Journal of Cleaner Production 262 (2020) 121145

Contents lists available at ScienceDirect

Journal of Cleaner Production

journal homepage: www.elsevier.com/locate/jclepro

Economic analysis of electrodialysis, denitrification, and anammox for

nitrogen removal in municipal wastewater treatment
Donald Vineyard a, Andrea Hicks a, K.G. Karthikeyan b, Phillip Barak c, *
a
University of WisconsineMadison, Department of Civil and Environmental Engineering, 1415 Engineering Dr, Madison, WI, 53706, USA
b
University of WisconsineMadison, Department of Biological Systems Engineering, 228 Agricultural Engineering Building, 460 Henry Mall, Madison, WI,
53706, USA
c
University of WisconsineMadison, Department of Soil Science, 1525 Observatory Drive, Madison, WI, 53706e1299, USA

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

Article history: Technologies to remove nitrogen from wastewater are employed to preserve drinking water and prevent
Received 20 November 2019 environmental damage. Nitrification/denitrification and partial nitrification-anammox are two accepted
Received in revised form nitrogen removal techniques for wastewater treatment plants. These processes require energy for
12 February 2020
aeration and can release fugitive greenhouse gases in the form of nitrous oxide. Electrodialysis could
Accepted 15 March 2020
Available online 18 March 2020
potentially be used as an alternative to remove ammonium from waste streams, but previous experi-
mentation has concluded that concentrate streams experience rapid scaling and fouling of membranes.
Handling Editor: Panos Seferlis This analysis compiles literature values to compare the state-of-the-art of municipal nitrogen removal to
the new electrodialysis technique on an economic basis. Results show that nitrogen removal and re-
Keywords: covery by electrodialysis is estimated to lower both initial capital costs and subsequent operation costs
Nutrient recovery than traditional N removal technologies.
Nutrient recycling © 2020 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license
Tertiary treatment (http://creativecommons.org/licenses/by/4.0/).
Ammonium
Reject water
Optimization

1. Introduction decreased property values and increased respiratory illness to

The United States consumes in excess of 21 million tons of
nutrient fertilizer each year, 12 million tons of which is nitrogen (N)
(USDA, 2018). The bulk of US nutrient pollution originates from
these agricultural fertilizers, manure, and eroded runoff but a
sizeable portion also stems from wastewater (USEPA, 2015b). The
nutrients in wastewater are derived from, among other things,
biological waste from households and industrial chemicals (USEPA,
2015b). Environmental policy is shifting toward discouraging
nutrient pollution of waterways (USEPA, 2009) as mitigation
technologies mature and the nation becomes more conscious of the
deleterious effects of eutrophication. Point sources such as
municipal wastewater utilities are common targets for regulation
and control. The United States Environmental Protection Agency
(2015) reviews a variety of means by which nutrient pollution
impacts local and national economies; these impacts range from
* Corresponding author.
E-mail address: phillip.barak@wisc.edu (P. Barak).
increased cost of drinking water treatment.
Partial removal of N at wastewater treatment plants is a com-
mon practice (U.S. EPA, 2007, 2007). In addition to preventing
eutrophication of surface water bodies, N removal in-plant can
reduce the occurrence of ammonium-toxicity in municipal sludge
digesters (Desloover et al., 2012, 2015); improved digester perfor-
mance can reduce end waste mass and cost of disposal while
yielding higher biogas recovery for on-site energy generation
(Karim et al., 2005; Gungor and Karthikeyan, 2008; Holly et al.,
2017). Existing N removal methods focus primarily on encour-
aging bacteria to decompose complex nitrogenous compounds into
atmospheric nitrogen gas (N2) (Clark et al., 2010); nitrification/
denitrification and the partial nitrification-anammox processes are
the two most common examples (Clark et al., 2010). A new tech-
nology of interest is the use of electrodialysis with monovalent-
selective membranes to recover ammonium (NHþ 4 ) directly from
wastewater and concentrate it for re-use (Barak, 2018).
The nitrification/denitrification process is typically a two-part
process that encourages microorganisms to off-gas organic-N and
NHþ4 -N as N2 after a series of oxidation/reduction reactions (USEPA,

https://doi.org/10.1016/j.jclepro.2020.121145
0959-6526/© 2020 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
2 D. Vineyard et al. / Journal of Cleaner Production 262 (2020) 121145
1973; Clark et al., 2010). Nitrification/denitrification is a mature
technology with a variety of reactor designs that utilize specialized
bacteria strains (Clark et al., 2010). As wastewater enters the reactor,
it is aerated to increase the dissolved oxygen content. NHþ 4 oxidizing
bacteria use this oxygen to nitrify NHþ 
4 to nitrite (NO2 ) and then
nitrate (NO 3 ). Once sufficient oxidation is achieved and microbe
populations are high, aeration ceases and the oxygen-starved
microbiota begins denitrification, converting soluble NO 3 to N2
gas through a series of intermediates using the various N oxides as
electron acceptors coincident with anaerobic metabolism of dis-
solved organic carbon (C). The nitrification/denitrification process
requires energy in the form of electricity for aeration and mixing and
requires a C source for energy and biomass; this C is typically derived
from organic molecules extant in the process water and thus serves
to decrease biological oxygen demand (USEPA, 1973). Reactors can
add an external organic substrate, such as methanol, to improve N
removal by providing an easily reacted C source to be oxidized by
NO 3 (USEPA, 1973; Purtschert et al., 1996; Theis and Hicks, 2012).
Anaerobic ammonium oxidation (anammox) is a recent dis-
covery and a relative newcomer to the wastewater N management
industry (Lackner et al., 2014). Similar to nitrification/denitrifica-
tion, anammox relies on microbial communities to convert NHþ 4 to
N gas (USEPA, 2007). In this process, NHþ 
4 reacts with NO2 to pro-
duce N2 and water (USEPA, 2007; Lu et al., 2011). Stoichiometrically,
anammox requires only 50% of NHþ 
4 to be oxidized to NO2 but
around 57% is considered ideal (Magri et al., 2012). Anammox re-
quires just over one-third of the oxygen of nitrification/denitrifi-
cation and no C amendment for the same NHþ 4 removal but is
incapable of processing NO 3 (Eini, 2012). The process is compli-
cated by the dual need to assure that NHþ 
4 oxidizes to NO2 without
continuing to NO 3 and to maintain a reactor favorable to the
growth of anaerobic ammonium oxidizers (Nifong et al., 2013).
Therefore, a partial nitrification-anammox reactor generally re-
quires specific ranges of pH, C, retention time, and temperature for
the slow-growing anammox bacteria to overcome their microbial
competitors (Eini, 2012). Reactors can take years to set up while
simultaneously being vulnerable to population collapse if condi-
tions are disturbed and competing bacteria populations bloom (Hu
et al., 2013). Anammox reactors are well-suited for side stream
processing (USEPA, 2007) of anaerobic digester reject water and
can achieve high N removal rates (Lu et al., 2011; Nifong et al.,
2013). Reactors can vary by batch vs. plug-flow design, hydraulic
retention time, oxygenation regime, and partial nitrification pre-
cursor (Nifong et al., 2013; USEPA, 2015a). More detailed explana-
tions of the anammox process and its implementation can be found
in van Dongen et al. (2001), Fux et al. (2002), and Eini (2012).
Electrodialysis is a mature technology typically used in industry
for desalination through the use of an electric current and ion-
permeable membranes (Pruyn et al., 1969). The electrodialysis
stack typically consists of alternating cation- and anion-exchange
membranes (CEM, AEM) separated with spacers, with inert elec-
trodes on each end to which an electrical field is applied to the
stack. As aqueous solutions flows between these specialized
membranes, cations are drawn from a diluate stream toward the
cathode through a cation exchange membrane into the concentrate
stream while anions are drawn from the diluate stream on the
other side toward the anode through anion exchange membranes.
The ions of the influent stream are thereby separated into an
enriched concentrate stream and a partially desalinated diluate
stream. A set of one CEM, one AEM, and their accompanying
concentrate and diluate streams comprise a cell pair. Electrodialysis
can be an energy-efficient method of concentrating ions but solu-
tions with high electrical resistance or large desalting goals can
affect stack performance; longer process times equate to higher
equipment costs to meet required capacity.
Electrodialysis has been previously explored as a method of
removing NHþ 4 from municipal treatment streams (Eisenmann,
1967) but the process was hampered by accumulation of calcium
(Ca) and magnesium (Mg) scale (presumably carbonates and
phosphates) on membranes that halted ion flow. Given that both Ca
and Mg are divalent ions, the intention of this new strategy is
twofold: prevent their accumulation beyond solubility and subse-
quent precipitation on membrane surfaces in the concentrate
stream as various carbonate and phosphate minerals (Cifuentes-
Araya et al., 2012) and to avoid wasted energy on moving low
value ions such as Ca and Mg in lieu of NHþ 4 . The tendency toward
fouling caused by interaction of complex organic molecules with
the anion exchange surface is also expected to be inhibited by
monovalent-selective anion exchange membranes. To further
alleviate this, we envisage a regime of frequent electrodialysis
reversal (EDR) and flushes of cleaning solution as necessary. The
scope and scale of these cleaning procedures that will be necessary
are specific to the stack and the solution. Chao and Liang (2008)
successfully used electrodialysis to desalinate raw municipal
wastewater using a sand filter pretreatment, an hourly polarity
reversal, and a once-per-week 30-min 5% hydrochloric acid treat-
ment. The American Water Works Association (1995) recommends
a faster 15- to 30-min reversal cycle. An electrodialysis system
processing wastewater may wish to keep diluate and concentrate
channels completely isolated, with wastewater as the diluate and a
receptor solution as the concentrate influent; this would necessi-
tate short, aggressive EDR cycles during which stack effluent would
be untreated and need to be disposed or diverted back to headflow.
In contrast to other N removal technologies, electrodialysis may
enable recover of an aqueous NHþ 4 product that has potential use as
a fertilizer. Electrodialysis will also recover other monovalent ions
such as potassium, sodium, and chloride e each of which may have
its own implication for fertilizer use depending on the destination
e in amounts that will vary depending on the composition of the
feed and selectivity of the ion exchange membranes.
This study seeks to economically compare an electrodialysis
strategy utilizing monovalent-selective membranes to recover NHþ 4
from digester reject water to established nitrogen removal tech-
niques, i.e., nitrification and anammox. We propose an optimiza-
tion strategy for determining ideal amperage for lowest cost N
removal from an electrodialysis stack using a multivariate model
accepting highly customizable influent and stack qualities. We
compare this theoretical optimized electrodialysis regime to liter-
ature surveys and estimates of costs for existing nitrification/
denitrification and anammox reactors to estimate potential cost
savings. As the analysis does not capture the full complex in-
teractions of an electrodialyzed municipal waste stream, this serves
as a first step for characterizing a potential emerging technology.
2. Methods
Our analysis assumes that the new technology will be imple-
mented with the intention of meeting a nutrient removal goal
rather than a volumetric influent goal and as such have chosen a
mass basis for N removed; this allows decision-makers to directly
compare the estimated impacts incurred by meeting their goals and
then size their system accordingly. Our functional unit for com-
parison is a single kilogram of NHþ
4 -N removed. We assess only the
marginal impacts of nutrient removal; scalability and total removal
potential of technologies are outside our scope.
When wastewater specifics are necessary, data are taken from
the Nine Springs Wastewater Treatment Plant in Madison, Wis-
consin, a mid-size treatment plant that serves a population of
360,000 over a 180 sq. mile (466 km2) service area, with a design
capacity of 50 million gallons per day (50 MGD; 189270 m3/day).
 D. Vineyard et al. / Journal of Cleaner Production 262 (2020) 121145
This plant anaerobically digests a 50:50 mixture of primary sludge
and waste activated sludge in a mesophilic digester generating
biogas and then dewaters the sludge using a gravity belt thickener
(GBT) (AECOM, 2015) to produce biosolids and reject water. The
GBT reject water is a high-nutrient stream that is currently recycled
to the headworks for reprocessing but which is well-suited for a
side-stream treatment process such as anammox or electrodialysis
for N removal. The total GBT filtrate flow is 0.25 MGD, or 946 cubic
meters per day, and averaged 1437 mg/L NHþ þ
4 -N, 214 mg/L K , and
256 mg/L Naþ for 103 meq/L, 5.47 meq/L, and 11.1 meq/L, respec-
tively (Barak, 2019, unpublished data).
We consider three N management strategies: (1) Nitrification/
denitrification; (2) Anammox; and (3) Electrodialysis for NHþ 4 re-
covery. The first two scenarios were selected to summarize the
current state of targeted N management and the last was to
compare an emerging technology.
All economic values are converted to year 2018 USD using the
United States Bureau of Labor Statistics’ Consumer Price Index
(USBL, 2019). Euro values are first converted to USD using the
average exchange rate of that year according to the XE Currency
Converter (Xe Corporation, 2019). Costs independent of operation
such as capital costs and some maintenance costs are amortized to
a monthly rate using the annuity formula in Eq. (1). The monthly
rate is then adjusted to an hourly rate based on the expected hours
of operation in a month, with a month totaling 730.5 available
hours. This rate is compared to expected N removal per hour of
operation to find the operation independent portion of the cost of N
removal.
P*i
A¼
1  ð1 þ iÞn
where A ¼ periodic payment amount; P ¼ amount of principal;
I ¼ periodic interest rate; n ¼ total number of payments.
In 2017, average electricity prices for industrial consumers in the
US ranged from $46/Megawatt-hour (MWh) in Washington to
$229/MWh in Hawaii. To calculate a generic price for electricity in
the US, we assume a wastewater treatment plant pays as an in-
dustrial customer as listed by the US Energy Information Admin-
istration (USEIA) in the year 2017. Rather than use the listed
national average electric price, we use a weighting scheme based
on state populations obtained from the 2017 US census (Supple-
mentary Information, Table 1A). The purpose of this weighting is to
calculate a national average electricity price based on population
(and municipal sewage production) rather than on total industrial
energy consumption for all sectors. We find that this weighting
scheme increases the average price to industrial customers from
$68.77/MWh to $77.31/MWh (Supplementary Information,
Table 1A).
2.1. Nitrification/denitrification
This traditional N removal technology is capital intensive,
requiring the construction of large cisterns and pumping systems.
Though older cost estimates exist (Smith, 1968), the largest and
most comprehensive assessment found by our literature review
was performed by the Chesapeake Bay Program (CBP). The CBP
surveyed the costs associated with nutrient removal in the Ches-
apeake Bay watershed, acquiring data from 539 point sources with
a cumulative designed limit of roughly 2.8 billion gallons of
discharge per day (CBP, 2002). Of these facilities, 304 were
municipal treatment plants with a combined design flow of 2336
MGD and an average discharge of 1550 MGD. One significant result
was an estimation curve for the capital cost of implementing N
removal technologies found in Eq. (2).
3
Y ¼ 2023829 þ 704350:8039*Xe5986:7330*X 2 (2)
where Y ¼ capital cost in 2000$; X ¼ plant design MGD, between
0.5 and 30 MGD.
This model is calculated by the CBP from what were judged the
66 most reliable of 154 data points for constructing an N removal
system to reach a 12-month rolling average of 8 ppm N in the
effluent. Data were primarily accounting of real construction costs
(52 of 66) with the remaining 14 being engineering design esti-
mates and facilities plans. Variability was considerable, with costs
per MGD ranging from $15,081 to $11,683,297 based on factors
such as total influent N, costs of local industry and real estate,
chosen technology, and contingency costs. Total N removal by these
systems is not reported directly by the survey, but a brief reading of
data for upgrades to obtain 8 ppm-N suggests that even the lowest
estimates for large facilities with concentrated influent streams can
expect minimum capital costs of ~$4.00/kg N or total costs of
>$5.15/kg N after amortization at 3% and adjusting for inflation.
Of greater use to this study, the CBP additionally estimated the
marginal costs of curtailing effluent N content from 8 to 5 ppm and
then to 3 ppm. The methodology for these estimates was generic
and less rigorous due to paucity of available data and is intended to
be a loose estimate at the watershed level. The estimated costs to
upgrade for 8 to 5 ppm average effluent were calculated by Eq. (3):
Y ¼ 864; 830 þ 386; 010*X (3)
where Y ¼ capital cost in 2000$; X ¼ plant design MGD, between 1
and 30 MGD.
(1)
These costs were estimated using several assumptions: a
methanol-amended secondary anoxic zone would be constructed at
the end of an existing nitrification/denitrification system; aeration
capacity would need upgrading at a cost nonlinearly related to ca-
pacity; an additional 25% clarification capacity would be installed
using USEPA cost curves for clarifiers; and these upgrades would
have an additional 30% implementation cost for engineering, man-
agement, legal, and administrative fees (CBP, 2002). The CBP also
estimated operating costs for these upgrades. All aerators were
assumed to be fine bubble diffusers with a standard oxygen transfer
efficiency of 25%, an actual oxygen requirement and standard oxy-
gen requirement of 0.33, 0. 0.2776 kg O2/m3, 1.053 kW per standard
cubic meter per minute, and 4.57 kg O2/kg N. From 8 to 5 ppm N, each
kg of NO 3 -N removed from the waste stream was assumed to cost
3.1 kg of methanol (at $0.33/kg) and generate 0.372 kg of additional
solids to be disposed at $331/dry metric ton. This value is similar to
Purtschert et al. (1996), who found a yield of 0.4 kg of solids per kg of
methanol. Unknown pumping requirements used $0.05/kilowatt-
hour (kWh). A yearly maintenance cost of 2% of capital cost accounts
for roughly two-thirds of the projected operation costs. Additional
labor was assumed to be unnecessary.
Assuming a 15 MGD plant reducing N marginally from 8 to
5 ppm for ~170 kg of N per day, the CBP estimated a cost of
$6,654,980 with estimated yearly costs of operation and mainte-
nance at $219,766 in year 2000 USD. Amortizing capital costs to 3, 4,
and 5% interest rates compounded daily affects the final prediction.
In all cases, amortized capital costs are at least double the operation
costs (Table 1). The CBP’s costs estimations were higher than other
literature sources (Fux and Siegrist, 2004; Ostace et al., 2011), but
were collected from a far larger sample size. Increased costs esti-
mates stem largely from the 2% maintenance cost and the predicted
consumption of approximately a gallon of methanol per kilogram of
nitrate removed.
The CBP estimated that energy costs would account for less than
10% of total operation, assuming a continuous 30-HP pump would
4 D. Vineyard et al. / Journal of Cleaner Production 262 (2020) 121145
Table 1
Chesapeake Bay Program estimates of upgraded N removal capital costs.
MGD CAPITAL YEARLY OPERATION
0.1 $241,000 $7046
1 $1,112,000 $29,218
10 $4,927,000 $157,469
30 $12,383,000 $293,938
be required for aeration (as assumed in an example within the EPA
nitrogen control manual (USEPA, 1993); and adding a 33% pumping
and lighting cost for a total of 3.54 kWh/kg N, or 2.65 kWh/kg for
aeration alone. This energy consumption is in agreement with
other literature estimates (Supplementary Information, Table 2A).
The CBP model assumed no oxygen recovery. Nitrification/Deni-
trification is estimated to cost $15.56/kg N removed (Table 2).
2.2. Anammox
Anammox cost data are taken from Fux and Siegrist (2004), and
environmental and economic assessment using simulated reactors
and industry standard price estimations. Amortized capital was
estimated to be more than half of the total cost for removal. Energy
costs were estimated to be about 12.5% of the remaining costs, or 6%
of the total costs, with the rest stemming from operation and repair
plus a very small budget for chemical amendments. The study
estimated 1.5 kWh/kg N electricity consumption, a survey of liter-
ature data (Supplementary Information, Table 3A) on anammox
reactor performance estimated 1.44 kWh/kg N. These results were
similar in proportion to the CPB study. Anammox is estimated to
cost $4.09/kgN removed (Table 3).
2.3. Electrodialysis
2.3.1. Capital and maintenance
The most comprehensive analysis we found for the installation
of a utility-scale ED system for municipal wastewater was a bid
made by General Electric (GE) to the Madison Municipal Sewerage
District included in a public report on alternative strategies for
chloride removal from its 15 MGD throughput (AECOM, 2015).
These capital estimates are similar to the predictions of McGovern
et al. (2014) and offer a comprehensive estimate of the capital cost
of a wastewater electrodialysis installation generalizable to other
purposes. The total cost estimate included the cost of 288 stacks,
pumps, reservoirs, chemical feeds, a dedicated building, and an
engineering and contingency budget proportional to the con-
struction cost. At ~$85.6 million in total cost (~$81M in 2015), the
system was expected to remove 19 mol of charge per kWh and
consume 4.05 kWh per thousand gallons for a total of nearly 1.2
million moles of charge per day.
We repurpose parts of this proposal for treating a portion of the
0.25 MGD GBT reject water stream for NHþ 4 removal with a single
stack. Scaled down, a single stack and its supporting pumps and
piping should occupy less than six square meters of floor space.
Table 2
Operation costs, capital costs amortized at 3%, and energy consumption per kilogram
of N for Nitrification/Denitrification.
Cost
Capital Operation Total
UNIT 2018$ USD 2018$ USD 2018$ USD
NITRO/DENITRO 8 TO 5 10.40 5.16 15.56
SOURCE CBP, 2002 CBP, 2002
Assuming a linear scale rate, a single stack would cost ~$77,000 and
the surrounding infra-structure would cost ~$220,000 for a total of
~$297,000 (~$85.6M divided among 288 stacks). Though a linear
scale rate will likely underestimate the cost of smaller installations
in isolation, it also does not account for the possibility of savings
through infrastructure overlap such as fitting a smaller installation
in a preexisting building. The net should be somewhere in the
middle and our sensitivity analysis covers the possibility of
dramatically higher capital costs. Amortizing to 20 years at 3% in-
terest compounded monthly, we find a monthly cost of capital of
~$1650 per stack. General Electric suggests 320 h per week to
maintain a 288-stack system, or 1.11 h per week per stack (~$243/
month at the suggested ~$50.20/hr in 2018 USD). They also sug-
gested 5% equipment/year maintenance costs, or ~$322 monthly
cost. Membrane replacement (~$38,100) every 10 years has a
monthly cost equivalent of ~$287 when discounted at a net 2%.
Given the small fraction of costs represented by chemical use in the
original analysis and tendency of monovalent-selective membranes
to inhibit scaling and fouling, we assume no chemical costs. We
assume no disposal cost due to the creation of a marketable product
from the concentrate. All inflexible costs sum to a total ~$2500
monthly. If we assume 10% downtime for cleaning, the inflexible
cost of the stack per hour of operation becomes ~$3.80 in 2018 USD,
with ~$2.51 stemming from upfront capital and ~$1.29 stemming
from maintenance.
2.3.2. Electricity consumption
In an electrodialysis stack, the total energy consumption is the
product of the voltage and the electric current or, by Ohm’s law, the
square of the current times the resistance. Resistance is an additive
property of both the cell membranes and the ionic solutions; dilute
solutions have a higher electrical resistance. The number of cell
pairs both increases the stack resistance and increases the number
of ions transported across membranes, with a net neutral effect.
The most important characteristics for predicting the energy con-
sumption for an ED stack are the conductivities and flow rates of
the concentrate and diluate streams, the resistance of the mem-
branes, and the desired operating current. The voltage per cell pair
is an emergent property from these characteristics and, together
with the current efficiency, can be used to directly calculate the
energy efficiency of the system. The relationship between these
stack conditions and resultant voltage is complicated. The most
profitable setting for an electrodialysis stack will be dependent
both on the relationship between voltage and current for the spe-
cific system and on the values of energy and NHþ 4 removal itself.
The Tanaka (2015) electrodialysis stack model is a calculator in
the form of a Microsoft Excel spreadsheet designed to account for a
number of electrochemical and fluid dynamic effects. To facilitate
our use of the spreadsheet, two modifications were made: the
addition of a macro to implement Solver root finding and automatic
copy-and-paste where necessary and the modification of a number
of cell formulas to avoid zero or negative values upon recalculation.
These alterations improved stability, speed, and ease of use of the
model without affecting the final results.
The GE stack with GBT filtrate and higher resistance
monovalent-selective exchange membranes to reduce scaling and
fouling was modeled using the Tanaka (2015) model with 600 cells,
44.1 cm by 92.8 cm membranes, a 0.0772 cm spacer, ~29500 L/h
flow rate, 23 U cm2 membranes, and a ~0.119 eq/L influent con-
Energy
ductivity to predict stack voltage as a function of stack current. The
current/voltage relationship predicted by the Tanaka model is
kWh mildly nonlinear and the predicted stack resistance, embodied by
3.54 the slope of the voltage to current relationship, can be estimated
CBP, 2002
(r2 ¼ 0.9995) by the polynomial:
 D. Vineyard et al. / Journal of Cleaner Production 262 (2020) 121145 5

Table 3
Operation costs, capital costs amortized at 3%, and energy consumption per kilogram of N for Anammox.

Cost Energy

Capital Operation Total

UNIT 2018$ USD 2018$ USD 2018$ USD kWh

ANAMMOX 2.12 1.96 4.09 1.44
SOURCE Fux and Siegrist (2004) Fux and Siegrist (2004) Supplementary Information, Table 3A

$0.00007731); P ¼ pump power (W, 947); A ¼ NHþ 4 sale price or

disposal cost ($/kg N, $0.53); Q ¼ NHþ4 removal rate (kg N/(ampere *
11:641 hr), 0.256); B ¼ number of cell pairs per stack (600); G ¼ fraction of
R ¼ 6:3197 þ 0:0264*I þ (4)
I monovalent cations as NHþ 4 (0.861); N ¼ molar mass of nitrogen
(kg/mol, 0.014); E ¼ current efficiency (0.95); F ¼ Faraday constant
where I ¼ current applied (amps) and R ¼ stack resistance (U).
(coulomb/mol, 96485.3329); T ¼ seconds per hour (3600).
In the GBT reject water of the Nine Springs Wastewater Treat-
The cost per hour in Eq. (5) can be divided by the amount of N
ment Plant, ~86.1% of monovalent cations are NHþ 4 . Tanaka (2003),
removed, Q * I, to calculate the cost per kilogram of NHþ 4 -N
McGovern et al. (2014), and AECOM (2015) all suggest a near 100%
removed:
current efficiency for electrodialysis below limiting current density,
indicating both a lack of water hydrolysis and a lack of back- ,

diffusion through membranes on a significant scale; the C þ M* I 2 *R þ P
Cost kgN ¼ A (9)
McGovern et al. (2014) diffusion estimates support these assump- ðQ *IÞ
tions when our operating conditions are entered and the AWWA
(1995) reports that such back diffusion is negligible below a 1:150
,

concentration gradient in typical stacks. Assuming no ion selec- 3:8 þ 0:00007731* I 2 *R þ 947
tivity beyond polyvalent exclusion and a 95% current efficiency, a Cost kgN ¼  0:53 (10)
0:256*I
600-cell pair stack will pull 0.256 kg of NH4eN per hour per ampere
of current applied. We use our weighted estimate of $77.31/MWh By inserting Eq. (4) into Eq. (9) and Eq. (10), we can find costs as
for energy cost. As electrodialysis is also an N recovery technology, a function of the stack current (Fig. 1) according to the Tanaka
it can produce fertilizer-quality NHþ 4 . We assume a sale price of model predictions:
$0.24/lb ($0.53/kg) of NH4eN (USGS, 2019). No weight, positive or
,

negative, was given to potassium, sodium, bicarbonate or chloride C þM* 0:0264*I 3 þ6:3197*I 2 þ11:641*I þP
byproducts. According to the GE proposal, the stack is estimated to Cost kgN ¼ A
cause ~80.8 kPa head loss, which translates to 663 W for a ðQ *IÞ
29.525 m3/hr pump at 100% efficiency and 947 W at 70% efficiency. (11)
The net costs and yields of operation for the stack can be
calculated by the following equations:

,

3:8 þ 0:00007731* 0:0264*I 3 þ 6:3197*I 2 þ 11:641*I þ 947
Cost kgN ¼  0:53 (12)
ð0:256*IÞ

Cost$

¼ C þ M* I 2 *R þ P  A*Q *I (5)
hr

where:

kgN B*T*G*N*E
Q¼ ¼ (6)
amp*hr F
or numerically:

Cost$

¼ 3:8 þ 0:00007731* I 2 *R þ 947  0:53*0:256*I (7)
hr

kgN 600*3600*0:861*0:014*0:95
Q¼ ¼ ¼ 0:256 (8)
amp*hr 96485:3329

where I ¼ current applied (amp); R ¼ stack resistance (U); Fig. 1. Cost of nitrogen removal as a product of stack amperage for the modeled
C ¼ hourly capital ($/hr, $3.80); M ¼ electricity price ($/Wh, electrodialysis stack.
6 D. Vineyard et al. / Journal of Cleaner Production 262 (2020) 121145
To find the optimum, or most profitable, current we solve for the
first derivative of Eq. (11) using current as the variable and then find
the root:
 
M* 2*0:0264*I þ 6:3197  ðPþC=MÞ
I2
0 ¼
Q
Energy consumption per unit N is calculated by omitting capital
and maintenance costs and NHþ4 sales in Eq. (11), yielding
, upfront capital costs and in-operation and maintenance costs.
0:0264*I 3 þ 6:3197*I 2 þ 11:641*I þ P
kWh kgN ¼
ðQ *IÞ
The percentage of NHþ
4 removed can be calculated as follows:
NHþ
4 removed ðQ *IÞ*1000
¼ (15)
NHþ 4 total N*D*S
where I ¼ current applied (amps); Q ¼ NHþ 4 removal rate (kg N/
(ampere*hr), 0.256); D ¼ influent NHþ 4 concentration (meq/L,
102.6); N ¼ molar mass of nitrogen (kg/mol, 0.014); S ¼ influent
flow rate (L/hr, 29525);
With the above data, we find the optimum stack current is ~70.6
amps, where the stack is estimated to consume 2.36 kWh/kg, cost
~$0.392 per kilogram of N removed (Table 4), and net a total of
~$0.138 gained after NHþ 4 sale. This value represents an optimal
profit method of current selection and is predicted to remove 42.6%
of NHþ4 from the stream in a single stage of treatment, but the cost
curve is shallow and values from 48 amps to 102 amps remain
above the $0.10/kWh gain mark. The modeled stack was predicted
to become cost neutral at just over 33 amps. Alternatively, an
operator may choose to run a higher but less profitable current to
increase N removal and recovery.
2.3.3. Limiting current
A practical constraint on electrodialysis arises upon increased
transport of ions through the ion exchange membranes from the
diluate stream leading to layers of ion-depleted water with high
electrical resistance on the membrane surface. The current at which
these depleted zones become significant is referred to as the
limiting current density. Current attempting to pass through ion-
depleted zones can be spent first splitting water molecules into
hydroxide and hydronium ions and then trans-porting those new
ions across the membrane. Both steps represent a low-value use of
energy intended to transport target ions. Industry typically rec-
ommends that an operator not exceed the limiting current density
of a given stack. The Tanaka (2015) model has an additional
calculator to predict limiting current density for a given stack. Using
this method, the Tanaka model predicts the limiting current density
for a GE stack processing GBT filtrate to be 102 stack amps.
Increasing the stack current to 102 amps, approximately the pre-
dicted limiting current density, increases the NHþ 4 removal by
nearly 50% at a net cost per kilogram of only ~$0.037 more, for a
Table 4
Operation costs, capital costs amortized at 3%, and energy consumption per kilogram of N for Electrodialysis.
Cost
Capital
UNIT 2018$ USD
ELECTRODIALYSIS 0.139
SOURCE AECOM (2015)
total cost of ~$0.429/kg N and a theoretical profit of ~$0.101/kg N
after sale. This increased current incurs a predicted environmental
cost of consuming a total of ~3.63 kWh/kg N. However, this
increased current also results in a removal of 62% of stream NHþ4.
3. Results
(13)
Removal of N by electrodialysis (Table 4) appears to be an order of
magnitude less expensive on a per kilogram N removed basis than
nitrification/denitrification (Table 2) and anammox (Table 3), both in
Nitrification/denitrification was predicted to be the most expensive
(14) treatment option considered, with roughly five times the upfront
capital investment of an equivalent capacity anammox system and
more than double the operation and maintenance costs. Capital
differences stem primarily from the difference in cistern scale be-
tween engineering for a whole-plant throughput and a side stream.
Operation costs stem from estimations of maintenance derived from
capital costs and from the purchase of methanol to serve as a C
source for denitrification. As competing side-stream technologies,
electrodialysis was predicted to be an order of magnitude cheaper
than anammox reactor. An electrodialysis stack is relatively inex-
pensive to purchase and install and has low requirement of gross
area, resulting in lower capital costs which offset the higher main-
tenance costs. Anammox was predicted to use 40% of the electricity
of an equivalent nitrification/denitrification system on a per-
kilogram N basis. By comparison, electrodialysis was predicted to
use 64% more energy per kilogram than an anammox reactor, but
33% less per kilogram than a nitrification/denitrification reactor.
4. Discussion
Nitrification/denitrification had a range of alternative energy
consumptions reported by literature and was found to have a
geometric mean near the highest energy usage, suggesting that
while most facilities have a high consumption some facilities can
meet their goals on lower consumption. The potential energy sav-
ings of even a best case scenario reactor were not enough to offset
the high amortized capital costs of even the cheapest nitrification/
denitrification systems reported in literature. A nitrification/deni-
trification basin that did not require methanol as a C substrate
might see a substantial reduction in operating cost predictions. If
such a reactor can use an alternative carbon source such as food
waste or sludge from another reactor, it might even be allocated an
offset of the disposal costs for such a substrate. Anammox had a
distribution of literature values similar to nitrification/denitrifica-
tion. Anammox showed a geometric mean near the lowest reported
energy consumption, while a few facilities consumed a large
amount of electricity for their nitrogen removal. This suggests that
reactor performance is of particular importance to anammox re-
actors as poor function can eliminate expected technological gains
from both environmental and energy saving perspectives.
Electrodialysis was predicted to be an extremely cost efficient
method of NHþ 4 removal for side-stream treatment for N removal.
Energy
Operation Total
2018$ USD 2018$ USD kWh
0.253 0.392 2.36
AECOM (2015) Tanaka (2015)
Tanaka (2015)
 D. Vineyard et al. / Journal of Cleaner Production 262 (2020) 121145 7

The modeled electrodialysis stack was predicted to have signifi-
cantly lower costs than the nearest competitor technology, the
anammox system. Price differences stem mostly from the reduced
upfront capital cost of an electrodialysis system versus the
construction-intensive aeration basin systems; even assuming a
best case sale price, NHþ 4 sale revenues were a small portion of
overall cost savings. The shallow cost curve around the economic
optimum suggests that electrodialysis stacks may be economically
forgiving of non-optimized amperages.
Our results are supported, in part, by multiple models.
McGovern et al. (2014), in a similar analysis, predicted cost of $0.60/
kg of salt removed for hydrofracking reject water of similar salin-
ities to our side stream; most costs were predicted to come from
amortized capital in an unfavorable installation environment.
Substituting the McGovern et al. (2014) stack resistance estimate
for the Tanaka estimate yields lower resistances and greater energy
efficiency for lower costs. Cost estimates per kilogram of NH4eN
removed were much lower than suggested by the GE proposal for
equivalent ion removal because the increased stack current lead to
greater removal which reduced average cost of capital and pump-
ing per unit removed. Even if increased electric currents prove
infeasible in practice and NHþ 4 has a less seller-favorable market,
the stack remains the estimated lowest cost option even at high of
20 stack amps used by the GE-model. Electrodialysis operated at
optimal cost efficiency was projected to consume more electricity
per kilogram of N than an anammox system but less than a nitri-
fication/denitrification system.
Electrodialysis with monovalent selective membranes appears
economically favorable in comparison to traditional aeration based
technologies. Comparisons are robust against changes in underly-
ing assumptions behind the electrodialysis model such as capital
cost, electric resistance, and operation conditions (Table 5); addi-
tional details of alternative assumptions are supplied in the sup-
plemental information. Creating an artificial amperage ceiling of
19.5 stack amps, a condition applied in the AECOM (2015) project,
more than doubles the price of removal. Tripling electricity cost, as
in the isolated Hawaiian island grids, adds roughly two thirds to the
cost while Washington electric prices reduce it by about 20%.
Quadrupling the stack resistance, a situation that could occur in
more dilute waste streams or with fouled membranes, less than
doubles the cost. Quintupling pumping costs causes only a minor
increase. Operation-independent cost perturbations, such as a 5%
amortization rate, 25% downtime, daily hour long technician
cleanings, or doubling capital or repair costs, all have smaller effects
than increased energy prices. Even an aggregation of doubled
capital costs, 25% operational downtime, daily hourly technician
cleaning cycles, quadruple stack resistance, quintuple pumping
costs, and an artificial current ceiling of 19.5 stack amps brings the
estimated cost of removal to only $2.24 per kilogram; this estimate
is still nearly half of the expected cost of an anammox reactor.
The product of an electrodialysis stack used for side stream NHþ 4
recovery will require yet unknown post processing to form a
marketable fertilizer product, but there is considerable margin in
the budget to include this processing and remain favorable. An
electrodialysis stack could theoretically concentrate NHþ 4 with
equimolar chloride and bicarbonate up to about 4 mol/L, or 6%
nitrogen by mass and roughly the limit of solubility, before pre-
cipitation and scaling begin at 20  C (Barak, 2019, unpublished
data). Compared to anhydrous ammonia (82%), urea (46%), or even
aqua ammonia (20%), a 6% solution would incur greater trans-
portation, storage, and application costs as a fertilizer. These costs
might be offset by shorter transportation distance between pro-
duction and application. It remains possible that post-processing
might include further concentration by volatilizing and
condensing ammonia or ammonium bicarbonate or cooling to
induce crystallization. Though these products will avoid typical
waste-derived fertilizer concerns about the over-application of
phosphorus (P), they will introduce concerns about application of
chlorides.
Because of interactions with organics and precipitates, it is likely
that a stack used to treat municipal waste will not have a voltage/
amperage relationship summarized by a simple quadratic equation.
Future work on pilot scale reactors should attempt to better char-
acterize the relationships mathematically so that operators can
utilize more robust optimization methods for current selection and
EDR scheduling.
Given the limited scope of this project and the complex in-
terrelations of wastewater treatment processes, there may be un-
foreseen externalities to implementation of these technologies not
captured in our methodology. Some wastewater treatment systems
may not have suitable influent for processing, may have side stream
processes incompatible with new nutrient removal methods or
may have existing products degraded by nutrient removal. One
such possibility is biosolids production, a product that is frequently
limited in agricultural use by an excess of P in relation to N; up-
stream removal of N may exacerbate P/N ratio concerns. Current
allocation methods disregard any costs or benefits of pathogen
reduction or C removal in the reaction. A greenhouse gas emissions
trading regime may economically favor electrodialysis over aera-
tion technologies.
The costs of the nitrification/denitrification and partial
nitrification-anammox processes have previously been quantified

Table 5
Sensitivity of electrodialysis optimization and predicted costs to altered assumptions.

CAPITAL $/WH (M) PUMP NHþ4 SALE OPTIMAL COST

SCENARIO (C) WATTAGE (P) PRICE (A) AMPERAGE (I) $/KG

DEFAULT 3.80 0.00007731 947 0 70.6 $0.392

MCGOVERN RESISTANCE 3.80 0.00007731 947 0 77.0 $0.352
19.5 AMPERE CAP 3.80 0.00007731 947 0 19.5 $0.820
5X PUMP ENERGY 3.80 0.00007731 4735 0 72.8 $0.408
HAWAII ELECTRIC 3.80 0.00022920 947 0 44.9 $0.662
WASHINGTON ELECTRIC 3.80 0.00004600 947 0 87.4 $0.309
25% DOWNTIME 4.57 0.00007731 947 0 76.2 $0.433
1HR/DAY CLEANING 5.83 0.00007731 947 0 84.2 $0.495
2X CAPITAL COST 6.39 0.00007731 947 0 87.4 $0.520
2X REPAIR COST 4.79 0.00007731 947 0 77.7 $0.444
5% AMORTIZATION 4.33 0.00007731 947 0 74.5 $0.421
DOUBLE RESISTANCE 3.80 0.00007731 947 0 52.5 $0.540
QUADROUPLE RESISTANCE 3.80 0.00007731 947 0 38.7 $0.748
2X CAPITAL COST, 25% DOWNTIME, 1HR/DAY CLEANING, QUADROUPLE RESISTANCE, 5X 9.93 0.00007731 4735 0 19.5 $2.24
PUMP ENERGY, 19.5 AMPERE CAP
8 D. Vineyard et al. / Journal of Cleaner Production 262 (2020) 121145
and shown to be facility-dependent, but the costs of municipal-
scale electrodialysis have not. McGovern et al. (2014), the GE pro-
posal (AECOM, 2015), and this research have attempted to estimate
costs with a useful methodological framework, but data are not
robust. Necessary cost data include equipment purchase, installa-
tion, operation and maintenance, disposal, and lifespan and remain
uncertain until deployment at scale. Installation and energy costs
are expected to be low, but uncertainties currently exist for esti-
mating maintenance costs and costs of product post-processing
and transportation. Future work should assess pilot scale electro-
dialysis installations and the post-processing required to make
fertilizer from concentrate streams.
5. Conclusions
Based on our study results, a theoretical electrodialysis stack
appears to be the preferable choice for nutrient removal technol-
ogies from an economic standpoint. Electrodialysis is predicted to
consume more electricity than an equivalent capacity anammox
system but less than an nitrification/denitrification system.
Reduced costs for an electrodialysis system stem largely from
greatly reduced upfront capital costs and low system footprint.
Side-stream electrodialysis as modeled does not have the N
removal rate to meet typical N regulations, but does represent a
low-cost method to remove NHþ 4 from N-rich streams that would
reduce the need for expansion of other reactors to meet new
regulation.
Author contribution statement
Donald Vineyard: Conceptualization, Methodology, Validation,
Formal Analysis, Investigation, Data Curation, Writing e Original
Draft, Writing e Review and Editing, Visualization.
Andrea Hicks: Writing e Review and Editing.
KG Karthikeyan: Writing e Review and Editing.
Phillip Barak: Conceptualization, Resources, Writing e Review
and Editing, Visualization, Supervision, Funding Acquisition.
Acknowledgements
One of the authors, Phillip Barak, is a co-owner of Nutrient Re-
covery and Upcycling, LLC, which is rights holder to US Patent
10,125,428 (issued 13 Nov. 2018) for the removal of nitrogen from
wastewater using monovalent-selective electrodialysis, with in-
ternational patent applications pending. This work was funded by
United States Department of Agriculture/National Institute of Food
and Agriculture Hatch Project No. WIS01920. We thank the Nine
Springs Wastewater Treatment Plant and the Madison Metropol-
itan Sewerage District for providing plant and reject water
specifications.
Appendix A. Supplementary data
Supplementary data to this article can be found online at
https://doi.org/10.1016/j.jclepro.2020.121145.
References
AECOM, 2015. Chloride Compliance Study Nine Springs Wastewater Treatment
Plant Final Report. Prepared for: Madison Metropolitan Sewerage District, Nine
Springs Wastewater Treatment Plant, Madison, WI. https://www.madsewer.
org/Portals/0/ProgramInitiatives/ChlorideReduction/MMSD%20Chloride%
20Compliance%20Study%20Report%20-%20Final%206-19-15bookmarks.pdf.
American Water Works Association, 1995. M38 Electrodialysis and Electrodialysis
Reversal. Am. Water Works Assoc., Washington DC, ISBN 9781613001677
(Chapter 2).
Barak, P. 2018. Electrodialysis stacks, systems, and methods for recovering ammonia
and monovalent salts from anaerobic digestate. U.S. Patent 10,125,428, 13 Nov
2018. (US Patent Application 20150308001. Filed: 24 Apr 2015, Publication date:
29 Oct 2015)..
Barak, P., 2019. Testing the Performance of an Electrodialysis Stack Extracting Ni-
trogen from the Gravity Belt Thickener Reject Water of the Nine Springs
Wastewater Treatment Plant in Madison. WI. Unpublished raw data.
Chao, Y.-M., Liang, T.M., 2008. A feasibility study of industrial wastewater recovery
using electrodialysis reversal. Desalination 221, 433e439. https://doi.org/
10.1016/j.desal.2007.04.065.
Chesapeake Bay Program, 2002. Nutrient Reduction Technology Cost Estimations
for Point Sources in the Chesapeake Bay Watershed. Prepared by: The Nutrient
Reduction Technology Cost Task Force e A Stakeholder Group of the Ches-
apeake Bay Program.
Cifuentes-Araya, N., Pourcelly, G., Bazinet, L., 2012. Multistep mineral fouling
growth on a cation-exchange membrane ruled by gradual sieving effects of
magnesium and carbonate ions and its delay by pulsed modes of electrodialysis.
J. Colloid Interface Sci. 372, 217e230. https://doi.org/10.1016/j.jcis.2011.12.067.
Clark, D.L., Hunt, G., Kasch, M.S., Lemonds, P.J., Moen, G.M., Neethling, J.B., HDR
Engineering Inc, 2010. Water Environment Research Foundation Nutrient
Management Report, 978-1-84339-617-8/1-84339-617-3.
Desloover, J., Woldeyohannis, A.A., Verstraete, W., Boon, N., Rabaey, K., 2012. Elec-
trochemical resource recovery from digestate to prevent ammonia toxicity
during anaerobic digestion. Environ. Sci. Technol. 46, 12209e12216. https://
doi.org/10.1021/es3028154.
Desloover, J., De Vrieze, J., Van de Vijver, M., Mortelmans, J., Rozendal, R., Rabaey, K.,
2015. Electrochemical nutrient recovery enables ammonia toxicity control and
biogas desulfurization in anaerobic digestion. Environ. Sci. Technol. 49,
948e955. https://doi.org/10.1021/es504811a.
Eini, E.J., 2012. Reducing Operating Cost with Anammox in Wastewater Treatment e
A Simulation Study. Chemical Engineering. Ryerson Univ., Engineering Com-
mons, Toronto, Canada, p. 113.
Eisenmann, J.D.S.J.L., 1967. Electrodialysis in Advanced Waste Treatment. Environ-
mental Protection Agency Old Reports, p. 218.
Fux, C., Siegrist, H., 2004. Nitrogen removal from sludge digester liquids by nitri-
fication/denitrification or partial nitritation/anammox: environmental and
economical considerations. Water Sci. Technol. 50, 19e26. https://doi.org/
10.2166/wst.2004.0599.
Fux, C., Boehler, M., Huber, P., Brunner, I., Siegrist, H., 2002. Biological treatment of
ammonium-rich wastewater by partial nitritation and subsequent anaerobic
ammonium oxidation (anammox) in a pilot plant. J. Biotechnol. 99, 295e306.
https://doi.org/10.1016/S0168-1656(02)00220-1.
Gungor, K., Karthikeyan, K.G., 2008. Phosphorus forms and extractability in dairy
manure: a case study for Wisconsin on-farm anaerobic digesters. Bioresour.
Technol. 99, 425e436. https://doi.org/10.1016/j.biortech.2006.11.049.
Holly, M.A., Larson, R.A., Powell, J.M., Ruark, M.D., Aguirre-Villegas, H., 2017.
Greenhouse gas and ammonia emissions from digested and separated dairy
manure during storage and after land application. Agric. Ecosyst. Environ. 239,
410e419. https://doi.org/10.1016/j.agee.2017.02.007.
Hu, Z., Lotti, T., de Kreuk, M., Kleerebezem, R., van Loosdrecht, M., Kruit, J.,
Jetten, M.S.M., Kartal, B., 2013. Nitrogen removal by a nitritation-anammox
bioreactor at low temperature. Appl. Environ. Microbiol. 79, 2807e2812.
https://doi.org/10.1128/aem.03987-12.
Karim, K., Hoffmann, R., Thomas Klasson, K., Al-Dahhan, M.H., 2005. Anaerobic
digestion of animal waste: effect of mode of mixing. Water Res. 39, 3597e3606.
https://doi.org/10.1016/j.watres.2005.06.019.
Lackner, S., Gilbert, E.M., Vlaeminck, S.E., Joss, A., Horn, H., van Loosdrecht, M.C.M.,
2014. Full-scale partial nitritation/anammox experiences e an application sur-
vey. Water Res. 55, 292e303. https://doi.org/10.1016/j.watres.2014.02.032.
Lu, H., Ding, S., Zheng, P., 2011. Central metabolism of anammox bacteriaea review.
Acta Microbiol. Sin. 51, 1014e1022.
Magri, A., Vanotti, M.B., Szogi, A.A., Cantrell, K.B., 2012. Partial nitritation of swine
wastewater in view of its coupling with the anammox process. J. Environ. Qual.
41, 1989e2000. https://doi.org/10.2134/jeq2012.0092.
McGovern, R.K., Weiner, A.M., Sun, L., Chambers, C.G., Zubair, S.M., Lienhard, V.J.H.,
2014. On the cost of electrodialysis for the desalination of high salinity feeds.
Appl. Energy 136, 649e661. https://doi.org/10.1016/j.apenergy.2014.09.050.
Nifong, A., Nelson, A., Johnson, C., Bott, C.B., 2013. Performance of a full-scale
sidestream DEMON® deammonification installation. Proc. Water Environ. Fed.
3686e3709. https://doi.org/10.2175/193864713813685700, 2013.
Ostace, G.S., Cristea, V.M., Agachi, P.Ş., 2011. Cost reduction of the wastewater
treatment plant operation by MPC based on modified ASM1 with two-step
nitrification/denitrification model. Comput. Chem. Eng. 35, 2469e2479.
https://doi.org/10.1016/j.compchemeng.2011.03.031.
Pruyn, K.T., Harrington, J.J., Smith, J.D., 1969. Mathematical Model of the Electro-
dialysis Process. Department of the Interior. Federal Water Quality Admin.,
Cincinnati, Ohio.
Purtschert, I., Siegrist, H., Gujer, W., 1996. Enhanced denitrification with methanol
at WWTP Zürich-Werdho € lzli. Water Sci. Technol. 33, 117e126. https://doi.org/
10.1016/0273-1223(96)00465-9.
Smith, R., 1968. Cost of conventional and advanced treatment of wastewater.
J. Water Pollut. Control Fed. 40, 1546e1574.
Tanaka, Y., 2003. Mass transport and energy consumption in ion-exchange mem-
brane electrodialysis of seawater. J. Membr. Sci. 215, 265e279. https://doi.org/
10.1016/s0376-7388(03)00020-6.
Tanaka, Y., 2015. Ion Exchange Membranes: Fundamentals and Applications.
 D. Vineyard et al. / Journal of Cleaner Production 262 (2020) 121145
Elsevier, Amsterdam. https://doi.org/10.1016/B978-0-444-63319-4.00001-8.
Theis, T., Hicks, A., 2012. White Paper: Methanol Use in Wastewater Denitrification.
Exponent, Inc. for the Methanol Institute, Alexandria, Virginia. Doc. no.
1105602.000 0101 0712 PT01.
United States Geological Survey, 2019. Mineral Commodity Summaries 2019. Min-
eral Resources Program P 116. United States Department of the Interior. https://
doi.org/10.3133/70202434.
U.S. Environmental Protection Agency, 2007. Biological Nutrient Removal Processes
and Costs. Office of Water. NEPIS publication number 823R07002. https://nepis.
epa.gov/Exe/ZyPURL.cgi?Dockey¼ 60000G2U.txt.
United States Bureau of Labor (USBL), 2019. The Consumer Price Index: Concepts
and Content over the Years. The Bureau, Washington, 1977.
United States Department of Agriculture (USDA), 2018. Fertilizer use and price. In:
Workbook, All Fertilizer Use and Price Tables in a Single Workbook. https://
www.ers.usda.gov/data-products/fertilizer-use-and-price.aspx.
United States Environmental Protection Agency (USEPA), 1973. Nitrification and
Denitrification Facilities Wastewater Treatment. Technology Transfer. August.
Revised February 1974. NEPIS publication number 625473004A. https://nepis.
epa.gov/Exe/ZyPURL.cgi?Dockey¼20008RFG.txt.
United States Environmental Protection Agency (USEPA), 1993. Nitrogen Control
Manual. Office of Water, Washington DC. NEPIS publication number 625R93010.
https://nepis.epa.gov/Exe/ZyPURL.cgi?Dockey¼30004MI0.txt.
9
United States Environmental Protection Agency (USEPA), 2007. Wastewater Tech-
nology Fact Sheet: Side Stream Nutrient Removal. Office of Water. NEPIS pub-
lication number 832F07017. https://nepis.epa.gov/Exe/ZyPURL.cgi?
Dockey¼P100IL7Z.txt.
United States Environmental Protection Agency (USEPA), 2009. An urgent call to
action e report of the state-EPA nutrient innovations task group. State-EPA
Nutrient Innovations Task Group. https://www.epa.gov/nutrient-policy-data/
nutrient-innovations-task-group-documents.
United States Environmental Protection Agency (USEPA), 2015a. Case studies on
implementing low-cost modifications to improve nutrient reduction at waste-
water treatment plants. Jointly supported by the office of wetlands oceans and
watersheds office of science and technology and office of wastewater man-
agement 1200 Pennsylvania ave., NW Washington, DC 20460. EPA-841-R-15-
004. http://www2.epa.gov/nutrient-policy-data/reports-andresearch#reports.
United States Environmental Protection Agency (USEPA), 2015b. A Compilation of
Cost Data Associated with the Impacts and Control of Nutrient Pollution. Office
of Water. EPA-820-F-15-096.
van Dongen, U., Jetten, M.S., van Loosdrecht, M.C., 2001. The SHARON-Anammox
process for treatment of ammonium rich wastewater. Water Sci. Technol. 44,
153e160. https://doi.org/10.2166/wst.2001.0037.
Xe Corporation, 2019. XE Currency Converter. Newmarket, Ontario, Canada. https://
www.xe.com/currencyconverter/.