electronic supplementary material
THE ECOINVENT DATABASE V3
The ecoinvent database version 3 (part II): analyzing LCA results and comparison to version 2
Bernhard Steubing1 • Gregor Wernet2 • Jürgen Reinhard3 • Christian Bauer4 • Emilia Moreno-Ruiz2
Received: 21 Juen 2015 / Accepted: 18 January 2016
© Springer-Verlag Berlin Heidelberg 2016
Responsibleeditor: Rainer Zah
1Institute of Environmental Engineering, Swiss Federal Institute of Technology (ETH) Zürich, Schafmattstr. 6, 8093 Zürich, Switzerland
2ecoinvent, Technoparkstrasse 1, 8005 Zurich, Switzerland
3University of Zurich, Department of Informatics, Informatics and Sustainability Research Group, Binzmuehlestrasse 14, 8050 Zurich, Switzerland
4Technology Assessment Group, Paul Scherrer Institute, PSI, 5232 Villigen, Switzerland
Bernhard Steubing
1Methodology
The average number of supply chain inputs can be obtained by dividing the number of product systems by the number non-zero elements in the inverted technosphere matrix, as shown in Table 1.
Table S1: Average amount of supply chain processes and degree of integration for version 2.2 and all three system models of version 3.1
v2.2 / v3.1 Cut-off / v3.1 APOS / v3.1 Consequential[E] Amount of non-zero elements in A-1 / 9,915,062 / 88,978,452 / 102,799,348 / 76,622,876
[P] Available product systems / 4,087 / 11,304 / 11,332 / 10,305
[E/P] Mean amount of inputs used per product system / 2,426 / 7,871 / 9,072 / 7,436
[(E/P)/P] Degree of integration / 0.59 / 0.70 / 0.80 / 0.72
2Historical perspective: comparison of v2.2 and v1.3
From version 1.3 to version 2.2 the new datasetswere added (the number of activities increased from 2632 to 4087). Table 2 shows the database and dataset statistics for this transition.
Table S2: Database and dataset statistics for the comparison of version 2.2 of the ecoinvent database to version 1.3
v2.2 / v1.3GWP / ReCiPe / ES'06
Median database deviation (MDB) / 1.0% / 1.3% / -3.1%
Median dataset deviation (MDS) / 3.6% / 4.3% / 5.7%
datasets deviating > 20% / 9.5% / 13.2% / 22.0%
datasets deviating > 50% / 3.3% / 4.8% / 5.9%
datasets deviating > 100% / 0.8% / 1.4% / 2.8%
2.1Midpoint results
Fig.S1: Midpoint comparison of version 2.2 and 1.3
3Version 3.1 Cutoff compared to version 2.2
3.1Midpoint results
Fig.S2: Midpoints comparison of version 3.1 Cut-off and version 2.2
3.2Dataset statistics and contribution by geography
Table S3: Overview of production and market datasets for the ten most important geographies in ecoinvent 3.1 (cut-off version)
Version 2.2 / Version 3.1 cut-offGeography / Datasets / Total / Production datasets / Market datasets
All / 4087 / 11301 / 8105 / 3196
GLO / 367 / 3187 / 631 / 2556
RoW / - / 2683 / 2574 / 109
RER / 1295 / 1209 / 1155 / 54
CH / 1384 / 1204 / 1149 / 55
CA-QC / - / 296 / 285 / 11
DE / 107 / 152 / 143 / 9
Europe without Switzerland / - / 143 / 117 / 26
US / 72 / 96 / 87 / 9
CN / 56 / 89 / 81 / 8
SE / 39 / 75 / 67 / 8
Fig.S3: Process contributions for the top 8 native ecoinvent geographies (boxplot: red line: median; bar: 25th and 75th percentiles; whiskers: 1.5*IQR; crosses: outliers) [redo in two separate plots!]
3.3Updated transport data
Fig.S4: Contribution of transport (direct emissions) to ReCiPe (top) and Ecological Scarcity (bottom) of all matched datasets (Boxplots: red line: median; black circle: mean; bar: 25th and 75th percentiles; whiskers: 1.5*IQR; crosses: outliers)
3.4Electricity contribution by native ecoinvent geography
Fig.S5: Process contributions (GWP100a) from electricity production by native ecoinvent geography for all matched datasets in two sample regions: global (GLO) and Region Europe (RER). Upper figures: v2.2. Lower figures: v3.1 Cut-off. Abbreviations: CN=China, CZ= Czech Republic, DE=Germany,ES=Spain, GB=Great Britain, GR=Greece, IN=India, IT=Italy, JP=Japan, NL=The Netherlands, PL=Poland, RFC=ReliabilityFirst Corporation, RoW=Rest-of-the-World, RU=Russia, SERC= SERC Reliability Corporation(Boxplots: red line: median; black circle: mean; bar: 25th and 75th percentiles; whiskers: 1.5*IQR; crosses: outliers)
3.5Electricity process contribution for ReCiPe and Ecological Scarcity
Fig.S6:Process contributions (ReCiPe Endpoint (H/A)) from electricity production by country for all matched datasets in two sample regions: global (GLO) and Region Europe (RER). Upper figures: v2.2. Lower figures: v3.1 Cut-off. Abbreviations: AU=Australia, BR=Brazil, CN=China, Europe=all geographies within Europe, IN=India, JP=Japan, RoW=Rest-of-the-World, RU=Russia, USA=United States.(Boxplots: red line: median; black circle: mean; bar: 25th and 75th percentiles; whiskers: 1.5*IQR; crosses: outliers)
Fig.S7:Process contributions (Ecological Scarcity 2006) from electricity production by country for all matched datasets in two sample regions: global (GLO) and Region Europe (RER). Upper figures: v2.2. Lower figures: v3.1 Cut-off. Abbreviations: AU=Australia, BR=Brazil, CN=China, Europe=all geographies within Europe, ID=Indonesia, IN=India, RoW=Rest-of-the-World, RU=Russia, USA=United States.(Boxplots: red line: median; black circle: mean; bar: 25th and 75th percentiles; whiskers: 1.5*IQR; crosses: outliers)
3.6Comparison when electricity inputs are removed
Fig.S8: Deviation and cumulative absolute deviation of v3.1 cutoff from v2.2 results for database versions where electricity inputs were set to zero
Table S4: Database and dataset statistics for the comparison of version 3.1 (Cutoff) to version 2.2 where electricity inputs were set to zero
without electricity: v3.1 cutoff / v2.2GWP / ReCiPe / ES'06
Median database deviation (MDB) / 1.1% / 6.8% / 18.9%
Median dataset deviation (MDS) / 13.5% / 13.8% / 25.3%
datasets deviating > 20% / 33.4% / 36.2% / 57.3%
datasets deviating > 50% / 11.9% / 13.3% / 29.0%
datasets deviating > 100% / 3.7% / 6.4% / 9.8%
4Comparison of version 3 system models
4.1Cut-off vs. APOS
Fig.S9: Midpoints comparison of the Cut-off and APOS versions
4.2Cut-off vs. Consequential
4.2.1Midpoint results
Fig.S10: Midpoints comparison of the Cut-off and Consequential versions
4.2.2Marginal suppliers replaced by average suppliers
Fig.S11: Comparison of v3.1 cut-off with v3.1 consequential when marginal suppliers are replaced by average suppliers as in the cut-off system model
Table S5: Database and dataset statistics for the comparison of v3.1 cut-off with v3.1 consequential when marginal suppliers are replaced by average suppliers as in the cut-off system model
v3.1 Cut-off / v3.1 Consequential with average suppliersGWP / ReCiPe / ES'06
Median database deviation (MDB) / 0.1% / 0.0% / 0.1%
Median dataset deviation (MDS) / 2.3% / 2.4% / 2.6%
datasets deviating > 20% / 13.6% / 13.7% / 14.9%
datasets deviating > 50% / 8.6% / 8.8% / 8.7%
datasets deviating > 100% / 4.4% / 5.4% / 5.5%
4.2.3Substitution and constrained markets
For the other two modeling principles, substitution and constrained markets, no direct comparisons with the Cut-off versions were possible. However, in order to assess the importance of these mechanisms within the Consequential version itself, the following experiments were performed: a database copy was created where all substitution flows (i.e. negative inputs of allocatable by-products that are not material for treatments) were set to zero and thus no credits were given for by-products. Further, a database copy was created where inputs from constrained markets were set to zero, i.e. constrained products come burden-free and there is no additional demand for primary production. Results show, that within the Consequential version, the median database deviation increased by 9%, 7%, and 5% for GWP, ReCiPe and ES’06, respectively, if no credits for substitution are given (see Table 6). Further, the if constrained products were obtained burden-free,the median database deviation would decrease by 2%, 1%, and 1% for GWP, ReCiPe and ES’06. These results suggest that of those two mechanisms, substitution is the more relevant one in explaining the remaining median dataset deviation after taking into account the effects due to marginal suppliers. Moreover, as the choice of marginal suppliers explains the median database deviation, neither substitution nor constrained markets lead – at average – to higher or lower LCIA results.
Fig.S12: Comparison of Consequential system model database to versions without substitution (top) and constrained markets (bottom)
Table S6: Database and dataset statistics for the comparison of the Consequential system model with/without substitution and constrained markets
v3.1 Consequential with/without substitution / v3.1 Consequential with/without constrained marketsGWP / ReCiPe / ES'06 / GWP / ReCiPe / ES'06
median of deviation / 8.6% / 7.2% / 5.4% / 4.1% / -0.3% / -0.2%
median of absolute deviation / 7.7% / 6.5% / 5.4% / 9.0% / 6.8% / 6.9%
datasets deviating > 20% / 26.5% / 28.8% / 25.5% / 31.9% / 26.2% / 28.0%
datasets deviating > 50% / 12.5% / 12.1% / 13.6% / 15.2% / 10.5% / 12.1%
datasets deviating > 100% / 7.7% / 7.3% / 8.5% / 7.8% / 5.6% / 7.1%
4.3APOS vs. consequential
Fig.S13:Deviation of v3.1 APOS from v3.1 consequential results
Table S7: Database and dataset statistics for the comparison of the APOS and Consequential system models
v3.1 APOS / v3.1 consequentialGWP / ReCiPe / ES'06
Median database deviation (MDB) / 1.6% / -2.2% / -0.8%
Median dataset deviation (MDS) / 8.2% / 6.9% / 7.3%
datasets deviating > 20% / 33.3% / 31.3% / 34.7%
datasets deviating > 50% / 15.5% / 12.9% / 14.7%
datasets deviating > 100% / 8.8% / 8.0% / 9.2%