Scope and functional unit
The unit of analysis is 1 kg of polymer prepared for further use in the targeted application, i.e. rigid packaging or car interior panels. This is done for the purpose of simplification, despite the fact that the authors are well aware that different polymers may require slightly different amounts to fulfil specific product properties or services (for this, see Nessi et al. 2019). The method used for the GW impact assessment is the Myhre et al. (2013), as suggested in the Product Environmental Footprint (“Climate Change” indicator as in EF3.0; European Commission, 2013). The focus is on GW as this serves the purpose of illustrating and discussing the methodological issues mentioned earlier. A detailed description of the individual issues tackled and how they are addressed in this study is provided in Sects. 2.3, 2.4, and 2.5.
System boundary and inventory
The system boundary encompasses the life cycle activities undergone from the production of the feedstock material up to the End-of-Life treatment of the polymer, excluding the use phase, which is assumed to be the same for all polymers belonging to each group (Fig. 1). The stages are as follows: (I) feedstock supply including extraction, transport, and refining of crude oil and natural gas; crop cultivation and transport; collection, transport and sorting of recyclate; transportation of this feedstock to downstream conversion processes, e.g. naphtha cracking, sugarcane fermentation, and wet milling of starch crops; (II) polymer production including conversion of feedstock materials into the final polymer including any transport involved; (III) manufacture operations to prepare the polymer for further use in the sector of application alongside distribution, i.e. extrusion for rigid packaging and injection for panels; plus, distribution to retailer and final consumer; and (IV) End-of-Life including collection, sorting, recycling, incineration, or landfilling of the article after use, including any substituted process, e.g. virgin material or energy production. The electricity mix substituted by the energy generated through incineration was assumed to be equal to the residual EU electricity grid mix, composed by 57% fossil fuel, 37% nuclear, and 7% renewables (GW 0.14 kg CO2-eq. MJ−1 electricity; Nessi et al. 2019). For heat, we used an EU average heat mix consisting of 42% natural gas, 31% hard coal, 22% biomass, and 5% heavy fuel oil (GW 0.07 kg CO2-eq. MJ−1 heat). Any multifunctional process was resolved following PEF guidelines. These prioritise subdivision and system expansion over allocation; yet, wherever an unambiguous market substitution could not be identified, economic allocation was applied instead. The EU-average EoL split for rigid packaging consists of 60% recycling, 21% incineration, and 19% landfilling, based on most recent data on PET bottles recycling (ICIS & Petcore Europe, 2018) and management of plastic packaging waste in Europe (EUROSTAT, 2019). For car panels, the EoL consists of 97% recycling, 2% incineration, and 1% landfilling (EUROSTAT, 2018). The life time of the products was assumed to 1 year for flexible packaging and 20 years for car panels. Notice that (i) landfilling here is modelled specifically for non-biodegradable plastics, i.e. carbon degradation is negligible; 9ii) recycling rates exclude possible future improvements following behavioural or technological changes; and (iii) the presence (and effects) of additives/contaminants in the recycling value chain is not addressed. The reader is referred to the Online Resource for a detailed description of the datasets used to represent the individual life cycle stages. Additional information may also be found in Nessi et al. (2019).
Recycled content and recyclability
We define the recycled content (R1) as the input of recycled material divided by the reference flow (i.e. the amount needed to fulfil the functional unit). We define the recycling output rate (R2) as the amount of recycled material supplied at the End-of-Life divided by the total waste generated after the use phase. R2 shall therefore consider the inefficiencies in the collection, sorting and recycling (or reuse) processes. We call the fraction of the total generated waste that is collected and sent to energy recovery the energy recovery rate (R3). Several mathematical formulas have been proposed in the literature to address the contributions of the recycled content, recyclability, and recoverability on the overall impacts. Such formulas describe the life cycle environmental impact of a product from the provision of feedstock (i.e. virgin plus recycled content) to the EoL. While it is beyond the scope of this paper to describe the formulas in detail, in this study we refer to and use as a starting point two key studies that compared multiple existing formulas, i.e. Allacker et al. (2017) and Schrijvers et al. (2016). Allacker et al. (2017) proposed a formula as the most suitable for the European Commission Product Environmental Footprint (PEF). This approach models a 50:50 share of burdens and benefits from waste management between two subsequent life cycles and was the basis for the version presented in the PEF method (Zampori et al. 2016). Zampori et al. (2016) further introduced a market factor (here named Aef) to replace the 50:50 share assumption of Allacker et al. (2017) to adjust for the market conditions of the recycling (Aef 0 for high quality and/or high demand, Aef 1 for low quality and/or low demand). The adapted formula of Zampori et al. is referred to as the circular footprint formula. A similar suggestion came from Schrijvers et al. (2016) that defined Ap as the ratio between the market price of the recycled and the displaced virgin material (Ap 1 when prices are the same, Ap 0 when the recycled material has a comparatively low value). The formula of Schrijvers et al. (2016) is updated in Schrijvers et al. (2020a, b), where Ap 1 and Ap 0 reflect high and low demand for the recycled material, respectively. Figure 2 introduces the conceptual framework for the three alternative approaches assessed in this study to account for the burden associated with the recycled content and the recyclability. The three formulas are presented in Table 1. In mathematical terms, for the way these factors are defined in Zampori et al. (2016) and Schrijvers et al. (2020a, b), when Aef tends to 1, Ap tends to 0. Compared with the original formula from Allacker et al. (2017), the PEF formula presented by Zampori et al. (2016) has been simplified deleting the term Ed* (avoided disposal of the material from which the recycled content is taken), see Table 1. This is also a substantial difference compared with the formula presented in Schrijvers et al. (2020a, b). The latter aims to reflect that the use of a recycled material that is already fully absorbed by the market diverts this material from other potential users that have to use an alternative, virgin, material instead. This leads not only to the induced production of this virgin material but also to other potential differences in the overall life cycle, such as differences in distribution, the use phase, and disposal. These additional differences between the expected life cycle and the substituted life cycle are also modelled in the formula (see ∆Eo for the recycled content and ∆E* for the recyclability in Table 1).
We apply the formulas from Allacker et al. (2017), Schrijvers et al. (2020a, b), and Zampori et al. (2016) to the case studies to illustrate the effect of the different assumptions on the contributions of the recycled content, the recyclability, and the recoverability to the final GW result (Table 1). In particular, we focus on the influence of the terms related to disposal (Ed, Ed*, Edo), which represent a substantial difference between the three formulas analysed. Besides the effect of the formula selection, we also investigate the influence of the market parameter A in the formula I and III (Schrijvers et al. 2020a, b and Zampori et al. 2016). We assess four scenario variants to illustrate the effects of varying the A value. While a general description of the implications of these values is given in Table 2, the example of using OW-PLA instead of virgin HDPE is illustrated as follows. A = Ap = Aef = 0.5 reflects the situation where the production of the OW-PLA is optimized for a defined demand. There is room for upscaling of the recycling process, although users of the recycled material could switch to an alternative material if prices were to increase (Ekvall, 2000), which is an effect that could be relevant in the short term. In a first scenario variant we set A = Ap = 0 (Aef 1), implying that there is a very low demand for OW-PLA. Stimulating the demand for OW-PLA could avoid the food waste, which is used as feedstock, being disposed of as waste (here assumed to be incineration with energy recovery). This reflects the current EU situation, i.e. most of the organic waste is still being disposed of (or incinerated) and the use of the recycled material is occasional rather than mainstream. In the second variant A = Ap = 1 (Aef 0), all the available waste is exploited and the maximum amount of OW-PLA is used, and further upscaling of the recycling process is not possible due to a lack of available food waste.
Biogenic carbon accounting
Bio-based products are seen as an opportunity to substitute finite fossil sources with renewable and carbon-neutral products, and as such are at the centre of the EU bioeconomy strategy (EU bioeconomy strategy, 2018). However, experience within the LCA community with bioenergy has shown that the climate change mitigation potential of bio-based commodities cannot be taken for granted, but rather needs to be assessed with care (European Commission, 2019). Especially critical is the way in which biogenic carbon (biogenic C) emissions are calculated along the life cycle of the bio-based product. Agostini et al. (2020), for instance, show that 71 out the 100 most cited LCA papers on bioenergy have applied a carbon neutrality assumption, meaning that both the used bio-based product and the emitted CO2 at the end of life are considered to have zero effect on global warming. Through this assumption the LCA practitioner avoids accounting for the biogenic carbon cycle by assuming that the carbon emitted from biomass combustion or decomposition will be reabsorbed by the growing plants on a time scale significantly shorter than the relevant temporal scale of the analysis. However, the biogenic carbon neutrality assumption is often misleading as the dynamics of the biogenic C cycle may be significant in the short/medium term (Agostini et al. 2013). In order to avoid potential burden shifting along the temporal scale, a growing number of LCA studies have been accounting explicitly for the time-dependent impacts of biogenic C cycle for at least 10 years (e.g. Levasseur, 2010; Cherubini et al. 2011; Brandao et al. 2013; Giuntoli et al. 2016, Levasseur, 2016; UNEP-SETAC, 2016, Breton et al. 2018). When assessing the climate change impact of bio-based products, the biogenic C cycle has a strong time-dependent nature: the biomass is harvested and transformed into a product, and while the biogenic C is stored in the technosphere during the use phase of the bio-product, the biomass re-grows sequestering atmospheric CO2 through specific dynamic trajectories (i.e. annual crops will re-sequester harvested CO2 every year, while a forest stand will take decades to regrow). The way these phenomena are characterized differs across various studies, based on the choice of climate metrics (instantaneous vs. cumulative; Giuntoli et al. 2015), the choice of absolute or normalised metrics (Cherubini et al. 2013), the choice of reference system (Koponen et al. 2018), and the choice of temporal boundaries of the analysis (Brandao et al. 2013; Levasseur, 2016; Breton et al. 2018). While discussing all these aspects in details is beyond the scope of this paper, it is important to notice that no choice is right or wrong, but that any choice will embed value judgement, thus we transparently report our assumptions in the Online Resource.
Testifying to the complexity of the issue, no normative agreement on how to account for biogenic C in LCA exists yet. For instance, the CEN (2015) suggests two different approaches for forest C-accounting: if the analysis is performed at landscape level, then a carbon neutrality assumption is advised, while a calculation of dynamic C-cycle is recommended if the analysis is performed at stand level. Similarly, the BSI (2011) instructs that, where some or all removed carbon will not be emitted to the atmosphere within the 100-year assessment period, the portion of carbon not emitted to the atmosphere during that period shall be treated as permanently sequestered carbon. European Commission (2013) and (ISO, 2013), conversely, instruct the user to characterise all emissions and sequestration of biogenic CO2 with a GWP = 0.
We consider the default choice of carbon neutrality (i.e. to characterize biogenic C sequestration and emissions with a GWP = 0) to be equivalent to taking a long-term perspective (e.g. > 100 years) in which the impact of the biogenic C cycle is irrelevant compared with the impact of the permanent release of fossil CO2. However, we reckon that evaluating short-term climate impacts is essential to inform decisions on a temporal scale more in tune with the urgency required by the climate crisis. Therefore, similar to the method presented by Cherubini et al. (2011) and Guest et al. (2013), we provide GWP factors for biogenic C and for fossil C that take into account the effects of time. For biogenic C, we provide characterisation factors (GWPbio) valid for biomass feedstocks with different rotation periods as well as with two different EoL scenarios: (i) incineration after the use period and (ii) landfilling after the use period. Consistently, for the case of 100% recycling at EoL, we considered the credits from temporary biogenic C storage incurred by extending the lifetime of the material in the technosphere (one additional lifecycle). This corresponded to an additional 20 years for the case of car panels (in total 40 years). The GWPbio factors range from net negative values to positive values (albeit still lower than 1). Consistently, we also provide GWPfossil values that take into account the effect of delayed emissions of fossil C compared with an emission taking place at t = 0. It should be noticed that there is a significant difference between the negative GWPbio values assigned to biogenic C and the reduced positive values for fossil CO2. The former represents a credit for temporary storage in the technosphere, while the biomass is allowed to regrow, while the latter is a decreased burden for delayed emission with respect to the fixed time horizon chosen. Indeed, all the GWP values used are reported at a fixed time horizon of 100 years, as it is common practice for LCA. All details of the calculations performed are presented in the Online Resource. The values in Tables S7 and S8 (Online Resource) are used in the following sections to evaluate the impact of explicitly considering biogenic C flows and time-dependent dynamics. Notice that while we use a dynamic approach solely for biogenic CO2, biogenic CH4 is accounted for using the proper GWP(100) factors within the EF3.0 method. In the case of products with significant, time-dependent emission profiles of biogenic- CH4, such as it could be for the landfilling of biodegradable bioplastics, we recommend that a similar dynamic approach is also taken for biogenic CH4, as illustrated in Giuntoli et al. (2016). Notice also that credits from temporary storage/delayed emissions are shared between life cycles conforming to the formulas considered (Table 1), consistently with the other GHG emissions and in line with the recommendations from Finkbeiner et al. (2012).
Land use changes
While dLUC refers to the changes occurring on the same land where the land use for the product under assessment takes place, iLUC refers to market-driven consequences incurred by the demand for land occurring in the first place (Schmidt et al. 2015; Valin et al. 2015). The point of departure for iLUC to occur is when arable land, already in use for cropping or grazing activities, is used for supplying the feedstock under assessment. The pre-condition for iLUC is that the global agricultural area is still expanding because of several confounding factors (such as increased population and GDP increase of some countries) and that its capacity is inherently limited/constrained (Schmidt et al. 2015). The main underlying postulate of iLUC is that this relative drop in availability is likely to cause a relative increase in agricultural prices, which in turn provides incentives to increase production elsewhere. This increase in production in principle can incur: (i) agricultural land expansion at the expenses of forest or grassland, (ii) production intensification, and (iii) crop-displacement mechanisms (reduced consumption). The latter is supported by some studies arguing that in the short-to-medium term not all of the displaced feedstock may need to be compensated by increased production as reduced consumption may also occur (e.g. Edwards et al. 2010). This hypothesis is however contradicted by other authors (e.g. Schmidt et al. 2015; Searchinger et al. 2015) arguing that this effect should not be included in LCAs, since it is the long-term effect of the demand that should guide decisions (Weidema et al. 2013). According to this and assuming no consumer’s behavioural changes (e.g. diet), the supply of goods and services should be assumed to be fully elastic, i.e. an increase in demand is to be met by a corresponding (1:1) increase in supply.
For biobased polymers we assess the contribution of various LUCs emission factors on the Global Warming results. Most of the iLUC factors derived with global equilibrium models already include dLUC, i.e. changes in C stock are estimated not only for (additional) natural land clearing (e.g. forest, grassland, pasture, savannah) but also for cropland, as stressed in Valin et al. (2015). Therefore, in this study we opt not to distinguish between dLUC and iLUC and instead simply refer to LUC. We quantify the LUC contribution using three different LUC models available from the literature. The first is the biophysical model proposed by Schmidt et al. (2015) used as the default, the second builds on the LUC factors derived with an economic equilibrium model (Valin et al. 2015) and the third is a normative-based approach that applies the LUC factors suggested by European Parliament and Council of the European Union (2015) (annex V and VIII; also obtained through economic modelling). Among the three models, Valin et al. (2015) strives to include all C-stock changes due to the increased biomass demand, i.e. natural land clearing and cropland. The LUC GW contribution is calculated as follows: the specific land demand for crop production is converted into a demand per functional unit, based on the specific consumption of crop for polymer production (kg crop kg−1 polymer, consistent with LCI modelling). The LUC GW contribution is finally calculated by multiplying the land demanded by the LUC GW emission factor obtained from the model chosen.
Along with the three LUC emission factors above, we also present the direct LUC contribution as quantified following the standard BSI (2011) complemented with BSI (2012) for illustrative purposes. The standard suggests the following approach to quantify dLUC: two main types of land transformation are considered, i.e. transformation from grassland and transformation from forest (to annual or perennial crop). The emissions arising from the product are assessed on the basis of the default land use change values provided in PAS 2050:2011 or using the relevant sections of the IPCC Guidelines for National Greenhouse Gas Inventories (IPCC, 2006) for countries and LUC types not covered by the former.