Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

Int J Life Cycle Assess (2016) 21:1218–1230

DOI 10.1007/s11367-016-1087-8

THE ECOINVENT DATABASE V3

The ecoinvent database version 3 (part I): overview


and methodology
Gregor Wernet 1 & Christian Bauer 2 & Bernhard Steubing 3,4 & Jürgen Reinhard 5 &
Emilia Moreno-Ruiz 1 & Bo Weidema 6

Received: 17 June 2015 / Accepted: 6 March 2016 / Published online: 21 April 2016
# Springer-Verlag Berlin Heidelberg 2016

Abstract modeling choices, or system models, to the same raw data


Purpose Good background data are an important requirement with little effort. This includes one system model for
in LCA. Practitioners generally make use of LCI databases for Consequential LCA. Flow properties were added to all ex-
such data, and the ecoinvent database is the largest transparent changes in the database, giving more information on the in-
unit-process LCI database worldwide. Since its first release in ventory and allowing a fast calculation of mass and other
2003, it has been continuously updated, and version 3 was balances. With version 3.1, the database is generally water-
published in 2013. The release of version 3 introduced several balanced, and water use and consumption can be determined.
significant methodological and technological improvements, Consumption mixes called market datasets were consistently
besides a large number of new and updated datasets. The aim added to the database, and global background data was added,
was to expand the content of the database, set the foundation often as an extrapolation from regional data.
for a truly global database, support regionalized LCIA, offer Results and discussion In combination with hundreds of new
multiple system models, allow for easier integration of data unit processes from regions outside Europe, these changes
from different regions, and reduce maintenance efforts. This lead to an improved modeling of global supply chains, and a
article describes the methodological developments. more realistic distribution of impacts in regionalized LCIA.
Methods Modeling choices and raw data were separated in The new mixes also facilitate further regionalization due to the
version 3, which enables the application of different sets of availability of background data for all regions.
Conclusions With version 3, the ecoinvent database substan-
tially expands the goals and scopes of LCA studies it can
Responsible editor: Rainer Zah
support. The new system models allow new, different studies
to be performed. Global supply chains and market datasets
* Gregor Wernet
wernet@ecoinvent.org significantly increase the relevance of the database outside
of Europe, and regionalized LCA is supported by the data.
Datasets are more transparent, include more information,
1
Ecoinvent, Technoparkstrasse 1, 8005 Zurich, Switzerland and support, e.g., water balances. The developments also sup-
2 port easier collaboration with other database initiatives, as
Technology Assessment Group, Paul Scherrer Institute,
5232 Villigen, Switzerland demonstrated by a first successful collaboration with a data
3 project in Québec. Version 3 has set the foundation for
Institute of Environmental Engineering, Swiss Federal Institute of
Technology (ETH) Zürich, Schafmattstr. 6, 8093 Zürich, Switzerland expanding ecoinvent from a mostly regional into a truly global
4 database and offers many new insights beyond the thousands
Institute of Environmental Sciences (CML), Leiden University,
2300, RA Leiden, The Netherlands of new and updated datasets it also introduced.
5
Department of Informatics, Informatics and Sustainability Research
Group, University of Zurich, Binzmuehlestrasse 14,
8050 Zurich, Switzerland Keywords Ecoinvent version 3 . Life cycle assessment
6
Department of Development and Planning, Aalborg University, (LCA) . Life cycle inventory (LCI) database .
Skibbrogade 5, 9000 Aalborg, Denmark Parametrization . Regionalization . System model

Content courtesy of Springer Nature, terms of use apply. Rights reserved.


Int J Life Cycle Assess (2016) 21:1218–1230 1219

1 Introduction 2014; Masanet et al. 2013; Scharlemann and Laurance 2008;


Simons and Bauer 2012; Sternberg and Bardow 2015; Treyer
Life cycle assessment (LCA) is a very data-intensive method- et al. 2014; Turconi et al. 2014; Volkart et al. 2013; von der
ology. A typical life cycle of a product or service covers thou- Assen et al. 2013; Wernet et al. 2010, 2012, 2011; Yue et al.
sands of human activities, each of which needs to be under- 2014), setting precedence in the environmental assessment of
stood and documented in terms of environmentally relevant different product groups and economic sectors. Additionally, first
material and energy flows. This information can usually not policy instruments were designed based on LCA and referred to
be gathered within each specific LCA project due to the high ecoinvent as standard background data (Swiss Confederation
cost of primary data collection. It is therefore common prac- 2014). Increasingly, it was also used beyond its original intended
tice to focus data collection efforts on selected activities that scope of Switzerland and Europe and has become a reference
reflect the immediate space for action—these activities are database for studies and LCA work in North and South
together called the foreground system—and to use generic America, Asia, Africa, and Australia (Hou et al. 2015).
data from Life Cycle Inventory (LCI) databases to model the Due to this increasing international and global use, it was de-
remaining activities, called the background system (Bourgault cided to further develop the database into the direction of a glob-
et al. 2012; Tillman 2000). We estimate that the background ally valid database instead of simply maintaining its status as a
system usually covers up to 99 % of the unit processes in the Swiss database. The increasingly international use of the database
product system; we believe that only in rare cases do the conflicted with some decidedly Swiss supply chains in the data-
number of unit processes modeled explicitly in the foreground base. Users were found to commonly apply Swiss or European
system exceed 5 %, for example in large-scale studies of the datasets to other regions due to lack of geographically appropriate
energy sector (Hertwich et al. 2014; Laurent and Espinosa data. This led to inconsistencies on several levels: first of all,
2015; Treyer and Bauer 2013; Treyer et al. 2014). Bearing conditions and supply chains for important factors, such as elec-
this in mind, background LCI databases can be considered tricity mixes, vary with geography for many activities and, sec-
the backbone of any LCA study. They provide the dominant ondly, with the increasing evidence of the importance of region-
share of the building blocks required for any LCA: aggregated alized impact assessment (Mutel et al. 2013, 2012; Mutel and
and/or disaggregated unit process data. Therefore, the avail- Hellweg 2009; Pfister et al. 2009; Potting and Hauschild 1997,
able quantity and quality of unit process data provided by LCI 2006; Wegener Sleeswijk and Heijungs 2010), these localized
databases are of utmost importance. supply chains would not be appropriately modeled with Swiss
The ecoinvent LCI database was first published in 2003 with or European data.
its version 1 (Frischknecht et al. 2005). It represented the results of At the same time, the continued growth of the database and its
many person-years of work and built on top of several existing user base challenged some aspects of the technical foundation of
LCI data collection efforts in Switzerland of the 1990s. The com- the database in terms of the required maintenance effort in the long
mon goal of the ecoinvent project partners was to facilitate the term. First of all, the management of the database and updating
application of LCA through the availability of consistent generic required a significant amount of manual work, which was man-
background LCI data. Lack of background data can severely im- ageable during the early days of the database, but grew to be pro-
pact the quality of LCA studies, and there was a general interest hibitive as the database expanded in size and scope. Secondly, it
among both research institutions and Swiss government offices to wasbecomingmoreandmoreobviousthat,asLCAdevelopedasa
applyLCAmoreconsistently.Theprojectscopewasintentionally tool, its use in policy and standards was increasing, and such ap-
limited to Life Cycle Inventories. While life cycle impact assess- plications often require specific modeling rules to be applied. The
ment (LCIA) methods were applied to the cumulative inventory approach of having all modeling choices fixed within the datasets
results to support users, LCIA method development was and is therefore restricted the usefulness of the data to a limited applica-
still outside of the scope of ecoinvent. While developed for appli- tion range. There was a need for a system that harmonizes some set
cation in Switzerland and with Swiss supply chains in mind, ver- modeling rules, allows a choice between different sets of rules, and
sion 1 was also frequently applied in studies with a primary scope applies them to the whole database consistently. Such an approach
outside of Switzerland. would also simplify some aspects of data entry, as data providers
With the release of version 2 in 2007, the data basis was sig- would no longer have to apply system modeling rules themselves.
nificantly expanded and enriched, adding several new economic Finally, several options for improvements in functionality and data
sectors such as biofuels, consumer electronics, and others to the coverage had been identified and a new technological foundation,
database. In addition, the geographical coverage increased, with including an update of the established ecospold data format
more data covering areas outside of Europe. Sectors such as elec- (Meinshausen et al. 2014) was required for further development
tronics were covered with data of production in Asia. Version 2 of the database. Several of these features were developed specifi-
has been used over the years in many relevant LCA studies (Amor cally to facilitate the collaboration and integration of international
et al. 2014; Arvesen and Hertwich 2015; Bauer et al. 2015; data and to facilitate cooperation with a broader range of data
Bouman et al. 2015; Henriksson et al. 2015; Hertwich et al. providers.

Content courtesy of Springer Nature, terms of use apply. Rights reserved.


1220 Int J Life Cycle Assess (2016) 21:1218–1230

Therefore, the plan for ecoinvent version 3 was developed, 2.2 Global background data
with the ambition to provide a technological and methodological
foundation for an LCI database with modern capabilities, flexi- The decision to support regionalized assessment with version
bility and room for future developments as well as the structure 3 brought with it the consequence that impacts needed to be
necessarytogrowintoatrulyglobaldatabase.Thispaperprovides adequately distributed globally. In versions 1 and 2, the data-
an overview of the most important new developments introduced base was focusing on Swiss and European production, while
with version 3, focusing on new dataset categories and the differ- datasets from other regions were often lacking (Steubing et al.
ent system models used for linking the unit processes. 2016). Therefore, users commonly applied the European
datasets also to other regions. In these cases, local conditions
(e.g., in terms of technologies used) were often inappropriate-
2 Developments ly reflected by the LCI data used and—in this context more
important—direct and upstream impacts were thus caused in
2.1 From multi-output activities to single product datasets inappropriate geographies. To introduce a more appropriate
representation of global supply chains, to support regionalized
One major change introduced with version 3 is that the data- impact assessment, and to support a more international user
base no longer stores data and modeling choices intertwined. base, global coverage for activities was consistently intro-
Data is stored first as raw, Bundefined^ datasets, which contain duced in version 3. There exists a global dataset for all activ-
unit process data without influences of, e.g., allocation ities in version 3, and in addition, there may be one or more
choices. This version of the database is best suited for under- local (non-global) datasets representing the production activ-
standing a process or activity in its original isolated state since ities for specific regions. Geographies in version 3 are de-
it shows all exchanges, i.e., inputs, co-products, and emissions scribed in the KML format (OGC 2014) to support interaction
as originally collected, in one overview. However, calculation with other machine-readable spatial data, but are also de-
of aggregated LCIs and LCIA results requires the application scribed using country names and codes to simplify human
of system modeling, i.e., linking and allocating the unit pro- interactions with the geography data.
cesses according to a distinct set of rules. Version 3 offers the All activities include data on annual production volume,
choice of more than one system model, and the application of and in the system model databases, the global datasets are
these models results then in several system model databases, used to approximate the activities in regions not covered by
each in the form of fully linked unit processes. Aggregated local data. For this, the global dataset is used as the basis for a
LCI and LCIA results are also available for each system model BRest-of-the-World^ (RoW) dataset applicable to the areas not
database. Thus, version 3 is available as a primary database covered by local data. As the local geographical coverage
with an Bundefined^ system model and, currently, three sys- differs from one activity to another, RoW datasets do not
tem model databases, which have different LCIA results due always describe the same geography. Using the KML geogra-
to the different modeling choices, despite being based on the phies, the geography of each RoW dataset can be described
same unlinked, raw unit process data (Fig. 1). individually. This geography is then also used to determine the
suppliers for the inputs of the dataset. In versions 3.0 and 3.1,
RoW datasets were generated by using a production-volume-
System Model X System Model Y
weighted average of all local datasets and calculating a differ-
LCIA results ence dataset to the global average available in the global
dataset. The RoW dataset was then created as a difference
Applicaon of dataset so that the weighted sum of all datasets resulted in
LCIA factors the values of the global average. However, the chosen imple-
mentation of this procedure demanded significant extra effort
Aggregated LCI data
from data providers, and results were generally not affected
Calculaon much compared to a simpler approach, which has now been
implemented with version 3.2. In this new approach, the RoW
Single output unit processes datasets are identical in flow values to the global dataset, albeit
still with supply chains adapted to its geography. A sensitivity
Model X Model Y analysis of the change in approach revealed generally minimal
changes in results due to the transition.
Physical process descripons Undefined A RoW dataset is only generated if the sum of all local
(mul-output) system production volumes is lower than the global production vol-
Fig. 1 Flowchart of generating system model results out of the primary ume. Therefore, a RoW dataset is not created if: first, all re-
raw unlinked data gions of the world are adequately covered with regional

Content courtesy of Springer Nature, terms of use apply. Rights reserved.


Int J Life Cycle Assess (2016) 21:1218–1230 1221

datasets (i.e., full geographical coverage), or second, an activ- provider will leave the dataset unlinked, in which case the
ity takes place in only parts of the world (i.e., the complete linking is carried out by the database service layer during the
global production is localized in one region, which is covered linking of the unlinked activities into system model databases.
by a local dataset). In both cases, the sum of the local produc- The input is then linked to the geographically appropriate
tion volumes is equal to the global production, and no RoW market or markets. In this Bindirect^ linking, a dataset can at
dataset is generated. first always rely on the existence of a global market providing
The inclusion of consistently available, global datasets fa- a global supply chain. Should more local markets be present,
cilitates the expansion of the database into different regions. or be added to the database for a future calculation, the same
The global datasets mean that the existence of background dataset would then link to this more specific market, or even to
supply chains can always be relied on, no matter for which a set of markets within its region, without requiring any mod-
region a dataset is created. Datasets can therefore be created ification of the dataset itself. In this way, data providers can
without requiring local supply chain data to also be available ensure that their supply chains will be updated to the most
immediately. Such supply chains may still become available appropriate data available in the database in the future, thus
later and will then be utilized through the linking algorithm reducing the updating and maintenance work between data-
(see below). base versions.
In a similar fashion, market datasets themselves are only
2.3 Market datasets completed during the linking of the unlinked activities into
system models. One reason is that different system models
When a product is produced by more than one activity (i.e., by have different rules for the composition of consumption mixes
different production technologies or by one technology, but in (see below). Another reason is that, in this approach, con-
more than one region each covered by an individual dataset), it sumption mixes are updated whenever new producers are
is important in activities consuming the product to know added to the database. Markets are completed using annual
which production contributes how much to the supply chain production-volume-weighted averages of the relevant pro-
of the consumer. For this reason, versions 1 and 2 of ecoinvent ducers and importers in the geography covered by the market.
contained a number of production, supply, and consumption The annual production volumes therefore determine the shares
mixes, but only for a selected number of products. Version 3 of different activities providing the same product to a certain
introduces independent product names, which are no longer market. Production output that is already used specifically for
tied to specific process names. The aim is to more easily iden- a particular consumer and for export does not contribute to
tify activities with the same product as output. With version 3 markets (Fig. 2). The approach of completing market datasets
also aiming to introduce more geographical coverage, an in- during linking was chosen on the basis that by default, a new
creased need for geography-specific consumption mix production activity should contribute towards the average
datasets was foreseen. For this reason, version 3 introduces consumption mix in a region. The impacts of new datasets
market datasets representing the consumption mixes for a giv- on existing market datasets therefore need to be observed
en region and product (so that consumption is supplied by the and reviewed.
production within the given geography plus imports minus re- With version 3 and the introduction of market datasets, the
export). Market datasets also contain additional exchanges standard estimations for transport within the database (default
related to the transport of the good from the producer to the values used to estimate transport distances when no specific
consumer, as well as losses of goods and emissions during data were available) were also updated and expanded to cover
transport (e.g., in the case of perishable fruit and vegetables, more forms of transport consistently and with sector-specific
leaking natural gas pipelines, or transmission and distribution data based on global transport statistics (ecoinvent 2014).
losses for electricity). For most other datasets, the market With these statistical data, market datasets describe the aver-
datasets only reflect a generic model for supply and transport age transport modes and distances for each product, including
of the products. A market dataset can be used as an input by shipping, road, and rail transport as well as, if appropriate, air
consuming activities that are lacking more specific supply transport and other, specialized transport methods, e.g.,
chain information. However, while market datasets represent pipelines.
a useful and convenient option in the building of a consistent Together with RoW datasets, market datasets represent the
network structure across the ecoinvent database, LCA practi- central building blocks for a more realistic representation of
tioners should consider using inputs directly from production global supply chains. Market datasets and the consistent con-
activities whenever more specific knowledge is available re- sumption mixes represent a useful tool both for users of the
garding product sourcing, losses and transport distances. database and for the operation of the ecoinvent database itself.
In ecoinvent, a data provider who has sufficient specific However, it should be noted that in foreground systems, built
information may link an input directly to specific activities on top of ecoinvent data, it is not necessary to use market
or markets. Without such specific information, the data datasets or a structure based on consumption mixes.

Content courtesy of Springer Nature, terms of use apply. Rights reserved.


1222 Int J Life Cycle Assess (2016) 21:1218–1230

Fig. 2 An illustration of linking via market datasets using animal feed are supplied by their local electricity markets, the RoW dataset by all
production as a theoretical example (** PV stands for production other electricity markets using a production-volume-weighted average
volume). At the bottom of the graph, electricity markets supply for their contributions. All three soybean datasets supply the global
electricity mixes without the need for choosing between individual market for soybeans. However, the US production does not supply all
producers such as coal or hydropower. Soybeans, one of the of its production to the market, as some of it is specified as directly linked
components of animal feed, are produced in two local datasets (USA by the US production of animal feed. This direct consumption does not
and Brazil). A RoW dataset covers the remaining global production of contribute to the global market share of US soybeans, as the market is a
soybeans that is not covered by the two local datasets. The local datasets consumption mix, not a production mix

2.4 System models in version 3 classified into either Ballocatable byproducts,^ Brecyclable
materials,^ or Bwastes.^ The classification is based on expert
Versions 1 and 2 of ecoinvent applied one system mod- judgment of an exchange’s value, use potential, and predicted
el to the database, which largely followed a recycled fate. Depending on this classification, byproducts (i.e., prod-
content or Bcutoff^ approach (Frischknecht et al. ucts that are not the reference flow) in multi-functional activ-
2005). Modeling choices were directly integrated into ities are handled differently (Figs. 3 and 4): Allocatable
the datasets by the authors of datasets. Therefore, the byproducts are allocated against other products based on the
consistent application of this modeling choice needed allocation factors specified for the specific system model, e.g.,
to be reviewed in each dataset individually, and some on physical relations, exergy, price, or mass.
inconsistencies remained in the database despite the re- Recyclable materials are cut off from the producing
view efforts. With the option of separating inventory product system, i.e., they are removed burden-free from
data and modeling choices, version 3 allows for more the producing activity, and no impacts or benefits are al-
than one system model and is indeed so far currently located to them. They are then available burden-free to
available in three system models. The modeling rules any consuming activities, e.g., recycling activities.
are defined once and then applied consistently to all Secondary materials thus bear only the burden of the
datasets. The application of the model therefore only recycling process and no burdens from the primary pro-
needs to be reviewed once, and inconsistencies are duction of the material. To introduce the cutoff, the system
much less likely to occur. The two system models ap- model introduces special datasets which are named
plying allocation rules are intended for use as back- Bexchange name, recycled content cutoff^ and which con-
ground data in attributional LCA, and the consequential tain no flows besides the reference product of the ex-
system model is intended for application in consequen- change itself. These are used to show that the cutoff is
tial LCA (Earles and Halog 2011; Ekvall and Weidema happening in the dataset and to maintain mass balances.
2004; Weidema et al. 2009; Zamagni et al. 2012). The cutoff is taking effect through the lack of inputs or
outputs to these special datasets (Fig. 4).
2.4.1 Allocation and cutoff by classification Wastes on the other hand are disposed of in a treatment
activity, and the burden of the disposal is completely at-
This model follows the same recycled content approach as tributed to the activity producing the waste. This is
versions 1 and 2. At the database level, all intermediate ex- achieved by first listing the produced wastes as a negative
changes (i.e., exchanges within the technosphere) are process input in the producing activity using the logic that

Content courtesy of Springer Nature, terms of use apply. Rights reserved.


Int J Life Cycle Assess (2016) 21:1218–1230 1223

Fig. 3 The cutoff process in the Bcutoff by classification^ system model treatment fully contribute to the burdens of the reference product of the
for waste treatment byproducts. The treatment byproduct is available waste-producing activity
burden-free in the database, while the inputs and emissions of the waste

a positive output is equal to a negative input. Waste treat- 2.4.2 Allocation at the point of substitution
ment activities have the reference product of removing
wastes, which is recorded as a negative product. The pro- The allocation at the point of substitution (APOS) system
cesses are therefore linked correctly: The burdens of waste model is based on the premise that allocation of materials that
treatment increase the burdens of the producing activity require further treatment before they can be of immediate val-
(Fig. 3). This approach is identical in results to the ap- ue is fundamentally challenging. Neither the price, mass,
proach of versions 1 and 2, where disposal of wastes was exergy, nor any other physical property is a useful allocation
an input to datasets. Listing the wastes as negative inputs parameter, e.g., waste glass. The APOS model is therefore
maintains the mass balance of the dataset, an important performing an expansion of the allocation system to include
validation tool and helpful in understanding the physical all treatment processes required for any byproducts (Fig. 5),
reality of the activity. As the burdens of waste treatment be they wastes or recyclable, and in fact, the system model
are to be fully attributed to the producing activity of the makes no such distinction (please note that this is unlike the
waste, any non-waste byproducts of waste treatments process some refer to as Bsystem expansion^ that is used in
(e.g., heat from waste incineration) are available burden- consequential LCA, a process we refer to as Bsubstitution^ in
free in this system model. this article). All products, which are not produced as a positive

Fig. 4 The upper half shows how Recyclable Materials are cut off in the difference is whether disposal of the material brings with it additional
Bcutoff by classification^ system model in the producing activity (labelled burdens, as in Fig. 3, or no additional burdens, as in this case. In the
BActivity^) through the use of the special, empty cutoff dataset (dotted bottom half, it can be seen how consumers of recyclable materials
arrows indicate additional inputs and emissions that are not relevant to the receive the material burden-free at the site of the producer. Therefore,
allocation process, as in Fig. 3). The producing activity lists the cut transport and collection are already contributing to the supply chain of
byproduct as a negative input to make the cutoff process explicit and to the secondary material. They are then consumed in other activities which
respect mass balances. Within the scope of the producing activity, Wastes describe the recycling process and produce secondary goods or products
and Recyclable Materials are therefore treated the same way. The that are partially based on recyclable materials

Content courtesy of Springer Nature, terms of use apply. Rights reserved.


1224 Int J Life Cycle Assess (2016) 21:1218–1230

Fig. 5 Overview of the expanded


allocation system in the APOS
model. While the cutoff model
allocates the production process
only to its allocatable byproducts,
the APOS model expands the
allocation system to include both
the production and the treatment
processes. Impacts of both are
combined and allocated to both
the product A and the treatment
byproduct E

reference product by any other activity, are defined as mate- required in this system model. Allocation on wastes and waste
rials for treatment (MFT). The MFTs are largely equal to the products is often a controversial topic in LCA studies
products classified as recyclable materials and wastes in the (Cherubini et al. 2009; Finnveden 1999; Heijungs and
cutoff model. Allocation systems are expanded to include all Guinée 2007), so avoiding this issue is a benefit of the
treatment steps necessary for conversion of the MFTs into APOS model. To allow an analysis of the unit process contri-
products that are not MFTs. The expanded system is then butions, the byproducts are allocated in the main producing
allocated to all its co-products based on the allocation factors activity, and the treatment functionality of the treatment pro-
in the database. Thus, the exchanges of both production and cesses is part of the supply chain with all the impacts of the
treatments are combined and then allocated to both the pro- treatment (Fig. 6). The allocation then results in datasets for
duction products and the treatment byproducts (Fig. 5). treatment byproducts that are specific to the activity producing
As the allocation system is expanded to cover the whole the MFT. To simplify the database and the use of such prod-
treatment process, any allocation within the treatment pro- ucts, all instances of a specific byproduct from a specific treat-
cess(es) can be avoided in this system model. Therefore, the ment are aggregated into one dataset. For example, the
distinction between wastes and recyclable materials is not datasets for electricity from incineration of municipal solid

Fig. 6 Modeling of joint


production activities in the APOS
system model (adapted from
Weidema et al. 2013). The upper
half shows the original process
structure. To allocate all 3
products X, Y, and Z against each
other, they need to be grouped in
one place at the end of the
expanded product system. The
bottom half shows how the
process structure is rearranged
after MFT byproducts are
considered as negative inputs and
how the non-MFT recycling
byproduct is moved to the joint
production activity to allow
allocation within a single activity

Content courtesy of Springer Nature, terms of use apply. Rights reserved.


Int J Life Cycle Assess (2016) 21:1218–1230 1225

waste is listed in the APOS model as one dataset. However, as Conditions can vary over time, and this may become relevant
explained above, each instance of municipal solid waste especially if data are directly used in foreground modeling.
(MSW) output results in an individual dataset for electricity, The current implementation of consequential modeling is
e.g., one from treatment of MSW from farming, one from rather basic since many datasets only exist as averages, and
treatment of MSW from machinery production, or one from constraints are determined in a basic way but the implemen-
treatment of MSW from other waste treatments. All of these tation is sufficient to fulfill the purpose of a background data-
datasets are aggregated into one production-volume-weighted base in many consequential studies, as it offers the core con-
average of all such treatment datasets. This dataset then rep- cepts of a consequential model, i.e., use of substitution to
resents the production of electricity using the average waste avoid allocation completely and use of marginal suppliers
origins within the database. instead of average suppliers, and as individual processes can
The two allocation models only differ in the approach to- be adapted if needed to include more specific data, e.g., from
wards allocating recycling and waste treatment products. The forecasting studies.
differences between the two system models in terms of LCA When all suppliers to a market are constrained, for example
results are therefore strongest for such products and often when the market product is exclusively produced as a
negligible otherwise (Steubing et al. 2016). byproduct, the consequential model applies the concept of
constrained markets (Fig. 7). An example is sodium hydroxide
2.4.3 Consequential, long-term, small scale (NaOH), which is currently only produced as a byproduct
(NaOH is a byproduct of chlorine production, which drives
The consequential system model follows a substitution-based the demand of their joint production). Activities in the data-
approach and is intended to serve as a background database in base that require NaOH therefore depend on a product that is
consequential LCA studies (Earles and Halog 2011; Ekvall constrained by the demand for chlorine. A marginal consump-
and Weidema 2004; Pehnt et al. 2008; Reinhard and Zah tion activity is then determined, which reduces its consump-
2009; Suh and Yang 2014; Tonini et al. 2012; Weidema tion of NaOH when consumption is increased elsewhere.
et al. 2009; Zamagni et al. 2012). In this system model, sub- Therefore, the production of chlorine as an activity receives
stitution is used to resolve multi-functionality in datasets in- a credit for the production of NaOH corresponding to this
stead of allocation. A reference product for an activity is al- reduction in marginal consumption. In the case of NaOH,
ways burdened with the full impacts of all inputs and emis- the actual credit is given for a reduction of production of
sions, but is credited with benefits for any byproducts it pro- sodium carbonate (which can substitute for NaOH as a neu-
duces that can substitute other productions. Byproducts are tralizing agent), which does not need to be produced due to the
recorded as negative inputs in the datasets and are linked to increased availability of NaOH. Similarly, consumers of
the production of the good or service they are substituting. The NaOH are burdened with the increase in sodium carbonate
negative sign results in an environmental credit if the production resulting as the indirect consequence of their
substituted activity has an environmental burden. As previ- NaOH consumption.
ously stated, the term Bsystem expansion^ is also sometimes
used for this approach; this article uses the term Bsubstitution.^
In the consequential model, only unconstrained suppliers
(those who can respond to changes in demand by adapting
their production) are taken into account.
Byproducts are always constrained in the Consequential
system model. The model defines that, as they are not the
driver for an activity, they cannot react in an unconstrained
way to changes in demand. That means that, in this system
model, market datasets generally only have inputs from activ-
ities where the product is the reference product. Any direct
links to such byproducts are also substituted with the input
from the corresponding unconstrained market.
Furthermore, activities can be constrained due to the tech-
nology level as defined in the dataset: only the most up-to-date
technologies are considered unconstrained. The electricity
sector (i.e., the markets) can be regarded as an example for
the implementation of these constraints (Treyer and Bauer Fig. 7 Example of a BConstrained market^; modeling according to the
2014). Constraints are therefore mostly based on current con- BConsequential, long-term, small scale^ system model (adapted from
ditions and not the result of specific forecasting projects. (Weidema et al. 2013))

Content courtesy of Springer Nature, terms of use apply. Rights reserved.


1226 Int J Life Cycle Assess (2016) 21:1218–1230

2.5 Other developments vegetables were added as well (Stoessel et al. 2012). Version
3.1 introduced, e.g., global data on aluminium production,
Several other changes were introduced with ecoinvent version new road freight transport data, forestry data, and tap water
3. First of all, the data format was updated to ecospold 2 production. Version 3.2 included among other data another
(Meinshausen et al. 2014). The new format improves the de- update and expansion of the electricity and heat sectors, data
scription of exchange data independent of process data. It also on refrigerated transport, updated cement and concrete data,
includes significantly increased options for documentation di- and data on European aluminum production.
rectly in the dataset, such as additional documentation fields
or the option to include images in dataset documentation
fields. In this way, more of the documentation of a dataset
can be included directly with the data. The format also in- 3 Discussion
cludes fields for exchange properties. Each exchange can be
defined and described by a number of properties, e.g., fuels Much of the development work for version 3 aimed to prepare
can have a heating value property. All exchanges in version 3 the database for continued growth in the following years. This
have at least the following properties: wet and dry mass, water resulted in new conceptual designs that are visible in the da-
content, and carbon content (fossil and non-fossil). The con- tabase structure. At the same time, version 3.0 was not re-
sistent availability of these properties also allows a simple leased with the ambition to be a complete database, and the
calculation of an activity’s wet and dry mass balance, its water database remains far from being complete. This is apparent in
balance, and its carbon balance. These balances are an impor- several ways. One is that the global datasets are very often not
tant validation and review tool and are also reported for each reflecting truly global data collection but are often extrapolat-
activity in the online database. With the release of version 3.1, ed from regional data for one or a few regions only. While
the database has achieved a water balance sufficient for de- clearly not the best option, this can be an adequate approach to
tailed and accurate water use and consumption assessments approximate inventories. Extrapolation from single regions, or
(Levova 2013). The new data format also allows for specifi- even single locations, to larger or other geographies has been
cation of parameters and for direct integration of mathematical an established practice within the database since its inception
relations within the dataset. For example, the carbon dioxide and is a common practice in LCA in general. The benefits of
emissions of an incineration can be calculated from a fuel the global datasets are nevertheless apparent in several ways.
input, the carbon content of the fuel, and any alternative car- Firstly, due to the geography-specific linking of activities,
bon fates such as soot or carbon monoxide. The fuel input such datasets have supply chains appropriate for their regions,
itself can be calculated from the required demand for energy be it local, global, or RoW datasets. Secondly, the pedigree
and the heating value of the fuel. In this way, calculated values information for geographical representativeness is adapted in
can be easily reviewed as well as understood and reproduced such datasets, which increases the uncertainties in the dataset
by practitioners; modifications to datasets can be carried out (Ciroth et al. 2013; Muller et al. 2014), thus revealing the
faster, and datasets can be updated more consistently. importance of further improvements in these data. Finally,
Price data was also collected for numerous products to emissions and resource uses are correctly located for region-
improve the allocation basis for many datasets. This has re- alized impact assessment. The main drawback of such extrap-
sulted in a significant improvement in results for many olated datasets is that they can be misinterpreted as actual
datasets (Steubing et al. 2016). While not all products within global data. Good and transparent documentation of the ex-
the database have price data available, most do, and this is trapolation process and the associated limitations is therefore
another useful tool for validation purposes within the data- important. The documentation of datasets includes comments
base. The existing price data is also provided to the users for on the data origin, and the datasets contain an explicit
any other uses they may see in it. BExtrapolations^ field. In addition, the pedigree for geograph-
Version 3.0 also introduced many new datasets and updates ical representativeness communicates the adequacy of data
to many sectors in the database. For example, the electricity sources. As the database is increasingly filled with regional
sector was updated and expanded substantially (Treyer and data, extrapolated global background datasets will decrease in
Bauer 2013, 2014). Emissions in agriculture were likewise relevance.
updated to reflect an improved understanding of emission As described above, the method for generating the RoW
profiles (Nemecek et al. 2014). Road passenger transport datasets out of global datasets was changed for version 3.2.
was completely updated and expanded with modern technol- The reason was that the weighted-average calculation used at
ogies, such as electric mobility and alternate fuel sources (Del first was putting significant strain on data providers for activities
Duce et al. 2014; Simons 2013), and many datasets for various where processes vary between regions. The consequences of
North American regions were added, with a focus on Québec changing to a simpler algorithm in terms of LCA results were
(Lesage and Samson 2013; Suh et al. 2013). Fruits and negligible in most cases, and the new algorithm minimizes

Content courtesy of Springer Nature, terms of use apply. Rights reserved.


Int J Life Cycle Assess (2016) 21:1218–1230 1227

artifacts and allows for an easier understanding of values in different questions, or compare results for sensitivity analyses
RoW datasets. towards modeling choices.
The introduction of consistent market datasets expands on The APOS system model, as a new way of looking at
the already useful mixes present in versions 1 and 2. As the allocation of waste treatment and recycling products, is
ecoinvent database grows, the likelihood increases that new differing from the established cutoff model in several ways.
datasets are not necessarily describing a new product, but The cutoff model uses a simple but fundamental decision to
rather an alternative production route for an already existing separate primary and secondary use stages. The approach
product, or the production in a different region. The creation strongly incentivizes the use of secondary materials. It does
of a distinct product list for version 3 allows users to more not however incentivize waste producers to maximize reuse of
easily understand which activities have the same products. At waste materials, as no benefits are given for any useful
the same time, practitioners are increasingly required to treatment products. APOS offers a different perspective, in
choose between several suppliers of a product. Markets as which waste producers are incentivized to assess recycling
consumption mix datasets can offer valuable support in cases and reuse possibilities due to the partial allocation of impacts
where a product is not explicitly sourced from a specific sup- to useful treatment products. The model can therefore be used
plier but consumed Bfrom the market^, hence the name market in studies where the question of waste disposal method is a
datasets. At the same time, it is critical to understand that topic, or as a counterpoint to the cutoff model in a sensitivity
practitioners should carefully choose between market inputs analysis. As Steubing et al. (2016) show, the two allocation
and direct inputs from producers, as relevant. When using the models generally show only insignificant differences in re-
market datasets, the included transport, losses, and emissions sults, except in the case of waste treatment and recycling
simplify the application of the background data. products. Therefore, it is either relatively unimportant which
Market datasets are also beneficial for the maintenance and model is chosen or it is of added value to assess results with
expansion of the database, as they allow for a simple introduc- both systems and discuss the difference.
tion of new producers into existing supply chains throughout In practical terms, the APOS model is more difficult to
the database, instead of a time-consuming, manual effort to implement in foreground systems than the cutoff model, and
introduce new producers into potentially thousands of con- treatment byproduct datasets are more challenging to analyze
suming datasets. Similarly, practitioners can adapt or modify and adapt by practitioners than in the cutoff model. This is
large sections of their background system with relatively little partially a consequence of the cutoff’s disconnection of prod-
effort by changing market compositions in a local copy of the uct paths—the disconnected systems are easier to assess and
database. modify than the expanded systems of the APOS model.
The combination of global supply chains and consistent Furthermore, the consolidated datasets for treatment products,
mix datasets for centralized updating of supply chains allows while still unit processes, are the result of the merging of
for the integration of regional database projects into the sometimes numerous datasets with diverse supply chains.
ecoinvent framework. One example is the Quebec LCI data- They can therefore reach a level of complexity that makes
base (Lesage and Samson 2013). Due to the structure of the an assessment of the reasons for impacts challenging. The
version 3 system, a database can be built first on the backbone complexity is a consequence of the varied supply chains of
of the global supply chains. This avoids one major problem many waste materials—the origin of, e.g., municipal solid
many new data initiatives face: a critical mass of background waste in an incineration plant is indeed very diverse. At the
data is needed to complete supply chains for any dataset, and a moment, a weakness of the model is the fact that waste
project to build an interconnected database will not yield streams are not complete within the database, which influ-
meaningful results until a core set of background data can be ences the impacts allocated to the treatment products: only
adequately covered (Bourgault et al. 2012). Later additions to waste supply chains recorded in the database are taken into
the database can then substitute the global background data account. Wastes from activities not included in the database
with local suppliers with little effort, as links to a global mar- should have an effect as well but do not. Therefore, the careful
ket can be substituted with local markets if necessary. Such comparison of results between the two allocation models is
collaborations between data networks can increase efficiency recommended when using the APOS model.
and productive use of resources in LCI collection. The consequential model offers a very different perspective
With the possibility to apply different modeling choices to on the database and is intended for use with goals and scopes
the same background database, version 3 introduces a novel of consequential LCA studies. In its current form, it is correct-
set of tools for LCA practitioners: being able to use the same ly reflecting the core consequential principles of substitution
data in different goal and scope situations, the effects of and the use of marginal suppliers. However, the differentiation
modeling choices can be observed much more easily. A first between constrained and unconstrained suppliers is limited
application was carried out in (Steubing et al. 2016). and the approach to determine marginal suppliers of electricity
Practitioners can use the different system models to answer is a basic one. The use of technology levels to determine

Content courtesy of Springer Nature, terms of use apply. Rights reserved.


1228 Int J Life Cycle Assess (2016) 21:1218–1230

marginal producers could be improved upon with, e.g., results geographies or carrying out sensitivity analysis.
of predictive studies regarding future energy use scenarios. It Unfortunately, implementation of the properties and parame-
is recommended that users of the system model carefully as- trization data of version 3 is still missing in most commercial
sess the datasets that are significant for their results and deter- LCA software tools at the moment.
mine whether more detailed information on the predicted sit-
uation within their scope is available to improve the reliability
of results. Nevertheless, the consequential model can 4 Conclusions
serve as a background database in many applications,
and this is the first and at the moment only large LCI The development of ecoinvent version 3 aimed to establish a
database with a consequential perspective. It is hoped robust, flexible data system for the management of a global
that it will form the core of increased research into inventory database. Key features envisioned were the support
and development of consequential LCI. for regionalized LCI data and regionalized LCIA, application
In general, the introduction of multiple system models has of multiple system models and the flexibility to introduce new
increased interest and discussion of system modeling choices models, complex supply chain data, increased transparency in
significantly as practitioners now face a real choice of back- documentation and flow properties, and integration of data
ground data modeling and have to make an informed decision. models in datasets among others. While the development of
There are different opinions on how modeling of systems version 3 was complex and time-consuming, the core goals
should be carried out (EC 2010; ISO 2006a, b), and there were achieved. At the same time, the many changes now need
are sometimes conflicting interpretations of the same set of to demonstrate their benefits in application, and a careful as-
rules (Weidema 2014). It seems generally agreed that the goal sessment of the value versus the cost for some complex fea-
and scope of the study should affect the system model, e.g., tures will be carried out. Version 3 was a revolution in the
the ILCD handbook (EC 2010) acknowledges the need for development of the ecoinvent database, and it will be followed
different system models in different use cases, and the system by a process of gradual evolution and consolidation to capi-
models ecoinvent now provides cover several of the ILCD talize on the capabilities and benefits it offers. With multiple
scenarios. The consistent separation of inventory data from system models, consistent consumption mixes, global supply
modeling choices is beneficial in this context. While only chains, support for regionalized LCIA, support for water
three system models are provided at the moment, the datasets footprinting, improved transport modeling, increased trans-
before application of any system model choices are also avail- parency through parametrization and improved documenta-
able, and it is envisioned that eventually practitioners can cre- tion in the datasets, and more flow information in the form
ate and modify system models within their LCA software of properties, version 3 offers many possibilities on the tech-
tools to adapt their data models to specific goals and scopes nical level that previous versions did not.
for specialized applications. For example, substitution can
also be applied to solve the allocation problem in an Acknowledgments The authors would like to thank Tereza Levova,
Sofia Parada Tur, Thomas Nemecek, Karin Treyer, Carl Vadenbo and
otherwise attributive approach with average suppliers,
especially Roland Hischier for their many contributions towards the de-
and such a system model could be created using the velopment of ecoinvent version 3, as well as Gabor Doka for the original
existing technology and data. design of Fig. 5, which is reused here with permission.
The consistent implementation of flow properties al-
lows for more insight and transparency of the flows in Compliance with ethical standards
an LCI, as well as increased validation options during Conflict of interest The authors declare that they have no conflict of
dataset creation and analysis. Together with the new trans- interest.
parent and systematic review process, these automatized
validations will increase data quality and reduce errors in
LCI data. Mass and content balances, such as water bal- References
ances or carbon balances, can be controlled and are report-
ed for the datasets during, e.g., data entry. In this way, the Amor MB, Gaudreault C, Pineau P-O, Samson R (2014) Implications of
integrating electricity supply dynamics into life cycle assessment: a
database has been water balanced within adequate margins for
case study of renewable distributed generation. Renew Energ 69:
consistent water footprinting (Levova 2013). Properties also 410–419
offer further development possibilities in the future, such as Arvesen A, Hertwich E (2015) More caution is needed when using life
developing dataset models based on input properties or LCIA cycle assessment to determine energy return on investment (EROI).
based in part on flow property data. The new option of param- Energ Policy 76:1–6
Bauer C, Hofer J, Althaus H-J, Del Duce A, Simons A (2015) The envi-
etrization of unit processes can be regarded as powerful tool ronmental performance of current and future passenger vehicles: life
for user-friendly modification of LCI data according to spe- cycle assessment based on a novel scenario analysis framework.
cific conditions, e.g., creation of processes for new Appl Energy

Content courtesy of Springer Nature, terms of use apply. Rights reserved.


Int J Life Cycle Assess (2016) 21:1218–1230 1229

Bouman EA, Ramirez A, Hertwich E (2015) Multiregional environmen- Mutel CL, Hellweg S (2009) Regionalized life cycle assessment: compu-
tal comparison of fossil fuel power generation—assessment of the tational methodology and application to inventory databases.
contribution of fugitive emissions from conventional and unconven- Environ Sci Technol 43:5797–5803
tional fossil resources. Int J Greenh Gas Con Mutel C, Pfister S, Hellweg S (2012) GIS-based regionalized life cycle
Bourgault G, Lesage P, Samson R (2012) Systematic disaggregation: a assessment: how big is small enough? Methodology and case study
hybrid LCI computation algorithm enhancing interpretation phase in of electricity generation. Environ Sci Technol 46:1096–1103
LCA. Int J Life Cycle Assess 17:774–786 Mutel C, de Baan L, Hellweg S (2013) Two-step sensitivity testing of
Cherubini F, Bargigli S, Ulgiati S (2009) Life cycle assessment (LCA) of parametrized and regionalized life cycle assessments: methodology
waste management strategies: landfilling, sorting plant and inciner- and case study. Environ Sci Technol 47:5660–5667
ation. Energy 34:2116–2123 Nemecek T, Schnetzer J, Reinhard J (2014) Updated and harmonised
Ciroth A, Muller S, Weidema B, Lesage P (2013) Empirically based greenhouse gas emissions for crop inventories. Int J Life Cycle
uncertainty factors for the pedigree matrix in ecoinvent. Int J Life Assess. doi:10.1007/s11367-014-0712-7
Cycle Assess. doi:10.1007/s11367-013-0670-5 OGC (2014) KML—Keyhole Markup Language version 2.2. Open
Del Duce A, Gauch M, Althaus H-J (2014) Electric passenger car trans- Geospatial Consortium, Wayland
port and passenger car life cycle inventories in ecoinvent version 3. Pehnt M, Oeser M, Swider DJ (2008) Consequential environmental sys-
Int J Life Cycle Assess. doi:10.1007/s11367-014-0792-4 tem analysis of expected offshore wind electricity production in
Germany. Energy 33:747–759
Earles J, Halog A (2011) Consequential life cycle assessment: a review.
Pfister S, Koehler A, Hellweg S (2009) Assessing the environmental impacts
Int J Life Cycle Assess 16:445–453
of freshwater consumption in LCA. Environ Sci Technol 43:4098–4104
EC (2010) International Reference Life Cycle Data System (ILCD)
Potting J, Hauschild M (1997) Spatial differentiation in life-cycle assess-
Handbook—general guide for life cycle assessment—detailed guid-
ment via the site-dependent characterisation of environmental im-
ance. European Commission, Joint Research Centre, Institute for
pact from emissions. Int J Life Cycle Assess 2:209–216
Environment and Sustainability, Luxembourg
Potting J, Hauschild M (2006) Spatial differentiation in life cycle impact
Ekvall T, Weidema BP (2004) System boundaries and input data in conse- assessment: a decade of method development to increase the envi-
quential life cycle inventory analysis. Int J Life Cycle Assess 9:161–171 ronmental realism of LCIA. Int J Life Cycle Assess 11:11–13
Finnveden G (1999) Methodological aspects of life cycle assessment of Reinhard J, Zah R (2009) Global environmental consequences of in-
integrated solid waste management systems. Resour Conserv creased biodiesel consumption in Switzerland: consequential life
Recycl 26:173–187 cycle assessment. J Clean Prod 17(suppl 1):S46–S56
Frischknecht R et al (2005) The ecoinvent database: overview and meth- Scharlemann JPW, Laurance WF (2008) Environmental science: how
odological framework. Int J Life Cycle Assess 10:3–9 green are biofuels? Science 319:43–44
Heijungs R, Guinée J (2007) Allocation and ‘what-if’ scenarios in life Simons A (2013) Road transport: new life cycle inventories for fossil-
cycle assessment of waste management systems. Waste Manage 27: fuelled passenger cars and non-exhaust emissions in ecoinvent v3.
997–1005 Int J Life Cycle Assess. doi:10.1007/s11367-013-0642-9
Henriksson P, Zhang W, Guinée JB (2015) Updated unit process data for Simons A, Bauer C (2012) Life cycle assessment of the European pres-
coal-based energy in China including parameters for overall disper- surized reactor and the influence of different fuel cycle strategies.
sions. Int J Life Cycle Assess 20:185–195 Proc Inst Mech Eng, Part A: J Power Energy 226:427–444
Hertwich E et al (2014) Integrated life-cycle assessment of electricity- Sternberg A, Bardow A (2015) Power-to-What?—environmental assess-
supply scenarios confirms global environmental benefit of low- ment of energy storage systems. Energy Environ Sci 8:389–400
carbon technologies. Proc Natl Acad Sci U S A. doi:10.1073/pnas. Steubing B, Wernet G, Reinhard J, Bauer C, Moreno E (2016) The
1312753111 ecoinvent database version 3 (part II): analyzing LCA results and
Hou Q, Mao G, Zhao L, Du H, Zuo J (2015) Mapping the scientific comparison to version 2. Int J Life Cycle Assess. doi:10.1007/
research on life cycle assessment: a bibliometric analysis. Int J s11367-016-1109-6
Life Cycle Assess. doi:10.1007/s11367-015-0846-2 Stoessel F, Juraske R, Pfister S, Hellweg S (2012) Life cycle inventory
ISO (2006a) ISO 14040. Environmental management—life cycle assess- and carbon and water foodprint of fruits and vegetables: application
ment—principles and framework. International Organisation for to a Swiss retailer. Environ Sci Technol 46:3253–3262
Standardisation (ISO) Suh S, Yang Y (2014) On the uncanny capabilities of consequential LCA.
Int J Life Cycle Assess 19:1179–1184
ISO (2006b) ISO 14044. Environmental management—life cycle assess-
Suh S, Leighton M, Tomar S, Chen C (2013) Interoperability between
ment—requirements and guidelines. International Organisation for
ecoinvent ver. 3 and US LCI database: a case study. Int J Life Cycle
Standardisation (ISO)
Assess. doi:10.1007/s11367-013-0592-2
Laurent A, Espinosa N (2015) Environmental impacts of electricity gener-
Swiss Confederation (2014) SR 641.611 - Mineralölsteuerverordnung.
ation at global, regional and national scales in 1980–2011: what can we
Swiss Confederation, Bern
learn for future energy planning? Energy Environ Sci 8:689–701
The ecoinvent LCA database, v3.1, Bcut-off by classification^ (2014) The
Lesage P, Samson R (2013) The Quebec Life Cycle Inventory Database ecoinvent center. www.ecoinvent.org
Project. Int J Life Cycle Assess. doi:10.1007/s11367-013-0593-1 Tillman A-M (2000) Significance of decision-making for LCA method-
Levova T (2013) Water Use Modelling With ecoinvent v3 Opens New ology. Environ Impact Assess Rev 20:113–123
Possibilities. LCA XIII, Orlando, September 30th - October 3rd 2013 Tonini D, Hamelin L, Wenzel H, Astrup T (2012) Bioenergy production
Masanet E et al (2013) Life-cycle assessment of electric power systems. from perennial energy crops: a consequential LCA of 12 bioenergy
Annu Rev Environ Resour 38:107–136 scenarios including land use changes. Environ Sci Technol 46:
Meinshausen I, Müller-Beilschmidt P, Viere T (2014) The EcoSpold 2 13521–13530
format—why a new format? Int J Life Cycle Assess. doi:10.1007/ Treyer K, Bauer C (2013) Life cycle inventories of electricity generation
s11367-014-0789-z and power supply in version 3 of the ecoinvent database—part I:
Muller S, Lesage P, Ciroth A, Mutel C, Weidema B, Samson R (2014) electricity generation. Int J Life Cycle Assess. doi:10.1007/s11367-
The application of the pedigree approach to the distributions fore- 013-0665-2
seen in ecoinvent v3. Int J Life Cycle Assess. doi:10.1007/s11367- Treyer K, Bauer C (2014) Life cycle inventories of electricity generation
014-0759-5 and power supply in version 3 of the ecoinvent database—part II:

Content courtesy of Springer Nature, terms of use apply. Rights reserved.


1230 Int J Life Cycle Assess (2016) 21:1218–1230

electricity markets. Int J Life Cycle Assess. doi:10.1007/s11367- Weidema B, Ekvall T, Heijungs R (2009) Guidelines for application of
013-0694-x deepened and broadened LCA. ENEA, The Italian National Agency
Treyer K, Bauer C, Simons A (2014) Human health impacts in the life cycle of on new technologies, energy and the environment
future European electricity generation. Energ Policy 74:S31–S44 Weidema BP et al (2013) Overview and methodology. Data quality
Turconi R, Tonini D, Nielsen CFB, Simonsen CG, Astrup T (2014) guideline for the ecoinvent database version 3. The ecoinvent
Environmental impacts of future low-carbon electricity systems: de- Centre, St. Gallen
tailed life cycle assessment of a Danish case study. Appl Energy Wernet G, Conradt S, Isenring H, Jiménez-González C, Hungerbühler K
132:66–73 (2010) Life cycle assessment of fine chemical production: a case study
Volkart K, Bauer C, Boulet C (2013) Life cycle assessment of carbon of pharmaceutical synthesis. Int J Life Cycle Assess 15:294–303
capture and storage in power generation and industry in Europe. Int J Wernet G, Mutel C, Hellweg S, Hungerbühler K (2011) The environmental
Greenh Gas Con 16:91–106 importance of energy use in chemical production. J Ind Ecol 15:96–107
von der Assen N, Jung J, Bardow A (2013) Life-cycle assessment of Wernet G, Hellweg S, Hungerbühler K (2012) A tiered approach to esti-
carbon dioxide capture and utilization: avoiding the pitfalls. mate inventory data and impacts of chemical products and mixtures.
Energy Environ Sci 6:2721–2734 Int J Life Cycle Assess 17:720–728
Wegener Sleeswijk A, Heijungs R (2010) GLOBOX: a spatially differ- Yue D, You F, Darling SB (2014) Domestic and overseas manufacturing
entiated global fate, intake and effect model for toxicity assessment scenarios of silicon-based photovoltaics: life cycle energy and envi-
in LCA. Sci Total Environ 408:2817–2832 ronmental comparative analysis. Sol Energy 105:669–678
Weidema B (2014) Has ISO 14040/44 failed its role as a standard for life Zamagni A, Guinée J, Heijungs R, Masoni P, Raggi A (2012) Lights and
cycle assessment? J Ind Ecol 18:324–326 shadows in consequential LCA. Int J Life Cycle Assess 17:904–918

Content courtesy of Springer Nature, terms of use apply. Rights reserved.


Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:

1. use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
2. use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
3. falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
4. use bots or other automated methods to access the content or redirect messages
5. override any security feature or exclusionary protocol; or
6. share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at

onlineservice@springernature.com

You might also like