You are on page 1of 12

Journal of Cleaner Production 355 (2022) 131720

Contents lists available at ScienceDirect

Journal of Cleaner Production


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

Application and adaptation of a scale-up framework for life cycle


assessment to resource recovery from waste systems
Nilay Elginoz a, b, *, Isaac Owusu-Agyeman a, Göran Finnveden c, d, Roland Hischier e,
Tomas Rydberg b, Zeynep Cetecioglu a
a
KTH Royal Institute of Technology, Dept. of Chemical Engineering, Sweden
b
IVL Swedish Environmental Research Institute, Life Cycle Management, Sweden
c
KTH Royal Institute of Technology, Dept. of Sustainable Development, Environmental Science and Engineering, Sweden
d
Luxembourg Institute of Science and Technology, Environmental Sustainability Assessment and Circularity, Esch-sur-Alzette, Luxembourg
e
Empa Swiss Federal Laboratories for Material Science and Technology, Technology and Society Laboratory, Lerchenfeldstrasse 5, 9014, St. Gallen, Switzerland

A R T I C L E I N F O A B S T R A C T

Handling Editor: Bruno Silvestre Prospective life cycle assessment (LCA) studies are widely used for evaluating emerging resource recovery sys­
tems. Simulations, engineering-based process calculations and stoichiometric methods are frequently used
Keywords: methods to generate life cycle inventory (LCI) in prospective LCAs. The engineering-based upscaling calculation
Upscaling is an efficient method for LCI generation requiring fewer resources than simulations. This study aims to test an
Laboratory-scale data: pilot-scale data
engineering-based upscaling method for LCI generation and adapt it to biochemical resource recovery processes.
Full-scale data
The method’s validity for biochemical resource recovery processes was tested using data for biogas generation by
Environmental sustainability
Life cycle inventory generation anaerobic digestion in laboratory, pilot, and full scales, and using a combination of lab-scale data and kinetic
Conceptual design equations. Biogas generation was chosen for two reasons: (1) there are several emerging technologies based on
anaerobic digestion with products other than biogas, and (2) data is available for different scales. The results
showed, a substantial difference between the methane production amount in actual and conceptual plants, is an
important cause of the variation in impact category results. Different estimations of fugitive emissions have an
important impact on the global warming potential results. Combination of lab-scale data and kinetic equations
approximates best with the actual plant for the abiotic depletion, eutrophication, freshwater aquatic ecotoxicity,
global warming and photochemical ozone creation potentials. The results are sensitive to biogas generation
amount in several categories.

engineering calculations (using mass and energy balances with process


calculations) and stoichiometric approaches are methods that can be
1. Introduction applied to generate inventory data in prospective LCAs (Parvatker and
Eckelman, 2019). Using process simulation tools for life cycle inventory
Introducing Life Cycle Assessment (LCA) in the design stage enables (LCI) generation gives similar results compared to plant-based data
the determination of hotspots in emerging technology and may lead to (Parvatker and Eckelman, 2019), however, this method requires much
process improvements (Ögmundarson et al., 2020). LCA is a more time and expertise compared to engineering-based process calcu­
well-established methodology (Rebitzer et al., 2004) but due to its na­ lations (Yao and Masanet, 2018).
ture, detailed information about material and energy flows is needed. Upscaling based on laboratory data using detailed process calcula­
Unlike mature technologies, full-scale plant data is not yet available for tions is widely used to generate LCIs for emerging technologies (Mor­
emerging technologies that are at low technology readiness levels (TRL), gan-Sagastume et al., 2016; Tsoy et al., 2020). There are however only a
and this lack of data availability is one of the major challenges of few studies testing the engineering process calculations based LCI gen­
early-stage LCAs (Hetherington et al., 2014; Moni et al., 2020). eration methods using data of a mature technology in different scales.
In the absence of plant-based data, process simulation (using process After reviewing prospective LCAs, Tsoy et al. (2020) reported that
simulators to model full-scale plants in virtual environment), process

* Corresponding author. IVL Swedish Environmental Research Institute, Life Cycle Management, Sweden.
E-mail addresses: nilay.elginoz.kanat@ivl.se (N. Elginoz), isaacoa@kth.se (I. Owusu-Agyeman), goran.finnveden@abe.kth.se (G. Finnveden), roland.hischier@
empa.ch (R. Hischier), tomas.rydberg@ivl.se (T. Rydberg), zeynepcg@kth.se (Z. Cetecioglu).

https://doi.org/10.1016/j.jclepro.2022.131720
Received 17 June 2021; Received in revised form 11 March 2022; Accepted 8 April 2022
Available online 15 April 2022
0959-6526/© 2022 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
N. Elginoz et al. Journal of Cleaner Production 355 (2022) 131720

Abbreviations GWP Global Warming Potential


HTTP Human Toxicity Potential
TRL Technology readiness level MAETP Marine Aquatic Ecotoxicity Potential
LCA Life cycle assessment ODP Ozone Layer Depletion Potential
LCI Life cycle inventory POCP Photochemical Ozone Creation Potential
WWTP Wastewater treatment plant TETP Terrestrial Ecotoxicity Potential
AD Anaerobic digestion SR Sensitivity ratio
LCIA Life cycle impact assessment DM Dry material
CML Centrum voor Milieuwetenschappen TS Total solids
ADP Abiotic depletion potential VS Volatile solids
AP Acidification Potential HRT Hydraulic retention time
EP Eutrophication Potential VFA Volatile Fatty Acid
FAETP Freshwater Aquatic Ecotoxicity Potential PHA Polyhydroxyalkanoates

prospective LCA results were most of the time not validated after along their entire life cycle to avoid burden-shifting. When the biomass
application of the assessed technologies at industrial scale, and they used for the biochemical production is agricultural feedstock, land-use
recommended validation of these results in order to analyse the effi­ impacts are much higher than for petrochemicals (Ögmundarson
ciency of the upscaling methods. Few examples investigated such vali­ et al., 2020). Thus, recovering resources to obtain bio-products from
dation. Fernandez-Dacosta et al. (2019) compared early-stage waste streams seems a promising option for producing bio-based ma­
assessments, prospective LCA and LCA based on commercial scale data terials. However, the consumption of energy and chemicals during the
using lactic acid production as a case study. Tsalidis and Korevaar various production and separation steps of such resource recovery
(2022) compared LCAs of torrefaction technology based on data in routes leads to environmental impacts that need to be assessed. This
different scales (bench scale, lab scale, process simulations, pilot scale assessment then enables a comparison of these emerging technologies
and commercial scale). with their incumbent counterparts.
There are several suggested upscaling frameworks in the literature. The current study aims to test and adapt the framework proposed by
Shibasaki (2009) proposed a method for upscaling from pilot to full Piccinno et al. (2016) to resource recovery technologies by comparing
scale. A weakness of this method is the need for the technology in LCI and life cycle impact assessment (LCIA) results based on actual
question to be in a higher TRL compared to laboratory scale. Piccinno plant-based data with results based on up-scaled data. Biogas generation
et al. (2016) and Simon et al. (2016) proposed frameworks for upscaling from wastewater sludge data was collected from laboratory experi­
from lab to full scale based on engineering process calculations and Tsoy ments, a pilot plant and an actual full-scale AD plant. The main inno­
et al. (2020) proposed another framework. The framework proposed by vation point of this study is the validation of LCI and LCIA results
Simon et al. (2016) consists of three steps which are analysis of (1) generated by the chosen up-scaling framework for a bio-chemical
functions, (2) dimensions and (3) similarities. In the first step, labora­ resource recovery system. The second innovation point of this study is
tory procedures are resembled to industrial processes, in the “analysis of the adaptation of the framework to such systems.
dimensions” step, production of the new material is quantitatively The specific objectives addressed in the study include:
described, and in the last step, analogies with large scale processes are
analysed. The framework published by Tsoy et al. (2020) has also three 1) Applying the upscaling framework from Piccinno et al. (2016) to
steps; first the technology developer designs a conceptual scenario for assess potential environmental impacts based on lab-scale experi­
the new technology, second LCA and technology experts develop a flow mental results and pilot plant data;
chart for LCA and finally they estimate data for unit processes. The core 2) Comparing the results obtained by this upscaling framework with
idea in these frameworks is to design a conceptual upscaled system for actual data from an industrial plant; and
the new technology based on similar existing industrial processes. 3) Refining the upscaling method for biochemical resource recovery
However, validation and verification of these scale-up frameworks is the processes, as a result of this comparison.
common weakness.
The upscaling framework tested in the current article is from Pic­ 2. Material and method
cinno et al. (2016). It is for upscaling of chemical processes from lab to
full scale and has been applied in four different case studies in the 2.1. Modelling approach
literature (Piccinno et al., 2016, 2018; Elginoz et al., 2020a, 2020b). The
framework starts with a laboratory protocol covering all the steps in the LCA studies of four biogas production pathways from wastewater
lab-scale. A conceptual simple plant is designed based on the lab pro­ sludge are performed for four scenarios: (S1) Full-scale system using
tocol, and then each step is scaled-up separately. LCA is conducted after measured data, (S2) Calculated conceptual up-scaled system based on
linking process steps. The framework is developed for chemical pro­ laboratory data, (S3) Calculated conceptual up-scaled system based on
cesses and in the scope of this study, the framework’s validity for pilot-scale data, (S4) Calculated conceptual up-scaled system based on a
biochemical resource recovery processes is tested using an established combination of laboratory data and kinetic equations.
biochemical process. Biogas generation by anaerobic digestion (AD) is For S1, plant-based data was collected from the sludge line of Hen­
used to test and adapt the framework for emerging biochemical pro­ riksdal Wastewater Treatment Plant (WWTP), which includes AD re­
cesses. This is relevant since AD is adapted for the production of bio­ actors for biogas generation. S2 is based on an AD system that has been
diesel, bioethanol, volatile fatty acid (VFA), Polyhydroxyalkanoates set up to produce biogas using batch reactors in the laboratory and run
(PHA), biohydrogen and long-chain fatty acids, respectively, from waste until maximum biogas production is reached. The influent of this
streams. laboratory-scale system was collected from the Henriksdal WWTP
Bio-based chemicals are often seen as a more sustainable option sludge line at the same time when plant data was collected. For S3,
compared to their fossil-based counterparts. But it is important to sludge line data from Hammarby Sjöstadsverket pilot plant, which is a
evaluate and, as far as possible, reduce their environmental impacts model version of Henriksdal WWTP where new processes or changes in

2
N. Elginoz et al. Journal of Cleaner Production 355 (2022) 131720

the processes are applied first, was collected. For S4, the biogas forma­ major and minor contributors to the total burdens and the effect of
tion rate was calculated based on experimental data and kinetic equa­ avoided burdens on the impact category results. Perturbation analysis
tions. For the second, third and fourth LCAs three conceptual AD plants was performed to assess the impact of parameter uncertainties on the
were designed, following the upscaling procedure in Piccinno et al. LCIA results following Clavreul et al. (2012). In this analysis, for each
(2016). The applied procedure is explained in detail in sections 2.4 to analysed parameter impact category results are recalculated for +10%
2.7 and the scenarios are summarized in Table 1. and − 10% variation of the initial parameter amount. Sensitivity ratios
All LCAs were performed in line with the ISO 140040-series (Inter­ (SR) for each parameter are calculated for each impact category using
national Organization for Standardization, 2006a; 2006b), using the Equation (1). If SR value of one parameter is 2, it means that the 10%
GaBi software version 9 (Thinkstep, 2020). Background data are mainly variation in the parameter cause a 20% change in the total impact
taken from the Ecoinvent database version 3 (Wernet et al., 2016). category result (Clavreul et al., 2012).
Chosen Ecoinvent datasets and the reasons for these choices are given in Δresult
supplementary material (Table S1). The functional unit of all four LCAs SR = initial result
(1)
Δparameter
is management of 1 ton of sludge in dry basis (1 tonDM). The investi­ initial parameter
gated system boundaries for the AD plant are described in Fig. 1. Sludges
enter the system using the zero-burden approach (c.f. Finnveden, 1999) 2.4. Scenario 1 – actual, full scale biogas generation in Henriksdal
similar to waste management studies. System boundary includes WWTP
collection of sludges from different parts of the WWTP, biogas genera­
tion in AD reactors, upgrading the biogas, dewatering the sludge, Henriksdal WWTP, operated by the Stockholm Water and Waste
transportation and incineration of the dewatered digestate. system Company, is located in Stockholm, Sweden. All data for the Henriksdal
boundary is expanded to include avoided heat production due to WWTP is collected through personal communications. The WWTP treats
incineration of the digestate and avoided diesel production replaced by approximately 250 000 m3 of municipal wastewater per day using me­
biogas. Construction of the facility is excluded from the system due to chanical, chemical and biological processes. The sludges from treatment
the uniqueness of Henriksdal WWTP. The facility was constructed un­ processes are fed into AD reactors to produce biogas. The digesters of
derground by blasting and excavation of rocks. Rock sides are also used Henriksdal, consist of two 7000 m3 and five 5000 m3 anaerobic reactors,
as walls for the AD reactors. This unique construction makes it hard to which are continuously operated under mesophilic conditions
collect construction data and less relevant in relation to the aim of the (35–37 ◦ C). The sludge line of the WWTP also covers centrifuges for
study. thickening excess sludge from secondary clarifiers and for dewatering
digested sludge and sludge tanks. Biogas upgrading is operated by
2.2. Impact assessment Scandinavian Biogas in the premises of the WWTP.
The AD reactors were fed by 91.3 × 104 m3 sludge in total during
For the comparison of full-scale data and up-scaled data, both LCI 2018. 81% of this total sludge was primary sludge from primary clari­
and LCIA results are used. For the LCIA step, the January 2016 version of fiers, 14% was thickened secondary sludge from secondary clarifiers, 5%
the so-called CML-method 2001 (Guinée et al., 2002) is used, as was grease separator sludge. Dry material (DM) content of the input
implemented in GaBi. The considered CML impact categories are the sludge (total solids, TS %) was 4.03% in 2018, hence total mixed sludge
abiotic depletion potential (ADP) the acidification potential (AP), the has a DM content of 37.9 × 103 ton, considering the density of mixed
eutrophication potential (EP), the freshwater aquatic ecotoxicity po­ sludge is 1030 kg/m3 59.4 × 102 tons crude glycerol was added to the
tential (FAETP), the global warming potential (GWP), the human sludge.
toxicity potential (HTP), the marine aquatic ecotoxicity potential The sludge line consumed 10 GWh of electricity, which is withdrawn
(MAETP), the ozone layer depletion potential (ODP), the photochemical from the grid, and 22 GWh of heat energy during 2018. 98% of heat
ozone creation potential (POCP) and the terrestrial ecotoxicity potential consumption was supplied from district heating, and the rest was sup­
(TETP). plied internally from heat transfer and boilers. Chosen background
datasets for energy inputs are explained in the supplementary material
2.3. Contribution and perturbation analysis (Table S2). 16.3 × 106 Nm3 biogas with 63% methane content was
generated in the digesters. 97.6% of the generated biogas was used as
Contribution analysis was performed using LCIA results to assess the fuel in biogas busses after upgrading. Excessive biogas sent to the torch
was 2.1%, and 0.3% of biogas was used in boilers. For simplification it is
assumed that all generated biogas is upgraded to replace diesel.
Table 1
Summary of the implied scenarios and their main characteristics. Generated biogas amount to upgraded biogas ratio is 1.86, and ac­
cording to Swedish Gas Technology Centre (SGC, 2012) 1 Nm3 upgraded
Scenarios Main characteristics
biogas corresponds to 0.99 L diesel. Fugitive methane emission from the
S1 – Actual, full scale system in Two 7000 m3 and five 5000 m3 AD sludge line was estimated as 5.7% of the generated biogas.
Henriksdal WWTP reactors
Digested sludge sourced from ADs was dewatered in centrifuges with
Commercial scale collected data
collected 196.5 tons of polymer addition and stored in tanks before transferring
431.4 Nm3 biogas generation per 1 ton for land reclamation and other purposes. The total amount of drained
DM digestate was 61.9 × 103 tons. Dry material content of the dewatered
S2 – Conceptual system based on One 5500 m3 AD reactor digestate was 28.3% TS, hence 17.5 × 103 ton DM. Dewatered digestate
laboratory data Lab-scale data upscaled to full-scale
59.74 Nm3 biogas generation per 1 ton
from WWTPs can be disposed in several ways like land reclamation,
DM incineration, spreading on agricultural areas etc. Here it is assumed that
S3 – Conceptual system based on pilot- One 5500 m3 AD reactor the digestate is transported to a nearby incineration plant, which has a
scale data Pilot-scale data upscaled to full-scale 12 km distance from Henriksdal WWTP. The recovered heat is calcu­
307.64 Nm3 biogas generation per 1 ton
lated as 0.523 MJ per 1 kg digestate based on the Ecoinvent process used
DM
S4 – Conceptual system based on One 5500 m3 AD reactor (treatment of digester sludge, municipal incineration) and it is assumed
laboratory data and kinetic equations Lab scaled data coupled with kinetic that it will replace district heating. Reject water from the centrifuges
equations upscaled to full scale goes back to primary clarifiers. The studied system is described in Fig. 1.
343.83 Nm3 biogas generation per 1 ton
DM

3
N. Elginoz et al. Journal of Cleaner Production 355 (2022) 131720

Fig. 1. System boundary of Henriksdal WWTP sludge line under investigation. Red dotted line depicts investigated system and black dotted line depicts expanded
system boundaries. Solid boxes are the processes which are included in the system boundary and dashed boxes are the excluded ones. The green box covers avoided
burden processes.

2.5. Scenario 2 – conceptual biogas generation system based on


laboratory data

Lab-scale experiments for biogas generation were conducted, at the


Hammarby Sjöstadsverk, which is run in collaboration by Swedish
Environmental Research Institute and Royal Institute of Technology.
Primary, secondary and fat sludge was collected from Henriksdal WWTP
and a 1000 ml mixture was prepared following the ratios 81%, 14%, 5%
respectively in line with the year 2018 amounts. 6.5 g glycerol was
added to the mixture. The volatile solids (VS) content of primary sludge,
secondary, fat, and seed sludges were 8.47, 27.12, 36, and 15.21 g/L
respectively. The experiment was performed with an automatic methane
potential test system (AMPTS II, Bioprocess Control) with glass bottles
(3 triplicate) connected to a CO2-capturing unit as previously described.
To each 500 ml batch reactor, 320 ml sludge mixture as substrate and
130 ml seed sludge as inoculum were fed. The batch reactors were Fig. 2. Average CH4 accumulation in batch reactors.
placed in a hot water bath to control the temperature at 35 ◦ C (similar to
the full-scale reactors). Accumulated CH4 was continuously measured (HRT) is assumed to be 20 days because the accumulated CH4 reached
for 40 days using a gas volume measuring device. The average CH4 its equilibrium on the 20th day of the experiments. The flow rate of the
accumulation time series of the three reactors is given in Fig. 2. Ac­ influent to the digester was calculated as 250 m3/day. Methane pro­
cording to the experimental results, the specific methane production in duction of the conceptual system was calculated as 391 m3 CH4/day
the batch reactors was 54.5 ml CH4/g VS. based on the average volatile solids (VS) amount of the actual sludge
The upscaling framework proposed by Piccinno et al. (2016) was influent and specific methane production value obtained in the experi­
followed to upscale lab-scale biogas generation to a conceptual full ments. Daily biogas production was calculated as 620.5 m3 considering
scale. The first step of the framework is the lab protocol of the investi­ 63% of the biogas is methane.
gated system which is defined in the paragraph above. Second, a con­ For third and fourth steps in the framework, each process is scaled up
ceptual plant flow chart is designed. This conceptual system includes a separately, and then they are linked by inputs and outputs of each
centrifuge for thickening secondary sludge, an AD reactor with 5000 m3 process. The details of each process and the flow amounts between each
reaction volume, a dewatering centrifuge for digested sludge, and stor­ unit were calculated and given in the supplementary material in detail.
age tanks similar to the actual plant. The Hydraulic Retention Time Energy and material consumption and fugitive emissions from the

4
N. Elginoz et al. Journal of Cleaner Production 355 (2022) 131720

conceptual system were calculated using engineering-based equations. chain fatty acids.
Energy for heating (Qreact ), stirring (Estir ), and pumping (Epump ) was The relationship between the substrate (S) utilization rate and the
calculated using Equations (2)–(4) (Piccinno et al., 2016). Energy for product (P) formation rate is expressed as (Manchala et al., 2017):
thickening and dewatering per ton dry material was estimated as 10
dP dS
kWh (Piccinno et al., 2016). Lubricating oil and fugitive CO2 and CH4 =− %YP/S (5)
dt dt
amounts were calculated using the amounts in the AD process in
Ecoinvent 3 database. The polymer amount used for dewatering the dS 1 dP
digestate was estimated as 4.6 g/kg TS. TS contents of all sludge flows − = % (6)
dt YP/S dt
within the system are given in supplementary material.
YP/S is the product yield coefficient which is the ratio between the
Cp × mmix × (Tr − T0 ) + A × ksa × (Tr − Tout ) × t change in product and change in biomass
Qreact = (2)
ηheat
ΔP
YP/S =
3
Np × ρmix × N × d × t 5 ΔS
Estir = (3)
ηstir Microbial growth is given by
dX
m × g × Δh = μX (7)
Epump = (4) dt
ηpump
In these equations Cp, mmix and ρmix are specific heat capacity, mass X = the biomass concentration [g/L]
and density of the sludge and Tr, T0 and Tout are reaction, starting and μ = specific growth rate [d− 1]
outside temperatures, respectively. A is the surface area of the reactor, ka
and s are the thermal conductivity and thickness of the insulation ma­ And items of the rate of substrate utilization and microbial growth
terial. Np and d depict the power number and diameter of the impeller, N
is the stirring velocity, t is the reaction time, g is gravitational acceler­ −
dS 1 dX
= % (8)
ation, Δh is the pumping height and η stands for efficiency of the device. dt Y dt
The conceptual design of the processes and energy calculations are given
in detail in the supplementary material.
Biogas upgrading and digestate incineration facilities are external
Y is the growth yield coefficient;
facilities and are not included in the laboratory protocol. Therefore, they
were not upscaled to full scale using the upscaling framework. They are
where Y = ΔX
assumed to function similar to S1 in all scenarios and in the LCA model ΔS
Specific growth rate μ is given by:
they were integrated to the conceptual upscaled system as full-scale
processes using relevant Ecoinvent datasets. S
μ = μmax (9)
Ks + S
2.6. Scenario 3 – conceptual biogas generation system based on pilot-scale
data μ = Specific growth rate
μmax = maximum specific growth rate [d− 1]
The pilot plant is operated continuously. It was constructed in 2013 S = substrate concentration [g/L]
for research and development of Henriksdal WWTP’s planned mem­ Ks = half-saturation constant [g/L]
brane bioreactors, and in 2017, the sludge line was added to the pilot
plant for investigation of digestion processes (Andersson et al., 2020). The growth rate and the rate of substrate utilization can therefore be
The sludge treatment line of the pilot plant processed mixed sludge given by
varying between 30 and 80 L/h during the year 2018. The average
dX SX
specific gas production in the pilot plant (280.7 ml CH4/g VS) was = μmax (10)
dt Ks + S
calculated based on measurements in weekly reports of the plant.
Methane production of the conceptual system in S2 was calculated as dS SX
2013 m3 CH4/day based on specific gas production of the pilot plant and = − km (11)
dt Ks + S
VS content of the sludge. For S3, the conceptual design explained in
section 2.5 was updated using methane production value. Combining equations (4) and (9), the product formation rate is given
as:

2.7. Scenario 4 – conceptual biogas generation system based on dP


= YP/S km
SX
(12)
laboratory data and kinetic equations dt Ks + S

km = maximum specific substrate utilization rate [d− 1]


The methane production per input amount of sludge varies consid­
Based on Equation (12), the formation of a bio-product (in this case
erably between the three preceding scenarios (431.4, 59.7 and 307.6
biogas) depends on factors such as product coefficient (YP/S) and
Nm3 CH4/ton DM sludge, respectively). For S4, we use a combination of
maximum substrate utilization rate (km). In order to estimate the rate of
theoretical production yield equations and experimental data for
product formation, each step of the AD process needs to be considered
biochemical resource recovery processes instead of using experimental
leading to the final stage. The substrates utilization rate depends on rates
data alone. In this section, equations are proposed for calculating pro­
of the various steps of the AD process including the rate of disintegra­
duction yield in AD processes in the upscaling procedure. Combining
tion, hydrolysis, uptake of monomers, update of volatile fatty acids and
these equations with experimental data, the production amount in
hydrogen which depends on the kind of substrate and the intermediate
conceptual cases can be estimated more precisely. The kinetic equations
products. The rate equations and suggested parameter values are given
which are given below can be used to calculate production yield and
in the AD model no.1 (ADM1) (Batstone et al., 2002). It is noteworthy to
amount for bioprocesses that produce items such as biodiesel, bio­
mention that parameters such as maximum specific substrate utilization
ethanol, VFA, Polyhydroxyalkanoates (PHA), biohydrogen and long-

5
N. Elginoz et al. Journal of Cleaner Production 355 (2022) 131720

rate and yield coefficients depend on the operating temperature and cases, the sludge and glycerol input ratios are the same as explained
reactor type. The given equations (5)-(12) can be used to calculate the above, therefore they have the same amounts per FU. Between the
product yield based on experimental data. The coefficients which are conceptual scenarios, the differing parameter is the generated biogas
needed in the equations can be found in the literature. For example, the amount, therefore calculated electricity, district heating, polymer,
coefficients for productions of biohydrogen by dark fermentation, digester sludge incineration and transportation amounts are the same
methane, and VFA can be obtained from ADM1 (Batstone et al., 2002) since the calculations are based on sludge input. The biogas related
and other recent literature (Aldin et al., 2011; Arudchelvam et al., 2010; parameters (biogas upgrading and avoided diesel) are different amongst
Momoh et al., 2013; Sun et al., 2021). Also, coefficients could be all scenarios. Avoided heat amount is the same in S2-4 since it is
retrieved from literature for PHA production (Chakravarty et al., 2010; correlated to dewatered digestate amount. The results for all parameters
Fortela et al., 2019; Gujer et al., 1999), biodiesel (Surendhiran et al., are discussed below.
2015), as well as for bioethanol (Srimachai et al., 2015). Calculated electricity consumption (11.72 kWh/1 ton DM), is much
From the lab scale experiments (Fig. 2), the product yield coefficient, less than the collected amount (267.47 kWh/1 ton DM). In the con­
YP/S is 54.634 mLCH4/g. ceptual cases electricity consumption was calculated for pumping, stir­
ring, thickening and dewatering. For solid-liquid separation processes
S = 15 g/L (thickening and dewatering) 10 kWh per ton DM is used, which is listed
X = 6.0 g/L as an upper limit by Piccinno et al. (2016). In the actual case, the
Ks = 0.16 g/L electricity consumption in the sludge line is given as a summation of all
electricity-consuming processes, therefore it is hard to detect where the
From Fig. 2, the maximum production rate was obtained at t = 0.125 difference is sourced.
h, Thus for km calculation, dP Heating energy for the conceptual cases is more than twice the
dt = 11.87 mLCH4/h. Therefore km can be
calculated as 2 d− 1. In the conceptual full-scale, using the calculated km amount used for the actual plant. It is calculated in two parts, Qheat and
and Ks , full-scale parameters below: Qloss, Oheat is the energy needed to increase the temperature of the re­
action to a certain level and Qloss is for compensating the energy loss
S = 33 g/L from the reactor surface. In this case, 46.4% of the calculated energy is
X = 21 g/L Qloss. In the Qloss calculation, the reactor is considered to be in direct
Ks = 0.16 g/L contact with air, having a glass fibre insulation material with a thickness
YP/S = 0.22 LCH4/g of 75 mm and the ambient temperature is estimated as 7 ◦ C which is the
yearly average temperature in Stockholm. However, as mentioned
− 1 − 1 above, ADs in Henriksdal have a different design and they are sur­
The rate of product formation, dP dt is calculated as 9.0 L CH4 L d
3 3 rounded by rock. This might be preventing heat losses, and this can
which corresponds to 2250 m CH4/day and 343.8 Nm CH4/ton DM
therefore cause the difference between the conceptual and actual cases.
sludge in the conceptual full-scale design. For S4, the conceptual design
Polymer dosage to the dewatering centrifuge is calculated as 4.6 g
explained in section 2.5 was updated using this production amount.
polymer per kg TS and the calculated amount is 58% of the actual
amount. Therefore, a higher amount of polymer is suggested to be used
3. Results
in the future upscaling applications. Biogas send to upgrading is the
lowest in S2 and highest in S1, and there is a substantial difference be­
3.1. Comparison of LCI results
tween all scenarios. The possible causes of this difference is discussed in
detail in the discussion section. Calculated dewatered digestate amount
LCIs of actual and conceptual cases are summarized in Table 2 per 1
in S2-4 is 80% of the actual case.
ton of sludge in dry basis (1 ton DM) and compared in Fig. 3 by
In the actual case, fugitive emissions are estimated as 5.7% of biogas
normalizing the input/output amounts in S2-4 against S1. In all four
production (personal communication), in the conceptual case 0.047 kg
abiotic CO2 (from abiotic sources such as detergents in the wastewater)
Table 2 and 0.022 kg biotic CH4 per 1 m3 biogas is estimated as fugitive emis­
Inventory data per management of 1 ton of sludge in dry basis (1 ton DM) for sions based on the Ecoinvent dataset.
actual plant (S1) and conceptual cases (S2-4).
Inputs and outputs Unit S1 S2 S3 S4
3.2. Comparison of LCIA results

Grease separator sludge m3/FU 1.13 1.13 1.13 1.13


The LCIA results of all scenarios are explored in this section (Fig. 4).
Thickened excess sludge m3/FU 3.34 3.34 3.34 3.34
Primary sludge m3/FU 19.60 19.60 19.60 19.60 The results are normalized against the actual case. For the ADP ele­
Glycerol kg/FU 156.00 156.44 156.44 156.44 ments, FAETP, GWP and POCP categories, results of S4 approximates
Electricity kWh/ 264.47 11.72 11.72 11.72 best with the actual plant case, and the result of S2 has the widest gap.
FU
For the AP category, the result of S2 is only 7% higher than S1, and S4 is
District heating kWh/ 570.62 1318.22 1318.22 1318.22
FU 17% higher than the actual case. For the EP and TETP categories, results
Polymer kg/FU 5.18 2.99 2.99 2.99 of the conceptual scenarios have insignificant differences (1–6%) when
Biogas to upgrading Nm3/ 431.40 59.74 307.64 343.83 compared to the actual case, and for the HTP category, S3 and S4 fits
FU very well with the actual case. ADP fossil and ODP, which are the cat­
Digester sludge kg/FU 1633.09 1299.67 1299.67 1299.67
egories that are affected most by the avoided burdens have a different
incineration
Fugitive CH4 emissions kg/FU 10.32 1.34 1.34 1.34 pattern. In both categories, the actual case has the highest environ­
Fugitive CO2 emissions kg/FU – 2.80 2.80 2.80 mental gain, followed by S4 and S3 but in S2 the avoided burdens could
Lubricating oil kg/FU 0.11 0.11 0.11 0.11 not compensate the burdens. The total LCIA results for all categories are
Digester sludge t.km/ 19.60 15.60 15.60 15.60
given in supplementary material.
transportation FU
Biogas upgrading Nm3/ 231.94 32.12 165.40 184.85
FU 3.3. Contribution analysis
Avoided heat kWh/ 237.25 188.81 188.81 188.81
FU In S1, glycerol is the most important contributor in 5 out of 10 impact
Avoided diesel kg/FU 187.14 25.91 133.45 149.15
categories (ADP elements, AP, EP, FAETP and TETP) and an important

6
N. Elginoz et al. Journal of Cleaner Production 355 (2022) 131720

Fig. 3. LCIs of all scenarios normalized to S1. Fugitive CO2 emissions are excluded from the graph since there is no data for S1.

Fig. 4. Comparison of CML 2001 impact category results of all scenarios. All scenario results are normalized to S1.

contributor for the rest of the categories, see Fig. 5. Biogas upgrading is Scenarios 3 and 4 (see Figs. 7 and 8), which produce similar amount
also an important contributor and it surpasses the avoided burdens of biogas (307.6 and 343.8 Nm3 per FU respectively) have very similar
provided by biogas replacing diesel in all categories except ADP fossil contribution patterns. Glycerol, district heating and biogas upgrading
and ODP. In these two categories total impacts are negative showing that are important contributors in 8 out of 10 impact categories. For the ADP
the gain due to avoided diesel is higher than the burdens of the system. fossil and ODP categories the gain provided by diesel replacement is
District heating is a minor contributor in 8 out of 10 categories and for higher than the burdens hence total impact category results are
ADP elements and TETP, its contribution is insignificant. Incineration of negative.
digester sludge is an important contributor only for the EP and FAETP Taking a closer look at GWP, which is an important impact category,
categories and avoided heat has an insignificant impact in all categories. and assessing the source of variation between the conceptual cases and
The fugitive emissions are significant for the GWP results. the actual case, is useful. The total impact of scenario 1 is 472 kg CO2-eq
In S2, glycerol is the biggest contributor for the ADP elements, AP, higher than scenario 2. 47% of this difference is caused by the difference
EP, FAETP, HTP, ODP and TETP categories, see Fig. 6. For the ADP fossil, in fugitive emission estimations and the rest is due to the higher amount
GWP and POCP categories, the impact of glycerol is close to the impact of biogas upgrading in S1. The GWP results of scenarios 3 and 4 are
of district heating which is also an important contributor. The positive closer to the actual case since the biogas generation are closer and hence
impact of avoided diesel and the negative impact of biogas upgrading the biogas upgrading amount, compared to S2. If the fugitive emissions
are less in this scenario compared to S1. Since the generated biogas in S2 in the conceptual cases would have been estimated closer to the actual
is only 14% of S1, this is an expected result. case, the GWP of S4 would be 5% higher than actual case instead of

7
N. Elginoz et al. Journal of Cleaner Production 355 (2022) 131720

Fig. 5. Contribution analysis of S1 CML 2001 LCIA results.

Fig. 6. Contribution analysis of S2 CML 2001 LCIA results.

being 13% lower. The emissions from the avoided diesel use plays a 4. Discussion
limited role in all scenarios (Figs. 5–8).
4.1. Biogas production amount and influent characterization
3.4. Perturbation analysis
The difference between the biogas yields from the lab-scale experi­
The results of perturbation analysis, which are plotted for each ments and the full-scale reactor can largely be attributed to the different
impact category in Fig. 9, show that ADP elements, AP, GWP, HTP and reactor configurations. The lab-scale experiment was done in a batch
POCP categories are sensitive to biogas upgrading amount. When mode of operation whereas the full-scale reactor is operated in a
increasing the amount of biogas upgrading by 10%, in mentioned cat­ continuous mode. It is known that batch reactors result in more product
egories SRs are 0.4–0.7. The ADP fossil and ODP categories are highly and substrate inhibitions than continuously stirred tank reactors. In the
sensitive to avoided diesel, 10% increase in the parameter results in SRs batch reactors, products (such as VFAs) that have inhibitory properties
of 2.4 and 1.9. The GWP category results are also sensitive to fugitive stay within the reactor for a much longer time, whereas in a continuous
CH4 emissions and district heating amount (0.2 and 0.12) since SR reactor, they are continuously removed with the effluents. In terms of
values higher than 0.1 are considered significant here. kinetics, it is known that continuous reactors are more efficient in
operating close to the steady-state than batch reactors. For instance, the
rate of disintegration and hydrolysis of substrates are known to be
higher in continuous reactors than batch reactors. It is estimated that
during digestion of primary sludge, the protein hydrolytic rate constant

8
N. Elginoz et al. Journal of Cleaner Production 355 (2022) 131720

Fig. 7. Contribution analysis of S3 CML 2001 LCIA results.

Fig. 8. Contribution analysis of S4 CML 2001 LCIA results.

in a batch reactor is only 0.2 d− 1 whereas in a continuous reactor at a conceptual design. For example, in this study, the specific biogas pro­
steady-state, it is 0.58 d− 1 (Batstone et al., 2002). duction amount (54.5 ml CH4/g VS) obtained in the experiments and 5-
Another important point for resource recovery systems is the varia­ year average VS value of Henriksdal WWTP sludge line influent were
tion in the lab-scale experiments’ results. As an example, AD is applied used to calculate potential biogas production of the conceptual plant for
to a wide range of waste streams or combinations of waste streams, S2.
including food waste, wastewater sludge, agricultural waste, industrial In the industrial biotechnology applications of resource recovery
wastewater with high organic loads, etc. Waste streams can be charac­ from waste, the inoculum is typically sourced from waste and therefore
terized by TS and VS contents and other parameters like chemical oxy­ it contains a mixture of organisms. This also leads to a large variation in
gen demand. Due to high variations in waste compositions, production results. For example, VFA recovery from waste streams is an emerging
amounts should be determined as specific production amounts per VS or technology in the research phase. According to the literature (Atasoy
other characterization parameters rather than per mass or volume. et al., 2020a, 2020b) even with the same operating conditions and same
Although the wastes with the same VS amount do not provide the same substrate, production yield varies based on different microorganism
amount of product due to variations in environmental and operational compositions in the influent sludge.
conditions and the variation in the biodegradability of organic matter During the research phase of new biotechnology, batch experiments
content, it approximates better to the reality than volumetric amounts. If are normally conducted first to optimize operational and environmental
the waste samples in the lab experiments do not reflect long-term vari­ conditions. For the best conditions, continuous long-term experiments
ations in the waste composition, the specific production per related are then conducted. If the full-scale plant shall be designed as a
characterization parameter obtained in the experiments should be continuous reactor, the results of this second group of experiments
coupled with the long-term characterization of the waste stream in the should be used for conducting LCA calculations instead of batch

9
N. Elginoz et al. Journal of Cleaner Production 355 (2022) 131720

Fig. 9. Parameter sensitivity ratios with respect to impact categories for S1.

experiments. line. However, due to the abovementioned variations in the conditions,


In the present study, the laboratory scale experiments were con­ the specific biogas production in the experiments did not match the
ducted for forty days in triplicate batch reactors for the same substrate, actual biogas production. Taking weekly composite samples and con­
which was collected as a grab sample from the Henriksdal WWTP sludge ducting the experiments for a year (if possible) would give closer results

10
N. Elginoz et al. Journal of Cleaner Production 355 (2022) 131720

Fig. 9. (continued).

to an industrial plant’s potential production. different kinds of data (pilot scale, upscaled pilot scale, data on different
TRLs etc.).

4.2. Comparison with previous studies 5. Conclusion

There are a few studies comparing different LCI inventory generation The transition to using waste streams as valuable sources leads to a
methods. According to Parvatker and Eckelman (2019) a process-based great deal of scientific research on resource recovery processes and there
method combined with advanced calculations approximates best with is a need for an established method to assess their sustainability. An
the plant-based data for GHG emissions, although they overestimated engineering-based upscaling framework for chemical processes devel­
the emissions. After comparing 18 impact category results, they suggest oped by Piccinno et al. (2016) was tested. Application of this framework
that an engineering-based LCI generation method should be used only to a biochemical case showed that the main difference between
for two impact categories: GWP and cumulative energy demand. How­ plant-based data and upscaled experimental data is sourced from the
ever, according to the results of this study, the ratio between LCIA re­ difference in the production yield. It is suggested that the framework
sults based on calculated data and plant data is quite close for the EP and developed by Piccinno et al. (2016) is further developed so that pro­
HTP categories (Fig. 3). In this study, TETP category also gives very duction yields are calculated using kinetic equations and experimental
similar results in all scenarios but it is caused by the sole contributor data. For future studies, validation of upscaling methods using a wide
being glycerol in this category. Tsalidis and Korevaar (2022) claim that range of case studies and data in different scales is suggested.
for the torrefaction process, the difference between GWP results based Additionally, in future studies, it is recommended to study the effect
on pilot scale and commercial scale data is negligible, according to this on the LCA results of using long-term batch and continuous experiment
study, the GWP based on upscaled pilot scale data is 17% lower than results instead of short-term batch experiment results.
commercial scale LCA. According to Fernandez-Dacosta et al. (2019)
except for freshwater ecotoxicity and land use categories, results of LCA CRediT authorship contribution statement
based on commercial data is notably lower than LCA based on pre­
liminary process design in TRL 5–6. This shows the importance of con­ Nilay Elginoz: Conceptualization, Methodology, Formal analysis,
ducting such validation studies for different kinds of processes and

11
N. Elginoz et al. Journal of Cleaner Production 355 (2022) 131720

Writing – original draft, Funding acquisition, Visualization. Isaac Finnveden, G., 1999. Methodological aspects of life cycle assessment of integrated solid
waste management systems. Resour. Conserv. Recycl. 26, 173–187.
Owusu-Agyeman: Investigation, Formal analysis, Resources. Göran
Fortela, D.L.B., Farmer, K., Zappi, A., Sharp, W.W., Revellame, E., Gang, D., Zappi, M.,
Finnveden: Conceptualization, Writing – review & editing, Supervision. 2019. A methodology for global sensitivity analysis of activated sludge models: case
Roland Hischier: Conceptualization, Writing – review & editing, Su­ study with activated sludge model No. 3 (ASM3). Water Environ. Res. 91, 865–876.
pervision. Tomas Rydberg: Project administration, Supervision. Zey­ Gujer, W., Henze, M., Mino, T., van Loosdrecht, M., 1999. Activated sludge model No. 3.
Water Sci. Technol. 39, 183–193.
nep Cetecioglu: Conceptualization, Resources, Project administration, Guinée, J.B., Gorrée, M., Heijungs, R., Huppes, G., Kleijn, R., de Koning, A., van Oers, L.,
Funding acquisition. Wegener Sleeswijk, A., Suh, S., Udo de Haes, H.A., de Bruijn, J.A., van Duin, R.,
Huijbregts, M.A.J., 2002. Handbook on Life Cycle Assessment: Operational Guide to
the ISO Standards. In: Series: Eco-Efficiency in Industry and Science. Kluwer
Declaration of competing interest Academic Publishers, Dordrecht.
Hetherington, A.C., Borrion, A.L., Griffiths, O.G., McManus, M.C., 2014. Use of LCA as a
development tool within early research: challenges and issues across different
The authors declare that they have no known competing financial sectors. Int. J. Life Cycle Assess. 19 (1), 130–143.
interests or personal relationships that could have appeared to influence International Organization for Standardization (ISO), 2006a. International Organization
for Standardization (ISO) (2006a) Environmental Management - Life Cycle
the work reported in this paper.
Assessment - Principles and Framework. ISO 14040:2006, second ed. 2006-06,
Geneva.
Acknowledgement International Organization for Standardization (ISO), 2006b. International Organization
for Standardization (ISO) (2006b) Environmental Management - Life Cycle
Assessment - Require-Ments and Guidelines. ISO 14044:2006; First edition 2006-07-
The authors would like to thank; Sofia Andersson and Andreas 01, Geneva.
Carlsson from Stockholm Vatten och Avfall for providing data for Hen­ Manchala, K.R., Sun, Y., Zhang, D., Wang, Z.-W., 2017. Chapter two - anaerobic digestion
modelling. Adv. Bioenergy 2, 69–141.
riksdal WWTP and answering continuous questions without hesitation,
Momoh, O.L.Y., Anyata, B.U., Saroj, D.P., 2013. Development of simplified anaerobic
Mayumi Narongin from IVL for sharing all requested information for the digestion models (SADM’s) for studying anaerobic biodegradability and kinetics of
pilot plant at Hammarby Sjöstadsverk and finally Christian Baresel and complex biomass. Biochem. Eng. J. 79, 84–93.
Moni, S.M., Mahmud, R., High, K., Carbajales-Dale, M., 2020. Life cycle assessment of
Klara Westling from IVL for kindly helping to make arrangements for
emerging technologies: a review. J. Clean. Prod. 24, 52–63.
reaching the people and data needed. This study is funded by the Morgan-Sagastume, F., Heimersson, S., Laera, G., Werker, A., Svanström, M., 2016.
Swedish Research Council Formas as part of a project entitled “Meth­ Techno-environmental assessment of integrating polyhydroxyalkanoate (PHA)
odology for environmental sustainability assessment in the early design production with services of municipal wastewater treatment. J. Clean. Prod. 137,
1368–1383.
stage of a resource recovery system” (MENToR) 2018-01297. Ögmundarson, Ó., Herrgård, M.J., Forster, J., Hauschild, M.Z., Fantke, P., 2020.
Addressing environmental sustainability of biochemicals. Nat. Sustain. 3, 167–174.
Parvatker, A.G., Eckelman, M.J., 2019. Comparative evaluation of chemical life cycle
Appendix A. Supplementary data
inventory generation methods and implications for life cycle assessment results. ACS
Sustain. Chem. Eng. 7, 350–367.
Supplementary data to this article can be found online at https://doi. Piccinno, F., Hischier, R., Seeger, S., Som, C., 2016. From laboratory to industrial scale: a
org/10.1016/j.jclepro.2022.131720. scale-up framework for chemical processes in life cycle assessment studies. J. Clean.
Prod. 135, 1085–1097.
Piccinno, F., Hischier, R., Seeger, S., Som, C., 2018. Predicting the environmental impact
References of a future nanocellulose production at industrial scale: application of the LCA scale-
up framework. J. Clean. Prod. 174, 283–295.
Rebitzer, G., Ekvall, T., Frischknecht, R., Hunkeler, D., Norris, G., Rydberg, T.,
Aldin, S., Nakhla, G., Ray, M.B., 2011. Modeling the influence of particulate protein size
Schmidt, W.-P., Suh, S., Weidema, B., Pennington, D., 2004. Life cycle assessment
on hydrolysis in anaerobic digestion. Ind. Eng. Chem. Res. 50, 10843–10849.
Part 1: framework, goal and scope definition, inventory analysis, and application.
Andersson, S.L., Westling, K., Karlsson, J., Narongin, M., Andersson, S., Persson, G.,
Environ. Int. 30, 701–720.
2020. Longterm Trials with Membrane Bioreactor for Enhanced Wastewater
SGC, 2012. Basic data on biogas. Serviceförvaltningen i Lunds kommun, Lund.
Treatment Coupled with Compact Sludge Treatment, Pilot Henriksdal 2040, Results
Shibasaki, M., 2009. Methode zur Prognose der Okobilanz einer Großanlage aufBasis
from 2018. IVL, Stockholm, Sweden. Report 2388.
einer Pilotanlage in der Verfahrenstechnik: ein Beitrag zur
Arudchelvam, Y., Perinpanayagam, M., Nirmalakhandan, N., 2010. Predicting VFA
GanzheitlichenBilanzierung [PhD thesis]. Univ Stuttgart, Stuttgart (DE) [German].
formation by dark fermentation of particulate substrates. Bioresour. Technol. 101,
Simon, B., Bachtin, K., Kilic, A., Amor, B., Weil, M., 2016. Proposal of a framework for
7492–7499.
scale-up life cycle inventory: a case of nanofibers for lithium iron phosphate cathode
Atasoy, M., Eyice, Ö., Cetecioglu, Z., 2020a. Volatile fatty acid production from semi-
applications. Integrated Environ. Assess. Manag. 12 (3), 465–477.
synthetic milk processing wastewater under alkali pH: the pearls and pitfalls of
Srimachai, T., Nuithitikul, K., O-thong, S., Kongjan, P., Panpong, K., 2015. Optimization
microbial culture. Bioresour. Technol. 297, 122415.
and kinetic modeling of ethanol production from oil palm frond juice in batch
Atasoy, M., Eyice, Ö., Cetecioglu, Z., 2020b. A comprehensive study of volatile fatty acids
fermentation. Energy Proc. 79, 111–118.
production from the batch reactor to anaerobic sequencing batch reactor by using
Sun, H., Yang, Z., Zhao, Q., Kurbonova, M., Zhang, R., Liu, G., Wang, W., 2021.
cheese processing wastewater. Bioresour. Technol. 311, 123529.
Modification and extension of anaerobic digestion model No.1 (ADM1) for syngas
Batstone, D.J., Keller, J., Angelidaki, I., Kalyuzhnyi, S.V., Pavlostathis, S.G., Rozzi, A.,
biomethanation simulation: from lab-scale to pilot-scale. Chem. Eng. J. 403, 126177.
Sanders, W.T., Siegrist, H., Vavilin, V.A., 2002. Anaerobic digestion model No 1
Surendhiran, D., Vijay, M., Sivaprakash, B., Sirajunnisa, A., 2015. Kinetic modeling of
(ADM1). Water Sci. Technol. 45, 65–73.
microalgal growth and lipid synthesis for biodiesel production. 3 Biotech 5,
Chakravarty, P., Mhaisalkar, V., Chakrabarti, T., 2010. Study on poly-hydroxyalkanoate
663–669.
(PHA) production in pilot scale continuous mode wastewater treatment system.
Thinkstep, A.G., 2020. Leinfelden-Ecterdingen GaBi Software-System and Database for
Bioresour. Technol. 101, 2896–2899.
Life Cycle Engineering 1992-2020 ©.
Clavreul, J., Guyonnet, D., Christensen, T.H., 2012. Quantifying uncertainty in LCA-
Tsalidis, G.A., Korevaar, G., 2022. Environmental assessments of scales: the effect of ex-
modelling of waste management systems. Waste Manag. 32, 2482–2495.
ante and ex-post data on life cycle assessment of wood torrefaction. Resour. Conserv.
Elginoz, N., Atasoy, M., Finnveden, G., Cetecioglu, Z., 2020a. Ex-ante life cycle
Recycl. 176, 105906.
assessment application of volatile fatty acid production from dairy wastewater.
Tsoy, N., Steubig, B., van der Giesen, C., Guinee, J., 2020. In: Upscaling methods used in
J. Clean. Prod. 269, 122267.
ex ante life cycle assessment of emerging technologies: a review, 25, pp. 1680–1692.
Elginoz, N., Khatami, K., Owusu-Agyeman, I., Cetecioglu, Z., 2020b. Life cycle
Wernet, G., Bauer, C., Steubing, B., Reinhard, J., Moreno-Ruiz, E., Weidema, B., 2016.
assessment of an innovative food waste management system. Front. Sustain. Food
The ecoinvent database version 3 (part I): overview and methodology. Int. J. Life
Syst. 4. Article 23.
Cycle Assess. 21 (9), 1218–1230.
Fernandez-Dacosta, C., Wassenaar, P.N.H., Dencic, I., Zijp, M.C., Morao, A., Heugens, E.
Yao, Y., Masanet, E., 2018. Life-cycle modelling framework for generating energy and
H.W., Shen, L., 2019. Can we assess innovative bio-based chemicals in their early
greenhouse gas emissions inventory of emerging technologies in the chemical
development stage? A comparison between early-stage and life cycle assessments.
industry. J. Clean. Prod. 172, 768–777.
J. Clean. Prod. 230, 137–149.

12

You might also like