S1.Detailson Conversion of Impact Scores Into Common Metrics

Electronic Supplementary Material

Content:

S1.Detailson conversion of impact scores into common metrics

S2. Details on case study

S3. Details and additional results for substance contributionanalysis

S4. Details on identifying reasons for differences in impact scores

S5. Additional results: impacts scores per life cycle stage

S6. References

S1

S1. Details on conversion of impact scores into common metrics

The following sections rely on documentation available for ILCD 2009 (EC-JRC, 2011, 2012; Hauschild et al. 2013), ReCiPe 2008 (Goedkoop et al. 2009) and IMPACT 2002+ (Jolliet et al. 2003). As the ILCD 2009 was not implemented into GaBi at the time of the study, characterization factors for the ILCD methods(version 1.0.3, 01 March 2012) were downloaded from the Life Cycle website of the European Commission ( and were imported into the software.For those ILCD 2009 methods where ReCiPe 2008 is the recommended method, impact scores were calculated using the original set of ReCiPe (version 1.05) characterization factors as implemented in GaBi. The same set of ReCiPe (version 1.05) characterization factors, and a set of IMPACT 2002+ (version 2.1) characterization factors implemented in GaBi, were employed to calculate impact scores for ReCiPe 2008 and IMPACT 2002+, respectively. For freshwater ecotoxicity, impact scores were calculated off-line using emission inventory data exported from GaBi and characterization factors from USEtox, version 1.01 (Rosenbaum et al. 2008). Wherever updated CFs were needed (e.g. for discussion), original sources were used, i.e. Jolliet et al. (2005) and Goedkoop et al. (2012) for IMPACT 2002+ and ReCiPe 2008, respectively.

S1.1. Climate change

• ReCiPe 2008 and ILCD 2009 use the recommended global warming potentials (GWP) provided in the Fourth Assessement Reportby the IPCC (Forster et al. 2007) with a time horizon (TH) of 100 years
• IMPACT2002+ use GWP with a TH of 500 years. Those are not up-to-date and originate from the AR3 (IPCC 2001).
• Units are all the same: mass of CO2eq

S1.2. Stratospheric ozone depletion

• All methods use the recommended ozone depletion potentials (ODP) provided by WMO (
• Units are the same for all methods: mass of CFC-11eq

S1.3. Photochemical ozone formation

• ILCD 2009 consists of ReCiPe 2008 method; unit is in mass of NMVOC equivalent. IMPACT2002+ use CFs expressed in mass of ethylene equivalent (kg C2H4eq)
• All units are brought to mass of NMVOC equivalent. Conversion from the IMPACT 2002+ unit to the reference unit was done using

S1.4. Terrestrial acidification and eutrophication

• ReCiPe 2008 addresses only terrestrial acidification (mass of SO2eq). IMPACT2002+ addresses both impact categories in an aggregated form using SO2 as reference substance. ILCD 2009 addresses both separately using the same unit Accumulated Exceedance (AE).
• SO2 does not contribute to terrestrial eutrophication. Because it is used as reference substance in IMPACT2002+ for both acidification and terrestrial eutrophication, an overall comparison is only possible if the indicator scores can be expressed in reference to a substance common to characterization models of both impact categories. Ammonia (NH3) was thus selected as reference substance.
• Conversion from the ILCD 2009 unit to the reference unit was done using:

and for acidification and terrestrial eutrophication, respectively.

• Conversion from the IMPACT2002+ unit to the reference unit was done using

for terrestrial acidification

• Conversion from theReCiPe 2008 unit to the reference unit was done using

for terrestrial acidication

• Comparisons can be done for both terrestrial acidification and eutrophication altogether between ILCD and IMPACT 2002+. A separate comparison for acidification is also made between ReCiPe 2008 and ILCD 2009.

S1.5. Aquatic eutrophication

• ReCiPe 2008 is used in ILCD 2009 and distinguishes impacts on freshwater and marine ecosystems. This distinction is not done in IMPACT2002+, which aggregates impacts into one single score (i.e., aquatic eutrophication).
• All indicator scores are converted into unit of kg PO4eq, which is the unit used by IMPACT 2002+: (freshwater eutrophication) and (Redfield conversion ratio taken from ReCiPe 2008 (Goedkoop et al. 2009); also used in IMPACT 2002+).
• Comparisons are made at the aggregated level of both marine and freshwater eutrophication.

S1.6. Ecotoxicity

• ILCD 2009 is based on USEtox model and expresses its CFs for freshwater ecotoxicity at midpoint level in PAF·m3·d. ReCiPe 2008 and IMPACT2002+ use reference substances, i.e. 1,4-dichlorobenzene (1,4-DB) and triethylene glycol (TEG), respectively.
• To obtain a reference unit matching the ICLD 2009, a conversion of all scores to PAF·m3·d is possible but introduces more uncertainties as units for ReCiPe 2008 and IMPACT2002+ need to be translated to endpoint unit (straightforward conversion), and then from endpoint unit back to the midpoint unit PAF·m3·d. The latter includes some degree of uncertainties, which justified the abandon of this approach. Therefore, for simplicity purposes, the reference substance was chosen to be 1,4-dichlorobenzene (see background CF for conversions in Table S11).
• In both ReCiPe 2008 and IMPACT 2002+, different emission compartments for the reference substance are considered in the expression of the CFs across the different ecotoxicity media (freshwater, marine, terrestrial). The reference unit for freshwater ecotoxicity, marine ecotoxicity and terrestrial ecotoxicity was taken as the mass of 1,4-DB emitted to freshwater (no CF for 1,4-DB emitted to industrial soil for ReCiPe 2008 was available in the used GaBi version to keep the original ReCiPe 2008 unit for terrestrial ecotoxicity).

Table S11. Characterization factors used to determine the unit conversion factors for ecotoxic impact categories

S1.7. Ionizing radiation (impacts on human health)

• ReCiPe 2008 is used in ILCD 2009 and expresses scores in U235equivalent (kBqU235 eq; releases to air), which is chosen as a reference.
• Scores in IMPACT 2002+ are expressed in BqC14eq (releases to air). Accounting for an additional factor of 1000, convertion was done using .

S1.8 Respiratory inorganics/particulate matters

• ILCD 2009 and IMPACT2002+ express the impact score in the unit of mass PM2.5eq; ReCiPe 2008does it in mass of PM10eq.
• Conversion was done using .

S1.9. Human toxicity

• Cancer and non-cancer effects are distinguished in both ILCD 2009 (USEtox) and IMPACT2002+. ReCiPe 2008 has an aggregated perspective. Translating the units into comparative toxic units (CTUh) is not possible here unless IMPACT2002+ and ReCiPe 2008 scores are converted into DALYs (endpoint) and brought back into CTUh using the severity factors from Huijbregts et al. (2005). This last step would however bring uncertainties in the conversion, which is not desirable. Therefore, a basic translation using a reference substance, i.e. that of ReCiPe 2008 (mass of 1,4-dichlorobenzene, 1,4-DBeq), was performed.
• Conversion from the IMPACT2002+ unit into the reference unit was done using

(to air), and (to air) for cancer and non-cancer effects, respectively.

• Conversion from the for ILCD 2009 unit into the reference unit was done using , and for cancer and non-cancer effects, respectively. CFs for continental and urban air were weighted equally (50:50) to reach these values (corresponding to ‘air, unspecified’ used in LCA software).

S1.10. Land use

• Land use in IMPACT 2002+ is only assessed with regard to land occupation. The midpoint impact score is expressed in m2·yr eq of organic arable land. In ILCD 2009, the method by Milà i Canals (2007) is used, in which the characterization model quantifies the soil quality, expressed in soil organic matter (kgC·yr), from both land occupation and transformation. Conversion from the ILCD 2009 unit to the reference unit was done using
• A comparison between IMPACT 2002+ and ILCD 2009 can be attempted although land transformation is not included in the former.
• The land use midpoint assessment in ReCiPe 2008 assumes competitiveness between land types within the 3 separate categories agricultural land occupation, urban land occupation and natural land transformation. All land occupation/transformation are ranked equally and summed up within those categories. As all CFs are equal to 1, the sum of agricultural and urban land occupation could thus be assimilated to m2·yr eq of organic arable land. A comparison is thus made along with IMPACT 2002+ and ILCD 2009. Note however that, like IMPACT 2002+, only land occupation is taken into account.

S1.11. Resource depletion (minerals/metals, fossils)

• Both ReCiPe 2008 and IMPACT2002+ distinguish the midpoint assessments of minerals/metals and fossils depletion. In both, the units used are the same, i.e., in unit of mass iron ore equivalent (minerals) and in MJsurplus or kgoil eq (fossils). Direct comparisons for each category are thus possible between the two methods although depletion scores for minerals and fossils are not comparable to one another. For assessment of fossils depletion, the considered conversion factor is: . Notethat uranium is considered as fossils in IMPACT2002+ whereas it is regarded as a metal in ReCiPe 2008.
• In the version of GaBi used (4.3), the midpoint assessment of minerals depletion for IMPACT 2002+ has been erroneously implemented and actually is the assessment at endpoint level (expressed in energy surplus). Conversionfrom the resulting endpoint indicator score into midpoint CF for iron ore, which is the reference resource, was done using .
• ILCD 2009 recommends the application of the CML methodology with use of characterization factors relying on the reserve bases (Guinee et al. 2002; Oers et al., 2002; EC-JRC, 2011). Metals and fossils are assessed simultaneously, and indicator scores are expressed in kg-antimony (Sb) equivalents. Two approaches exist to allow comparisons with ReCiPe 2008 and IMPACT 2002+: (i) splitting the CML method into two categories, i.e. one for metals/minerals and the other for fossils, or (ii) reuniting metal/minerals and fossils depletion indicators for ReCiPe 2008 and IMPACT 2002+. The former approach requires redefining new CFs, and was therefore not considered in this study. The latter approach can easily be done by using the ratio of CFs between iron and oil in CML and applying it to ReCiPe 2008 and IMPACT 2002+ methods. This was believed to induce large uncertainties as part of the modelling approach of CML would thus be implanted into IMPACT2002+ and ReCiPe 2008. Therefore, it was decided not to include this possible conversion in the conversion framework used in the manuscript, and the comparisons are only made possible between scores obtained in ReCiPe 2008 and IMPACT2002+.

S1

S2.Details on case study

S2.1. Scope

The scope concerns four window alternatives of parameters presented in Table 3 in the main part. Reference flows to fulfil the functional unit are shown in Table S1. Process diagrams are shown in Fig S1. As a general rule, capital equipment such as buildings or machines is not included in life cycle boundaries. Some of the processes used for modelling inventories are aggregated and might include capital equipment; it is estimated that capital equipment can contribute to 10% of total impact.

Table S1. Life times and reference flows needed to fulfil the functional unit for each window type.

S1

a

b

S1

c

d

Fig S1. Process diagrams for each window type: (a) wood; (b) wood/aluminum, (c) PVC, and (d) wood/composite.

S1

S2.2. Life cycle inventories

Details on life cycle inventories are presented in the below.Unit processes are summarized in Tables S8 and S9. Overview of life cycle emissions is shown in Table S11.

Materials and manufacturing.Full bills of materials and energy were collected from window manufacturers and literature (Weir et al. 1998; Krogh et al. 2003; Salazar and Sowlati 2008a, 2008b; Tarantini et al. 2011) and were mapped to processes in the Ecoinvent database, version 2.2 (Hischier 2010) combined with the PlasticsEurope (PE) database (Plastics Europe 2010). The following cut-off criteria were applied to model product systems: (i) materials contributing to less than 5% of total mass are cut-off, except of substances for surface treatment of the window frame which can be toxic; and (ii) cleaning of window pane is cut-off. Bills of materials and energy for a single window are provided in Tables S2 and S3.

Table S2. Bill of materials (in kg) for a single window of each type.

Table S3. Bill of energy in the manufacturing stage (in MJ)for a single window of each type.

S1

Use.Danish district heating mixis based on 34% biomass (8% straw, 16% wood, 10% renewable waste), 30% natural gas, 24% coal, 6% non-renewable waste, and 3% oil (Danish Energy Agency2011). Other energy sources contribute to less than 4% of the total mix and were not modeled.Unit processes for heat mix are shown in Table S10.

Heat loss is directly proportional to the U-value and the window area, and depends on the temperature difference between the indoor and outdoor environment. In Denmark, the average outdoor temperature is 7.7 °C. The average indoor temperature is set to 17 °C, and is a combination of daily temperature recommended in the building, and the temperature in the building when little or no heating takes place (in the night).Heat losseswerecalculated using eq S1 and are summarized in Table S4.

(eq S1)

where Φt(W) is the heat loss; U (W·m-2·K-1) is the U-value; and θiand θe (K) are the indoor and outdoor temperatures.

During the use stage, the W window is painted. Given that about 1 L of paint is needed for a window frame area of 1 m2, 0.6 kg of paint is required to paint the W window once.

Table S4. U-values and heat losses calculated using eq S1 for each window type per year.

Disposal. All windows are disposed of according to the Danish waste policy, where glass, aluminum and steel are mainly recycled, and wood is incinerated. Other materials are either incinerated or landfilled. PVC is technically recyclable but not to the extent as for other plastics (30%). The remaining part of PVC is landfilled. The composite (glass fiber/polyamide) is technically difficult to recycle, and it is assumed to be 100% incinerated. Energy recovery from incineration is credited for using average Danish heating and electricity mix. All recycled materials replace virgin materials in the market, i.e. glass cullets, aluminum ingot, steel billet, and PVC granulate mix. Summary of end-of life scenarios is given in TablesS5 and S6.

Table S5. Bill of energy in the disposal stage (in MJ)for a single window of each type.

S1

Table S6. Disposal routes (recycling/incineration/landfilling, in %) for window materials.

Transportation.Transportation means and distances were assumed. They are shown in Table S7.

Table S7. Transportation means and distancesfor each window type.

S1

Unit processes.

Table S8. Unit processes used to model life cycle inventories.