 237-245.

Costerton, J.W., Caldwell, D.E., Korber, D.R., and Lappin-Scott, H.M. (1995) Microbial
biofilms. Annual Reviews in Microbiology 49: 711-45.

Costerton, J.W. (2007) The Biofilm Primer, Springer, Berlin, 199 p.

Cousin, C.P. and J.J. Ganczarczyk. (1998) Effects of salinity on physical characteristics of
activated sludge flocs. Water Quality Research Journal of Canada. 33 (4) 565-587.

Davey, M.E., and OToole, G.A. (2000) Microbial biofilms: from ecology to molecular genetics.
Microbiology and Molecular Biology Reviews, 64 (4) 847-867.

Davies, D.G., Chakrabarty, A.M., and Geesey, G.G. (1993) Exopolysaccharide production in
biofilms: substratum activation of alginate gene expression by Pseudomonas aeruginosa. Applied
Environmental Microbiology. 59, 1181-1186

Davies, D.G., Parsek, M.R., Pearson, J.P., Iglewski, B.H., Costerton, J.W., and Greenberg, E.P.
(1998) The involvement of cell-to-cell signals in the development of a bacterial biofilm. Science.
280, 295-298.

de Beer, D., Stoodley, P., Roe, F., and Lewandowski, Z. (1994) Oxygen distribution and mass
transport in biofilms. Biotechnology and Bioengineering. (43) 1131-1138.

de Beer,. D., and Stoodley, P. (1995) Relation between the structure of an aerobic biofilm and
transport phenomena. Water Science and Technology. 32 (8) 11-18.

Derjaguin, B.V. Landau, L. (1941) Theory of the stability of strongly charged lyophobic sols and
of the adhesion of strongly charged particles in solutions of electrolytes, Acta Physico Chemica
URSS 14:633.
180

Dignac, M.-F., Urbain, V., Rybacki, D., Bruchet, A., Snidaro, D., and Scribe, P. (1998) Chemical
description of extracellular polymers: implication on activated sludge floc structure. Water
Science and Technology. 38 (8-9) 45-53.

Dominguez, D.C. (2004) Calcium signaling in bacteria. Molecular Microbiology 54 (2) 291297.

Dow, J.M., Crossman, L., Findlay, K., He, Y.-K., Feng, J.-X., and Tang, J.-L. (2003) Biofilm
dispersal in Xanthomonas campestris is controlled by cell cell signaling and is required for full
virulence to plants. Proceedings of the National Academy of Sciences. 100 (19) 1099511000.

Dubois, M.K., Gilles, K.A., Hamilton, J.K., Rebers, P.A., and Smith, F. (1956) Colorimetric
Method for Determination of Sugars and Related Substances. Analytical Chemistry. 28 (3) 350-
356.

Eastment, H. and W. Krzanowski. (1982) Crossvalidatory choice of the number of components


from a principal component analysis. Technometrics. 24, 73-77.

Eberl, H.J., Picioreanu, C., Heijnen, J.J., and van Loosdrecht, M.C.M. (2000) A three-
dimensional numerical study on the correlation of spatial structure, hydrodynamic conditions,
and mass transfer and conversion in biofilms. Chemical Engineering Science. 55, 6209-6222.

Eisenmann, H., Harms, H., Meckenstock, R., Meyer, E.I., and Zehnder, A.J.B. (1998) Grazing of
a tetrahymena sp. on adhered bacteria in percolated columns monitored by in situ hybridization
with fluorescent oligonucleotide probes. Applied and Environmental Microbiology. 64 (4) 1264-
1269.

Embley, D. (2001) Moving bed bio-reactor pilot system at Irving Pulp & Paper, Ltd.
International Environmental, Health & Safety Conference and Exhibit, Charlotte, NC, United
States, Apr. 22-25, pp 146-154.

Eriksson, L., and B. Alm. (1991) Study of flocculation mechanisms by observing effects of a
complexing agent on activated sludge properties. Water Science and Technology. 24 (7) 21-28.

Eriksson, L., Hagberg, P., Johansson, E., Rnnar, S., Whelehan, O., Astrm, A., and Lindgren, T.
(2001) Multivariate process monitoring of a newsprint mill. Application to modeling and
predicting COD load resulting from de-inking of recycled paper. Journal of Chemometrics. 15,
337-352.

Firenci, T. (1999) Growth of bacterial cultures 50 years on: towards an uncertainty principle
instead of constants in bacterial growth kinetics. Research in Microbiology 150: 431-438.

Flemming, H.-C., Wingender, J., Moritz, R., Borchard, W., and C. Mayer. (1999) Physico-
chemical properties of biofilms a short review. In: Biofilms in the aquati environment, Keevil,
C.W., Godfree, A.F., Holt, D.M., and C.S. Dow (eds). Royal Society of Chemistry. pp 1-12.
181

Flemming, H.-C. and J. Wingender (2001a) Relevance of microbial extracellular polymeric


substances (EPSs) part 1: structural and ecological aspects. Water Science and Technology. 43
(6) 1-8.

Flemming, H.-C. and J. Wingender. (2001b) Relevance of microbial extracellular polymeric


substances (EPSs) part 2: technical aspects. Water Science and Technology 43 (6) 9-15.

Flemming, H.-C., Neu, T.R., and Wozniak, D.J. (2007) The EPS matrix: the house of biofilm
cells Journal of Bacteriology. 189 (22) 7945-7947.

Fouad, M. and Bhargava, R. (2005) A simplified model for the steady-state biofilm-activated
sludge reactor. Journal of Environmental Management. 74, 245-253.

Fried, J., Mayr, G., Berger, H., Traunspurger, W., Psennere, R., and Lemmerf, H. (2000)
Monitoring protozoa and metazoa biofilm communities for assessing wastewater quality impact
and reactor up-scaling effects. Water Science and Technology

Frlund, B., Palmgren, R., Keiding, K., and Nielsen, P.H. (1996) Extraction of extracellular
polymers from activated sludge using a cation exchange resin. Water Research. 30 (8) 1749-
1758.

Fryer, H.J., Davis, G.E., Manthorpe, M., and Varon, S. (1986) Lowry protein assay using an
automatic microtiter plate spectrophotometer. Analytical Biochemistry. 153, 262-266.

Geladi, P. and Kowalski, B.R. (1986) Partial least squares regression: a tutorial. Analytica
Chemica Acta.185, 1-17.

Galil, N.I., Schwartz-Mittelman, A., and Saroussi-Zohar, O. (1998) Biomass deflocculation and
process disturbances exerted by phenol induced transient load conditions. Water Science and
Technology. 38 (8-9) 105-112.

Garney, K., Horn, H., and Neu, T.R. (2008) Interaction between biofilm development, structure
and detachment in rotating annular reactors. Bioprocess and Biosystems Engineering. 31, 619-
629.

Gaul, T., Mrker, S., and Kunst, S. (2005) Start-up of moving bed biofilm reactors for
deammonification: the role of hydraulic retention time, alkalinity and oxygen supply. Water
Science and Technology.

Gieseke, A., Purkhold, U., Wagner, M., Amann, R., and Schramm, A. (2001) Community
structure and activity dynamics of nitrifying bacteria in a phosphate-removing biofilm. Applied
and Environmental Microbiology 67 (3) 1351-1362.

Gjaltema, A., Arts, P.A.M., van Loosdrecht, M.C.M., Kuenen, J.G., and Heijnen, J.J. (1994)
Heterogeneity of biofilms in rotating annular reactors: occurrence, structure, and consequences.
Biotechnology and Bioengineering, 44, 194-204.
182

Gjaltema, A., Tijhuis, L., van Loosdrecht, M.C.M. and Heijnen, J.J. (1995) Detachment of
biomass from suspended nongrowing spherical biofilms in airlift reactors. Biotechnology and
Bioengineering. 46 (3) 258-269.

Gjermansen, M., Ragas, M.P., Sternberg, C., Molin, S., and Tolker-Nielsen, T. (2005)
Characterization of starvation-induced dispersion in Pseudomonas putida biofilms.
Environmental Microbiology. 7 (6), 894904.

Hall-Stoodley, L. and Stoodley, P. (2002) Developmental regulation of biofilms. Current


Opinion in Biotechnology.13 228233.

Hall-Stoodley, L., Costerton, J.W., and P. Stoodley (2004) Bacterial biofilms: from the natural
environment to infectious diseases. Nature Reviews: Microbiology. 2 (2) 95-108.

Hammer, B.K., and Bassler, B.L. (2003) Quorum sensing controls biofilm formation in
Vibrio cholerae. Molecular Microbiology. 50 (1) 101114.

Harremoes, P., Jansen, J.C., and Kristensen, G.H. (1980) Practical problems related to nitrogen
bubble formation in fixed film reactors. Progress in Water Technology. 12, p 253-269, 1980.

Haznedaroglu, B.Z., Bolster, C.H., and Walker, S.L. (2008) The role of starvation on
Escherichia coli adhesion and transport in saturated porous media. Water Research. 42: 1547-
1554.

Higgins, M.J., Tom, L.A., and Sobeck, D.C. (2004a) Case study i: application of the divalent
cation bridging theory to improve biofloc properties and industrial activated sludge system
performancedirect addition of divalent cations. Water Environment Research. 76 (4) 344-352.

Helness, H. and H. degaard. (2001) Biological phosphorus and nitrogen removal in a


sequencing batch moving bed biofilm reactor. Water Science and Technology. 43 (1) 233-240.

Hermanowicz, S.W. (1998) A model of two-dimensional biofilm morphology. Water Science


and Technology. 37, 219-222.

Higgins, M.J., Sobeck, D.C., Owens, S.J., and Szabo, L.M. (2004b) Case study ii: application of
the divalent cation bridging theory to improve biofloc properties and industrial activated sludge
system performanceusing alternatives to sodium-based chemicals. Water Environment
Research. 76 (4) 353-359.

Higgins, M.J. and J.T. Novak. (1997a) The effect of cations on the settling and dewatering of
activated sludges: laboratory results. Water Environment Research 69, 215-224.

Higgins, M.J. and J.T. Novak. (1997b) Dewatering and settling of activated sludges: the case for
using cation analysis. Water Environment Research. 69, 225-232.
183

Horn, H. and Hempel, D. (1997) Growth and decay in an auto-/heterotrophic biofilm. Water
Research. 31 (9) 2243-2252.

Horn, H., Neu, T.R., and Wulkow, M. (2001) Modelling the structure and function of
extracellular polymeric substances in biofilms with new numerical techniques. Water Science
and Technology. 43 (6) 121-127.

Horn, H., Reiff, H., and Morgenroth, E. (2003) Simulation of Growth and Detachment in Biofilm
Systems Under Defined Hydrodynamic Conditions. Biotechnology and Bioengineering. 81 (5)
607-617.

Howell, J.A., and Atkinson, B. (1976) Sloughing of microbial film in trickling filters. Water
Research. 10, 307-315.

Huang, J., and Pinder, K.L. (1995) Effects of Calcium on Development of Anaerobic Acidogenic
Biofilms. Biotechnology and Bioengineering. 45, 212-218.

Hunt, S.M., Werner, E.M., Huang, B., Hamilton, M.A., and Stewart, P.S. (2004) Hypothesis for
role of nutrient starvation in biofilm detachment. Applied and Environmental Microbiology. 70
(12) 7418-7425.

Huws, S.A., McBain, A.J., and Gilbert, P. (2005) Protozoan grazing and its impact upon
population dynamics in biofilm communities. Journal of Applied Microbiology. 98, 238-244.

Hvala N., Vreko, D., Burica, O., Straar, M., and Levstek, M. (2002) Simulation study
supporting wastewater treatment plant upgrading. Water Science and Technology. 46 (4-5) 325-
332.

Jahren, S. and degaard, H. (1999) Treatment of thermomechanical pulping (TMP) whitewater


in thermophilic (55oC) anaerobic-aerobic moving bed biofilm reactors. Water Science and
Technology. 40 (8) 81-89.

Jahren, S. J. Rintala, J. A., and degaard, H. (2002) Aerobic moving bed biofilm reactor treating
thermomechanical pulping whitewater under thermophilic conditions. Water Research 36 (4)
1067-1075.

Jennings, P.A., Snoeyink, V.L., and Chian, E.S.K. (1976) Theoretical model for a submerged
biological filter. Biotechnology and Bioengineering. 18, 1236-1240.

Kmpfer, P., Weltin, D., Hoffmeister, D., and Dott, W. (1995) Growth requirements of
filamentous bacteria isolated from bulking and scumming sludge. Water Research, 29 (6) 1585-
1588.

Kaplan, J.B., Ragunath, C., Ramasubbu, N., and Fine, D.H. (2003) Detachment of Actinobacillus
actinomycetemcomitans biofilm cells by an endogenous -hexosaminidase activity. Journal of
Bacteriology. 185 (16) 4693-4698.
184

Kaplan, J.B., Ragunath, C., Velliyagounder, K., Fine, D.H., and Ramasubbu, N. (2004)
Antimicrobial Agents and Chemotherapy. 48 (7) 2633-2636.

Keiding, K. and Neilsen, P.H. (1997) Desorption of organic macromolecules from


Activated sludge: effect of ionic composition. Water Research, 31 (7) 1665-1672.

Keller, L., Surette, L.G. (2006) Communication in bacteria: an ecological and evolutionary
perspective. Nature Reviews Microbiology, 4, 249-258.

Kierek, K. and Watnik, P.I. (2003) The Vibrio cholerae O139 O-antigen polysaccharide is
essential for Ca2+-dependent biofilm development in sea water. PNAS 100 (24) 1435714362.

Kinner, N.E. and Curds, R. (1987) Development of protozoan and metazoan communities in
rotating biological contactor biofilms. Water Research 21 (4) 481-490.

Kjelleberg, S., and Givskov, M. (2007) The biofilm mode of life: mechanisms and adaptations.
Horizon Biosciences. Wymondham. 248 p.

Kjelleberg, S., and Molin, S. (2002) Is there a role for quorum sensing signals in bacterial
biofilms? Current Opinion in Microbiology 5 (3) 254-258.

Konopka, A. (2000) Theoretical analysis of the starvation response under substrate pulses.
Microbial Ecology. 38: 321-329.

Konopka, A., Zakharova, T., and Nakatsu, C. (2002) Effect of starvation length upon microbial
activity in a biomass recycle reactor. Journal of Industrial Microbiology & Biotechnology. 29:
286-291.

Krstgens, V., Flemming, H.-C., Wingender, J., and W. Borchard. (2001) Influence of calcium
ions on the mechanical properties of a model biofilm of mucoid pseudomonas aeruginosa. Water
Science and Technology. 43 (6) 49-57

Kourti, T. and MacGregor J.F. (1995) Process analysis, monitoring and diagnosis, using
multivariate projection methods. Chemometrics and Intelligent Laboratory Systems. 28, 3-21.

Kreft, J.-U., Booth, G. and Wimpenny, J.W. T. (1998) BacSim, a simulator for individual-based
modeling of bacterial colony growth. Microbiology 144, 32753287

Kresta, J.V., MacGregor, J.F., and Marlin T.E. (1991) Multivariate statistical monitoring of
process operating performance. The Canadian Journal of Chemical Engineering. 69, 35-46.

Kwok, W.K., Picioreanu, C., Ong, S.L., van Loosdrecht, M.C.M., Ng, W. J., and J.J. Heijnen.
(1998) Influence of biomass production and detachment forces on biofilm structures in a biofilm
airlift suspension reactor. Biotechnology Bioengineering 58 (4) 400-407.
185

LaPara, T.M., Konopka, A., Nakatsu, C.H., and Allenman, J.E. (2000) Effects of elevated
temperature on bacterial community structure and function in bioreactors treating a synthetic
wastewater. Journal of Industrial Microbiology & Biotechnology. 24, 140145.

Laspidou, C. S. and Rittmann, B. E. (2002) A unified theory for extracellular polymeric


substances, soluble microbial products, and active and inert biomass. Water Research 36, 2711
2720

Laspidou, C., and Rittmann, B.E. (2003) Modeling the development of biofilm density including
bacteria, inert biomass, and extracellular polymeric substances. Water Research 38, 3349-3361.

Lappin-Scott, H.M., and Costerton, J.W. (1995) Microbial Biofilms. Cambridge University
Press. Cambridge. 324p.

Lazarova, V. and Manem, J. (1995) Biofilm characterization and activity analysis in water and
wastewater treatment. Water Research 29 (10) 2227-2245.

Lazazzera, B.A. (2000) Quorum sensing and starvation: signals for entry into stationary phase.
Current Opinion in Microbiology. 3:177182.

Lee, N.M. and Welander, T. (1996) Reducing sludge production in aerobic wastewater treatment
through manipulation of the ecosystem. Water Research. 30 (8) 1781-1790.

Lewandowski, Z., Beyenal, H., and Stookey, D. (2004) Reproducibility of biofilm processes and
the meaning of steady state in biofilm reactors. Water Science and Technology. 49 (11-12) 359-
364.

Lewandowski, Z. and Bayenal, H., (2007) Fundamentals of biofilm research. CRC Press. Boca
Raton. 452p.

Li, Z.H., Kuba, T., and Kusuda, T. (2006) The influence of starvation phase on the properties
and the development of aerobic granules. Enzyme and Microbial Technology. 38: 670-674.

Liao, B.Q., Allen, D.G., Droppo, I.G., Leppard, G.G., and S.N. Liss. (2001) Surface properties of
sludge and their role in bioflocculation and settleability. Water Research. 35 (2) 339-350.

Liao, B.Q. and Liss, S.N. (2007) A comparative study between thermophilic and mesophilic
membrane aerated bioreactors. Journal of Environmental Engineering and Science 6: 247-252.

Liu, H. and Fang, H.H.P. (2002) Extraction of extracellular polymeric substances (EPS) of
sludges. Journal of Biotechnology. 95, 249-256.

Liu, Y.-Q., Liu, Y., and J.-H. Tay. (2004) The effects of extracellular polymeric substances on
the formation and stability of biogranules. Applied and Environmental Microbiology 65 (2) 143-
148.
186

Liu, Y.-Q. and Tay, J.-H. (2007) Characteristics and stability of aerobic granules cultivated with
different starvation time. Applied Microbiology and Biotechnology. 75: 205-210.

Lofqvist , A. and Welander, T. (2007) Method for the biological purification of water using a
carrier material. United States Patent #7189323B2. Assignee: AnoxKaldnes AS.

Lowry, O.H., Rosebrough, N.H., Farr, A.L., and Randall, R.J. (1951) Protein measurement with
the folin phenol reagent. Journal of Biological Chemistry. 193, 265-275.

Lynch, M.J., Swift, S., Kirke, D.F., Keevil, C.W., Dodd, C.E.R., and Williams, P. (2002) The
regulation of biofilm development by quorum sensing in Aeromonas hydrophila. Environmental
Microbiology. 4(1), 1828.

Martn-Cereceda, M., Prez-Uz, B., Serrano, S., and Guinea, A. (2002) An integrated approach
to analyse biofilms of a full scale wastewater treatment plant. Water Science and Technology. 46
(1-2) 199-206.

Malmcrona-Friberg, K., Goodman, A., and Kjelleberg, S. (1990) Chemotactic Responses of


Marine Vibrio sp. Strain S14 (CCUG 15956) to Low-Molecular-Weight Substances under
Starvation and Recovery Conditions. Applied and Environmental Microbiology. 56 (12) 3699-
3704.

Malmcrona-Friberg, K., Tunlid, A., Marden, Kjelleberg, S., and Odham, G. (1986) Chemical
changes in cell envelope and poly--hydroxybutyrate during short term starvation of a marine
bacterial isolate. Archives of Microbiology. 144: 340-345.

Mason, C.A., Hamer, G., and Bryers, J.D. (1986) The death and lysis of microorganisms in
environmental processes. FEMS MicrobiologyReviews. 39: 373-401.

Massol-Dey, A.A., Whallon, J., Hickey, R.F., and Tiedje, J.M. (1995) Channel structures in
aerobic biofilms of fixed-film reactors treating contaminated groundwater. Applied and
Environmental Microbiology. 61 (2) 769-777.

Masuko, T., Minamib, A., Iwasakib, N., Majimab, T., Nishimurac, S.-I., and Leea, Y.C. (2005)
Carbohydrate analysis by a phenolsulfuric acid method in microplate format. Analytical
Biochemistry. 339 59-72.

Matsumoto, S., Terada, A., and Tsuneda, S. (2007) Modeling of membrane-aerated biofilm:
Effects of C/N ratio, biofilm thickness and surface loading of oxygen on feasibility of
simultaneous nitrification and denitrification. Biochemical Engineering Journal. 37 98-107.

Matz, C. and Kjelleberg, S. (2005) Off the hook how bacteria survive protozoan grazing.
TRENDS in Microbiology. 13 (7) 302-307.

Maurer, M., Abramovich, D., Siegrist, H., and Gujer, W. (1999) Kinetics of biologically induced
phosphorus precipitation in waste-water treatment. Water Research. 33 (2) 484-493.
187

Mayer, C., R. Moritz, C. Kirschner, W. Borchard, R. Maibaum, J. Wingender, and H. C.


Flemming. (1999). The role of intermolecular interactions: studies on model systems for
bacterial biofilms. International Journal of Biological Macromolecules 26:316.

McLean, K.H., Winson, M.K., Fish, L., Taylor, A., Chhabra, S.R., Camara, M., Daykin, M.,
Lamb, J.H., Swift, S., Bycroft, W.B., Stewart, G.S.A.B., and Williams, P. (1997) Quorum
sensing and Chromobacterium violaceum: exploitation of violacein
production and inhibition for the detection of N-acylhomoserine lactones. Microbiology. 143,
37033711.

Mtris, A., Gerrard, A.M., Cumming, R.H., Weigner, P., and Paca, J. (2001) Modelling shock
loadings and starvation in the biofiltration of toluene and xylene. Journal of Chemical
Technology and Biotechnology. 76: 565-572.

Meyer, B., (2003) Approaches to prevention, removal and killing of biofilms. International
Biodeterioration & Biodegradation 51 (4) 249-253.

Miettinen, T., Hurse, T.J., Connor, M.A., Reinikainen, S-P., and Minkkinen, P. (2004)
Multivariate monitoring of a biological wastewater treatment process: a case study at Melbourne
Waters Western Treatment Plant. Journal of Chemometrics and Intelligent Design. 73, 131-138.

Moreno-Andrade, I., Lopez-Vidal, Y., and Buitrn, G. (2006) Effect of starvation on activity and
viability of Pseudomonas aeruginosa ATCC 10145 degrading 4-chlorophenol. Water Science &
Technology. 54 (10) 163-168.

Morgenroth, E. and P.A. Wilderer. (2000) Influence of detachment mechanisms on competition


in biofilms. Water Research. 34 (2) 417-426.

Morgan-Sagastume, F. and Allen, D.G. (2003) Effects of temperature transient conditions on


aerobic biological treatment of wastewater. WaterResearch. 37 (15) 3590-3601.

Morgan-Sagastume, F. and Allen, D.G. (2005) Activated sludge deflocculation under


temperature upshifts from 30 to 45 C. WaterResearch. 39 (6) 1061-1074.

Morgan-Sagastume, F., Boon, N., Dobbelaere, S., Defroirdt, T., and Verstraete, W. (2005)
Production of acylated homoserine lactones by Aeromonas and Pseudomonas strains isolated
from municipal activated sludge. Canadian Journal of Microbiology. 51, 924-933.

Murga, R., Stewart, P.S., and Dalg, D. (1995) Quantitative analysis of biofilm
thickness variability. Biotechnology and Bioengineering. 45, 503-510.

Muyzer, G., de Wall, E.C., and Uitterlinden, A.G. (1993) Profiling of complex microbial
populations by denaturing gradient gel electrophoresis analysis of polymerase chain
reaction-amplified genes coding for 16S rRNA. Applied and Environmental Microbiology. 59 (3)
695-700.
188

Nakhla, G., and M.T. Suidan. (2002) Determination of biomass detachment rate coefficients in
anaerobic fluidized-bed-GAC reactors. Biotechnology and Bioengineering 80 (6) 660-669.

Nealson K.H., Hastings J.W., (1979) Bacterial bioluminescence: its control and ecological
significance. Microbiological Reviews. 43 (4) 496518.

Neu, T.R., and Lawrence, J.R. (1997) Development and structure of microbial biofilms in river
water studied by confocal laser scanning microscopy. FEMS Microbiology Ecology. 24 91) 11-
25.

Nicolella, C., Chiarle, S., Felice, R.D., and Rovatti, M. (1997) Mechanisms of biofilm
detachment in fluidized bed reactors. Water Science and Technology. 36 (1) 229-235.

Nielsen, P.H., Jahn, A., and Palmgren, R. (1997) Conceptual model for production and
composition of exopolymers in biofilms. Water Science and Technology. 36 (1) 11-19.

Norris, T.B., Wraith, J.M., Castenhotz, R.W., and McDermott, T.R. (2002) Soil microbial
community structure across a thermal gradient following a geothermal heating event. Applied
and Environmental Microbiology. 68 (12) 6300-6309.

Novak, J.T., Higgins, M.J., and N.G. Love. (1999) Closure. Water Environment Research 71 (2)
252-254.

degaard, H., Rusten, B., and H. Badin. (1993) Small wastewater treatment plants based on
moving bed biofilm reactors. Water Science and Technology. 28 (10) 351-359.

degaard, H., Rusten, B., and Westrum, T. (1994) A new moving bed biofilm reactor
applications and results. Water Science and Technology, 29 (10-11) 157-165.

degaard, H. (1995) Method for purification of water. United States Patent #5458779. Assignee:
Kaldnes Miljoteknologi A/S.

degaard, H., Gisvold, B., and Strickland, J.. (2000) The influence of carrier size and shape in
the moving bed biofilm process. Water Science and Technology. 41 (4-5) 383-391.

degaard, H. (2006) Innovations in wastewater treatment: the moving bed biofilm process.
Water Science and Technology. 53 (9) 17-33.

Ohashi, A., and H. Harada. 1994. Characterization of detachment mode of biofilm developed in
an attached-growth reactor. Water Sci. Technol. 30:35-45.

Okabe, S., Hiratia, K., Ozawa, Y., and Watanabe, Y. (1996) Spatial Microbial Distributions of
Nitrifiers and Heterotrophs in Mixed-Population Biofilms. Water Science and Technology. 50,
24-35.
189

Okabe, S., Satoh, H., and Watanabe, Y. (1999) In Situ Analysis of Nitrifying Biofilms as
Determined by In Situ Hybridization and the Use of Microelectrodes. Applied and
Environmental Microbiology. 65 (7) 3182-3191.

Palmgren, R., Nielsen, P.H. (1996) Accumulation of DNA in the exopolymeric matrix of
activated sludge and bacterial cultures. Water Science and Technology. 34 (5-6) 233-240.

Park, C., and Novak, J.T. (2007) Characterization of activated sludge exocellular polymers using
several cation-associated extraction methods. Water Research. 41 (8) 1679-1688.

Parsek, M.R. and Greenberg, E.P. (2005) Sociomicrobiology: the connections between quorum
sensing and biofilms. TRENDS in Microbiology. 13 (1) 27-33.

Patrauchan, M.A., Sarkisova, S., Sauer, K. and Franklin, M.J. (2005) Calcium influences cellular
and extracellular product formation during biofilm-associated growth of a marine
Pseudoalteromonas sp. Microbiology (151) 28852897.

Peyton, B.M. and Characklis, W.G. (1993) A Statistical Analysis of the Effect of Substrate
Utilization and Shear Stress on the Kinetics of Biofilm Detachment. Biotechnology and
Bioengineering. 41 728-735.

Picioreanu, C., Kreft, J.-U. and van Loosdrecht, M.C.M. (2004) Particle-Based Multidimensional
Multispecies Biofilm Model. Applied and Environmental Microbiology. 70 (5) 3024-3040.

Picioreanu, C., van Loosdrecht, M. C. M. & Heijnen, J. J. (1998) A new combined differential
discrete cellular automaton approach for biofilm modeling: application for growth in gel beads.
Biotechnology and Bioengineering 57, 718731

Pizarro, G., Griffeath, D. & Noguera, D. R. (2001) Quantitative cellular automaton model for
biofilms. Journal of Environmental Engineering 127, 782789

Plattes, M., Henry, E., Schosseler, P.M., and Weidenhaupt, A. (2006) Modelling and dynamic
simulation of a moving bed bioreactor for the treatment of municipal wastewater Biochemical
Engineering Journal 32, 6168.

Powell, E.O. (1967) The growth rate of microorganism as a function of the substrate
concentrations. in Microbial Physiology and Continuous Culture. edited by E. D. Powell, C. G.
T. Evans, R. E. Strange, and D. W. Tempest, pages 34-56, Her Majesty's Stationery Office,
London.

Prats, C., Giro, A., Ferrer, J., Lopez, D., and Vives-Rego, J. (2008) Analysis and IbM simulation
of the stages in bacterial lag phase: Basis for an updated definition. Journal of Theoretical
Biology. 252 (1) 56-68.

Ramasamy, P., and Zhang, X. (2005) Effects of shear stress on the secretion of extracellular
polymeric substances in biofilms. Water Science and Technology. 52 (7) 217-223.
190

Reichert, P., AQUASIM - A tool for simulation and data analysis of aquatic systems, Water Sci.
Tech., 30(2), 21-30, 1994.

Rice, S.A, Koh,K.S., Queck, S.Y., Labbate, M., Lam, K.W., and Kjelleberg, S. (2005) Biofilm
Formation and Sloughing in Serratia marcescens are controlled by quorum sensing and nutrient
cues. Journal of Bacteriology. 187 (10) 3477-3485.

Rittmann, B.E. (1982) The effect of shear stress on biofilm loss rate. Biotechnology and
Bioengineering. 24 501-506.

Rittmann, B.E., and P.L. McCarty. (1980a) Model of steady-state-biofilm kinetics.


Biotechnology Bioengineering. 22 (11) 2343-2357.

Rittmann, B.E., and P.L. McCarty. (1980b) Evaluation of steady-state-biofilm kinetics.


Biotechnology Bioengineering. 22 (11) 2359-2373.

Rittmann, B. E. and P. L. McCarty. (2001). Environmental Biotechnology: Principles and


Applications. McGraw-Hill. New York.

Romeo, T. (2006) When the party is over: a signal for dispersal of Pseudomonas aeruginosa
biofilms. Journal of Bacteriology. 188 (21) 7325-7327.

Rosak, D.B., and Colwell, R.R. (1987) Survival strategies of bacteria in the natural environment.
Microbiology Reviews. 51:365-379.

Rosen, C. and Lennox, J.A. (2001) Multivariate and multiscale monitoring of wastewater
treatment operation. Water Research. 35 (14) 3402-3410.

Rusten, B., degaard, H., and Lundar, A. (1992) Treatment of dairy wastewater in a novel
moving bed biofilm reactor. Water Science and Technology 26 (3-4) 703-711.

Rusten, B., Mattson, E., Broch-Due, A., and Westrum, T. (1994) Treatment of pulp and paper
industry wastewaters in novel moving bed biofilm reactors. Water Science and Technology. 30
(3) 161-171.

Rusten, B., Kolkinn, O., and degaard, H. (1997) Moving bed biofilm reactors and chemical
precipitation for high efficiency treatment of wastewater from small communities. Water Science
and Technology. 35 (6) 71-79.

Rusten, B., Johnson, C.H., Devall, S., Davoren, D., and Cashion, B.S. (1999) Biological
pretreatment of a chemical plant wastewater in high-rate moving bed biofilm reactors. Water
Science and Technology. 39 (10-11) 257-264.

Rusten, B., Eikebrokk, B., Ulgenes, Y., and Lyrgen, E. (2006) Design and operation of the
Kaldnes moving bed biofilm reactors. Aquaculture Engineering. 34, 322-331.
191

Sanin, S., Sanin, F. D., and Bryers, J. D. (2003) Effect of starvation on the adhesive properties of
xenobiotic degrading bacteria. Process Biochemistry. 38 (6) 909-914.

Santegoeds, C.M., Schramm, A., and de Beer, D. (1998) Microsensors as a tool to determine
chemical microgradients and bacterial activity in wastewater biofilms and flocs. Biodegradation.
9 159-167.

Sarkisova, S., Patrauchan, M.A., Berglund, D., Nivens, D.E., and Franklin, M.J. (2005)
Calcium-induced virulence factors associated with the extracellular matrix of mucoid
pseudomonas aeruginosa biofilms. Journal of Bacteriology. 187 (13) 4327-4337.

Sawyer, L.K. and Hermanowicz, S. (1998) Detachment of biofilm bacteria due to variations in
nutrient supply. Water Science and Technology. 37 (4-5) 211-214.

Satoh, H., Onoa, H., Rulinb, B., Kamoc, J., Okabed, S., and Fukushia, K.-I. (2004). Macroscale
and microscale analyses of nitrification and denitrification in biofilms attached on membrane
aerated biofilm reactors. Water Research. 38 1633-1641.

Sauer, K., Camper, A.K., Ehrlich, G.D., Costerton, J.W., and Davies, D.G. (2002) Pseudomonas
aeruginosa displays multiple phenotypes during development as a biofilm. Journal of
Bacteriology. 184 (4) 11401154.

Sauer, K., Cullen, M. C., Rickard, A. H., Zeef, L.A.H., Davies, D.G. and Gilbert, P. (2004)
Characterization of nutrient-induced dispersion in Pseudomonas
aeruginosa PAO1 biofilm. Journal of Bacteriology 186 73127326.

Sez, P.B., and B.E. Rittmann. (1992) Accurate pseudoanalytical solution for steady-state
biofilms. Biotechnology and Bioengineering. 39 (7) 790-793.

Schiesser, W.E. (1994) Computational Mathematics in Engineering and Applied Science: ODEs,
DAEs, and PDEs. CRC Press. Boca Raton.

Schramm, A., Larsen, L.H., Revsbech, N.P., and Amann, R.I. (1997) Structure and function of a
nitrifying biofilm as determined by microelectrodes and fluorescent oligonucleotide probes.
Water Science and Technology. 36 (1) 263-270.

Schramm, A., de Beer, D., van den Huevel, J.C., Ottengraf, S., and Amann, R. (1999) Microscale
Distribution of Populations and Activities of Nitrosospira and Nitrospira spp. along a
Macroscale Gradient in a Nitrifying Bioreactor: Quantification by In Situ hybridization and the
Use of Microsensors. Applied and Environmental Microbiology. 65 (8) 3690-3696.

Sekiguchi, Y., Kamagata, Y., Syutsubo, K., Ohashi, A., Harada, H., and Nakamura, K. (1998)
Phylogenetic diversity of mesophilic and thermophilic granular sludges determined by 16S
rRNA gene analysis. Microbiology. 144: 2655-2665.
192

Shanahan, J.W., and Semmens, M.J. (2004) Multipopulation model of membrane-aerated


biofilms. Environmental Science and Technology. 38 (11) 3176-3183.

Smith, R.S. and Iglewski, B.H. () P. Aeruginosa quorum-sensing systems and virulence. Current
Opinion in Microbiology. 6, 56-60.

Sobeck, D.C., and Higgins M.J. (2002) Examination of three theories for mechanisms of cation-
induced bioflocculation. Water Research 36 (3) 527-538.

Speitel, G. E. Jr., DiGiano, F. A. (1987). Biofilm shearing under dynamic conditions. J. Environ.
Eng. 113: 464.

Starkey, M., Gray, K.A., Chang, S.I., Parsek, M.R. (2004) A sticky business: The extracellular
polymeric substance matrix of bacterial biofilms. In: Microbial biofilms. Ghannoum, M. and
G.A. O'Toole (eds). American Society for Microbiology. pp 174-191.

Steinberger, R.E., and Holden, P.A. (2005) Extracellular DNA in single- and multiple-species
unsaturated biofilms. Applied and Environmental Microbiology. 71 (9) 5404-5410.

Stewart, P. S. (1993) A model of biofilm detachment. Biotechnology and Bioengineering. 41


111-117.

Stoodley, P., de Beer, D., and Lewandowski, Z. (1994) Liquid Flow in Biofilm Systems. Applied
and Environmental Microbiology. 60 (8) 2711-2716.

Stoodley, P., Wilson, P., Hall-Stoodley, L., Boyle, J.D., Lappin-Scott, H.M., and J. W.
Costerton. (2001) Growth and detachment of cell clusters from mature mixed-species biofilms.
Applied and Environmental Microbiology 67 (12) 5608-5613.

Suidan, M.T., Ritmann, B.E., and Traegner, U.K., 1987. Criteria establishing biofilm-kinetic
types. Water Research. 21 (4) 491498.

Sutherland, I.W. (2001) Exopolysaccharides in biofilms, flocs and related structures. Water
Science and Technology. 43 (6) 77-86.

Swinnen, I.A.M., Bernaerts, K., Dens, E.J.J., Geeraerd, A.H., and Van Impe, J.F. (2004)
Predictive modelling of the microbial lag phase: a review. International Journal of Food
Microbiology. 94 (2) 137-159.

Tal, Y., Watts, J.E.M., Schreier, S.B., Sowers, K.R., and Schreier, H.J. (2003) Characterization
of the microbial community and nitrogen transformation processes associated with moving bed
bioreactors in a closed recirculated mariculture system. Aquaculture. 215, 187-215.

Tappe, W., Laverman, A., Bohland, M., Braster, M., Rittershaus, S., Groeneweg, J., and van
Verseveld, H.W. (1999) Maintenance energy demand and starvation recovery dynamics of
Nitrosomonas europaea and Nitrobacter winogradskyi cultivated in a retentostat with complete
biomass retention. Applied and Environmental Microbiology. 65 (6) 2471-2477.
193

Telgmann, U., Horn, H., and Morgenroth, E. (2004) Influence of growth history on sloughing an
erosion from biofilms. Water Research. 38 (17) 3671-3684.

Thormann, K.M., Saville, R.M., Shukla, S., and Spormann, A.M. (2005) Induction of rapid
detachment in Shewanella oneidensis MR-1 biofilms. Journal of Bacteriology. 187 (3) 1014
1021.

Tijhuis, L., van Loosdrecht, M.C.M., and J.J. Heijnen. (1995) Dynamics of biofilm detachment
in biofilm airlift suspension reactors. Biotechnology Bioengineering. 45 (6) 481-487.

Tomita, R.K., Park, S.W., and Sotomayor, O.A.Z. (2002) Analysis of activated sludge process
using multivariate statistical tools a PCA approach. Chemical Engineering Journal. 90, 283-
290.

Tripathi C.S. and Allen, D.G. (1999) Comparison of mesophilic and thermophilic aerobic
biological treatment in sequencing batch reactors treating bleached kraft pulp mill effluent.
Water Research. 33 (3) 836-846.

Trulear, M.G. and Characklis, W.G. (1982) Dynamics of Biofilm Processes Journal of the Water
Pollution Control Federation Vol 54, No 9, p 1288-1301

Tsushima, I., Ogasawara, Y., Kindaichi, T., Satoh, H., and Okabe, S. (2007) Development of
high-rate anaerobic ammonium-oxidizing (anammox) biofilm reactors. Water Research. 41
1623-1634.

Turakhia, M.H., Cooksey, K.E., and W.G. Characklis. (1983) Influence of a calcium-specific
chelant on biofilm removal. Applied and Environmental Microbiology. 46 (5) 1236-1238.

Turakhia, M.H., and W.G. Characklis. (1989) Activity of pseudomonas aeruginosa in biofilms:
effect of calcium. Biotechnology and Bioengineering. 33, 406-414.

Urbain, V., Block, J.C., and J. Manem. (1993) Bioflocculation in activated sludge: an analytic
approach. Water Research 27 (5) 829-838.

Valle, A., Bailey, M.J., Whiteley, A.S., and Manefield, M. (2004) N-acyl-L-homoserine lactones
(AHLs) affect microbial community composition and function in activated sludge.
Environmental Microbiology. 6 (4), 424433.

Van Alst, N.E., Picardo, K.F., Iglewski, B.H., and Haidaris, C.G. (2007) Nitrate sensing and
metabolism modulate motility, biofilm formation, and virulence in Pseudomonas aeruginosa.
Infection and Immunity. 75 (8) 3780-3790.

Vandevivere, P. and Kirchman, D.L. (1998) Attachment Stimulates Exopolysaccharide Synthesis


by a Bacterium. Applied and Environmental Microbiology. 59 (10) 3280-3286.
194

van Dongen, G. and Geuens, L. (1998) Multivariate time series analysis for design and operation
of a biological wastewater treatment plant. Water Research. 32 (3) 691-700.

van Loosdrecht, M.C.M., Picioreanu, C., and Heijnen, J.J. (1997) A more unifying hypothesis for
biofilm structures. FEMS Microbiology Ecology. 24 (2) 181-183.

van Loosdrecht, M.C.M., Heijnen, J.J., Eberl, H., Kreft, J. and C. Picioreanu. (2002)
Mathematical modelling of biofilm structures. Antonie van Leeuwenhoek 81 245256.

Verwey E.J.W., Overbeek J.Th.G., Theory of the stability of lyophobic colloids, Elsevier,
Amsterdam (1948)

Vogelaar, J.C.T., De Keizer, A., Spijker, S., and Lettinga, G. (2005) Bioflocculation of
mesophilic and thermophilic activated sludge. Water Research. 39: 37-46.

Wang, X.J., Xia, S.Q., Chen, L., Zhao, J.F., Renault, N.J., and Chovelon, J.M. (2006) Process
Biochemistry. 41, 824-828.

Wanner, O. and Gujer, W. (1986) A multispecies biofilm model. Biotechnology and


Bioengineering. 28 314-328.

Wanner O., Eberl H., Morgenroth, E., Noguera, D., Picioreanu, C., Rittmann, B. van Loosdrecht,
M.C.M., (2006) Mathematical Modeling of Biofilms. IWA Publishing, London, 178pp.

Wanner, O., and Reichert, P. (1996) Mathematical modeling of mixed-culture biofilms.


Biotechnology and Bioengineering. 49, 172-189.

Waters, C.M. and Bassler, B.L. (2005) Quorum sensing: cell-to-cell communication in bacteria.
Annual Review of Cell Development Biology. 21, 319-346.

West, S.A., Griffin, A.S., Gardner, A., and Diggle, S.P. (2006) Social evolution theory for
microorganisms. Nature Reviews Microbiology. 4 597-607.

Wik, T. (2003) Trickling filters and biofilm reactor modelling. Reviews in Environmental
Science and Bio/Technology. 2, 193-212.

Wilson, S., Hamilton, M.A., Hamilton, G.C., Schumann, M.R., and P. Stoodley. (2004)
Statistical quantification of detachment rates and size distributions of cell clumps from wild-type
(PAO1) and cell signaling mutant (JP1) Pseudomonas aeruginosa biofilms. Applied and
Environmental Microbiology 70 (10) 5847-5852.

Wimpenny, J.W.T., and Colasanti, R. (1997) A unifying hypothesis for the structure of microbial
biofilms based on cellular automaton models. FEMS Microbiology Ecology 22 (1) 1-16.

Wimpenny, J., Manz, W., and Szewzyk, U. (2000) Heterogeneity in biofilms. FEMS
Microbiology Reviews 24 (5) 661-671
195

Wingender, J., Neu, T.R. & Flemming, H.-C. (1999) What are bacterial extracellular polymeric
substances? In Microbial Extracellular Polymeric Substances: Characterization, Structure and
Function, pp. 119. Edited by J.Wingender, T.R. Neu and H.-C. Flemming. Berlin, Heidelberg:
Springer-Verlag.

Wirtz, K.W. (2002) A generic model for changes in microbial kinetic coefficients. Journal of
Biotechnology. 97 (2) 147-162.

Wold, S., Esbensen, K. and Geladi, P. (1987) Principal Component Analysis. Chemometrics and
intelligent laboratory systems. 2, 37-52.

Wood, B.D., Ginn, T.R., and Dawson, C.N. (1995) Effect of microbial metabolic lag in
contaminant transport and biodegradation modeling. Water Resources Research. 31 (3) 553-563.

Wright, C.T., and Klaenhammer, T.R. (1983) Influence of calcium and manganese on dechaining
of Lactobacillus bulgaricust. Applied and Environmental Microbiology. 46 (4) 785-792.

Wuertz, S., Okabe, S., and Hausner, M. (2004) Microbial communities and their interactions in
biofilm systems: an overview. Water Science and Technology. 49 (11-12) 327-336.

Xavier, J.B., Picioreanu, C., and van Loosdrecht, M.C.M. (2004) A modeling study of the
activity and structure of biofilms in biological reactors. Biofilms. 1: 377-391.

Zhang, T.C. and Bishop, P.L. (1994) Density, porosity, and pore structure in biofilms. Water
Research. 28 (11) 2267-2277.

Zhang, Y. and Allen, D.G. (2007) Strategies for minimizing deflocculation of biosolids due to
oxygen disturbances Water Science and Technology. 55 (6) 173180.

Zhang, X., Bishop, P.L., and Kinkle, B.K. (1999) Comparison of extraction methods for
quantifying extracellular polymers in biofilms. Water Science and Technology. 39 (7) 211-218.

Zhu, J., Miller, M.B., Vance, R.E., Dziejman, M., Bassler, B.L., and Mekalanos, J.J. (2002)
Quorum-sensing regulators control virulence gene expression in Vibrio cholerae. Proceedings of
the National Academy of Sciences. 99 (5) 31293134.

Zita, A. and M. Hermansson. (1994) Effects of ionic strength on bacterial adhesion and stability
of flocs in a wastewater activated sludge system. App. Environ. Micro. 60 (9), 3041-3048.
196

Appendix 1: Variable Identification for Multivariate Models

Table A1. Description of Variable Identifiers in Multivariate Statistical Analysis Models


1 Woodtype Pulped (Birch, Maple, Softwood) 30 COD total (kg/d) in MBBR Effluent
2 Temperature of MBBR Influent 31 COD soluble (mg/L) in MBBR Effluent
3 pH of MBBR Influent 32 COD soluble (kg/d) in MBBR Effluent
4 Conductivity of MBBR Influent 33 BOD total (mg/L) in MBBR Effluent
5 D0 stage flow 34 BOD total (kg/d) in MBBR Effluent
6 E0p stage flow 35 BOD soluble (mg/L) in MBBR Effluent
7 Ration of D to E 36 BOD soluble (kg/d) in MBBR Effluent
8 Cold water flow 37 COD total removal
9 Total flow 38 COD total loading
10 TSS (mg/L) in MBBR Influent 39 COD soluble removal
11 TSS (kg/d) in MBBR Influent 40 COD soluble loading
12 COD total (mg/L) in MBBR Influent 41 BOD total removal
13 COD total (kg/d) in MBBR Influent 42 BOD total loading
14 BOD total (mg/L) in MBBR Influent 43 BOD soluble removal
15 BOD total (kg/d) in MBBR Influent 44 BOD soluble loading
16 BOD soluble (kg/d) in MBBR Influent 45 Hydraulic Retention Time
17 Flow rate of nutrient feed 46 TSS in MBBR Effluent - TSS in MBBR Influent (mg/L)
18 Nitrogen flow (kg/d) in MBBR Influent 47 TSS in MBBR Effluent - TSS in MBBR Influent (kg/d)
19 Phosphorous flow (kg/d) in MBBR Influent 48 TSS % increase
20 Nitrogen to BOD soluble ratio 49 Yield (TSS / COD total)
21 Phosphorous to BOD soluble ratio 50 Yield (TSS / BOD total)
22 Residual Nitrogen 51 Yield (TSS / COD soluble)
23 Residual Phosphorous 52 Yield (TSS / BOD soluble)
24 Temperature of MBBR Effluent 53 Ratio of COD to BOD in MBBR Influent
25 pH of MBBR Effluent 54 Ratio of COD to BOD removed
26 Coductivity of MBBR Effluent 55 Ratio of COD to BOD in MBBR Effluent
27 TSS (mg/L) in MBBR Effluent 56 Ratio of COD to BOD soluble in MBBR Effluent
28 TSS (kg/d) in MBBR Effluent 57 Airflow to MBBR
29 COD total (mg/L) in MBBR Effluent 58-84 Upstream mill variables for pulping and bleaching
197

Appendix 2: Higher Organism Counts by Mass

Table A2 - Protazoa and metazoa counts for biofilm and suspended biomass referenced to biomass

[Calcium] 1 mg/L 50mg/L 100mg/L 200mg/L


Biofilm1 Suspended2 Biofilm1 Suspended2 Biofilm1 Suspended2 Biofilm1 Suspended2
Ciliates
Stalked 0.5 (1.48) 15.8 (10.0) 22.9 (15.2) 9.0 (12.4) 11.3 (6.4) 19.2 (10.7) 0.1 (0.2) 12.0 (12.8)
Free Swimming N.D. 11.3 (8.9) N.D. 33.0 (22.0) N.D. 4.8 (6.6) N.D. 3.0 (5.5)
Rotifers
Active Rotifers 2.4 (1.4) 4.5 (8.9) 2.5 (1.3) 9.0 (12.4) 5.7 (1.2) 9.6 (10.0) 9.6 (4.4) 78.0 (61.5)
Rotifer Cysts 5.7 (3.0) 15.8 (13.5) 20.6 (10.9) 19.5 (32.0) 13.0 (7.8) 2.4 (5.4) 8.5 (3.9) 27.0 (27.0)
Nematodes 4.5 (4.8) N.D. 1.6 (0.4) 6.0 (9.0) 2.7 (1.6) 4.8 (6.6) 1.9 (1.7) 58.5 (49.5)
Total Organisms 13.3 (8.8) 33.0 (3.2) 47.6 (26.4) 183 (125) 32.8 (15.2) 40.8 (13.7) 20.1 (8.1) 178.5 (112.5)

1 Biofilm counts are presented as counts (x1000) per mg dry weight biofilm mass (standard deviations for 8 measurements)
2 Suspended counts are presented as counts (x1000) per mg dry weight volatile suspended solids (standard deviations for 8 measurements)
N.D. None detected
198

Appendix 3: Scilab code for the theoretical MBBR model

The MBBR model involves several program scripts in the Scilab language that are run
from a central method. The Scilab code for the two principal methods is shown below, with
comments inserted to describe the function of the model. The theoretical basis for the model is
described in Chapter 5. Wanner and Gujer (1986) provides a useful reference with code for a
similar mathematical approach in FORTRAN.

Central Method

//Defining Constant and Model Data

Df = 0.64; //cm2d-1 (diffusion of organics in biofilm)


Dof = 1.75; //cm2d-1 (diffusion of oxygen in biofilm)
q = 8; //d-1
k = 0.01; //mgcm-3
ko2 = 0.0001; //mgcm-3
Xf = 15; //mgcm-3
kb = 0.08;//d-1
ki = 0.04;//d-1
D = 0.8; //cm2d-1 (diffusion of organics in water)
Do = 2.19; // cm2d-1 (diffusion of oxygen in water)
L = 0.01; //cm
areaguess = 5.00; //cm2cm-3, Constant area if
Soguess = 0.017; //mgcm-3, initial value of the bulk substrate concentration = effluent concentration.
Oguess = 0.008; //mgcm-3, initial value of dissolved oxygen in the biofilm. Also the constant value in the
bulk
Fguess=0.2; //unitless, initial fraction of inert material in the biofilm.
Lf = 0.1375; //cm, Initial biofilm thickness
kdet = 0.1; //d-1, Constant specific rate of detachment by mass of biofilm
growdet = 0.7; // Growth-associated detachment rate. Fraction of growth that is detached.
Y = 0.6; //mgXmgS-1
Xbo = 0.270; //mgXbulk/cm3
V = 1850; //cm3 total volume
fill = 1100; //cm3 of carriers per reactor
Scrit = k/10; //critical concentration below which growth is inhibited
qmin = 0.4;//absolute minimum q value
lda = 4; //d-1 rate constant governing switch-off of metabolic function. Used in logistic function in
PDE_1_mod_b
ldb = 0.5; //d-1 rate constant governing switch-on of metabolic function. Used in logistic function in
PDE_1_mod_b
qdown =10; //rate constant for rapid shift of qlim
HRTstead = 3/24; //hr, steady state hydraulic retention time.
Sstead = 0.550; //mg/cm3, steady state substrate concentration entering the reactor
timea = 5; //d, For starvation trials, the start time of starvation. Note, starvation is modelled by setting HRT
to 1000 d for the period
dur = 11/24; //d, For starvation trials, the duration of the starvation
amp = +0.2; //The percent change in feed concentration for square wave substrate profile manipulation

//Time dependant model specifications

Soselector = 0; //use to select type of So function. 1 = square wave, 2 = Sin wave, 3 = imported data, 4 =
starvation, 0 = SS
199

HRTselector = 4; //use to select type of HRT function. 1 = square wave, 2 = sin wave, 3 = imported data,
4 = starvation, 0 = SS
reactor = 2; //use to select data

Acclimate = 0; // use to decide if the inert fraction will be inputted from the last run. 0 = Fguess, 1 =
fractions from last run

//making variables global for use in other methods

global Df
global Dof
global q
global k
global ko2
global Xf
global kb
global ki
global D
global Do
global L
global areaguess
global HRTguess
global Soguess
global kdet
global growdet
global Xbo
global fill
global Soselector
global HRTselector
global reactor
global Fguess
global Scrit
global qmin
global lda
global ldb
global qdown
global HRTstead
global timea
global Sstead
global dur
global amp

//Definition of gridspace for discretization of dz.


//Note that the grids represent the ds/dz fragment while the three other equations represent dL/dt,
dSbulk/dt, dXsusp/dt, dAbulk/dt, dqlimbulk/dt

grids = 20; //0,1,2,3...19,20 = Lf


n= (5*grids) + 6;

global grids;

//Initial conditions

szo = biofilm3(Lf,Xf,q,k,D,L,Df,Soguess,HRTguess,areaguess);// guess function for substrate profile


200

u0(1,1)= Lf;
u0(2:(grids+1),1) = szo(1,:)';
u0((grids+2):((2*grids)+1)) = Oguess;
if (Acclimate == 0) then
u0(((2*grids)+2):(3*grids+1)) = Fguess;
elseif (Acclimate == 1) then
for j=1:grids
u0(j+(2*grids)+1)=F2dprof(j);
end
end
u0((3*grids+2):(4*grids+1),1)= q;
u0((4*grids+2):(5*grids+1),1)= 0;
u0(n-4,1) = q;
u0(n-3,1) = 0;
u0(n-2,1)= Soguess;
u0(n-1,1)= Oguess;
u0(n,1)= Xbo;
u0;

//Definition of the times of integration and the number of time steps


t0=0.0;
tf=24;
timedivs=10000;
tout=linspace(t0,tf,timedivs);
nout=n;

//ODE integration. Stiff integrator. See pde_1_mod for details on generation of ode's

[u]=ode(['stiff'],u0,t0,tout,pde_1_mod_b);

Function Defining Matrix of Discretized PDEs (function: pde_1_mod_b)


//Begin by importing parameters and defining some useful relationships

// Imported parameters

global Df,Dof,q,Xf,k,ko2,D,Do,L,Lf,Y,fill,Soselector,HRTselector;
global kdet,growdet,areaguess,kb,reactor,grids,ki,Scrit,qmin,lda;
global ldb,qdown,HRTstead,timea,Sstead,dur,amp;

//Defining bounds of theta, from 0 to 1

xl=0.0;
xu=1;

// Definition of PDE matrix size

n=length(u);

// The matrix column that represents the end of the MOL points for substrate

nS = grids + 1

//Defining distance between points in terms of theta (dimentionless x) and distance squared
201

dx= (xu-xl)/(grids);

dx2= dx^2;

//Definition of area as a funtion of thickness

r=0.5;
height = 0.7;
loss = 0.005;
Vcarrier = 0.970873786;//cm3/carrier based on packing factor

function a = area(lf)
a = 0.6*(((8*(r-(2*lf))+(2*%pi*(r-2*lf)))*height)+((%pi*r^2)-(%pi*(r-2*lf)^2)))/Vcarrier;
endfunction

//**************************************************************************

//RATE EQUATIONS

//Consumption of limiting substrate


function rS = rateH(s,so,qt)
rS = qt*((s/(k+s))*(so/(ko2+so)));
endfunction

//Growth of active biomass due to consumption of substrate


function uS = growH(s,so,qt)
uS = Y*rateH(s,so,qt);
endfunction

//Endogenous Respiration
function dc = decay(so)
dc = kb*(so/(ko2+so))
endfunction

//Conversion of active biomass to inactive biomass


growI = ki;

//Average specific growth rate


function ga = growAve(s,so,fi,qt)
ga = ((1-fi)*(growH(s,so,qt)-decay(so)-growI))+(fi*growI);
endfunction

//Uati, as defined in model description, chapter 5.

function Ui = Uati(L,i)

for j=1:(i-(2*grids)-1)
uintF(j) = growAve(u(j+1),u(j+grids+1),u(j+(2*grids)+1),u(j+(3*grids)+1));
end;

gridpointsF=linspace(xl,((i-(2*grids)-1)/(grids)),(i-(2*grids+1)));

Ui = L*(inttrap(gridpointsF,uintF));
endfunction
202

//**************************************************************************

//Begin setup of differential equations

for i=1:n

//BIOFILM THICKNESS

//Defining change in Lf (Lf = u(1)), note that j has been adjusted to skip Lf
if(i==1)

//Creating a matrix of substrate rates at each gridpoint for integration to determine the mass flux
for j=1:(grids)
uint(j) = (growAve(u(j+1),u(j+grids+1),u(j+(2*grids)+1),u(j+(3*grids)+1)));
end;

gridpoints=linspace(xl,xu,(grids));

ut(i) = (u(1)*(inttrap(gridpoints,uint)))-(kdet*u(1))-(growdet*(u(1)*(inttrap(gridpoints,uint)))) ;

// ^
// |
//definition of detachment rate with different terms for mass and growth based detachment.
//fraction of growdet will detach that percentage of the biomass growth. Set as 0 to only consider
//specific detachment.

//*****************************************************

//ORGANIC SUBSTRATE

//Defining substratum boundary

elseif(i==2) then ut(i)= -(Xf*rateH(u(i),u(i+grids),u(i+(3*grids))));

//Defining non-boundary MOL points (i=2:n-3) for substrate ds/dt = Df/L * FD2ndorder +
(z*Ul/L)*FD1storder - rateS

elseif [i<(grids+1), i>2]


ut(i)=(Df/((u(1))^2))*((u(i+1)-(2.0*u(i))+u(i-1))/(dx2))+ (((((i-2)*dx)*ut(1))/u(1))*((u(i+1)-u(i-1))/(2*dx))) -
(Xf*rateH(u(i),u(i+grids),u(i+(3*grids))));

//Defining Lf boundary condition

elseif(i==(grids+1))

//Boundary condition is a rearrangement of the finite difference element to meet the condition that
substrate flux
//is the same on both sides of the biofilm water interface.

bc = ((u(1)*D)/(L*Df))*((u(n-2)-u(i)));
sub = (2*dx*bc)+u(i-1);
203

ut(i)=(Df/(u(1)^2))*((sub-(2.0*u(i))+u(i-1))/(dx2))+ (((((i-2)*dx)*ut(1))/u(1))*(bc)) -
(Xf*rateH(u(i),u(i+grids),u(i+(3*grids))));

//****************************************************************************

//OXYGEN SUBSTRATE

//Defining substratum boundary for

elseif(i==(grids+2)) then
ut(i)= -(Xf*(1-Y)*(rateH(u(i-grids),u(i),u(i+(2*grids)))+decay(u(i))));

//Defining non-boundary MOL points (i=2:n-3) for substrate ds/dt = Df/L2 * FD2ndorder +
(z*Ul/L)*FD1storder - rateH
// note that oxygen depletion results from both heterotrophic and endogenous decay terms

elseif [i<((2*grids)+1), i>(grids+2)]


ut(i)=(Dof/((u(1))^2))*((u(i+1)-(2.0*u(i))+u(i-1))/(dx2))+ (((((i-2)*dx)*ut(1))/u(1))*((u(i+1)-u(i-1))/(2*dx))) -
((Xf*(1-Y)*(rateH(u(i-grids),u(i),u(i+(2*grids)))+decay(u(i)))));

//Defining Lf boundary condition

elseif(i==(2*grids+1))

//Boundary condition is a rearrangement of the finite difference element to meet the condition that
substrate flux
//is the same on both sides of the biofilm water interface.
// note that oxygen depletion results from both heterotrophic and endogenous decay terms

bc = ((u(1)*Do)/(L*Dof))*((u(n-1)-u(i)));
sub = (2*dx*bc)+u(i-1);

ut(i)=(Dof/(u(1)^2))*((sub-(2.0*u(i))+u(i-1))/(dx2))+ (((((i-2)*dx)*ut(1))/u(1))*(bc)) -(Xf*(1-Y)*(rateH(u(i-


grids),u(i),u(i+(2*grids)))+decay(u(i))));
// Df/lf2 FD2ndorder (z*ul/L) ds/dz rateS

//**************************************************************************************************************8

//BIOMASS FRACTION FOR INERT BIOMASS

elseif (i==((2*grids)+2))

//Substratum boundary layer

ut(i)=((growI-growAve(u(i-(2*grids)),u(i-grids),u(i),u(i+grids)))*u(i));

elseif[i>((2*grids)+2),i<((3*grids)+1)]

//All other MOL points for f


204

ut(i)=((growI-growAve(u(i-(2*grids)),u(i-grids),u(i),u(i+grids)))*u(i))-(((Uati(u(1),i)-(((i-(2*grids)-
2)*dx)*ut(1)))/u(1))*((u(i+1)-u(i-1))/(2*dx)));
// df/dt = (ui - uave) * Fi - (U - theta* Ul) /Lf * FDfirstorder

//Film water interface boundary. Note that I only changed the final derrivative to be a slope from one side.

elseif (i == ((3*grids)+1))

ut(i)=((growI-growAve(u(i-(2*grids)),u(i-grids),u(i),u(i+grids)))*u(i))-(((Uati(u(1),i)-(((i-(2*grids)-
2)*dx)*ut(1)))/u(1))*((u(i)-u(i-1))/(dx)));

//****************************************************************

//CHANGES IN BIOMASS ACTIVITY

// Matrix for activity values (qmax)

elseif [i >((3*grids)+1), i<(4*grids+2)];

if (u(i-(3*grids))<Scrit) then

qlim = (((u(i-(3*grids)))/Scrit)*(q-qmin))+qmin;
ut(i)= -(lda)*(u(i)-qlim)*(q+0.1-u(i)); //first order reduction in q down to qlim when s falls below scrit

else

ut(i)= ldb*(q-u(i))*(u(i)-u(i+grids)); //Logistic Function

end

// Matrix entries for qlim(t), the limiting value for umax during starvation

elseif [i >((4*grids)+1),i<(5*grids+2)]

if (u(i-(4*grids))<Scrit) then

ut(i)= (qdown*(u(i-grids)-u(i)-0.02))+ut(i-grids)//following function to determine qmax at S>Scrit

else

ut(i)= 0

end
//***************************************************************
//BULK PROPERTIES

// Bulk Activity (n-4 = qmax n-3 = qlim(t), )

elseif (i==n-4)

if (u(n-2)<Scrit) then
205

qlim = (((u(n-2))/Scrit)*(q-qmin))+qmin; //the limiting value of q defined as the ratio of s to Scrit


ut(i)= -(lda)*(u(i)-qlim)*(q+0.1-u(i)); //first order reduction in q down to qlim when s falls below scrit

else

ut(i)= ldb*(q-u(i))*(u(i)-u(i+1)); //Logistic Function

end

elseif (i==n-3)

if (u(n-2)<Scrit) then

ut(i)= (qdown*(u(i-1)-u(i)-0.02))+ut(n-4)//following function to determine qmax at S>Scrit

else

ut(i)= 0

end

//defining change in bulk S organics (n-2)


elseif(i==n-2)
ut(i)=
(1/HRT(t,HRTselector,reactor,HRTstead,timea,dur,amp))*(So(t,Soselector,Sstead,timea,dur,amp,reactor)
-u(i))-(area(u(1))*D*(u(i)-u(grids+1))/L)-((u(n)*rateH(u(i),u(n-1),u(n-4))));

//defining change in bulk Oxygen (n-1)


elseif(i==n-1)
ut(i)= 0;

//defining change in Xsusp (n)


elseif(i==n)
ut(i)= (Xf*area(u(1))*((kdet*u(1))+growdet*(u(1)*(inttrap(gridpoints,uint))))) + (u(i)*growH(u(n-2),u(n-
1),u(n-4))) - (u(i)*kb) - ((1/HRT(t,HRTselector,reactor,HRTstead,timea,dur,amp))*u(i))
// = 40*3*0.03*fill*0.1* = mgdet/d

end;
end;

//Note that ODE solver needs the matrix transposed

ut=ut'

endfunction;

Note that the functions HRT, HRTdata, So, and Sodata are simple methods describing the
input of experimental data or specified input conditions, such as for reproducing the starvation
period. They are not included for brevity.

S-ar putea să vă placă și