tI) Pergamon War. Sci. Tech. Vol 40. No.9. pp. 157-163.

1999
(0 1999
Published by Elsevier ScienceLtdon behalfof the IAWQ
Pnnted an GreatBntain. All rightsreserved
0273-1223199 $20.00 + 0.00
PIT: S0273-1223(99)OO652-6

REMOVAL OF NATURAL ORGANIC

MATIER (NOM) FROM DRINKING
WATER SUPPLIES BY
OZONE-BIOFILTRATION

Raymond M. Hozalski*, Edward 1. Bouwer** and

Sudha Goel**
• Department ofCivil Engineering, University ofMinnesota . Minneapolis. MN 55455. USA
.. Department ofGeography and Environmental Engineering. Johns Hopkins University.
Baltimore. MD 2 J2 J 8. USA

ABSTRACT

Removal of natural organic matter (NOM) in biofilters can be affected by many factors including NOM
characteristics, use of pre-ozona tion, water temperature. and biofilter backwash ing. Laboratory experiments
were performed and a biofilter simulation model was developed for the purpose of evaluating the effects of
each of these factors on NOM removal in biofilters. Four sources of NOM were used in this study to
represent a broad spectrum of NOM types that may be encountered in water treatment. In batch experiments
with raw NOM , the removal of organic carbon by biodegradation was inversely proportional to the UV
absorbance (254 nmj-to- TOC ratio and directly proportional to the percentage of low molecular weight
material (as determined by ultrafiltration). The extent and rate of total organic carbon (TOC) removal
typically increased as ozone dose increased, but the effects were highly dependent on NOM characteristics.
NOM with a higher percentage of high molecular weight material experienced the greatest enhancement in
biodegradability by ozonation. The performance of laboratory-scale continuous-flow biofilters was not
significantly affected by periodic backwashing, because backwashing was unable to remove large amounts of
biomass from the filter media . Model simulations confirmed our experimental results and the model was
used to further evaluate the effects of temperature and backwash ing on biofilter performance. © 1999
Published by Elsevier Science Ltd on behalf of the IAWQ. All rights reserved

KEYWORDS

Natural organic matter; ozone; biofiltration; backwashing; simulation model.

INTRODUCTION

Natural organic matter (NOM) in drinking water is not known to have any direct health effects; however, its
reactivity with dissolved and particulate species in water significantly impacts water quality and treatment.
The presence of NOM in water supplies can lead to impaired drinking water quality due to: 1) reaction with
chemical oxidants (e.g. chlorine) during disinfection to form halogenated disinfection by-products (DBPs)
(Rook, 1977),2) stabilization of particles to inhibit their removal during water treatment (Amirtharajah and
OMelia, 1990), and 3) the ability to serve as electron donor for the "regrowth" of heterotrophic bacteria in
distribution systems (LeChevallier et al., 1996). From a public health perspective, reducing NOM levels
prior to terminal disinfection and distribution is important for production of safe, high quality drinking
water.
157
 158 R. M. HOZALSKI et al.
There are several approaches for removing NOM from water to limit DBP formation and control microbial
regrowth including biologically-active filtration (i.e. biofiltration). Because the majority of NOM in natural
waters is recalcitrant, preceding biofiltration by ozonation typically enhances NOM biodegradability and
increases overall NOM removal as compared to biofiltration alone (van der Kooij et al., 1989). The
oxidation of NOM by ozone can enhance NOM biodegradability by reducing the size of NOM molecules
(Kaastrup and Halmo, 1987) and by increasing the number of oxygenated functional groups. Given the
potential benefits of ozone including enhanced NOM removal, inactivation of Cryptosporidium, and reduced
formation of halogenated DBPs, it is likely that ozone usage as a primary disinfectant will increase in the
United States and elsewhere. Therefore, it is important to understand the effects of ozone on NOM and the
impacts on downstream processes including biofiltration.

This paper will discuss the results from several studies that investigated biofiltration and ozone-biofiltration
for use in water treatment. Batch experiments were performed to evaluate the effect of NOM source (i.e.
NOM characteristics) and ozone dose on NOM biodegradation. Laboratory-scale column experiments were
performed to evaluate factors affecting NOM removal in biofilters such as periodic backwashing. In
addition, this paper also includes results from model simulations of the biofiltration process.

METHODS

Batch experiments

Batch experiments were performed to evaluate the effect of NOM source and ozone dose on NOM removal
by biodegradation. Four sources of natural organic matter were used in the batch experiments to represent a
wide spectrum of NOM compositions that might be encountered in drinking water treatment: 1) a high TOC
source rich in fulvic acids obtained from the Great Dismal Swamp in southeastern Virginia (DSW), 2) a
groundwater supply from Florida (FGW), 3) soluble exudate material produced by photosynthetic
cyanobacteria cultured in the laboratory (AE), and 4) commercial humic acid (CHA). In an attempt to
characterize the NOM sources, the molecular weight distribution of each of the NOM sources was
determined by ultrafiltration according to the method of Logan and Jiang (1990).

The batch biodegradation experiments were performed by placing raw or ozonated NOM (diluted to
approximately 4 mg TOCfL), inorganic mineral media (phosphate buffer and other salts, pH = 7.2), seed
bacteria, and oxygen (saturated at 7-9 mg/L) into thirty-two 160 mL serum bottles. Eight of the bottles were
spiked with 100 mg/L of HgCl2 to serve as killed controls. The serum bottles were incubated in the dark at
22.5°C for up to 24 days. Triplicate "active" bottles and 1 control bottle were sacrificed at the beginning and
end of the incubation period and at six sampling times in between to measure dissolved oxygen, total
carbon, total organic carbon (TOC), dissolved organic carbon (DOC), and UV absorbance at 254 nm. More
details on the batch experiments are provided by Goel et al. (1995). In addition, continuous-flow biofilter
results using the same four NOM sources are described in Hozalski et at. (1995).

Continuous-flow column experiments

Continuous-flow column experiments were performed to study the removal of attached biomass during
backwashing and the subsequent effect on NOM removal in the biofilters. The laboratory-scale biofiltration
apparatus consisted of a 2.54 em (inner diameter) x 61 ern column packed with 30.5 ern bed depth of porous
media (0.56 mm glass beads or 0.52 mm effective size sand). Peristaltic feed pumps were used to supply a
concentrated suspension of bacteria, concentrated mineral media solutions, and dilution water to the top of
the column. Most experiments were performed without growth substrate (i.e. NOM) to investigate biomass
deposition and removal without the complications of biofilm growth. The biofiltration apparatus was also
equipped with an additional peristaltic pump for backwashing (upflow). The filter hydraulic loading rate was
approximately 2.5 m/h and backwash loading rates ranging from 23 to 53 m/h were tested. The biofilter
columns were operated in a constant temperature room at 22.5°C. More details on the experimental setup are
provided by Hozalski and Bouwer (1998).
 Removal of NOM from drinking water supplies by ozone-bioflltration 159
Analytical methods

Aqueous carbon concentrations were measured with a Dohrmann DC-80 organic carbon analyzer equipped
with an autosampler (Rosemount Analytical Corp., Santa Clara, CA). TOC samples were prepared by
acidifying to pH < 2 with phosphoric acid and stripping with N2 gas to remove inorganic carbon prior to
injection. DOC samples were filtered through pre-rinsed 0.45 IJll1 pore size membrane filter units (Durapore,
Millipore Corp., Bedford, MA) prior to acidification, stripping, and analysis. UV absorbance was measured
using a spectrophotometer (Lambda-3, Perkin-Elmer Corp., Norwalk, CT) and matched 1 em quartz
cuvettes. Dissolved oxygen concentrations were measured using a dissolved oxygen probe and meter (Model
820, Orion Research Inc., Boston, MA). Bacteria were enumerated using the acridine orange direct count
method (Hobbie et al., 1977) and by plate counts (Standard Methods, 1989).

Simulation model

A biofilter simulation model (BIOFILT) was developed to simulate the non-steady state substrate
(biodegradable organic matter or BOM) and biomass (both attached and suspended) profiles in a biofilter
through incorporation of the fundamental processes of: (I) deposition of suspended biomass onto filter
media, (2) transport of substrate through the filter by advection and dispersion, (3) diffusion of soluble
substrate across a liquid film layer, (4) diffusion of substrate in the biofilm, (5) biodegradation of substrate
by immobilized (in biofilm matrix) bacteria, (6) biomass growth, (7) biomass decay, (8) biomass loss from
the biofilm due to fluid shear, and (9) biomass loss during a backwashing event. In addition, laboratory-scale
biofiltration experiments were performed to calibrate the model and to evaluate its ability to simulate
biofilter behavior.

The main goal of the BIOFILT model is to predict the substrate concentration (S) as a function of time and
distance along the reactor length (reactor depth). Therefore, the backbone of the model is the one-
dimensional advection-dispersion equation with reaction

as =D,a 2s _vas _ kSX,rap _!!"'J

at ax 2
ax K,+S & f

where D'= dispersion coefficient (L2rr), v = fluid velocity (LIT), t = time (T), x = distance along biofilter
length (L), k = maximum rate of substrate utilization (Ms!MxT), K. = Monod half-velocity coefficient
(MslL\ X susp = suspended cell concentration (MxlL\ a = specific surface area of the media (IlL), Jr= flux
of substrate into the biofilm (Ms/L 2T), and e = the bed porosity. This equation is coupled to a biofilm model
through the flux expression (Jr). The accumulation term on the left side of Equation I was not canceled with
a steady state assumption so that we could simulate the non-steady period during biofilter startup or ripening
and biofilter recovery following a backwashing event. The suspended growth term was assumed to be
insignificant for simulating drinking water biofilters since the short hydraulic residence time in rapid
filtration systems (typically 2 to 10 minutes) leads to washout, and the suspended cell concentrations are
relatively low.

The substrate advection-dispersion equation was formulated into a Crank-Nicolson finite difference scheme
and the resulting set of linear equations was solved using Gaussian elimination at each time step. The
boundary conditions (BC) and the initial conditions (IC) used for solution of the advection-dispersion
equation are shown below:

dS for x = 0 t ~ 0
BC 1: vSo = vS'-', - D' dx '
dS~
BC 2: - = 0 for x = 4, t ~ 0
dx

IC 1: S = So for 0:;; x s 4, t = 0
J60 R. M. HOZALSKI et al.
where Lb = filter bed depth (L). The substrate flux into the biofilm (Jr) was determined from a biofilm
modeling approach developed by Rittmann and McCarty (1981) and is based on Fickian diffusion of
substrate through a stagnant film layer and into the biofilm where diffusion and reaction (Monod kinetics)
are occurring simultaneously. Additional mass balances on suspended biomass and attached biomass (i.e.
biofilm bacteria) complete the model. More details on the model development and model simulations are
provided in Hozalski (1996), Hozalski and Bouwer (in preparation, a), and Hozalski and Bouwer (in
preparation, b).

RESULTS AND DISCUSSION

Batch experiments

Initial batch experiments compared the removal of TOC by biodegradation for the four raw NOM sources.
One of the main reasons for using multiple sources of NOM was to attempt to develop correlations between
NOM characteristics and NOM behavior (i.e. biodegradability). Two methods were used to characterize the
NOM samples in our study: UV-to-TOC ratio (also known as specific UV absorbance or SUVA) and
molecular weight distribution by ultrafiltration. The UV absorbance was measured at 254 nm in our studies,
although some researchers have used other wavelengths including 280 nm (Chin et al., 1994). The effects of
SUVA and the percentage of DOC smaller than 1000 Daltons on TOC removal by biodegradation are shown
in Table 1.

Table 1. Effects of SUVA and the percentage of DOC smaller than 1000 Daltons on NOM removal by
biodegradation in batch cultures

NOM Source SUVA, DOC < 1000 Daltons, TOe removal by

(L/m2)cm·1 0/0 biodegradation, %
CHA 0.091 10 0
FGW 0.039 47 21
DSW 0.031 22 22
AE 0.020 60 37

SUVA provides a measure of the aromaticity of NOM, so that sources with high SUVA are considered to
have a preponderance of unsaturated carbon bonds including aromatic structures (Chin et al., 1994). SUVA
correlates positively with the molecular weight of NOM (Chin et al., 1994), and SUVA correlates with some
measures of NOM reactivity including chlorine consumption (Krasner, 1997). As shown in Table I,
enhanced removal ofTOC by biodegradation was observed for the NOM sources with a lower SUVA value.
In addition, there was a reasonably good linear correlation between the two parameters:

TOC removal = -463.88(SUVA) + 40.991, R2 = 0.9217

Because of the low number of data points, more data are needed to determine if this correlation is generally
applicable to a wide variety of NOM sources.

The molecular weight distribution of the NOM is important for biodegradability because lower molecular
weight compounds are more easily transported across the cell membrane and attacked by metabolic enzymes
(Leisinger et al., 1981). The results in Table 1 are in agreement with this theory and indicate a general trend
of improved biodegradation with increasing percentage ofIow molecular weight material.

In addition to the experiments with raw NOM, we also investigated the impact of ozonation on NOM
biodegradation. Ozone was applied to the NOM sources at doses ranging from 1.3 to 7.3 mg Oymg TOC
(reacted dose) and then the ozonated NOM samples were subjected to the batch biodegradation assay as
described above. The ozonation process itself resulted in the direct mineralization of 6 to 27% of the initial
TOe. When combined with the amount of TOC removed by biodegradation, increasing ozone dosage
generally resulted in increasing NOM removal. However, in terms of enhancing biodegradation, the NOM
sources with a higher percentage of high molecular weight materials (i.e. DSW and CHA) experienced a
 Removal of NOM from drinking water supplies by ozone-b iofiltranon 161
substantial improvement in TOC removal by ozonation and subsequent biodegradation (Goel et al., 1995). It
appeared that ozone doses in the range of I to 2 mg Oj/mg TOC were optimal for enhancing the
biodegradation of these higher molecular weight NOM sources. There was very little benefit to ozonating
the NOM sources with a relatively high percentage of low molecular weight material (i.e. FGW and AE).
The exudate material (AE), which is likely to consist of predominantly lower molecular weight aliphatic
materials, was readily degradable without pre-ozonation, In addition, the low SUVA value for AE suggested
that this material lacked the unsaturated sites that serve as targets for attack by ozone.

Continuous-flow biofilter column experiments

A series of experiments were performed with a lab-scale biofilter to evaluate the removal of attached
bacteria during backwashing and the subsequent impact on NOM removal in the biofilters (Hozalski and
Bouwer, 1998). Backwashing has the potential to impair or temporarily reduce NOM removal in a biofilter
by removing biomass (along with particles and floc) from the filter media. The biofilters ean respond to this
"disturbance" by regenerating the attached biomass population during the next filter run to result in a
recovery of the NOM removal performance. This biomass loss and restoration cycle would be continuously
repeated. In initial experiments, bacteria were deposited on the filter media by filtration and then the filter
was suddenly backwashed to evaluate how many of the deposited bacteria would be removed. Since no
substrate was supplied, the adhesion of the bacteria in these experiments was primarily due to the initial
adhesion forces and is not due to production of extracellular polymers as in a biofilm. The fraction of
organisms removed ranged from 20 to 40% (i.e, 60 to 80% were retained), suggesting that backwashing is
not very effective at physically removing bacteria from the filter media surface.

An additional biofilter experiment was performed with ozonated NOM in the feed to evaluate the impact of
backwashing on removal of NOM (as TOC). After an initial inoculation ofthe filter bed (sand) with a mixed
population of bacteria from a freshwater reservoir, the biofilters continuously received a solution of
ozonated DSW (0.58 ± 0.12 mg Oymg TOC) and inorganic mineral media at a hydraulic loading rate of
2.44 mIh. The biofilter was backwashed for 10 minutes (36 mIh, bed expansion = 23%) every two days with
disinfectant-free water. Following a brief startup period of approximately 6 days, backwashing had no
significant impact on TOC removal in the biofilter (Hozalski and Bouwer, 1998). The effluent TOC
concentration did not vary significantly from the period immediately following a backwash to the next
backwash cycle. This result is encouraging for those water utilities that wish to operate existing particle
removal filters as biofilters.

BIOFILT simulation results

Simulations were performed to evaluate the effect of temperature (3 to 22.5°C) and backwashing on the
removal of biodegradable organic matter (BOM) in biofilters. The input parameters were selected to
represent the operation of a full-scale biofilter with periodic backwashing (lor 2-day intervals). The details
of these simulations are provided in Hozalski and Bouwer (in preparation, b).

Effect of Temperature. Simulations were performed to evaluate the effect of temperature on the removal of
BOM. Two BOM compositions were investigated: Composition A '" I0% acetate and 90% organic matter
with a maximum biodegradation rate one-tenth that of acetate (0.1Oklee); and Composition B '" I0% acetate
and 90% organic matter with a maximum biodegradation rate one-fourth that of acetate (0.25k ace) . The
effects of temperature deviation from the baseline case (22.5°C) on substrate degradation were simulated by
adjusting the maximum substrate utilization rate (k) and substrate diffusivity values as described in Hozalski
and Bouwer (in preparation, a). The simulation for Composition A at 22.5°C indicated that 55 % removal of
the BOM (Figure la) was attained after a 19-day startup period (Figure Ib). Approximately 87 % of the
more readily degradable BOM (Composition B) was removed after only a lO-day startup period at 22.5°C
(Figures la and Ib). Decreasing temperature resulted in a decrease in the steady state BOM removal and an
increase in the time to reach steady-state. The effects for Composition A were much more dramatic because
of the slower biodegradation kinetics for the bulk of the BOM. A 100-day simulation for biofilter operation
at 3°C indicated that the total BOM removal for Composition A remained less than 1% (Figure la). At
operating temperatures of 9°C and lower, the simulations indicate that the accumulation of biomass and
overall removal of BOM are likely to be severely inhibited by cold temperatures. However, significant
162 R. M. HOZALSKI et al.
removals of readily degradable substrates (e.g., acetate) is still possible at these lower temperatures once
pseudo-steady state conditions are obtained. The expected removals also depend strongly on the NOM
composition.

A -p:I;ll oi
'A
C
5.
.. 0 rr osn 1B C Cal-p:I; ll A
>-
"..'" o CallU5'1 B
iO ;;;
"".>.
>
0
E
."

~-
e ::l '"
c.
:E s:

[
0 u
CD
e'"
10 B
e
0 E
;::
22' 12'
Torrporature TempornlurC

Figure I. Effect of BaM composition and temperature on simulated BaM removal in a full-scale biofilter

Effect ofBackwashing. Several simulations were performed to evaluate the effect of biomass removal during
backwashing on the BOM removal performance ofbiofilters. The effect of biomass lost during backwashing
on the average BOM removal is illustrated in Figure 2a for different BOM compositions (% acetate vs. %
0.25k ucc fraction). The filter run time (i.e. time between backwashes) used in these simulations was 2 days.
The simulations illustrate that there is little effect on the average BOM removals until the amount of
biomass lost during backwashing exceeds 60%. The impact of biomass lost during backwashing on the
duration of the start-up period is also shown in Figure 2b. The simulations indicate that the start-up period is
relatively insensitive to backwashing conditions that remove less than 60% of the attached biomass in the
biofilter. Therefore, the simulation results con finn the lab-scale biofiltration results that demonstrated that
backwashing does not significantly impact biofilter performance.

45
.
~ 4O B
'C
80 A
~ ~ 35
>
iO 70 o . kLtare
> E
0
E 60 e
e 50
;l;
0 laY. Jla tar 1Y. 0
CD 40

e'"
a> 30
tale. 50% 025k
'"
>
et 20
... 10% I\c.elale . 025k
10

0 0 '- - - -
0 20 40 60 80 100 o 20 40 00
Bi OlmS S removal du ring bac kw:lshing. % BiOfIDSS rc rmval Wring backwashing. %

Figure 2. Effect of BaM composition and biomass loss during backwashing on simulated BaM removal in a full-
scale biofilter.

CONCLUSIONS

Laboratory-scale experiments and simulation modeling were used to evaluate the effects of NOM
characteristics, pre-ozonation, water temperature, and biofilter backwashing on the removal of NOM in
biofilters. The main conclusions from these research efforts are as follows:
 Removal of NOM from drink ing water supplies by ozone-biofiltration 163
• The removal of organic carbon by biodegradation was inversely proportional to SLNA and directly
proportional to the percentage of low molecular weight material (as determined by ultrafiltration).
• The extent and rate of total organic carbon (TOC) removal typically increased as ozone dose increased,
but the effects were highly dependent on NOM characteristics. NOM with a higher percentage of high
molecular weight material experienced the greatest enhancement in biodegradability by ozonation .
• The performance of laboratory-scale continuous-flow biofilters was not significantly affected by
periodic backwashing, because backwashing was unable to remove large amounts of biomass from the
filter media.
• Model simulations confirmed our experimental results concerning the effects ofbackwashing on biofilter
performance . The model was also used to evaluate the effects of temperature on biofilter operation. The
model suggested that biofiltrat ion can provide BOM removal at low water temperatures (i.e. 3 to 9°C)
but performance is significantly impaired. In addition, long biofilter startup times [i.e, from a clean filter
bed to a pseudo-steady state biofilm population) can be expected at these colder temperatures.

ACKNOWLEDGMENTS

The authors acknowledge the financial support provided by the American Water Works Association
(AWWA) through the Abel Wolman Doctoral Fellowship Program and by the AWWA Research Foundation
for Project 252. The authors also acknowledge the helpful comments of fellow team members on AWWARF
Project 252 including Peter Huck, Brad Coffey, Appiah Amirtharajah, Monica Emelko, and Daniel Urfer.

REFERENCES

Amirtharajah, A. and OMelia, C.R. (1990). Coagulation Processes: Destabilization, Mixing, and Flocculation. In Water Quality
and Treatment, 4$ edition, F.W. Pontius (ed.). McGraw-Hili, New York, NY, USA, pp . 269-365.
Chin, Y. P., Aiken, G. and O'Loughlin, E. (1994). Molecular We ight, Polydispersivity, and Spectroscopic Properties of Aquat ic
Humic Substances. Environ . Sci. Technol ., 28(1 I), 1853.
Goel, S., Hozalski, R.M ., and Bouwer, E.J. (1995). Biodegradation of NOM: Effect of NOM Source and Ozone Dose . Journal
AWWA, 87(1), 90-105.
Hobb ie, J.E., Daley , R.J., and Jasper, S . (1977). Use of Nucleopore Filters for Coun ting Bacteria by Fluorescence Microscopy.
Appl. Env iron. Microbiol., 33,1225.
Hozalski, R. M. (1996). Removal of Biodegradable Organic Matter in Drinking Water Biofilters: Experimental Studies and Model
Development. Ph.D . Dissertation, The Johns Hopkins University, Baltimore, MD, USA.
Hozalski, R. M. and Bouwer, E. J. (manuscript in preparation, a). Non -Steady State Simulation of BOM Removal in Dnnking
Water Biofilters: Model Development.
Hozalski, R. M. and Bouwer, E. J. (manuscript in preparation, b). Non-Steady State Simulation of BOM Removal in Dnnking
Water Biofilters: Results and Interpretation.
Hozalski, R. M. and Bouwer, E. J. (1998). Deposition and Retent ion of Bacteria in Backwashed Filters . Journal AWWA , 90(1) ,
71-85.
Hozalski, R. M., Gael, S. and Bouwer, E. 1. (1995). TOC Removal in Biological Filters . Journal AWWA , 87(12), 40-54.
Kaastrup, E. and Halmo, T.M . (1987). Removal of Aquatic Humus by Ozonation and Activated Carbon Adsorption. Proc . 19r
Mtg. Div. Envir. Chern.• ACS, 27(1), 355 .
Krasner, S. (1997). Oxidation: Chlorine Reactivity with NOM. Proc. AWWA Wat. Qual. Technol. Conf Denver, AWWA.
LeChevallier, M.W., Welch, N.J., and Smith, D.B. (1996). Full-Scale Studies of Factors Related to Coliform Regrowth in
Drinking Water. Appl. Environ . Microbiol., 62(7), 2201·2211.
Leisinger, T ., Cook, A.M., Hutter, R., and Nuesch, J. (1981). Microbial Degradation ofXenobiotics and Recalcitrant Compounds.
Academic Press, London, UK.
Logan, B.E., and Jiang, Q. (1990). Molecular Size Distribution of Dissolved Organic Matter. Jour . Envir. Engrg., 116(6), 1046-
1062
Rittmann, B.E. and McCarty, P.L. (1981). Substrate Flux into Biofilrns of Any Thickness. Jour. Environ Engrg ., 107(EE4), 831 .
Rook, J.J. (1977). Chlorination React ions of Fulvic Acids in Natural Waters. Environ . Sci. Technol ., 11(5),478.
Standard Methods for the Examination of Water and Wastewater (1989). 17'" edition, American Public Health
Association/American Water Works AssociationlWater Environment Federation, Washington, DC, USA.
van der Kooij, D., Hijnen, W.A.M., and Kru ilof, J.C. (1989). The Effects of Ozonation, Biolog ical Filtration, and Distribution on
the Concentration of Easily Assimilable Organic Carbon (AOC) in Drinking Water . Ozone Sci. and Engrg. , II , 297.