You are on page 1of 11

Water Research 199 (2021) 117180

Contents lists available at ScienceDirect

Water Research
journal homepage: www.elsevier.com/locate/watres

Review

Removal of effluent organic matter with biofiltration for potable


reuse: A review and meta-analysis
Eric S. Peterson∗, R. Scott Summers
University of Colorado Boulder, Department of Civil, Environmental, and Architectural Engineering, 607 UCB, Boulder, CO 80309, USA

a r t i c l e i n f o a b s t r a c t

Article history: Biofiltration, historically used for biodegradable organic matter (BOM) removal in drinking water treat-
Received 5 February 2021 ment, is being increasingly applied for potable reuse which requires unique characterization. This review
Revised 15 April 2021
and meta-analysis evaluates BOM occurrence as part of bulk wastewater effluent organic matter (EfOM),
Accepted 18 April 2021
quantifies the roles of operational parameters to achieve EfOM removal in biofilters, and identifies re-
Available online 22 April 2021
search gaps which may be fruitful for understanding reuse biofilter performance. Literature data (n = 76)
Keywords: indicates EfOM has a high biodegradable fraction (median 26%), which after typical ozone doses is higher
Empty bed contact time (57%). A biofiltration performance dataset (n = 160 across 42 WWTP effluents) shows that EfOM removal
Ozone of 35–40% can be expected when design parameters are optimized. Specifically, higher EfOM removal is
Biological activated carbon achieved by adding pre-ozonation and use of biological activated carbon (BAC) media, with comparatively
BDOC smaller impacts of increasing ozone dose or increasing empty bed contact time under typical scenarios.
Total organic carbon
Combined, these factors strongly correlate with observed EfOM removal (r2 = 0.64) after accounting for
confounding by adsorptive removal in BAC media with fewer than 20,0 0 0 bed volumes treated. Future
research that quantifies the occurrence of BOM, biomass activity on filter media, steady-state removal by
BAC, and impacts of longer empty bed contact times in potable reuse scenarios could impact optimization
strategies to meet or exceed biofilter performance observed to date.
© 2021 Elsevier Ltd. All rights reserved.

1. Introduction (DBP) precursors (Michael-Kordatou et al., 2015) and decreasing


microbial growth potential in distribution systems (Weinrich et al.,
Advanced treatment of municipal wastewater effluent is an in- 2010), and also indirectly by providing the carbon source needed
creasingly common strategy to reduce negative impacts of wastew- for metabolism of specific contaminants which occur at trace (i.e.,
ater treatment plant (WWTP) discharge on receiving waterbodies secondary substrate) concentrations (Zearley and Summers, 2012).
or augment drinking water supplies through potable reuse. Rapid- In potable reuse, removal of bulk EfOM is also a specific design
rate biologically active granular media filtration (biofiltration) is parameter because total organic carbon (TOC) concentrations in
well-established, especially in Europe, for removal of organic mat- common wastewater effluents (6–9 mg/L) (Krasner et al., 2009)
ter and contaminants in drinking water treatment (Terry and Sum- exceed typical regulatory limits (e.g., 0.5–3 mg/L) (U.S. Environ-
mers, 2018) and is becoming widespread in advanced systems mental Protection Agency, 2012; California Department of Public
to treat wastewater effluents (WWefs) (Gerrity et al., 2013). Or- Health, 2018; Florida Potable Reuse Commission, 2020). A range of
ganic matter (OM) in wastewater effluent includes both specific treatment technologies can be utilized to remove EfOM, including
compounds (e.g., pesticides, pharmaceuticals, personal care prod- reverse osmosis and activated carbon adsorption; however, biofil-
ucts, ozonation byproducts) and non-specific, bulk effluent organic tration is a comparatively lower cost, lower energy usage approach
matter (EfOM), a fraction of which is biodegradable OM (BOM) which is attractive for strategies seeking to optimize performance
(Shon et al., 2006). Bulk EfOM represents the primary substrate and target water quality that is fit for purpose. (Gerrity et al., 2014;
for aerobic heterotrophic growth in biofilters; its removal im- Schimmoller and Kealy, 2014).
proves water quality directly, by removing disinfection byproduct Removal of bulk EfOM with biofiltration has been reported
in a growing body of studies which collectively comprise differ-
ent source waters, pretreatments, and operating conditions; how-

Corresponding author.
ever, biofilter performance across these factors has not been well-
E-mail address: eric.peterson@colorado.edu (E.S. Peterson).

https://doi.org/10.1016/j.watres.2021.117180
0043-1354/© 2021 Elsevier Ltd. All rights reserved.
E.S. Peterson and R.S. Summers Water Research 199 (2021) 117180

characterized. Most studies are logistically constrained to one or in the meta-analysis dataset and recommendations for future re-
two source waters and only a few operational conditions, thus search are presented.
effects of water quality or biofilter operation can be site-specific
and difficult to generalize. Such generalization is needed to rec- 2. Methods
oncile the wide range of biofilter performance observed across
and within studies. For example, peered reviewed studies reporting 2.1. External data collection and screening criteria
biofilter performance, as TOC or dissolved organic carbon (DOC) re-
moval, include: 3–15% (Ho et al., 2011), 6–30% (Arnold et al., 2018), Three databases were developed to address the specific objec-
17–37% (Kalkan et al., 2011), 23–51%, (Farré et al., 2011), 22–63% tives. Both peer-reviewed and gray literature (e.g., conference pro-
(Reungoat et al., 2011) and 59–76% (Chuang and Mitch, 2017). In ceedings, dissertations, utility reports) were investigated. The first
these and other studies, EfOM removal not only exceeds the av- database assessed BOM occurrence in WWefs (n = 76 from 12
erage BOM removal achieved in drinking water rapid-rate biofil- studies, Table SI 4.1). The second database assessed the transition
ters (i.e., 10% TOC removal without ozone and 15% removal after from GAC, a condition in which DOM removal is dominated by GAC
ozone), but also the average total BOM fraction itself (20% and adsorption, to BAC, a condition in which DOM removal is domi-
30%) (Terry and Summers, 2018), and thus reuse biofilters require nated by biodegradation of BOM (n = 50 from 21 studies, Table SI
unique characterization. The study goal was to quantitatively as- 5.1). The most extensive database was developed from studies that
sess biofiltration performance for EfOM removal on a wide range reported BOM removal from EfOM by biofiltration and the asso-
of source waters and operational conditions to provide insight on ciated biofilter operating parameters (n = 286). Forty two studies
the resulting wide ranges in EfOM removal. on biofiltration of municipal wastewater EfOM that reported all of
Reuse biofilters treating municipal wastewater effluent are im- the following criteria were retained in this database: 1) biofilter
plemented using principles established for drinking water, given influent TOC or DOC concentration, 2) BOM removal, reported as
that the ratio of EfOM to drinking water DOM is only 2 to 5 and absolute or percent change in TOC or DOC concentration, 3) ozone
that many drinking water sources are impacted by wastewater dis- dose or specific ozone dose (O3 /TOC, mg/mg), 4) EBCT, 5) granular
charges. Thus, this analysis is focused on treatment of wastewa- media type, either inert (typically sand or anthracite) or BAC, and
ter effluents through the lens of well-established drinking water 6) for BAC media, the throughput (in bed volumes) treated before
biofiltration (Urfer et al., 1997; Zhu et al., 2010; Basu et al., 2016; sampling. Studies using BAC media for which prior throughput was
Brown et al., 2016; Terry and Summers, 2018). Briefly, biofilter per- unknown were also included if the media had been in use at field
formance is related to the amount and character of influent BOM scale for at least 2 years (i.e., about 35,0 0 0 bed volumes at 30 min
and the amount and activity of filter biomass, though in practice EBCT). Other parameters were considered as eligibility criteria, in-
rather inexact but more easily defined surrogates are used. BOM cluding temperature, the biodegradable EfOM fraction (BDOC or
is usually quantified as either biodegradable DOC (BDOC), a mea- AOC), and biomass quantification, but not used because they were
sure of the total, or ultimate, biodegradable fraction defined as not frequently reported.
the change in DOC after long-term exposure to biomass; or, as- It is common to locate additional (pre)treatment processes be-
similable organic carbon (AOC), a measure of readily biodegradable tween the WWTP secondary process and the biofilter, and in most
OM defined as the growth of specific microbial species relative to cases results from these studies were included in the analysis.
their growth on a reference substrate (Escobar and Randall, 2001). Examples of these pretreatments include tertiary filtration (rapid
Ozone pre-oxidation reacts with recalcitrant OM to increase the sand or cartridge), coagulation, and ozonation. When significant
BOM fraction (Huck, 1990). Biofilter performance is impacted by pretreatment was reported (e.g., ozonation at high dose and not
influent water quality conditions (type and amount of BOM and immediately preceding biofiltration, ion exchange for DOM re-
water temperature) which affect microbial activity and operating moval) those results were excluded.
conditions (empty bed contact time (EBCT) and media type) which
affect the amount of biomass. Use of granular activated carbon 2.2. Internal WWef treatability study
(GAC) media is beneficial but complicates the steady-state biofilter
performance assessment as it may hold residual adsorption capac- In addition to the external data, biofiltration results generated
ity. Psuedo-first-order kinetic steady-state behavior based on plug- by the authors as part of a larger study to evaluate DBP control
flow hydraulics is a simplified but useful model of biofilter perfor- in WWefs by coagulation, ozonation, biofiltration, and GAC adsorp-
mance. In this approach, once acclimated and at steady-state, BOM tion were included (Summers et al., 2020). One motivation for in-
removal is modeled to be independent of influent BOM concentra- cluding the data in this review was inert media was used, while
tion and thus can be reported on a relative (percent removal) basis most published WWef studies (88%) used BAC media. The labora-
(Terry and Summers, 2018). tory study evaluated six different WWefs for BDOC (Table SI 2.2)
The specific objective was to address the following questions: and biofiltration performance with inert anthracite media, specific
what level of EfOM removal can be expected with biofiltration, and ozone doses of 0 and 1.0 O3 /TOC (mg/mg), and EBCTs of 5, 15, and
what are the contributions of operating conditions? The approach 30 min (Table SI 2.1). The ozonation, biofiltration, and analytical
taken was to first review the occurrence of BOM in wastewater methods were reported previously (Summers et al., 2020). The TOC
treatment plant effluents, both with and without pre-ozonation, to removal data from this study (n = 36) was compared to the liter-
establish a treatability baseline. Second, a biofiltration performance ature data for inert media alone (n = 29) and no evidence was
database of EfOM removal paired to system operating parameters found to suggest that the laboratory data would introduced bias
was complied. The impacts of ozone dose, EBCT and media type in interpreting the literature dataset. A detailed description of this
were then evaluated in three ways; 1) critical review of studies analysis is in SI Section 3.
which report the impact of each parameter with paired data, 2)
systematic evaluation of biofiltration performance in an internal 2.3. Data pre-processing
six WWef lab study, and 3) meta-analysis of data across all stud-
ies. The role of non-steady state adsorptive processes when using The initial biofilter performance dataset contained 286 obser-
biological activated carbon (BAC) media is addressed. Finally, a de- vations of EfOM removal by biofiltration across 42 unique WWTP
scriptive model is used to evaluate the combined effects of the effluents from 43 studies. Unique WWTPs were identified by name
treatment parameters on EfOM removal across all source waters or by a combination of location and treatment process description.

2
E.S. Peterson and R.S. Summers Water Research 199 (2021) 117180

Table 1
Summary of studies which systematically evaluated biofiltration for EfOM removal.

Source Water(s) Scale Ozone/TOC (mg/mg) EBCT (min) Media BAC Throughput, 10 0 0-BVsa TOC Removal Reference

Adelaide, AU Bench, Pilot None 15, 30 BAC, Sand 23 - 30 (onsite) 3–15% Ho et al. (2011)
Albuquerque, USA Pilot 0.5–2.0 20 Anthracite N.A. 6–20% Lee et al. (2012)
Caboolture, AU Pilot, Full None, 0.5 18–120 BAC, Sand 7 - 68 (onsite) 22–63% Reungoat et al. (2011)
Caboolture, AU Pilot, Full None, 0.7 18, 60 BAC, Sand 25 - 67 (onsite) 23–51% Farre et al. (2011)
California, USA (2) Bench None, 0.4–1.0 15–60 BAC 4–5 (onsite) 59–76% Chuang & Mitch (2017)
Dubendorf, CH Pilot None, 0.6 10–18 BAC, Sand 28–37 (onsite & offsite) 13–20% Bourgin et al. (2018)
Golden, USA Pilot None 15–45 BAC 13 (onsite) 65–70% Vatankah et al. (2019)
Istanbul, TU Bench None 9, 18 BAC (2) 22 (onsite) 17–37% Kalkan et al. (2011)
Las Vegas, USA Pilot None, 1.0 2–20 BAC, Anthracite 5+ years (offsite) 9–40% Bacaro et al. (2019)
Las Vegas, USA Pilot 0.7–2.6 2–20 BAC, Anthracite 5+ years (offsite) 6–30% Arnold et al. (2018)
Las Vegas, USA Pilot None, 0.4–1.1 2–15 BAC, Anthracite 5+ years (offsite) 0–29% Gifford et al. (2018)
Reno, USA Pilot 0.7 3–30 BAC 22 - 200 (onsite) 0–33% Gerrity et al., 2011
Windhoek, NM Pilot None, 0.5 7–21 BAC 16–49 (onsite) 4–28% Olivier (2001)
USA (6) Bench None, 1.0 5–30 Anthracite N.A. 1–39% Table SI 2.1

AU: Australia; CH: Switzerland; TU: Turkey; NM: Namibia; USA: United States of America.
a
Throughput achieved prior to being reported as “exhausted” for TOC adsorption.

Some studies evaluated multiple WWTPs and some WWTPs were


evaluated in multiple studies; data were grouped by WWTP rather
than by study to isolate variations in treatability by water qual-
ity rather than research publication. The median number of obser-
vations per WWTP was 3 with a maximum of 86. To avoid over-
leveraging WWTPs with more available data, a preprocessing step
was used to consolidate the data for each WWTP effluent by bin-
ning across similar operating parameters. The following bins were
used: O3 /TOC (mg/mg; 0, >0 to <0.6, 0.6 to <0.9, 0.9 to < 1.2,
>1.2); EBCT (min; 0 to <8, 8 to <16, 16 to <24, 24 to <32, >32);
media type (inert, BAC). For each WWTP effluent bin combination,
the average of each inclusion criteria (i.e., TOC removal, O3 /TOC,
EBCT, and BAC throughput) were retained as single estimators for
that bin. The resulting biofiltration performance dataset, n = 160,
was used for further evaluation. Studies which systematically eval-
uated different operating conditions are summarized in Table 1
and the full dataset is available in Table SI 1.1

3. Results and discussion

3.1. Biofiltration dataset descriptive statistics

The distributions of biofilter influent EfOM TOC concentra-


tions and treatment operating parameters O3 /TOC, EBCT and BAC
throughput in the initial dataset are shown in Fig. 1. We evalu-
ated each distribution and determined that the dataset is a good
representation of typical field-scale conditions as follows. The TOC
median concentration, 7.5 mg/L, and interquartile range (IQR), 6.3
to 9.6 mg/L, are representative of typical WWTP effluents; for ex-
ample, these values are consistent with surveyed WWTPs (n = 38)
that practice nitrification (Krasner et al., 2009). The median spe-
cific ozone dose, 0.8 mg O3 / mg TOC, and IQR (0.6 to 1.0) coincide
with doses typically used to achieve multiple ozonation objectives
in potable reuse, including disinfection and trace organic contam-
inant mitigation while balancing formation of negative byproducts
and cost (Gerrity et al., 2012, 2014; Lee et al., 2013). In 35 cases,
41%, ozone was not utilized (Figure SI 1.1). The median EBCT of
15 min (IQR 10 to 30 min) is consistent to the median EBCT of
18 min in the full-scale facilities (n = 9) in the dataset. For BAC
media, the throughput (in number of bed volumes treated) prior to Fig. 1. Distribution of effluent organic matter levels and biofilter operating param-
sampling had a wide and distributed range, from below 20 0 0 bed eters evaluated in the biofilter performance dataset. Boxplots show the median and
volumes to higher than 10 0,0 0 0. Not shown in Fig. 1 are through- interquartile range with individual data points offset for clarity. For each parameter,
the unique values for each source water in each study are shown once. Data for
put values above 60,0 0 0 bed volumes (n = 16), which represents
no pre-ozonation, EBCT > 70 min, and BAC throughput exceeding 60,0 0 0 BVs are
BAC media from full-scale filters which had been in use for multi- shown in Figure SI 1.1.
ple years (i.e., 2+ years at the median 17 min EBCT). The impact of
residual adsorption at low throughput values was separately eval-
uated (Section 3.3).

3
E.S. Peterson and R.S. Summers Water Research 199 (2021) 117180

3.2. Biodegradable EfOM occurrence and removal O3 /TOC (mg/mg, units used throughout), is often used to nor-
malize ozone dose and exposure, reported as the nitrite-adjusted
Dissolved organic matter, in both WWef and DW source waters, dose to isolate reactions with organics. Despite variations in EfOM
consists of a range of heterogeneous compounds that have a dis- composition across different wastewater effluents, strong relation-
tribution of biodegradable behavior. A biodegradable fraction and ships have been reported between O3 /TOC and ozone exposure
a recalcitrant fraction are operationally defined based on labora- (Buffle et al., 2006) and change in UV absorbance (Wert et al.,
tory tests. The biodegradable fraction represents a continuum with 2009; Arnold et al., 2018) which support this approach.
some compounds being readily consumed - often term rapid, fast, Despite wide evaluation of ozone reactivity with dissolved com-
or easily biodegradable - and others being more recalcitrant, but pounds, relatively few studies have quantified the effect of ozone
ultimately biodegradable, given time and the method of charac- dose on changes in the biodegradable fraction of EfOM (SI Fig-
terization. Two measures of the biodegradable fraction, AOC and ure 4.2). Lee et al. (2012) systematically evaluated the change in
BDOC, commonly applied to natural waters and for treatment and BDOC resulting from specific ozone doses in the 0.3 to 3.0 range
distribution of conventional drinking water, have also been applied on an MBR effluent (TOC = 3.7 mg/L). Ozonation caused a sub-
to WWef (Servais et al., 1987; Khan et al., 1998b; Weinrich et al., stantial increase in BDOC/DOC, from about 14% with no ozonation
2009). up to 45% at O3 /TOC of 1.0 and a maximum of 61% at O3 /TOC
of about 2.0. The change in BDOC was approximately linear in
3.2.1. Occurrence of biodegradable EfOM without ozone the O3 /TOC range of 0 to 1 with decreasing BDOC yield at higher
Secondary treatment (biological) at the WWTP does not achieve doses. Li et al. (2006) found that an O3 /TOC dose of about 1.0 in-
complete BOM removal. In a survey of 21 WWTP effluents, an av- creased the BDOC/DOC from about 25% to 60% and in the internal
erage biodegradable EfOM fraction, BDOC/DOC, of 23% and a range six WWef study it increased from an average of 40% to 62% (n = 6,
of 4%−41% were reported (Weinrich et al., 2010). A similarly wide Table SI 2.1), while other studies report smaller effects after ozona-
range of results were found across other surveys: range 3%−66%, tion at similar doses (Amy et al., 1996; Wang et al., 2008). On an
n = 17 (Krasner et al., 2009), range 9%−33%, n = 13 (Khan et al., AOC basis, Wert et al. (2007) observed an increase from 5% to 16%
1998a), range 23–35%, n = 8 (Babcock et al., 2001) and range after ozonation of a WWTP effluent practicing tertiary coagulation,
33%−44%, n = 6 from the internal six WWef study (SI Table 2.2.). with the majority of the increase occurring below O3 /TOC of 0.5; in
The results from these and other studies are summarized in SI Ta- contrast, Li et al. (2017) reported a higher AOC increase (31%) from
ble 4.1, and taking all data together results in a median BDOC/DOC a secondary treated WWTP and with the majority of the increase
of 26%, interquartile range 18%−34%, and a minimum and maxi- occurring above O3 /TOC of 0.4.
mum of 3% and 70%, respectively (SI Fig 4.1). Lower effluent BDOC Taken together, EfOM ozonation at doses to achieve typical ob-
and TOC have been associated with WWTPs that provide higher jectives (O3 /TOC 0.6 to 1.0, described in Section 3.1) can be ex-
levels of biological treatment, either based on longer solids reten- pected to yield a substantial BOM fraction, measured either by
tion time (Khan et al., 1998a) or level of nitrification/denitrification BDOC or AOC, across many types of biological processes used at
achieved (Krasner et al., 2009). However, in this dataset there is WWTPs. An O3 /TOC of 1.0 yields a median BDOC/TOC of 59%,
no significant correlation between effluent TOC and the biodegrad- which is 50% higher than without ozone on a paired basis (39%,
able fraction when TOC is above 3 mg/L (p = 0.25, r2 = 0.02, SI n = 10) and more than double when compared to the dataset as a
Fig 4.3), which indicates that WWTP effluent TOC concentration whole, i.e., on an unpaired basis (26%, n = 61). However, the small
may not be a reliable surrogate for its biodegradability. The type BDOC dataset for ozonated EfOM (n = 14) indicates that research
of WWTP treatment process has also been associated with differ- is needed to better understand the dose-response relationship be-
ent levels of BDOC, with attached growth processes (i.e., trickling tween ozone and formation of BDOC and AOC in wastewater efflu-
filters) having higher BDOC/DOC, 35% on average (Weinrich et al., ents. Understanding the total biodegradable fraction, and develop-
2010) and up to 70% (Cha et al., 2004), while MBR effluents com- ment of more precise estimates of the biodegradable continuum, is
monly have low BDOC/DOC, about 10%−15% (Krasner et al., 2009; needed to properly evaluate biofilter performance and to account
Weinrich et al., 2010; Lee et al., 2012)(Figure SI 4.1). However, for the possibility of biofilm growth downstream in distribution
the extent to which these differences result from inherent effects systems or building premise plumbing as potable reuse becomes
of the treatment processes or other variables has not been well- more widely adopted.
defined. Weinrich et al. (2010) in their survey (n = 21) found
a median AOC/BDOC of 50% (IQR 35%−68%), indicating that the
typical biodegradable fraction of WWef is relatively balanced be-
tween readily assimilable compounds and more recalcitrant, but
ultimately biodegradable, constituents. 3.2.3. Removal of biodegradable EfOM during biofiltration
Incomplete BOM removal is observed during biofiltration un-
3.2.2. Occurrence of biodegradable EfOM after ozone der normal EBCTs (<60 min). BDOC removals of 30% (Lee et al.,
Ozone is widely acknowledged to substantially increase the 2012), 60% (Li et al., 2006), and 75% (Kalkan et al., 2011) have
biodegradability of DOM, as such biofiltration is often placed been reported from EfOM. In the internal six WWef study, the av-
downstream of an ozonation process (Gerrity et al., 2013). Ozone erage BDOC removal across the six wastewater effluents ranged
reacts quickly with unsaturated bonds in DOM, breaking aro- from 10% (no ozone, 5 min EBCT) to 50% (O3 /TOC 1.0, 30 min
matic rings and forming oxygenated low molecular weight byprod- EBCT). This range of BDOC removal is similar to that seen during
ucts including carboxylic acids, aldehydes, and ketones (von Gun- biofiltration of conventional drinking waters, 25–80% (Juhna and
ten, 2003). Ozonation byproducts are themselves ozone resistant Melin, 2006). Similarly, Terry and Summers (2018) in their meta-
and accumulate with increasing ozone dose, thus ozonation largely analysis of drinking water biofilter performance found that median
transforms, but does not mineralize DOM (von Gunten, 2003). In TOC removal was about 50% of median BDOC for both ozonated
the absence of nitrite or other reduced inorganics, ozone decom- and non-ozonated waters. Given that only 8% of the literature
position and exposure is directly controlled by reactions between treatment data found was expressed as BDOC or AOC, the results
ozone and EfOM (Buffle et al., 2006). The specific ozone dose, herein will be expressed on a TOC basis.

4
E.S. Peterson and R.S. Summers Water Research 199 (2021) 117180

Fig. 2. Effect of influent TOC and throughput on BAC biofilter steady-state performance. a) Filter run time required to reach steady state (i.e., fully biological mode) can vary
substantially based on influent TOC. A power relationship for n = 52 breakthrough curves demonstrates that progressively higher throughput is required to reach steady
state when influent TOC is lower (i.e., better pretreatment). b) The highest TOC removal in BAC biofilters was achieved only when media had low throughput. Data were
re-classified as “GAC” when steady state may not have been achieved (BVs treated were less than 20,0 0 0 and below the power relationship in Fig. 2a). c) TOC removal
was higher in “GAC” filters than in “BAC” filters (t-test, p < 0.001). Only data for “BAC” filters, which exhibit the characteristics of long-term performance, were retained in
subsequent biofilter analysis. Note: data shown for EBCT ≤ 60 min.

3.3. Determination of steady-state BAC biofilter performance the type and concentration of biomass needed to metabolize BOM
which takes a few weeks to a few months (Wang et al., 1995;
The use of GAC in biofilters has many benefits which include Greenstein et al., 2018; Sundaram et al., 2020) and in the adap-
residual adsorption capacity for strongly adsorbing compounds tation of the biomass to degrade more recalcitrant compounds
and the ability to attenuate influent spikes of adsorbing com- which can take longer (Zearley and Summers, 2012). . The tran-
pounds. The transition of fresh GAC into biological activated car- sition to steady state bioremoval is asymptotic and dependent on
bon (BAC) consists of decreasing adsorptive removal that over- water quality factors, such as temperature and BOM characteris-
laps with the establishment of bio-removal, which makes charac- tics. To best characterize long-term BAC performance the unsteady-
terization of biological performance difficult (Fig. 2 in Terry and state contribution of adsorption should be minimized. If the BAC
Summers, 2018). The adsorption capacity utilization phase can filter is not operated for enough time, an over-estimate of the
last for several months and longer depending on filter size and long-term BOM removal will be made. As such, an estimate of the
the physiochemical properties of both the adsorbent and adsor- time or throughput needed to allow adsorption to become a small
bate; for example, slow adsorption of large molecular weight DOM part of the overall BOM removal is needed. Zachman and Sum-
into GAC micropores can take months (Peel and Benedek, 1983; mers (2010) have systematically shown that adsorption capacity is
Summers and Roberts, 1987; Sontheimer et al., 1988) and more higher for low DOM concentration and pH, smaller GAC size and
strongly adsorbing compounds can take months to years to de- longer EBCT. In particular, the run time (throughput) to a target ef-
plete the adsorption capacity (Kennedy et al., 2015). This over- fluent concentration is inversely proportional to the DOM influent
laps with the bioacclimation phase, both in the establishment of concentration.

5
E.S. Peterson and R.S. Summers Water Research 199 (2021) 117180

We hypothesized that high TOC removal observed in some for which influent TOC was low (relative throughput less than 1),
studies with BAC media might be explained in part by residual indicating non-steady-state behavior even when all biofilters were
adsorption rather than biofiltration alone, overstating steady-state reported to be operating at steady state.
biofilter performance. A literature review of full- or pilot-scale GAC To account for the significant effect of throughput on TOC re-
application to WWEfs was conducted and the transition from GAC moval in this meta-analysis, data for adsorptive media filters were
adsorption to “steady state” (SS) TOC bioremoval, i.e., BAC, as a re-classified as “BAC” when throughput exceeded 20,0 0 0 bed vol-
function of influent TOC concentration was assessed for 26 GAC umes or when relative throughput was 1.0 or higher and as “GAC”
runs (Fig. 2a, summarized in Table SI 5.1). Other parameters known when neither threshold was achieved. Of 89 data points, 40 were
to impact steady state throughput, e.g., temperature, pH and GAC classified as “GAC”, indicating that a large portion of the dataset
type and size, were not reported with enough frequency to quan- represents biofilters for which adsorptive processes could be ex-
tify their impact. The average steady state removal after GAC tran- pected (Fig. 2c). TOC removal with “GAC” media was 29% percent-
sition to BAC was 25% ± 8%. It is often difficult to establish the age points higher on average compared to “BAC” media, even when
point at which “steady state” removal is achieved. When sufficient accounting for effects of EBCT and O3 /TOC (ANCOVA, p < 0.001). As
data were available a slope of 0.3% change / 10 0 0 bed volumes was a result, data for “GAC” media are excluded from further discussion
used as the steady state removal criterion. However, most stud- and the meta-analysis.
ies were plagued by insufficient data at long run times and varia-
tions in the influent concentration or upstream operating condi- 3.4. Effects of biofilter operating parameters on WWef TOC removal
tions. The relationship between influent TOC and throughput to
80% breakthrough from the meta-analysis of 26 GAC runs reported The effects of ozone, EBCT and media type, on TOC removal are
by Benstoem et al. (2017) is also shown in Fig. 2a. Again, other shown in Fig. 3, using bins to illustrate the dataset as a whole (i.e.,
parameters known to impact GAC performance were not reported. on an unpaired basis). For each bin no systematic differences in the
Throughput data from both the field studies evaluated at steady prevalence of the other parameters were found, thus the effects
state (average 25% removal) and Benstoem et al. (2017) study at of ozone, EBCT, media type, and water temperature are reviewed
80% breakthrough (20% removal) were combined and the overall individually in this section. Analysis of the operating parameters all
regression for these 52 data points is also shown in Fig. 2a. While together, with ozone dose and EBCT taken as continuous variables,
influent pH, EBCT and GAC type and size are not modeled by this will follow in Section 3.5.
relationship, it can serve as a guide for estimating the run length
of a GAC adsorber with coal based GAC (bituminous or lignite) to 3.4.1. Ozone and ozone dose
the point at which adsorption is not a major factor in TOC re- For the dataset as a whole, biofiltration without pre-ozonation
moval. The relationship can be further estimated by the product (n = 44) yielded a median TOC removal of 13% (Fig. 3) and with
of 90,0 0 0 to the inverse of the influent TOC. For lower pH (<7.5), pre-ozonation, which increases the BOM fraction, the TOC removal
longer EBCTs (>15 min) and smaller GAC particle size (12 × 40 increased to 21% (n = 73). Higher TOC removal after pre-ozonation
sieve), longer run times would be expected before adsorptive TOC is also reported in studies with paired data, and the magnitude of
removal is negligible. Thus, new pilot studies that are only inter- the difference is often higher than indicated in the whole dataset.
ested in long-term bioremoval assessment and not evaluating GAC These paired data cover a wide range of WWTPs and biofilter op-
to BAC transition should consider using GAC from other sites with erating parameters, summarized in Table SI 6.2. Most commonly,
similar water quality. This media should be re-acclimated to the TOC removal was 10–20 percentage points (p.p.) higher for biofil-
new source water until a steady state removal is established. ters after ozonation (O3 /TOC 0.5–1.1) compared to no ozonation
The effect of throughput on bulk organic matter removal (as (Olivier, 2001; Farré et al., 2011; Gifford et al., 2018), and the inter-
TOC) for adsorptive media in the biofiltration performance dataset nal six WWef study yielded 7–18 p.p. higher removal after ozona-
(n = 89) is shown in Fig. 2b using a bin width of 10,0 0 0 bed tion at O3 /TOC 1.0 (Table SI 2.1). The difference in TOC removal
volumes. Only steady state operation data, either as defined by ranged from as high as 30 p.p. (Bacaro et al., 2019) to as low as 4
the authors of these studies or as determined in this review by p.p. (Bourgin et al., 2018).
analysis of the endpoints of the TOC breakthrough curves when The impact of different ozone doses, however, is more var-
available, are shown in Fig. 2b. Under steady state conditions ied. Gifford et al. (2018) fit linear relationships between ozone
average TOC removal is independent of throughput, however in dose and TOC removal across O3 /TOC of 0.3–1.2 for BAC and an-
the dataset the median TOC removal for biofilters with less than thracite media on an MBR effluent. The fitted slopes were 1.2–1.4
10,0 0 0 bed volumes treated was nearly four-fold higher than for p.p. higher TOC removal per 0.1 O3 /TOC dose increase, indicating
those with greater than 40,0 0 0 bed volumes treated, 70% and that large changes in O3 /TOC were required to achieve substantially
18%, respectively. Compared to 40,0 0 0+ bed volumes, TOC removal higher TOC removal. Similarly, Lee et al. (2012) observed a 9 p.p.
was higher for 0–10,0 0 0 (p < 0.001) and 10,0 0 0–20,0 0 0 bed vol- increase in TOC removal when O3 /TOC was increased from 0.5 to
umes (p < 0.001) but was not significantly different for bins above 1.0, though the TOC removal was much smaller than the increase
20,0 0 0 bed volumes (p ≥ 0.3) (ANOVA and Tukey’s HSD). in O3 -derived BDOC (i.e., BDOC/TOC of 54%, compared to 14% with-
Additionally, the relative throughput, defined as the ratio of out ozone) which may be the result of media type (anthracite) or
throughput achieved to predicted throughput required for reach- EBCT (20 min). Applying even higher ozone doses (O3 /TOC ≥2.0
ing steady state (Fig. 2a) was calculated for the dataset. A relative vs ≤1.0) can further improve TOC removal, by about 5–10 p.p.,
throughput of 1.0 or greater indicates that sufficient bed volumes but the marginal benefit is lower (Lee et al., 2012; Arnold et al.,
were treated to reach steady state based on influent TOC, assum- 2018). In contrast, some studies report no substantial increase
ing the organic matter exhibits average treatability. For through- in TOC removal with increasing ozone dose across O3 /TOC 0.35–
put below 20,0 0 0 bed volumes, median TOC removal when rela- 1 (Chuang and Mitch, 2017) or 1.0–2.0 (Sundaram et al., 2020).
tive throughput was 1.0 or above (26%) was lower than for relative Zhu et al. (2015) reported small incremental improvements in DOM
throughput below 1.0 (68%), while for throughput above 20,0 0 0 removal on a COD basis across O3 /TOC 0.6 to 1.2; although re-
bed volumes median TOC removal (20%) was independent of rel- moval reached about 10 p.p. higher on average at these endpoints,
ative throughput (Figure SI 5.1). In combination, higher TOC re- the difference in means was smaller than the standard devia-
moval was associated with adsorptive media biofilters which were tions. Overall, pre-ozonation improves biofilter performance, but
evaluated at low throughput (less than 20,0 0 0 bed volumes) and only modest gains can be expected at incrementally higher ozone

6
E.S. Peterson and R.S. Summers Water Research 199 (2021) 117180

Fig. 3. Distributions of TOC removal across biofilter operating parameters. Bins correspond to the pre-processing method (i.e., each source water is shown a maximum of
once per bin). Each data point is shown once per panel (n = 118).

doses (e.g., from 0.8 to 1.0). Bigger performance gains are expected removal for a chemically-activated carbon. A challenge with inter-
with much larger ozone doses but, because of higher cost and chal- preting BAC results, even with paired data, is that each EBCT has
lenges with bromate formation, ozone dose will likely be deter- achieved a different throughput, therefore the role of EBCT can
mined by other treatment objectives in a potable reuse scenario be confounded with the role of residual adsorption. As an exam-
such as disinfection. ple, a GAC to BAC filter operated for 9 months will have treated
about 80,0 0 0 BVs at 5 min EBCT but just 13,0 0 0 BVs at 30 min;
3.4.2. Empty bed contact time it could be expected that the biomass profile throughout the fil-
Longer EBCTs can increase DOM removal as more exposure to ter has reached steady state and that minimal adsorption capacity
biomass is provided (Wang et al., 1995). For the whole dataset, in- remains in the upper filter, but the lower filter remains squarely
creasing the EBCT from below 8 min (n = 28) to 8 to <16 min in the adsorption mass transfer zone. While the above-referenced
(n = 44) increased the median TOC removal from 10% to 17% studies were screened for BAC throughput to minimize adsorptive
(Fig. 3). Incremental DOM removal decreases at longer EBCTs as confounding (Section 3.3) it is not possible to isolate biological
the more easily biodegradable substrate is preferentially removed processes within the data itself. The study by Vaidya et al. (2019),
in the upper filter leaving only more recalcitrant compounds in the however, reported 21% TOC removal in a BAC biofilter (5 min EBCT)
lower filter (Zearley and Summers, 2012). In drinking water treat- and an additional 20% TOC removal in a GAC adsorber (10 min
ment TOC removal typically begins to plateau at EBCTs in the range EBCT) which was operated to 40,0 0 0 bed volumes and exhibited
of 5 to 15 min (Terry and Summers, 2018), however as shown in pseudo-steady-state behavior. This indicates that EBCT played a
Fig. 3 no plateau in median EfOM removal is evident up through major role in the 16 p.p. increase in TOC removal at the total
30 min EBCT. This is consistent with higher overall BDOC in WWefs 15 min EBCT compared to 5 min for BAC media.
compared to typical drinking waters (Section 3.1) and suggests that Both the dataset as a whole and the paired-data studies show
a substantial portion of the BDOC is relatively recalcitrant, requir- that EBCTs above 8 min normally provide meaningfully higher TOC
ing longer EBCTs for removal. Biofilters treating WWef typically removal than lower EBCTs. In many cases, extending EBCT above
have utilized longer EBCTs, median 18 min in the nine full scale 20 or 30 min can also have a significant benefit. This supports the
facilities in the dataset, than those utilized in conventional drink- current practice of using longer EBCTs than in conventional drink-
ing water biofilters, median 7 min (Brown et al., 2016). ing water treatment and indicates that even longer EBCTs could be
In the most extensively investigated source water with paired useful. In practice, designers and operators must balance the in-
data, an MBR pilot plant using both anthracite and 5+ year old creased cost and operational complexity of longer EBCTs with po-
BAC, maximum TOC removal was typically achieved at or below tential negative impacts of a residual biodegradable EfOM being
10 min EBCT (Gifford et al., 2018; Bacaro et al., 2019). The excep- released to the subsurface or distribution system in potable reuse
tion was after high ozone dose (O3 /TOC 2.3), where TOC removal scenarios.
increased by 6 p.p. when EBCT was increased from 10 min to
3.4.3. Media type
20 min (Arnold et al., 2018). In contrast, in the internal six WWef
As stated in Section 3.3, GAC is often used for biofilter media as
study using anthracite media across six non-MBR WWTPs, higher
it offers several advantages over inert media including better TOC
TOC removal was consistently observed with increasing EBCT in
removal. However, the performance of biofilters that use GAC as
each water, both from EBCT 5 to 15 min and from 15 to 30 min,
media can be substantially higher due to residual TOC adsorption
and both with and without ozone (Table SI 2.1). In these waters,
capacity when prior throughput and influent TOC are low. The BAC
increasing EBCT from 5 to 15 min after pre-ozonation improved
data in Fig. 3 does not include that from biofilters with GAC that
TOC removal by an average of 10 p.p., and increasing EBCT to
did not meet the minimum throughput criteria (20,0 0 0+ BV) and
30 min improved TOC removal by an additional 10 p.p.
the TOC removal with BAC media was still higher than inert me-
Other studies which evaluated the effect of EBCT have used BAC
dia on average by 7 percentage points, 15% vs 23%. Similarly, paired
media which started as fresh GAC (Table SI 6.3). In these studies,
studies that evaluated both inert and BAC media (also 20,0 0 0+ BV)
extending EBCT to longer than 20 min improved TOC removal by
under the same feed water and operational conditions (i.e., ozone
up to 8 p.p. (Olivier, 2001), 12 p.p. (Gerrity et al., 2011), and 23
dose and EBCT) almost always report higher TOC removal with
p.p. (Farré et al., 2011), compared to performance below 20 min.
BAC media than inert media (SI Table 6.1). The largest differences
Kalkan et al. (2011) investigated two types of BAC media in paral-
in TOC removal were in the range of 21–27 percentage points
lel and reported a similar increase in TOC removal for a thermally-
higher in BAC filters in two studies that use long EBCTs, 60 min
activated carbon (7 p.p., EBCT 9 to 18 min) but no increase in TOC
7
E.S. Peterson and R.S. Summers Water Research 199 (2021) 117180

Table 2
Standardized linear regression coefficients by operating parameter.

Coefficient (β ) β Standard Error Mean (μ) Standard Deviation (σ 2 ) p

Intercept 13.2% 0.7% – – <0.001


Media Type (BAC) 8.7% 1.1% – – <0.001
O3 /TOC 5.3% 0.6% 0.48 0.44 <0.001
EBCT 3.9% 0.6% 14.4 min 8.5 min <0.001
(EBCT)∗ (O3 /TOC) 1.4% 0.5% – – 0.011

and 40 min, respectively (Farré et al., 2011; Reaume et al., 2015). jority of the dataset (n = 98 out of 118). Data outside the model
Smaller increases in BOM removal with BAC were reported in the space, plus one high-leverage value, were removed as outliers.
other studies. Ho et al. (2011) observed 10–12 p.p. higher TOC re- The regression model results are shown in Table 2 with simu-
moval in BAC under non-ozonated conditions, with the difference lated TOC removal represented by the standardized regression co-
driven by very low TOC removal in the inert media columns, 3– efficient (β ), and visualized in Fig. 4a with isoremoval lines as a
5%. In several studies with MBR pilot plant effluent the TOC re- function of O3 /TOC and EBCT for each media type. The regression
moval ranged from 6 to 40% and the BAC filter consistently out- intercept represents the simulated TOC removal when all parame-
performed the parallel anthracite filter by 1–12 p.p. (median 5.5 ters are at the dataset average (μ) and remaining coefficients rep-
p.p.) (Arnold et al., 2018; Gifford et al., 2018; Bacaro et al., 2019). resent the change in simulated TOC removal when the operating
Hooper et al. (2020) found a 7 percentage point effect for BAC parameter is changed by one standard deviation (σ 2 ). The TOC
compared to anthracite at low temperatures, 12–15 o C, despite a removal simulation range (5–32%) matches the observed 5th and
shorter EBCT (7.5 min vs 12 min). In contrast, Zhu et al. (2015) re- 95th percentiles in the dataset (3–31%) and each main effect cor-
ported no significant difference between anthracite and BAC on a related strongly with TOC removal (p < 0.001). TOC removal with
COD basis and (Bourgin et al., 2018) observed slightly higher TOC BAC media was substantially higher than for inert media, 8.7 per-
removal in a full scale sand filter compared to a pilot scale BAC centage points on average after controlling for adsorption, which
filter (20% vs 18%) and despite the BAC filter having a longer EBCT supports the use of BAC media in reuse biofilters. Ozone dose af-
(18 min vs 10 min). fected TOC removal by 1.2 p.p. per 0.1 change in specific dose,
In these paired studies, the advantage of BAC over inert me- similar to that found by Gifford et al. (2018) for one feed wa-
dia biofilters was greatest for systems that started with onsite GAC ter (avg 1.3 p.p. per 0.1 change in specific dose). For both ozone
and ran the filters long enough, 20,0 0 0+ BV, to yield BAC, com- dose and EBCT (i.e., 1 p.p. per 2.2 min change in EBCT) relatively
pared to studies that sampled used GAC from an offsite full-scale large operational changes are required to expect meaningful im-
facility or previous studies. In the latter case, in which the risk pact on TOC removal. Additionally, strong evidence was found of a
of residual non-steady-state adsorption is less likely, performance positive interactive effect between EBCT and O3 /TOC (p = 0.011);
with BAC media ranged from about 10 percentage points better to this suggests that longer EBCT may be more effective when more
no substantial effect. In total, there is strong evidence that BAC me- BOM is formed at higher ozone doses, however the actual effect
dia can be expected to outperform inert media, even when con- size (β = 1.4%) was relatively small. Such an interactive effect was
trolling for residual non-steady-state adsorptive capacity, however observed experimentally by Gifford et al. (2018) where increasing
very large differences may indicate non-steady state behavior in ozone dose also required a longer EBCT to maximize TOC removal.
the BAC which should be closely evaluated. Weaker evidence was found for a positive interactive effect be-
tween BAC media and longer EBCT (β = 2.0%, p = 0.18). Applying
these interactive effects is difficult because relatively few data ex-
3.4.4. Temperature
ist for high O3 /TOC or for BAC media with high EBCT; this analysis
Water temperature affects microbial activity and can signifi-
suggests that future research in these areas is warranted.
cantly affect the kinetics of BOM removal in biofilters. For exam-
The meta-analysis multiple linear regression was statistically
ple, in drinking water treatment achieving 50% BDOC removal re-
significant (p < 2E-16) and provides good correlation (adj-
quires a 14 min EBCT on average for temperatures below 10 o C,
r2 = 0.64) considering the scope and context of this analy-
while only 4 min EBCT is required for temperatures above 20 o C
sis, i.e., evaluating three operational parameters for total (not
(Terry and Summers, 2018). The effects of temperature on kinet-
biodegradable) organic carbon removal across a diverse range of
ics of EfOM removal, however, have been largely neglected. In the
studies. Incorporating more precise water quality parameters and
dataset, only 17 of 43 studies reported water temperature data dur-
standardization of biofiltration methods (e.g., ensuring scalability
ing their evaluation of biofiltration. Thirteen of these studies eval-
(McKie et al., 2019; Terry et al., 2019)) may help explain the re-
uated a single temperature, each within the narrow range of 22–25
o C, and the others reported a wider temperature range (e.g., 13–28 maining variance. The regression presented here was judged to
o C) across their testing, but did not isolate the role of temperature be an appropriate representation of the dataset based on residu-
als analysis (Figure SI 7.1), and though transformed and non-linear
on biofilter performance. Thus, the impact of temperature on EfOM
model forms were also considered, none presented a better fit to
removal, specifically any important deviations from that seen dur-
the data. Future research is needed to more completely describe
ing conventional drinking water biofiltration, remains unknown.
biofilter performance at higher ozone doses and EBCTs, and the
least-squares average TOC removal presented here should not be
3.5. Regression evaluation of biofiltration performance extrapolated beyond the model space.
The distribution of the regression model residuals gives insight
The combined impacts of ozone dose, EBCT, and media type on into expected performance variations that can benchmark future
biofilter performance were evaluated with a linear regression. This research and implementation of reuse biofilters (Fig. 4b). The resid-
descriptive model approach was used in order to estimate the rel- uals interquartile range was ± 3.5 percentage points, (i.e., half of
ative contributions of each operating parameter on observed EfOM the data was within this range) indicating that operational param-
removal and to identify the variance associated with these con- eters alone yield a suitably precise treatability estimate on average
trollable factors. The model space was defined as EBCT 5–30 min, to be useful for preliminary design and for designing bench and
O3 /TOC 0-1.2, and TOC removal ≥ 0% which encapsulates the ma- pilot scale studies for many WWTP effluents. Further, the stan-

8
E.S. Peterson and R.S. Summers Water Research 199 (2021) 117180

Fig. 4. Expected TOC removal (%) as a function of ozone dose, EBCT and media type evaluated from linear regression. a) Isoremoval lines show simulated TOC removal. b)
Observed vs simulated TOC removal shown with residuals interquartile range (dashed lines) and residual standard error (dotted lines). Simulated TOC removal was accurate
for both laboratory and field scale data.

dard deviation of the residuals (5.6 percentage points) indicates 4. Conclusions, recommendations, and areas for future
the likelihood of biofilter performance which diverges from the research
average. For example, TOC removal outside of ± 11 p.p. (2 stan-
dard deviations) of the regression relationship is statistically un- • A meta-analysis was used to evaluate the effects of the opera-
likely given the current dataset, and observations of such outlier tion parameters, ozone dose, EBCT, and media type, on effluent
data should be closely studied as fundamental insights into good organic matter removal by biofiltration. Each operating param-
(or bad) biofilter performance could provide a large step forward eter had a significant impact on observed TOC removal, which
for the field. were quantified with a regression model.

9
E.S. Peterson and R.S. Summers Water Research 199 (2021) 117180

• Performance of BAC media can be significantly influenced by Benstoem, F., Nahrstedt, A., Boehler, M., Knopp, G., Montag, D., Siegrist, H., Pin-
run time (throughput) prior to data collection, and the com- nekamp, J., 2017. Performance of granular activated carbon to remove microp-
ollutants from municipal wastewater—a meta-analysis of pilot- and large-scale
bination of slow adsorption and biodegradation combined with studies. Chemosphere 185, 105–118.
normal influent water quality perturbations makes it difficult to Bourgin, M., Beck, B., Boehler, M., Borowska, E., Fleiner, J., Salhi, E., Teichler, R., von
ascertain steady-state behavior during transition from “GAC ad- Gunten, U., Siegrist, H., McArdell, C.S., 2018. Evaluation of a full-scale wastew-
ater treatment plant upgraded with ozonation and biological post-treatments:
sorber” to “BAC biofilter”. A minimum of 20,0 0 0 bed volumes abatement of micropollutants, formation of transformation products and oxida-
of throughput prior to evaluating biofiltration is recommended tion by-products. Water Res. 129, 486–498.
when using BAC media, and even higher throughput may be re- Brown, J., Upadhyaya, G., Carter, J., Brown, T., Lauderdale, C., 2016. North American
Biofiltration Knowledge Base. Water Research Foundation.
quired if influent TOC is below about 5 mg/L.
Buffle, M.-O., Schumacher, J., Meylan, S., Jekel, M., von Gunten, U., 2006. Ozonation
• Relatively few studies have evaluated the impact of ozone on and advanced oxidation of wastewater: effect of O3 dose, PH, DOM and HO•
the biodegradable EfOM fraction. An O3 /TOC of 1.0 yields a me- - scavengers on ozone decomposition and HO• generation. Ozone Sci. Eng. 28
(4), 247–259.
dian BDOC/TOC of 59%, which is more than 50% higher than
California Department of Public Health, 2018. Regulations Related to Re-
without ozone on an unpaired basis (26%, n = 61) and very cycled Water https://www.waterboards.ca.gov/board_decisions/adopted_orders/
few studies reported filter biomass. Improved knowledge of as- resolutions/2015/rs2015_0 0 02.pdf.
similable and total biodegradable EfOM and filter biomass can Cha, W., Fox, P., Mir, F.N., Choi, H., 2004. Characteristics of biotic and abiotic re-
movals of dissolved organic carbon in wastewater effluents using soil batch re-
be used to generate more precise estimates of biofilter perfor- actors. Water Environ. Res. 76 (2), 130–136.
mance. Chuang, Y.H., Mitch, W.A., 2017. Effect of ozonation and biological activated car-
• The effects of WWTP treatment level upstream of biofiltration, bon treatment of wastewater effluents on formation of N-nitrosamines and
halogenated disinfection byproducts. Environ. Sci. Technol. Technol. 51 (4),
seasonality, and temperature for EfOM removal remain to be 2329–2338.
understood. Escobar, I.C., Randall, A.A., 2001. Assimilable organic carbon (AOC) and Biodegrad-
• Meta-analysis indicates that higher ozone doses and EBCTs can able dissolved organic carbon (BDOC): complementary measurements. Water
Res. 35 (18), 4444–4454.
provide meaningful gains in TOC removal, however data at high Farré, M.J., Reungoat, J., Argaud, F.X., Rattier, M., Keller, J., Gernjak, W., 2011. Fate of
operating levels is relatively scarce. Future studies should in- N-nitrosodimethylamine, trihalomethane and haloacetic acid precursors in ter-
vestigate these levels in more depth to more precisely deter- tiary treatment including biofiltration. Water Res. 45 (17), 5695–5704.
Florida Potable Reuse Commission, 2020. Framework for the Implementation of
mine where diminishing returns in TOC removal occur across
Potable Reuse in Florida. Alexandria, VA.
many different source waters. Since high EBCTs and high ozone Gerrity, D., Gamage, S., Holady, J.C., Mawhinney, D.B., Quiñones, O., Trenholm, R.A.,
doses are costly, improved understanding of possible benefits is Snyder, S.A., 2011. Pilot-scale evaluation of ozone and biological activated carbon
for trace organic contaminant mitigation and disinfection. Water Res. 45 (5),
needed to properly support design decision making.
2155–2165.
• Sequential use of biofilters followed by GAC adsorbers could Gerrity, D., Gamage, S., Jones, D., Korshin, G.V., Lee, Y., Pisarenko, A., Trenholm, R.A.,
provide de facto extended biofilter EBCT, as the GAC adsorbers von Gunten, U., Wert, E.C., Snyder, S.A., 2012. Development of surrogate cor-
acclimate to the BOM and transition from GAC to BAC. This relation models to predict trace organic contaminant oxidation and microbial
inactivation during ozonation. Water Res. 46 (19), 6257–6272.
could be especially effective when the GAC adsorber is operated Gerrity, D., Pecson, B., Shane Trussell, R., Rhodes Trussell, R., 2013. Potable reuse
in parallel blended effluent mode, which could prolong the run treatment trains throughout the world. J. Water Supply Res. Technol. - AQUA 62
life of GAC adsorbers. (6), 321–338.
Gerrity, D., Owens-bennett, E., Venezia, T., Stanford, B.D., Plumlee, M.H., Debroux, J.,
Trussell, R.S., 2014. Applicability of ozone and biological activated carbon for
Declaration of Competing Interest potable reuse. Ozone Sci. Eng. 36 (2), 123–137.
Gifford, M., Selvy, A., Gerrity, D., 2018. Optimizing ozone-biofiltration systems for
organic carbon removal in potable reuse applications. Ozone Sci. Eng. 40 (6),
The authors declare that they have no known competing finan-
427–440.
cial interests or personal relationships that could have appeared to Greenstein, K.E., Lew, J., Dickenson, E.R.V., Wert, E.C., 2018. Investigation of biotrans-
influence the work reported in this paper. formation, sorption, and desorption of multiple chemical contaminants in pi-
lot-scale drinking water biofilters. Chemosphere 200, 248–256.
von Gunten, U., 2003. Ozonation of drinking water: part I. Oxidation kinetics and
Acknowledgments product formation. Water Res. 37, 1443–1467.
Ho, L., Grasset, C., Hoefel, D., Dixon, M.B., Leusch, F.D.L., Newcombe, G., Saint, C.P.,
The first author was funded by a Dean’s Fellowship, Department Brookes, J.D., 2011. Assessing granular media filtration for the removal of chem-
ical contaminants from wastewater. Water Res. 45 (11), 3461–3472.
of Civil, Environmental, and Architectural Engineering, University of Hooper, J., Funk, D., Bell, K., Noibi, M., Vickstrom, K., Schulz, C., Machek, E.,
Colorado Boulder. The authors acknowledge Sierra Johnson, Steven Huang, C.H., 2020. Pilot testing of direct and indirect potable water reuse us-
Shiokari, Hannah Peterson, and Professor Julie Korak for their as- ing multi-stage ozone-biofiltration without reverse osmosis. Water Res. 169.
Huck, P.M., 1990. Measurement of biodegradable organic matter and bacterial
sistance in data collection and analysis.
growth potential in drinking water. J. /Am. Water Work. Assoc. 82 (7), 78–86.
Juhna, T., Melin, E., 2006. Ozonation and biofiltration. Water Treatment - Opera-
Supplementary materials tional Status and Optimization Issues Techneau D.5.3.1.B:80.
Kalkan, Ç., Yapsakli, K., Mertoglu, B., Tufan, D., Saatci, A., 2011. Evaluation of biolog-
ical activated carbon (BAC) process in wastewater treatment secondary effluent
Supplementary material associated with this article can be for reclamation purposes. Desalination 265 (1–3), 266–273.
found, in the online version, at doi:10.1016/j.watres.2021.117180. Kennedy, A.M., Reinert, A.M., Knappe, D.R.U., Ferrer, I., Summers, R.S., 2015. Full- and
pilot-scale GAC adsorption of organic micropollutants. Water Res. 68, 238–248.
References Khan, E., Babcock, R.W., Viriyavejakul, S., Suffet, I.M.H., Stenstrom, M.K., 1998a.
Biodegradable dissolved organic carbon for indicating wastewater reclamation
Amy, G., Debroux, J.F., Arnold, R., Wilson, L.G., 1996. Preozonation for enhancing plant performance and treated wastewater quality. Water Environ. Res. 70 (5),
the biodegradability of wastewater effluent in a potable-recovery soil aquifer 1033–1040.
treatment (SAT) system. Rev. des Sci. l’eau /J. Water Sci. 9 (3), 365–380. Khan, E., Babcock Jr., R.W., Suffet, I.H., Stenstrom, M.K., 1998b. Method development
Arnold, M., Batista, J., Dickenson, E., Gerrity, D., 2018. Use of ozone-biofiltration for for measuring biodegradable organic carbon in reclaimed and treated wastewa-
bulk organic removal and disinfection byproduct mitigation in potable reuse ap- ters. Water Environ. Res. 70 (5), 1025–1032.
plications. Chemosphere 202, 228–237. Krasner, S.W., Westerhoff, P., Chen, B., Rittmann, B.E., Nam, S.-N., Amy, G.L., 2009.
Babcock, R.W., King, S., Khan, E., Stenstrom, M.K., 2001. Use of biodegradable dis- Impact of wastewater treatment processes on organic carbon, organic nitro-
solved organic carbon to assess treatment process performance in relation to gen, and DBP precursors in effluent organic matter. Environ. Sci. Technol. 43,
solids retention time. Water Environ. Res. 73 (5), 517–525. 2911–2918.
Bacaro, F., Dickenson, E., Trenholm, R.A., Gerrity, D., 2019. N-nitrosodimethylamine Lee, C.O., Howe, K.J., Thomson, B.M., 2012. Ozone and biofiltration as an alterna-
(NDMA) formation and mitigation in potable reuse treatment trains employing tive to reverse osmosis for removing PPCPs and micropollutants from treated
ozone and biofiltration. Environ. Sci. Water Res. Technol. 5 (4), 713–725. wastewater. Water Res. 46 (4), 1005–1014.
Basu, O.D., Dhawan, S., Black, K., 2016. Applications of biofiltration in drinking water Lee, Y., Gerrity, D., Lee, M., Bogeat, A.E., Salhi, E., Gamage, S., Trenholm, R.A.,
treatment - a review. J. Chem. Technol. Biotechnol. 91 (3), 585–595. Wert, E.C., Snyder, S.A., Von Gunten, U., 2013. Prediction of micropollutant elim-

10
E.S. Peterson and R.S. Summers Water Research 199 (2021) 117180

ination during ozonation of municipal wastewater effluents: use of kinetic and Sundaram, V., Pagilla, K., Guarin, T., Li, L., Marfil-Vega, R., Bukhari, Z., 2020. Extended
water specific information. Environ. Sci. Technol. Technol. 47 (11), 5872–5881. field investigations of ozone-biofiltration advanced water treatment for potable
Li, L., Zhu, W., Zhang, P., Zhang, Z., Wu, H., Han, W., 2006. Comparison of AC/O3-BAC reuse. Water Res. 172.
and O3-BAC processes for removing organic pollutants in secondary effluent. Terry, L.G., Summers, R.S., 2018. Biodegradable organic matter and rapid-rate biofil-
Chemosphere 62 (9), 1514–1522. ter performance: a review. Water Res 128, 234–245.
Li, W.T., Cao, M.J., Young, T., Ruffino, B., Dodd, M., Li, A.M., Korshin, G., 2017. Appli- Terry, L.G., Pruisner, P., Peterson, E., Dickenson, E., Wang, J., Summers, R.S., 2019.
cation of UV absorbance and fluorescence indicators to assess the formation of Scale-up methodology for biological filtration removal of dissolved organic mat-
biodegradable dissolved organic carbon and bromate during ozonation. Water ter. Environ. Eng. Sci. 36 (4), 405–412.
Res. 111, 154–162. U.S. Environmental Protection Agency, 2012. Guidelines for Water Reuse Washing-
McKie, M.J., Ziv-El, M.C., Taylor-Edmonds, L., Andrews, R.C., Kirisits, M.J., 2019. Biofil- ton, DC.
ter scaling procedures for organics removal: a potential alternative to piloting. Urfer, D., Huck, P.M., Booth, S.D.J., Coffey, B.M., 1997. Biological filtration for BOM
Water Res. 151, 87–97. and particle removal: a critical review. J. /Am. Water Work. Assoc. 89 (12),
Michael-Kordatou, I., Michael, C., Duan, X., He, X., Dionysiou, D.D., Mills, M.A., Fat- 83–98.
ta-Kassinos, D., 2015. Dissolved effluent organic matter: characteristics and po- Vaidya, R., Buehlmann, P.H., Salazar-Benites, G., Schimmoller, L., Nading, T., Wil-
tential implications in wastewater treatment and reuse applications. Water Res. son, C.A., Bott, C., Gonzalez, R., Novak, J.T., 2019. Pilot plant performance com-
213–248. paring carbon-based and membrane-based potable reuse schemes. Environ. Eng.
Olivier, J., 2001. Granular Activated Carbon Performance at Three Southern African Sci. 36 (11), 1369–1378.
Water Treatment Plants. Rand Afrikaans University. Wang, J.Z., Summers, R.S., Miltner, R.J., 1995. Biofiltration performance: part 1, rela-
Peel, R.G., Benedek, A., 1983. Biodegradation and adsorption within activated carbon tionship to biomass. J. /Am. Water Work. Assoc. 87 (12), 55–63.
adsorbers. J. Water Pollut. Control Fed. 55 (9), 1168–1173. Wang, S., Ma, J., Liu, B., Jiang, Y., Zhang, H., 2008. Degradation characteristics of
Reaume, M.J., Seth, R., McPhedran, K.N., da Silva, E.F., Porter, L.A., 2015. Effect of me- secondary effluent of domestic wastewater by combined process of ozonation
dia on biofilter performance following ozonation of secondary treated municipal and biofiltration. J. Hazard. Mater. 150 (1), 109–114.
wastewater effluent: sand vs. GAC. Ozone Sci. Eng. 37 (2), 143–153. Weinrich, L.A., Giraldo, E., LeChevallier, M.W., 2009. Development and application of
Reungoat, J., Escher, B.I., Macova, M., Keller, J., 2011. Biofiltration of wastewater treat- a bioluminescence-based test for assimilable organic carbon in reclaimed wa-
ment plant effluent: effective removal of pharmaceuticals and personal care ters. Appl. Environ. Microbiol. 75 (23), 7385–7390.
products and reduction of toxicity. Water Res. 45 (9), 2751–2762. Weinrich, L.A., Jjemba, P.K., Giraldo, E., LeChevallier, M.W., 2010. Implications of or-
Schimmoller, L., Kealy, M.J., 2014. Fit For Purpose Water: The Cost of Overtreating ganic carbon in the deterioration of water quality in reclaimed water distribu-
Reclaimed Water. Alexandria, VA. tion systems. Water Res. 44 (18), 5367–5375.
Servais, P., Billen, G., Hascoet, M.-C., 1987. Determination of the biodegradable frac- Wert, E.C., Rosario-Ortiz, F.L., Drury, D.D., Snyder, S.A., 2007. Formation of oxidation
tion of dissolved organic carbon in waters. Water Res. 21 (4), 445–450. byproducts from ozonation of wastewater. Water Res 41 (7), 1481–1490.
Shon, H.K., Vigneswaran, S., Snyder, S.a, 2006. Effluent organic matter (EfOM) in Wert, E.C., Rosario-Ortiz, F.L., Snyder, S.A., 2009. Using ultraviolet absorbance and
wastewater: constituents, effects, and treatment. Crit. Rev. Environ. Sci. Technol. color to assess pharmaceutical oxidation during ozonation of wastewater. Envi-
36, 327–374 (June). ron. Sci. Technol. 43 (13), 4858–4863.
Sontheimer, H., Crittenden, J.C., Summers, R.S., 1988. Activated Carbon for Water Zachman, B.A., Summers, R.S., 2010. Modeling TOC breakthrough in granular acti-
Treatment, 2nd Edition DVWG Forschungsstelle, Karlsruhe, Germany. vated carbon adsorbers. J. Environ. Eng. 136 (2), 204–210.
Summers, R.S., Roberts, P.V., 1987. Rate of humic substance uptake during activated Zearley, T.L., Summers, R.S., 2012. Removal of trace organic micropollutants by
carbon adsorption. J. Environ. Eng. ASCE 113 (6), 1333–1349. drinking water biological filters. Environ. Sci. Technol. 46 (17), 9412–9419.
Summers, R.S., Shiokari, S.T., Johnson, S., Peterson, E., Yu, Y., Cook, S., 2020. Reuse Zhu, I.X., Getting, T., Bruce, D., 2010. Review of biologically active filters in drinking
treatment with ozonation, biofiltration, and activated carbon adsorption for to- water applications. J. /Am. Water Work. Assoc. 102 (12), 67–77.
tal organic carbon control and disinfection byproduct regulation compliance. Zhu, I.X., Wang, J., Wieland, A., 2015. Ozone-enhanced biologically active filtration
AWWA Water Sci 2 (5), 1–14. for wastewater reuse. J. /Am. Water Work. Assoc. 107 (12), E685–E692.

11

You might also like