PREDICTING THE DISINFECTION EFFICIENCY RANGE IN CHLORINE

CONTACT TANKS THROUGH A CFD-BASED APPROACH

Athanasios Angeloudis1,2, Thorsten Stoesser1, Roger A Falconer1,

1
Hydro-environmental Research Centre, School of Engineering, Cardiff University, The
Parade, Cardiff, UK
1
angeloudisa@cf.ac.uk

ABSTRACT

In this study three-dimensional computational fluid dynamics (CFD) models, incorporating

appropriately selected kinetic models, were developed to simulate the processes of chlorine
decay, pathogen inactivation and the formation of potentially carcinogenic by-products in
disinfection contact tanks (CTs). Currently, the performance of CT facilities largely relies on
Hydraulic Efficiency Indicators (HEIs), extracted from experimentally derived Residence
Time Distribution (RTD) curves. This approach has more recently been aided with the
application of CFD models, which can be calibrated to predict accurately RTDs, enabling the
assessment of disinfection facilities prior to their construction. However, as long as it
depends on HEIs, the CT design process does not directly take into consideration the
disinfection biochemistry which needs to be optimized. The main objective of this study is to
address this issue by refining the modelling practices to simulate some reactive processes of
interest, while acknowledging the uneven contact time stemming from the RTD curves.
Initially, the hydraulic performances of seven CT design variations were reviewed through
available experimental and computational data. In turn, the same design configurations were
tested using numerical modelling techniques, featuring kinetic models that enable the
quantification of disinfection operational parameters. Results highlight that the optimization
of the hydrodynamic conditions facilitates a more uniform disinfectant contact time, which
correspond to greater levels of pathogen inactivation and a more controlled by-product
accumulation.

Keywords: Disinfection Tanks, RANS Modeling, Disinfection by-products, Pathogen

Inactivation, Residence Time Distribution

1. INTRODUCTION

Disinfection is a process designed for the inactivation of pathogenic micro-organisms, thus

preventing the transmission of waterborne diseases. Disinfection of water supply occurs
through contact with suitable dose concentrations of disinfectant, and for sufficient time for
micro-organisms to be inactivated, in appropriately designed contact tanks (CTs). Chlorine
contact tank units suggest plug flow to be the optimal hydrodynamic condition at which
disinfection performance is maximized (Stamou, 2002; Falconer and Tebbutt, 1986; Marske

1
and Boyle, 1973). Under such flow conditions, disinfectant transport becomes ideal by
remaining in the tank for a uniform duration, whilst achieving the desired disinfection.

However, previous studies (e.g. Teixeira, 1993; Shiono and Teixeira, 2000) indicate that flow
exhibits a residence time distribution (RTD) which can be significantly different from that
dictated by plug flow. The shape of the tracer RTD curves can provide an insight into the
hydrodynamic and mixing conditions, as explained by Levenspiel (1999). Digression from
plug flow can be attributed to complex hydrodynamic processes, such as short-circuiting and
recirculation zone formation (Kim et al., 2010). Short-circuiting occurs when particles pass
through a reactor quicker than the theoretical hydraulic residence time (T). Recirculation
zones not only promote short-circuiting (since they occupy a considerable part of the tank
volume) but they also trap solutes and particles (or pathogens), which are then retained in the
tank for a longer period than T. The occurrence of such flow patterns has a detrimental effect
on the overall CT efficiency, because the exposure of pathogens with the disinfectant is either
too short (insufficient treatment) or too long, which can result in excessive disinfection by-
products.

Computational Fluid Dynamics (CFD) techniques have been implemented widely to simulate
flow conditions and mixing processes during the operation of CT facilities (Rauen et al.,
2012; Salzano and Gualtieri, 2014). Such computational models are normally set-up to firstly
reproduce available tracer experimental results and subsequently predict hydraulic efficiency
indicators (HEIs) for proposed CT designs (Kim et al., 2013a; Kim et al., 2013b, Zhang et al.,
2013a; Amini et al., 2011; Gualtieri, 2007; Gualtieri, 2006). Unfortunately, HEIs cannot
indicate disinfection specific parameters, such as optimum disinfectant dosage, pathogen
survival level or by-product formation potential, i.e. invaluable information for the operation
of CTs which are often determined empirically. Such parameters could be potentially
deduced through the integration of disinfection kinetics into computational models, a practice
which was scarcely reported in the literature until recently (Angeloudis et al., 2014a; Zhang
et al., 2014; Wols et al., 2010; Wang et al.,2003; Zhang et al., 2000).

Results are presented herein of 3D numerical model simulations of tracer transport and
disinfection processes in seven different design configurations of a small-scale contact tank
model. The core objectives of this investigation were to:

a) review the hydraulic efficiency of each individual design computationally and compare
the predictions against available experimental data,
b) study a range of inlet and baffle configurations for their capability in inactivating
pathogens and producing by-products under the same disinfection conditions, and
c) examine the by-product concentration and pathogen survival ratio range associated with
each design, based on their RTD.

2
2. RESEARCH METHODOLOGY
2.1 Contact Tank Configurations and Experimental Data

The contact tank configurations modelled were based on a small-scale disinfection tank,
which was studied at Cardiff University, which was 3.0 m long, 2.0 m wide and 1.2 m deep
(Figure 1). The baffling configuration inside the prototype was particularly flexible as shown
by Figure 1(a-b), where the internal baffle arrangement could be altered with relative ease. In
addition, the physical model featured two inlet options, an open channel entry or a pipe inlet.
The channel inlet was located in the northeast corner of the tank and consisted of an open
channel of width Wc = 0.365 m and depth Hi of 0.30 m, which is approximately 1/4 of the
tank depth. Honeycomb flow straighteners were located in the approach channel to promote
uniformity as the flow entered the system. In line with the centre of the shallow approach
channel, there was a 152 mm internal diameter plastic inlet pipe as indicated in the cross
section of Figure 1(c). A tracer injection mechanism was incorporated by Rauen (2005),
which consisted of a diffuser, a control valve with a connection for a syringe and three
injection needles placed around the inlet pipe. The water level, controlled via a rectangular
sharp crested weir at the outlet, was measured to be at Ht = 1.02 m during experimentation
(Angeloudis et al., 2014b).

Figure 1 (a) Schematic view of laboratory model tank configuration for CT1-C indicating the approach channel inlet;
(b) Alternative baffle configurations examined experimentally and computationally(MS1-P,MS2-P,MS3-P,MS4-P
and MS4-C); and (c) CT1-P compartment 1 cross-section showing the pipe inlet setup (dimensions in mm).

3
 The seven designs of CT1-C, CT1-P, MS1-P, MS2-P, MS3-P, MS4-C and MS4-P (Figure 1)
were selected due to the availability of experimental data from previous studies, thus enabling
the validation of hydrodynamic and solute transport simulations. CT1-C was an 8
compartment model which exhibited the standard features of a conventional baffled contact
tank, i.e. it was separated into a certain number of compartments of equal volume where the
flow meandered due to the baffles being arranged in an alternating fashion. CT1-P shared the
same baffle configuration, but the flow entered the system by means of the pipe instead of the
channel as indicated by Figures 1(a) and 1(c).The MS3-P setup was a highly inefficient
design paradigm as it featured no baffles to neutralize the inflow three-dimensionality
towards plug flow conditions. The MS1-P and MS2-P were combinations between the MS3-P
and CT1 baffle configurations to demonstrate how the gradual addition of baffles influenced
the disinfection efficiency (Figure 1(b)). In these three cases, only the pipe inlet was imposed.
The MS4 (i.e. both MS4-C and MS4-P) configurations on the other hand were characterized
by an optimized baffling configuration, which had been shown to outperform the hydraulic
efficiency results of CT1 (i.e. CT1-C and CT1-P).

Table 1 Contact tank model operational parameters during experimentation.

Theoretical
CT Flow rate Q Tank Capacity
Retention Time
Setup (l/sec) (m3)
(min)
CT1-C 4.72 5.98 21.11
CT1-P 3.70 5.98 26.93
MS1-P 3.70 6.06 27.29
MS2-P 3.70 6.02 27.11
MS3-P 3.70 6.12 27.56
MS4-C 3.70 5.99 26.98

All experimental data were acquired in the investigations of either Rauen (2005) or
Angeloudis et al. (2014b), where an experimental campaign was launched for the
investigation of hydrodynamics and tracer transport using Acoustic Doppler Velocimetry
(ADV) and tracer dye injection techniques. Operative conditions established during
laboratory experimentation for each design configuration are summarised in Table 1 for
completeness. The hydrodynamic measurements of Rauen (2005) focused on the observation
of some hydrodynamic aspects encountered in CT1, MS1-P, MS2-P, MS3-P and MS4,
whereas the measurements of Angeloudis et al. (2014b) aimed at the formation of a
comprehensive data set for the CT1-C design to be used as a benchmark for numerical
modelling studies. Of particular relevance to the present work were the pulse tracer
experiments of both studies, which involved Rhodamine WT injections at the inlet, where
submersible sensors monitored fluorescence levels at designated locations for the production
of normalized RTD curves and the derivation of HEIs.

4
Table 2 Hydraulic Efficiency Indicator Definition

HEI Definition Indicator

t10 Time to passage of 10% tracer mass Short-circuiting

tp Peak concentration time Short-circuiting
tg Mean residence time Short-circuiting
t90 Time to passage of 90% tracer mass Mixing Processes
t90/t10 Morrill Index (Mo) Mixing Processes
σ2 Dispersion Index Mixing Processes

The HEIs defined in Table 2 constituted short-circuiting and mixing indexes, obtained to
evaluate the hydraulic performance of the CT configurations. More information on the
adequacy and derivation of individual HEIs has been discussed further in the studies of
Teixeira and Siqueira (2008), Gualtieri (2010) and Marske and Boyle (1973).

2.2 Hydrodynamics and Solute Transport Modelling

The RANS simulations of the hydrodynamic processes were performed implementing a

finite-volume approach on a structured orthogonal grid. A Semi-Implicit Method for
Pressure-Linked Equations (SIMPLE) was applied, coupling the pressure to the velocity
field, with the inclusion of the standard k-ε turbulence closure model to compute the
Reynolds Stresses. Once the steady state flow field had converged, the transport of scalar
quantities were simulated by solving the three-dimensional Reynolds-averaged advection-
diffusion equation, by adopting the gradient-diffusion hypothesis with a Schmidt number of
0.7. Details of the numerical methods, computational model setup, grid independency study
and validation of the CFD model are not repeated here for brevity and the interested reader is
referred to previously published work (Angeloudis et al., 2014a; Kim et al., 2013).

The geometry and boundary conditions of the CFD simulations were chosen in accordance
with a scenario of the CT disinfection being accommodated with a mean retention time T of
35 min, i.e. a realistic estimate for chlorination at water treatment works. This translates to a
flow rate Q of 2.86 l/s, which was imposed for all of the CT configurations considered in this
study. At the inlet surface, a Dirichlet boundary condition was applied for the velocity and
turbulence parameters. The transversal V and vertical W velocity components were set to zero
and a uniform streamwise velocity U was assumed, with a magnitude dependent on the
disinfection scenario flow rate.

2.3 Disinfection Kinetics in Solute Transport

Reactive simulations were conducted for a chlorine disinfection scenario where, as soon as
chlorine was introduced, it reacted with both organic and inorganic substances, leading to a
process of decay. This decay rate is normally described by means of a first-order kinetic
model. However, previous studies (e.g. Brown et al., 2011) have illustrated that the initial
introduction of chlorine in CT systems is subject to contact with fast-reacting compounds.

5
This is associated with a period of more rapid chlorine consumption, typically within the first
5 minutes of chlorination, and has been reported to correspond to a 37-53% decrease from the
initial dosage concentration (Brown et al., 2010). In order to account for these effects, a
parallel decay model was adopted as a source term for the solute transport of chlorine, given
as:

(1)

where CCl is the chlorine concentration and kb is the disinfectant bulk decay rate. The value of
kb is generally contingent on water quality, as well as the disinfection conditions, and can
therefore vary significantly. In spite of this, a kb value equal to 2.77×10-4s-1 was set for all
simulations, i.e. a realistic estimate of decay rate for raw water in accordance with Brown et
al. (2010). CFR is the concentration of fast reacting compounds and kFR (= 4×10-3s-1) is the
consumption rate of chlorine due to these compounds, respectively. CFR itself is modelled to
follow a first-order decay and becomes negligible after 5 min of contact time, minimizing the
influence of fast reactants for the remainder of the process.

Ultimately, pathogen inactivation is the core objective of disinfection, and a quantitative

indication of the survival level expected across the CT domain could aide geometry
modifications, with a view to disinfection optimization. In addition, it is interesting to see the
effect of the geometry on the inactivation process under identical chlorine disinfection
operational conditions, a scenario which is difficult to accommodate for experimental
investigations. The inactivation rate within the current investigation is incorporated in the
pathogen transport simulation by differentiating the Chick-Watson law as:

(2)

where N is the bacteria population and n is a coefficient of dilution, approximately equal to 1

for chlorine disinfection (Zhang et al., 2000). The decay rate k’ is a complicated function that
is influenced by disinfection parameters such as microbe type, water chemical composition,
disinfectant, temperature and pH. This should be adjusted accordingly for practical situations.
This study considers the inactivation of the Giardia Lamblia protozoa, a particularly chlorine
resistant drinking water pathogen, commonly chosen as an indicator of pollution. For
G.Lamblia, k’ is given the value of 18.4 l mg-1h-1 at 25 oC and a pH value of 7.0 (Johnson,
1998).

Recent concerns over the formation of potentially carcinogenic by-products during

chlorination have led to practises requiring disinfection by-products, such as total
trihalomethanes (TTHM), to be constrained within certain limits (Maximum Contaminant
Levels, MCLs) at the final stage of treatment for water supply. Similarly to the previously
discussed kinetic processes of decay and micro-organism inactivation, a wide range of
mathematical models have been developed to predict the formation of by-products (Sadiq and
Rodriguez, 2004). The accumulation of TTHMs was considered in this study, by including

6
an appropriate model (Singer, 1994) when simulating their transport and formation through
the CT system:
. . . . . .
0.00306 2.6 1 (3)

where TTHM is the total trihalomethane concentration in μg/l, TOC is the total organic
carbon concentration in mg/l, UV254 is the ultraviolet absorbance at 254 nm in cm−1, Te is the
temperature in °C, Br is the bromide ion concentration in mg/l and t is the contact time at the
particular location. Table 3 depicts the water quality parameter input for the simulation of by-
product formation. Fluctuations in the system water quality are unavoidable but these were
not considered since in most cases the water has already undergone several treatment
processes before it is subjected to the disinfectant. Therefore, the incoming water quality was
assumed to remain unaltered during the simulations. Tracer transport was simulated prior to
the disinfection simulation to determine t across the computational domain. It was considered
that even for steady-state conditions t can be variable as dictated by the RTD at each
particular point. Therefore, it was estimated for several RTD landmarks at computational
points, such as for:

i. the passage of 10% tracer mass (t10)

ii. the peak concentration time (tp)
iii. the average residence time (tg)
iv. the passage of 90% tracer mass (t90)

The value of t was not only necessary for the TTHM source term in equation (3), but was also
invaluable to provide an overview of the disinfection progress in the post-processing analysis
expanded below.

Table 3 Water Quality Parameter Input

Parameters Values Units

Total Organic Carbon (TOC) 4.5 mg/l

Ultraviolet Absorbance at 254 nm (UV254) 0.06 cm-1

o
Water Temperature (Te) 25 C

Bromide Ion Concentration (Br) 0.036 mg/l

pH 7 -

With regard to the effects of turbulence on the chemical reactions, it was considered
negligible as the chemical time scales modelled for all the reactive processes were
significantly larger compared to the mixing and turbulence time scales (slow chemistry).
Under this scenario, micro-mixing is fast enough so that the composition variables can be
approximated by their mean values, meaning that the scalar covariance between reactants is
zero. For more information about the closure of chemical source terms the interested reader is
directed to the work of Fox (2003) and Pope (2000).

7
The inlet concentration boundary conditions remained unaltered during the disinfection
simulations, assuming CCl = 2 mg/l, CFR = 1 mg/l and CTTHM = 0 μg/l. For G.Lamblia, instead
of providing an actual concentration, a dimensionless value of 1 was set which represented
the population survival ratio (N/N0) of micro-organisms introduced in the system. In turn, the
simulations were run until steady-state conditions were established for all scalar quantities.
The reaction modelling methodology followed within this work was tailored to the transport
of free chlorine as a disinfectant, G.Lamblia pathogen population and Total THM (TTHM)
by-product transport. The same framework can be considered for other disinfectants,
pathogens or by-products by imposing appropriate kinetic models to the scalar transport
source terms.

3. RESULTS AND DISCUSSION

3.1 Hydraulic Efficiency Assessment of the Contact Tank Configurations

Due to the inlet configurations the flow entered the CT by means of a high momentum jet,
which resulted in a 2D profile at the vertical (x-z) plane downstream of the inlet. The jet was
deflected from the opposite wall and then again from the compartment bed, or the water
surface, based on whether the channel or the pipe option was implemented. In contrast, the
entire contact tank flow was meant to be 2-dimensional in the horizontal (x-y) plane, i.e. the
flow should move horizontally through the tank towards the outlet. Hence there was
significant three-dimensionality close to the inlet, which was where the flow recovered from
2D in the longitudinal, to 2D in the horizontal, plane (Angeloudis et al. 2014b). Depending
on the baffle configuration, this three-dimensionality could be neutralized early (e.g. MS4,
CT1) corresponding to a rapid transition towards more uniform flow conditions, or not, as in
the case of the MS3-P design. The accurate reproduction of these hydrodynamic effects
through numerical model simulations was considered of paramount importance for this
investigation and is reflected here by the agreement with experimentally derived RTDs and
HEIs in Figure 2 and Table 4, which were all obtained from the tracer gauge point at the
outlet.

Figure 2 presents normalized (θ = t/T) outlet RTD predictions produced computationally,

illustrating how the outlet curve shape could be improved once the design was gradually
optimised. The curves were normalised with respect to T, since the experimental flow rates
were different to the one modelled for the disinfection scenario. For the unbaffled model of
MS3-P, the flow was completely three-dimensional, with a very poor hydraulic efficiency as
the peak of the curve appeared unacceptably early (tp = 0.13 θ). The CT1 design contained the
initial three-dimensional flow structure caused by the inflow within the first 3 compartments
(Angeloudis et al., 2014b), rather than allowing a regime of complete mixing to dominate the
entire CT volume. This led to significantly reduced short-circuiting effects (tp = 0.86 θ) and
an improved hydraulic efficiency, as suggested by the shape of the associated RTD curve.
MS4 featured a baffling configuration that attempted to control the three-dimensionality
induced by the inlet water jet, even earlier than that acquired when the jet reached CT1. A
baffle opposite the approach channel confined the vertical recirculation zone to the width of
the first compartment, rather than allowing it to affect the entire length of the CT, in contrast

8
to other configurations. The tank was divided into 6 compartments instead of 8, as the cross-
baffling design enabled longer compartments which occupied more volume, thereby reducing
the compartment number, while preserving the meandering flow structure of a serpentine CT.

Figure 2 Predicted and experimental outlet: (a) RTD (Eθ), and (b) FTC (Fθ) curves for the different configurations
examined experimentally in the laboratory model (Rauen, 2005; Angeloudis et al.,2013).

Table 4 Experimental and Computational HEIs obtained at the outlet for different Contact Tank Designs

Hydraulic Efficiency Indicators

CT
σ2 t10/T tp/T t90/T Mo
Setup
EXP CFD EXP CFD EXP CFD EXP CFD EXP CFD
CT1-C 0.095 0.094 0.7 0.72 0.84 0.86 1.48 1.49 2.12 2.07
CT1-P 0.097 0.092 0.68 0.7 - 0.88 1.45 1.48 2.13 2.11
MS1-P 0.224 0.223 0.5 0.55 - 0.65 1.69 1.7 3.38 3.09
MS2-P 0.306 0.539 0.37 0.38 - 0.39 1.81 2.07 4.89 5.45
MS3-P 0.534 0.769 0.16 0.16 - 0.12 1.84 2.03 11.5 12.69
MS4-C 0.055 0.047 0.78 0.76 - 0.89 1.32 1.38 1.69 1.82
MS4-P 0.052 0.048 0.79 0.75 - 0.88 1.35 1.4 1.71 1.87

The capacity of the numerical models to reproduce the measured data for different baffle
configurations is summarised in Table 4, through comparisons of experimentally and
computationally derived HEIs. Firstly, the short-circuiting indicator t10 was predicted
accurately in all cases and deviations are typically below 5%. In terms of the mixing
indicators, the best agreement was observed for the optimal baffling configuration of MS4,
followed closely by CT1. However, for MS2 and MS3 despite the pronounced estimate of
short-circuiting (t10), the deviation in certain indexes like the dispersion index (σ2) became

9
substantial, a sign that the mixing processes for these geometries were not reproduced as
well.

It is speculated that for CT1, MS1-P and MS4 the initial flow unsteadiness was neutralized by
the baffles, which encouraged flow uniformity early in the flow-path that enabled the steady
state simulation to provide a good approximation of the flow and turbulence field. For MS2-
P and MS3-P no significant measures were taken to mitigate the flow unsteadiness which
corresponded to discrepancies from the steady-state CFD predictions. This also related to the
findings of Kim et al. (2013), where it was argued that employing the gradient diffusion
hypothesis to model an unsteady process through a steady state flow field, led to accurate
predictions of tracer RTD curves, but only for certain flows. In other cases, the turbulent Sc
required calibration, especially for flows that were dominated by large-scale turbulence
structures.

For narrow compartment tanks (e.g. CT1, MS4), the hydraulic efficiency findings were quite
accurate with the standard value of the Sc, whereas Sc had to be decreased substantially to
provide accurate results for wide compartment tanks. Superior approximations for MS2-P and
MS3-P, which featured wider compartments, could therefore have been achieved by a
calibration of the Sc number. Undertaking such investigations was not contemplated further,
following supplementary information provided in Kim et al. (2013b). A different
interpretation of these errors could rely on the inadequacy of the k-ε model to provide a more
accurate closure to the turbulence field in the configurations of MS2-P and MS3-P, due to the
low Reynolds numbers occurring in the large recirculation zones (Figure 3) which dominated
the flow structure. The particular turbulence model is more suitable for highly turbulent flow
regimes, such as the ones developed in the narrow compartment CT configurations, and
perhaps low-Reynolds number k-ε model variants could result in better approximations. An
alternative approach would be the selection of the k-ω model due to its promising
performance when modelling flow separation and recirculation in complex flows (Wilcox,
2004). Nonetheless, the shape of the RTD curves (Figure 2) predicted herein fits well with
the experimental data even for the wide-compartment configurations of MS1-P, MS2-P and
MS3-P.

3.2 Impact of Contact Tank Hydrodynamics on Disinfection Processes

Simulations of the disinfection processes were conducted for each of the CT configurations,
under an assumption that the water in the system was subjected to the same chlorine dosage
and identical kinetic characteristics, thereby enabling the impact of the tank geometry to be
assessed on the process performance. Figure 4 presents contour plots of average contact time,
CCl, G.Lamblia N/N0 and CTTHM at mid water depth, providing an appreciation of the strong
interconnection between the hydrodynamics and chemical reaction kinetics in the CT
configurations.

10
Figure 3 Contour plots of Estimated Average Contact Time, Disinfectant Distribution, G.lamblia Inactivation and
total Trihalomethane formation as predicted from the CFD simulation of the disinfection processes for each of the
CT designs (CT1-C,MS1-P,MS2-P,MS3-P and MS4-C). The plots are produced at half-depth, i.e. z/Ht = 0.50.

11
By observation of the MS1-P, MS2-P and MS3-P contour plots, large low-velocity horizontal
recirculations were formed in the absence of internal baffles. These areas were associated
with high contact times, that corresponded to low pathogen N/N0 and high concentrations for
by-products. On the other hand, the majority of the flow short-circuited around these regions,
which was reflected at the outlet by a low average contact time with lesser by-product
concentrations and greater pathogen survival ratios. Indicatively, the average G. Lamblia
N/N0 at the outlet of MS3-P was predicted at 7.6 × 10-2 whereas in certain internal
recirculation zones it was placed at 2.5 × 10-2. A similar pattern was observed for CTTHM,
which was estimated at 57 μg/l at the outlet and 65 μg/l at certain recirculation zones of the
horizontal plane. Even with more superior baffle configurations (CT1, MS4), some
recirculation was unavoidable. In spite of this, they are less prevalent to the tank volume and
disinfectants were not detained in these for long, due to greater velocity magnitudes, which
led to rapid mixing rather than dead zones. This was demonstrated by the average t contour
plots (Figure 3) of CT1 and MS4, where the average t values calculated in recirculation zones
was not substantially different compared to surrounding regions along the streamwise flow
path.

A holistic appreciation of the impact of flow conditions on kinetic processes can be obtained
from Figure 4 where CCl, N/N0 and CTTHM were plotted as a function of θ for all
computational cells (each data point represents a location in the tank). As expected, a distinct
pattern appeared between the consumption of CCl and the pathogen survival. The more
efficient designs led to a greater disinfectant consumption and a reduced N/N0 at higher θ.
This was in compliance with the theory that superior hydrodynamics facilitate optimized
contact between the disinfectant and reactants, which can either be pathogens or organic
material to produce by-products (Figure 4c). In contrast, for inefficient designs the pathogen
N/N0 predictions were more scattered in the presence of large recirculation zones; a
consequence of uneven mixing with the chlorine. In this manner, poor hydraulic efficiency
increases the uncertainty over the performance of the disinfection process.

Figure 4 Disinfection performance with respect to mean residence time, based on computational cells of CFD
simulations: (a) Chlorine Concentration (CCl), (b) G.Lamblia Survival Ratio (N/N0), and (c)Total Trihalomethane
(TTHM) formation. The figure indicates that despite disinfection occurring under identical conditions of operation
and water quality, the CT geometrical differences have a distinctive impact on the kinetics.

12
The production of TTHMs appeared consistent for all the geometries and seemed to be
almost linearly connected with the residence time at each computational cell (Figure 4c).
However, it should be noted that the current TTHM simulations assumed a constant
concentration of TOC (Table 3) throughout the disinfection reactions. According to Brown et
al. (2010), where THM formation was monitored at water treatment facilities, the by-product
production rate is not constant but flattens out because of the reduced availability of TOC as
it reacts with disinfectants over time. A better understanding of the relationship between
organic material and chlorine could improve the accuracy of these simulations.
Supplementary refinements were not tested within the scope of this investigation, due to the
lack of robust mathematical models linking the organic carbon with free chlorine that could
be integrated in the numerical methodology.

3.3 Effect of Outlet Residence Time Distribution to Reactive Processes

A particular disadvantage with Eulerian CFD simulations is that disinfection-related

processes are calculated based on an averaged residence time (tg), estimated for each
computational cell. Arguably, for a more accurate analysis, the full distribution of chlorine
exposure times should be taken into account to elucidate and quantify the consequences of
either short-circuiting or scalar (e.g. disinfectant) entrapment in the CT geometries. This was
accomplished herein by the examination of the disinfection progress, by analysing the
exposure times of t10, tp and t90, rather than simply providing an indication of the mean
residence time (tg). Through such an analysis it is intended to illustrate the extent to which the
disinfection performance of the CT designs may vary depending on the RTD.

Figure 5 Exposure time fields expected for: (a) t10, (b) tg, and (c) t90, are shown to demonstrate how the contact time
may vary due to the residence time distribution for each computational cell. The plots are obtained at mid-depth
(z/Ht = 0.50) of the CT1-C configuration.

In order to emphasize the significance of the RTD and establish how CFD findings for
reactive processes relying on a mean residence time may be limited, the contour plots of
Figure 5 were produced. Based on the CT1-C simulation, the disinfectant exposure times for
t10, tg and t90 for computational points at mid-depth (z/Ht = 0.50) are shown to exemplify how
much the contact time can deviate from its mean value. The validity of these predictions was
confirmed by the satisfactory agreement against tracer experimental data (Angeloudis et al,

13
2014b), where Rhodamine WT concentration gauging was undertaken at 24 sampling points
inside the physical model to obtain HEIs.

In consideration of the operational specifications of disinfection which aim towards pathogen

inactivation maximization and by-product formation minimization, using the mean residence
time to produce average values result in underestimating the pathogen survival ratio and the
by-product concentration occurrence. This can be justified since the pathogen ratio N/N0 at tg
would naturally be lower at the outlet compared to a scenario where pathogens short-circuit
(e.g. at t10). In a similar manner, CTTHM will be significantly greater at higher residence times
(such as t90) because of the production rate increase across the CT flow, due to the excessive
disinfectant exposure (Equation 3).

Figure 6 Comparison between peak concentration time (tp) and residence time calculated within computational cells.

Another argument which questions the suitability of using the mean residence time to
determine the performance of disinfection stems from its deviation from the tp parameter, i.e.
the most common residence time encountered at sampling points. Figure 6 graphically
demonstrates the correlation between tp and the residence time for the different CT
configurations examined. It was observed that due to the effects of short-circuiting, tp
highlighted the inefficiency of baffle configurations in MS1-P, MS2-P and MS3-P designs as
it was substantially different to tg. Taking MS3-P as an example, the tp residence time at the
outlet gauge point was approximately 4.2 min, while the flow rate in the tank should
accommodate a theoretical contact time of 35 min. The tp value was reported even earlier
than t10 (≈ 5.6 min) for that case. This might seem extraordinary as the mean time tg was
calculated to be considerably higher. This was explained by the averaging of the residence
times which for t90 exceeded 72.5 min. In contrast, the optimisations undertaken for the MS4-
C model resulted in a decrease in the residence time range at the outlet, as indicated by t10 =
26.6, tp = 31.2 and t90 = 48.3 min (normalised values were included in Table 4). These
statistics were meant to quantify the impact of poor hydraulic performance on the disinfectant
exposure uncertainty and how reducing the departure of the RTD from the mean residence
time should be of paramount priority for water treatment facilities.

14
In the study of Greene et al. (2007) it was discussed that the output given by experimental
RTDs is insufficient to calculate the range of disinfection, since the actual mixing state
cannot be discerned directly for the whole geometry. However, the simulation of the
disinfection through a CFD methodology does provide an insight on how the different mixing
regimes affect the kinetics with respect to contact time (Figure 4(b-c)). Figure 4(b) was used
to deduce the range of survival ratio at the outlet for the disinfectant exposures that
correspond to t10, tp, tg and t90. Results estimated for each configuration with regards to the
inactivation of the G. Lamblia protozoa are summarised in Table 5.

Table 5 Outlet survival ratio predictions for t10, tp, tg and t90 as a means to illustrate the variation of the disinfection
efficiency due to the residence time distribution.

G.Lamblia Survival Ratio

CT Setup
t10 N/N0 tp N/N0 tg N/N0 t90 N/N0
CT1-C 0.0166 0.0061 0.0014 0.0001
CT1-P 0.0197 0.0053 0.0014 0.0001
MS1-P 0.0470 0.0168 0.0040 0.0001
MS2-P 0.0905 0.0759 0.0156 0.0003
MS3-P 0.6491 0.6725 0.0757 0.0082
MS4-C 0.0088 0.0032 0.0012 0.0001
MS4-P 0.0091 0.0037 0.0012 0.0001

The most successful design was the MS4-C configuration, closely followed by MS4-P, where
in both cases the disinfection performance, despite short-circuiting (t10) effects, appeared to
be below a 2-log inactivation level (99%), i.e. a common compliance standard for the
disinfection of G.Lamblia. CT1-C also exhibited similar outcomes with CT1-P, thus
suggesting that the two particular inlet designs (i.e. a channel and pipe) did not have a
significant difference on their impact to the reactive processes. This could be attributed to the
fact that they induced similar mixing effects downstream, as in both cases the flow entered
the geometry by means of a high velocity water jet.

The impact of the RTD on the effectiveness of disinfection can be established when the
results (Table 5) between the different baffle configurations are compared. For the case of
MS4, a 2-log inactivation threshold was surpassed for all the examined contact time
parameters (t10, tp, tg, t90) demonstrating the good efficiency of this configuration. For CT1,
while a 2-log inactivation was expected for tp and tg, the ratio N/N0 at t10 failed to reach this
threshold. As hydraulic conditions deteriorated further in MS1-P, a 2-log inactivation was
only met for the mean residence time, but not for t10 or tp. The disinfection efficiency of
MS2-P and MS3-P was estimated to be even worse, exhibiting the need for geometry
retrofitting towards more efficient configurations. For t90 the N/N0 ratio values were
satisfactory for all the CT designs.

15
Table 6 Outlet disinfection by-product accumulation predictions for t10, tp, tg and t90

TTHM Concentration (μg/l)

CT Setup
t10 TTHM tp TTHM tg TTHM t90 TTHM
CT1-C 35.41 36.96 44.34 51.41
CT1-P 35.72 36.64 45.15 52.44
MS1-P 31.34 32.66 42.86 47.88
MS2-P 27.73 28.82 44.52 45.89
MS3-P 29.21 27.68 57.39 65.25
MS4-C 34.95 37.12 39.60 47.31
MS4-P 34.61 36.73 39.68 48.08

In contrast, the exposure associated with t90 was more prone to higher by-product
concentrations, posing an additional challenge for the design of such facilities. Predictions
with regards to the TTHM accumulation were obtained using the contact time fields (e.g.
Figure 5) produced for t10, tp, tg and t90 by incorporating these values into the source term of
the by-product scalar transport equation. Outlet results are presented in Table 6 showing the
variation of TTHMs according to the individual CT configuration contact tank exposure, with
the most undesirable conditions occurring at t90, as opposed to the N/N0 results. Due to the
more concentrated nature of the MS4 RTD, the range of TTHMs formed between t10 and t90 is
lower compared to the other designs, but since tp is marginally greater than in other cases,
CTTHM is quite high in comparison to the other models. The lowest and highest CTTHM
predictions at tp and t90 were reported for MS3-P; a result consistent with the combined
adverse effects of short-circuiting and disinfectant entrapment in dead zones, characteristic of
the unbaffled design.

With the growth of computing power and the popularisation of computational models,
numerical model studies have been increasingly undertaken as a means to replace empirical
approaches to assess the disinfection efficiency of water treatment facilities. The
development of such techniques is attributed to the capacity of CFD models to accurately
account for the mixing effects occurring due to the hydrodynamic complexity in CTs. Thus
far, such studies can be classified into three types:

a) simulation of tracer experiments to more efficiently determine HEIs for CTs, either a-
priori construction or as a means to avoid time-consuming or operation-invasive tracer
experimentation (e.g. Zhang et al, 2013b; Kim et al, 2010; Stamou, 2008; Khan et al,
2006);
b) simulation of the pathogen inactivation processes to produce an approximation of the
expected survival ratio for various microorganisms in the treated water (e.g. Angeloudis
et al., 2014b; Zhang et al., 2014;Wols et al, 2010; Greene et al, 2007);
c) simulation of average by-product concentrations accumulated during the disinfection
process (Angeloudis et al, 2014a; Zhang et al., 2000).

For the majority of studies undertaken for the second and third classifications, the impact of
the RTD on disinfection had not been considered, even though the relevant HEIs were
derived to validate the modelling strategy capability to reproduce solute transport results. An
exception to this rule was the study of Wols et al. (2010), aimed at the optimization of an

16
ozone contactor, where a Lagrangian particle tracking methodology was implemented to
evaluate the pathogen inactivation. Even though particle tracking was generally more
computationally demanding than the Eulerian modelling method contemplated in their study,
it yielded statistics of the full disinfectant exposures.

In practical conditions where numerical modelling techniques are not always employed, the
US EPA’s Surface Water Treatment Rule promulgated the Ct concept for the performance
assessment of disinfection facilities, according to which the water has to remain in contact
with a certain concentration of disinfectant (C) for a sufficiently long enough time t, so as to
provide the desirable level of inactivation for different water quality conditions and
disinfectants (Rauen et al., 2012). Disinfection is deemed to be achieved once the Ct product
exceeds reference values (AWWA, 1990) derived empirically. The use of T or tg as the
contact time for the Ct product calculation is not recommended, since this does not take into
consideration the hydraulic efficiency of the tank. Instead, guidelines recommend the use of
the parameter t10 (Johnson et al., 1998)

This suggests that there is a gap in current numerical modelling practise as the effects of
short-circuiting or dead zones must similarly be acknowledged for their impact on
disinfection efficiency and by-product formation respectively. In that respect, part of the
uncertainty linked with different designs could be established in order to increase the
reliability of the CT disinfection performance, with the implementation of a novel approach
contemplated in this investigation. However, more sophisticated decay, inactivation and by-
product kinetic models can be adopted to refine the current numerical modelling strategy.
Examples could include the modelling of the transport and the reactive behaviour of organic
materials or additional substances that may interfere with the progress of disinfection
processes at water treatment works.

4. CONCLUSIONS

In this study a RANS-based CFD model was employed to predict the mixing and disinfection
characteristics of chlorine contact tanks (CTs). In the validation phase it was found that the
RANS model was able to reproduce the time-averaged flow and solute transport processes in
the CT quite well, thereby providing a good prediction of scalar transport processes
throughout the tank. This was also reflected in the good match of simulated RTD, F curves
and HEIs with available experimental data.

The analysis was then extended to investigate the effects of various CT designs on the
hydrodynamics, tracer transport and disinfection processes, with the latter through direct
modelling of pathogen inactivation (G.lamblia) and by-product formation (TTHM). It is
shown that a 3-D CFD model can aide in and/or guide the design of contact tanks by
providing reliable predictions of complex flow, pathogen inactivation and a means to regulate
the formation of potentially carcinogenic disinfection by-products. The study shows that
disinfection models incorporated in computational modelling methodology could reinforce
the current practise of obtaining hydraulic efficiency indicators, which are often ambiguous

17
in terms of their significance in predicting the disinfection processes without an overview of
the hydrodynamic conditions also being developed.

The simulation results suggest that the design of disinfection tanks must take into account
two aspects: firstly, minimization of short-circuiting to ensure sufficient and uniform
disinfectant exposure for pathogens and secondly, prolonged residence time due to
disinfectant entrapment in dead zones, which will invariably promote the formation of
disinfection by-products. A CFD approach was proposed to demonstrate how RTD variations
affect both the concentrations of pathogens and by-products across the domain by calculating
their transport under contact time scenarios associated with short-circuiting (t10), peak values
of the RTD (tp) and extended mixing (t90).

Nevertheless, it is acknowledged herein that there is a need to further inform the numerical
methodology with more sophisticated kinetic models, which should be compared with
experimental studies focused on how the water chemistry is altered during disinfection; these
are the subject of on-going research as such refinements could provide a more accurate
assessment of newly designed or retrofitted CTs for a variety of circumstances.

ACKNOWLEDGEMENTS

The first author acknowledges the financial support of CH2M Hill.

REFERENCES

American Water Works Association, 1990. Water Quality and Treatment: a Handbook of
Community Water Supplies. McGraw-Hill.

Amini, R., Taghipour, R., Mirgolbabaei, H., 2011. Numerical Assessment of Hydrodynamic
Characteristics in Chlorine Contact Tank. International Journal for Numerical Methods in
Fluids, 67, 885-898.

Angeloudis, A., Stoesser, T., Kim, D., Falconer, R.A., 2014. Flow, Transport and
Disinfection Performance in Small- and Full- Scale Contact Tanks. Journal of Hydro-
Environment Research (under review).

Angeloudis, A., Stoesser, T., Kim, D., Falconer, R.A., 2014. Modelling of flow, transport and
disinfection kinetics in contact tanks. Proceedings of the Institution of Civil Engineers, Water
Management (in press). DOI: 10.1680/wama.13.00045.

Brown, D., West, J. R., Courtis, B. J., Bridgeman, J., 2010. Modelling THMs in Water
Treatment and Distribution Systems. Proceedings of the Institution of Civil Engineers, Water
Management, 163(WM4), 165-174.

18
Brown, D., Bridgeman, J., West, J. R., 2011. Predicting Chlorine Decay and THM Formation
in Water Supply Systems. Reviews in Environmental Science and Biotechnology, 10(1), 79-
99.

Falconer, R. A., Tebbutt, T. H. Y., 1986. A Theoretical and Hydraulic Model Study of a
Chlorine Contact Tank. Proceedings of the Institution of Civil Engineers, Part 2-Research and
Theory, 81, 255-276.

Fox, R. O., 2003. Computational Models for Turbulent Reacting Flows. Cambridge
University Press, Oxford, UK, 56-98.

Greene, D. J., Haas, C. N., Farook, B., 2007. Computational Fluid Dynamics Analysis of the
Effects of Reactor Configuration on Disinfection Efficiency. Water Environment Research,
78, 909-919.

Gualtieri, C., 2006. Numerical Simulation of Flow and Tracer Transport in a Disinfection
Contact Tank. Proceedings of the 3rd Biennial Meeting of the International Environmental
Modelling and Software Society, Vermont.

Gualtieri, C., 2007. Analysis of the Effect of Baffles Number on a Contact Tank Efficiency
with Multiphysics 3.3. Proceedings of the COMSOL User Conference 2007, Napoli.

Gualtieri, C., (2010). Discussion on E.C.Teixeira and R.N.Siqueira: Performance assessment

of hydraulic efficiency indexes. Journal of Environmental Engineering, 134(10) 851−859.
Journal of Environmental Engineering, 136(9), 1006−1007

Johnson, P., Graham, N., Dawson, M., Barker, J., 1998. Determining the Optimal Theoretical
Residence Time Distribution for Chlorine Contact Tanks. Journal of Water Services Research
and Technology – Aqua, 47(5), 209-214.

Khan, L. A., Wicklein E. A., Teixeira, E. C., 2006. Validation of a Three-Dimensional

Computational Fluid Dynamics Model of a Contact Tank. Journal of Hydraulic Engineering
132(7), 741-746.

Kim, D., Kim, D., Kim, J. H., Stoesser, T., 2010. Large Eddy Simulation of Flow and Tracer
Transport in Multichamber Ozone Contactors. Journal of Environmental Engineering, 136(1),
22-31.

Kim, D., Stoesser, T., Kim, J. H., 2013a. The Effect of Baffle Spacing on Hydrodynamics
and Solute Transport in Serpentine Contact Tanks. Journal of Hydraulic Research, 51(5),
558-568.

Kim, D., Stoesser, T., Kim, J. H., 2013b. Modeling aspects of flow and solute transport
simulations in water disinfection tanks. Applied Mathematical Modelling, 37(16), 8039-8050.

Levenspiel, O., 1999. Chemical Reaction Engineering. 3rd ed., John Wiley & Sons, New
York.
19
Marske, D. M., Boyle, J. D., 1973. Chlorine Contact Chamber Design – A Field Evaluation.
Water and Sewage Works, 120(1), 70-77.

Pope, S. B., 2000. Turbulent Flows. First edition, Cambridge University Press, Oxford, UK,
545-555

Rauen, W. B., 2005. Physical and Numerical Modelling of 3-D Flow and Mixing Processes
in Contact Tanks. PhD Thesis, Cardiff University.

Rauen, W. B., Angeloudis, A., Falconer, R. A., 2012. Appraisal of Chlorine Contact Tank
Modelling Practices. Water Research, 46(18): 5834-5847.

Sadiq, R., Rodriguez, M. J., 2004. Disinfection By-Products (DBPs) in Drinking Water and
the Predictive Models for their Occurrence: a Review. Science of the Total Environment,
321(1-3), 21-46.

Salzano, F., Gualtieri, C., 2014. The effect of baffle spacing on hydrodynamics and solute
transport in serpentine contact tanks, Journal of Hydraulic Research, 52(1), 152-154.

Shiono, K., Teixeira, E. C., 2000. Turbulence Characteristics in a Baffled Contact Tank.
Journal of Hydraulic Research, 38(4), 271-278.

Singer, P. C., 1994. Control of Disinfection By-Products in Drinking Water. Journal of

Environmental Engineering, 120(4), 727-744.

Stamou, A. I., 2002. Verification and Application of a Mathematical Model for the
Assessment of the Effect of Guiding Walls on the Hydraulic Efficiency of Chlorination
Tanks. Journal of Hydroinformatics, 4(4), 245-254.

Stamou, A. I., 2008. Improving the Hydraulic Efficiency of Water Process Tanks using CFD
Models. Chemical Engineering and Processing: Process Intensification, 47(8), 1179-1189.

Teixeira, E. C., 1993. Hydrodynamic Processes and Hydraulic Efficiency of Chlorine Contact
Units. PhD Thesis, University of Bradford.

Teixeira, E. C., Siqueira, R. N., 2008. Performance Assessment of Hydraulic Efficiency

Indexes. Journal of Environmental Engineering, 134(10), 851-859.

Wang, H., Shao, X., Falconer, R. A., 2003. Flow and Transport Simulation Models for
Prediction of Chlorine Contact Tank Flow-Through Curves. Water Environment Research,
75(5), 455-471.

Wilcox, D.C. (2004). Turbulence Modeling for CFD, ISBN 1-928729-10-X, 2nd Ed., DCW
Industries, Inc..

20
Wols, B. A., Hofman J. A. M. H., Uijttewaal W. S. J., Rietveld, L. C., van Dijk, J. D., 2010.
Evaluation of Different Disinfection Calculation Methods Using CFD. Environmental
Modelling & Software, 25 (2010): 573-582.

Zhang, G., Lin, B., Falconer, R. A., 2000. Modelling Disinfection By-products in Contact
Tanks. Journal of Hydroinformatics, 2(2), 123-132.

Zhang, J., Tejada-Martinez, A. E.,and Zhang, Q. 2013a. Hydraulic Efficiency in RANS of

the Flow in Multi-Chambered Ozone Contactors. Journal of Hydraulic Engineering,
139(11),1150-1157.

Zhang, J., Tejada-Martinez, A. E., Zhang, Q., 2013b. Reynolds-Averaged Navier-Stokes

Simulation of the Flow and Tracer Transport in a Multichambered Ozone Contactor. Journal
of Environmental Engineering, 139(3), 450-454.

Zhang, J., Tejada-Martinez, A. E., Zhang, Q., Lei, H., 2014. Evaluating Hydraulic and
Disinfection Efficiencies of a Full-Scale Ozone Contactor using a RANS-Based Modeling
Framework, Water Research, 52(2014), 155-167.

21
FIGURE CAPTIONS
Figure 2 (a) Schematic view of laboratory model tank configuration for CT1-C indicating the approach channel
inlet; (b) Alternative baffle configurations examined experimentally and computationally(MS1-P,MS2-P,MS3-
P,MS4-P and MS4-C); and (c) CT1-P compartment 1 cross-section showing the pipe inlet setup (dimensions in
mm).

Figure 2 Predicted and experimental outlet: (a) RTD (Eθ), and (b) FTC (Fθ) curves for the different
configurations examined experimentally in the laboratory model (Rauen, 2005; Angeloudis et al.,2013).

Figure 3 Contour plots of Estimated Average Contact Time, Disinfectant Distribution, G.lamblia Inactivation
and total Trihalomethane formation as predicted from the CFD simulation of the disinfection processes for each
of the CT designs (CT1-C,MS1-P,MS2-P,MS3-P and MS4-C). The plots are produced at half-depth, i.e. z/Ht =
0.50.

Figure 4 Disinfection performance with respect to mean residence time, based on computational cells of CFD
simulations: (a) Chlorine Concentration (CCl), (b) G.Lamblia Survival Ratio (N/N0), and (c)Total
Trihalomethane (TTHM) formation. The figure indicates that despite disinfection occurring under identical
conditions of operation and water quality, the CT geometrical differences have a distinctive impact on the
kinetics.

Figure 5 Exposure time fields expected for: (a) t10, (b) tg, and (c) t90, are shown to demonstrate how the contact
time may vary due to the residence time distribution for each computational cell. The plots are obtained at mid-
depth (z/Ht = 0.50) of the CT1-C configuration.

Figure 6 Comparison between peak concentration time (tp) and residence time calculated within computational
cells.

22