Accounting for effects of carbon flows in LCA of biomass-based products—exploration and evaluation of a selection of existing methods

The case study used for testing the different methods was packaging made from polyethylene (PE) based on biomass. Specifically, polyethylene produced from wood (from Swedish boreal forest) and from sugarcane (from Brazil) was considered (see Fig. 1). In total, three PE production routes were included: (1) two different hypothetical routes from wood to PE, via fermentation to ethanol followed by dehydration to ethylene and via gasification yielding syngas that is converted to methanol and subsequently to ethylene and (2) one existing route from sugarcane to PE, via fermentation to ethanol followed by dehydration to ethylene. The end-of-life scenario applied was incineration, i.e. complete oxidation of the PE packaging. All three routes were investigated in a series of preceding papers (Liptow and Tillman 2012, Liptow et al. (2015), Liptow et al. (2013)). These three studies included inventory data for biogenic CO₂ emissions, that is, CO₂ emissions originating from biomass, and were used as a database to quantify emissions of fossil and biogenic CO₂ in the life cycles of each case. In addition, data for the case of the fossil based PE production route were used for comparison. The selected assessment methods were then applied to these inventory data.

A literature review was done on methods for the assessment of global warming due to land use and biogenic CO₂ emissions in LCA. There are different indicators to express the impact on global warming of biogenic carbon (see, e.g. Ericsson et al. (2013)); the review however considered methods using global warming potential (GWP) only because of its ubiquitous use compared to other indicators such as global temperature potential (GTP). Furthermore, methods that are theoretically related but not directly applicable in an LCA context (e.g. Bernier and Paré (2013), Holtsmark (2012), Zanchi et al. (2012)) were outside the scope of this study. Such methods do not provide characterization factors for the impact due to biogenic carbon flows, for instance, but rather focus on the variation over time of these flows. This can be considered as a part of the inventory analysis in an LCA but not as part of an impact assessment method. Also, this study only considered the attributional assessment of global warming, and therefore, methods focusing on effects of indirect land use change (iLUC) (e.g. Kløverpris and Mueller (2013), Schmidt et al. (2015)) were outside the scope of this study.

Three carbon cycle methods (see Introduction), the GWP_bio method (Cherubini et al. 2011a), the GWP_netbio method (Pingoud et al. 2012) and the WF method (Väisänen et al. 2012), were selected. Of these three methods, two (GWP_bio and GWP_netbio methods) calculate characterization factors (CFs) for assessing the climate impact of biogenic CO₂ flows. Like for any other CF, these CFs are applied to inventory flows, in this particular case to the biogenic CO₂ flows per tonne of PE produced (which is the functional unit used in the LCA case study (see Sect. 2.1)). In contrast, the WF method calculates a weighting factor (WF) which is applied to the inventoried amount of biogenic CO₂ before applying the CFs commonly used for fossil CO₂ in the life cycle impact assessment. The three selected methods are now described in more detail.

The methods were analysed based on a framework developed by Helin et al. (2013) which can be summarized with the following five questions:

Helin et al. (2013) stated an additional sixth question, which investigates the type of indicator used to express the climate impact of GHG emissions. However, this question was outside the scope of this study (see also Sect. 2.2). Table 1 gives a comparison of the carbon cycle and land use methods that are discussed above, using these five questions.

In addition to addressing these five method-oriented questions in the analysis, practical aspects of applying the methods were also tested. These aspects included data availability, acceptance of the method and ease of application.

Table 2 presents the key inventory data used for testing the impact assessment methods, based on data from Liptow and Tillman (2012), Liptow et al. (2013) and Liptow et al. (2015). These inventory flows are also depicted in Fig. 1 (and their naming is given in Table 2). The life cycle inventory data were derived using several major modelling choices and assumptions. First of all, an attributional LCA approach was used. Next, all biogenic carbon emissions released during biomass harvest and production, and all carbon emissions related to land use, were accounted for and were assessed as CO₂. In the case of sugarcane, which is cultivated in Brazil, pre-harvest burning was included, and in the case of wood, which originates from managed boreal forest in Sweden, the inventory for its acquisition was modified to stem wood using data from Berg and Lindholm (2005). For comparison, the CO₂ emissions from a conventional fossil production route are also shown in Table 2. Furthermore, emissions from incineration of the PE (the method of disposal) were included and, for reasons of simplification, completely allocated to the PE.

As required by the tested methods, additional data were collected, e.g. on carbon stocks in soil and vegetation. These data are summarized in Table 3.

Considering only fossil CO₂ and GHGs other than CO₂ released along the life cycle (see Fig. 3, GWP_other (these values are based on the results reported in Liptow and Tillman (2012), Liptow et al. (2013) and Liptow et al. (2015)), the wood gasification route is the preferable option among the biomass and fossil routes (see Fig. 3, GWP_other and GWP_{fossil PE}). However, this changes dramatically when including the results from the GWP_bio and the WF methods. Both methods assess the impact of the route’s biogenic CO₂ emissions to be close to the overall impact of the fossil alternative. When adding the methods’ results for the impact of biogenic CO₂ to the route’s GWP from fossil CO₂ and other GHGs (GWP_other), its impact becomes even bigger than that of the fossil alternative (4.5 and 5 t CO_2,eq/t PE for the GWP_bio and WF methods, respectively, vs. 4.4 t CO_2,eq/t PE for the fossil alternative). A similar effect was found for the wood fermentation route (4.6 and 5 t CO_2,eq/t PE for the GWP_bio and WF methods, respectively). Only for the sugarcane route, the GWP does not increase using the GWP_bio or the WF methods and sugarcane becomes the preferable option in comparison to both the wood based and the fossil alternatives. The reason for the considerable impact of biogenic CO₂ emissions within the wood routes when applying the GWP_bio and the WF methods is the way the methods handle the release and uptake of CO₂ by biomass. They start with a full grown stand that is harvested and converted to products, which are eventually disposed of, causing emissions along the way. The re-growing plants eventually take up an equal amount of CO₂. However, since this takes time, the atmospheric concentration of CO₂ stays elevated and contributes to global warming. In the case of boreal wood, this re-growth takes around 100 years, causing an impact of 3–4 t CO_2,eq/t PE using the GWP_bio and the WF methods. In contrast, the sugarcane grows very fast, leading to an almost instantaneous uptake of emissions and hence an impact close to 0 t CO_2,eq/t PE.

Although the GWP_netbio method uses the same principle of temporal differences between CO₂ release and uptake, it leads to a smaller total GWP for the wood routes (2.9 t CO_2,eq/t PE for the wood fermentation route and 2.1 t CO_2,eq/t PE for the wood fermentation route). This is due to the consideration of avoided fossil emissions, which disguises the impact from the temporal differences in release and uptake of CO₂. As a consequence, despite their increase in GWP, the wood fermentation and gasification routes are still preferable in comparison to the fossil route. However, sugarcane is again the preferable option when applying the GWP_netbio method, and even leads to a negative total GWP (− 1.6 t CO_2,eq/t PE), again due to consideration of avoided fossil emissions.

To summarize, all three methods find the sugarcane route to be the better alternative for bio-based PE production. Furthermore, the results for the GWP_bio and WF methods are similar to one another and make the wood-based cases similar to or worse than the fossil alternative. This similarity is due to PE packaging being a short-lived product. However, for long-lived products, results will differ between the two methods—see the discussion on temporary carbon storage (see Sect. 4.5) for more details. Finally, the GWP_netbio method results in the same ranking for the biomass-based routes, but they have a significantly lower impact on global warming than the fossil route. This is due to the consideration of avoided fossil emissions in the GWP_netbio method. Moreover, the method also considers lost uptake, which together with avoided emissions cannot only lead to opaque results (if not well documented) but is also challenging to determine. This was particularly the case for the sugarcane based PE route. In our study, we estimated lost uptake for sugarcane to be close to zero. However, this estimate is not self-evident and could have been done differently, considering that the method was developed for forestry and provides no guidance on how to determine lost uptake for short rotation crops.

According to the CRP method, all of the bio-based routes perform better than the fossil route (see Fig. 3). However, the sugarcane route performs significantly better than the two wood routes (1.6 and 2.1 t CO_2,eq/t PE for the wood routes vs. 0.12 t CO_2,eq/t PE for the sugarcane route). This is in line with the results of the GWP_netbio carbon cycle method, which shows a similar impact for the wood based routes and a significantly smaller impact (even negative, in this case) for the sugarcane based route. The difference between the wood gasification and fermentation routes is explained by the fact that more wood is needed for the production of 1 t of PE via gasification than via fermentation. For the wood routes, the impact is fully due to occupational land use, since it was assumed that the transformation took place a long time ago and is already amortized. Furthermore, since there is no change in soil organic carbon during occupation (Chen et al. 2010; De Simon et al. 2012), all impact is related to changes in above ground carbon stocks. The higher impact values, when compared to the sugarcane route, can be explained by the slower growth rate of a boreal forest, which determines the requirement for land area (i.e. the land occupation inventory flow). The assumed carbon uptake rate in Swedish forest (1.1 t C/ha year) results in a land occupation flow of 0.9 ha year/t C. It should be noted however that regional differences occur: the actual carbon uptake rate (or forest growth rate) varies significantly between southern and northern Sweden (from 1.5 t C/ha.year in the south to 0.7 t C/ha.year in the north) leading to large differences in the resulting land occupation flow (Swedish Forest Agency 2014) and thus in the CRP results. The results shown in Fig. 3 reflect an average value. The CRP method points out that the sugarcane route is preferable over both wood routes with regard to land use. This is in spite of the carbon losses from vegetation and soil (58 t C/ha) during transformation (from Brazilian Cerrado, amortized over 20 year) and due to the fact that sugarcane has a much faster growth rate, resulting in a lower land occupation flow.

The results of the assessment with the GWP_soil method show that there is no impact for the wood-based routes and they are preferable over the fossil PE route (Fig. 3). This is due to the assumptions that land transformation took place a long time ago and that there is no change in the soil organic carbon content in a well-managed forest due to harvest. The sugarcane route was even found to result in carbon sequestration during land use at − 0.11 t CO_2,eq/t PE (transformation amortized over 20 years and occupation, taken together) (see Fig. 3), although the transformation from natural vegetation (Brazilian Cerrado) to sugarcane cultivation leads to a significant decrease in soil organic carbon content. However, due to carbon sequestration into the soil during the occupation (Anderson-Teixeira et al. 2009), this loss is compensated. Nevertheless, it would take close to 100 years to fully recover the soil organic carbon content due to this uptake. As a consequence, the choice of the amortization time of the land transformation impact may affect the impact result significantly. For example, if the land transformation impact was fully allocated to the first year of occupation, then the impact would be + 1.24 t CO_2,eq/t PE.

One of the methodological issues that has been brought forward for the assessment of land use and of time lags (as part of the carbon cycle methods) in LCA is the need to define a reference state. Reference states are used in all methods tested, however, more or less explicitly. In the methods underpinning the UNEP-SETAC guideline, here referred to as land use methods, the need for a reference state is made explicit and the guideline recommends the use of a biome-dependent (quasi-)natural state (Koellner et al. 2013) (p.1199). However, it is unclear whether this refers to Q_his (the historical land quality) or Q_rel (the land quality of the relaxed state) according to the terminology in Fig. 2. In comparison, Milà et al. (2007) recommend the relaxed state (Q_rel) as a reference state.

The determination of a reference state for land use was particularly challenging for the wood routes, since the carbon stock in the mature managed forest was found to be similar to the one for a mature natural forest. We found data stating that a mature, 100-year-old managed boreal forest has a carbon content of 40 t/ha, while a 200-year-old boreal forest has a carbon content of 50 t/ha (see Table 3). However, there is a variability in such data (not investigated in this paper). Since data for mature managed forests presumably are easier to find, it might be argued that the carbon stock in the mature managed forest could be used as a good enough approximation for the mature natural forest.

The carbon cycle methods also use reference states, even several of them. For instance, the GWP_netbio method uses two explicit references. One is the avoided alternative production of the product (see the description of the GWP_netbio method in Sect. 2.2), in which case, the practitioner is left with the decision on which avoided alternative to model, e.g. whether it is a crude oil-based or a coal-based alternative. The other one is the reference needed to calculate the lost uptake, which would have occurred had the feedstock been left standing. The definition of the latter was particularly challenging for the sugarcane case. When interpreting the GWP_netbio for sugarcane, we chose to include lost uptake for 1 year of occupation assuming that the occupation prevents the first year of re-evolution to Cerrado. Expressed in the terminology of the land use methods, occupation postpones the relaxation back to Cerrado. However, this interpretation of the GWP_netbio is not self-evident and could have been done differently. For example, the lost uptake due to missed re-evolution could have been assessed at a later point during the relaxation process. However, the method description provides no guidance on how to determine lost uptake for short rotation crops, and maybe the method was never even intended to be used for short rotation crops.

In addition to the reference states, there is also a less explicit point of reference in all of the carbon cycle methods. All three methods start their assessment with the harvesting of a mature stand and model the re-growth to this state. For forestry, we argue that this means that the fully grown forest stand is used as point of reference. However, as for example pointed out by Levasseur et al. (2013), this is not a self-evident choice for the point of reference, but rather depends on the goal and scope of the assessment. In their study, Levasseur et al. (2013) propose to use a ‘just beginning to grow’ forest as the reference point for the assessment of afforestation and related production. The methods we have investigated choose the mature forest as a point of reference, without making the choice explicit. This is in contrast to the land use methods, which are explicit about the need for and use of reference states. The value-based choice of a reference point for the carbon cycle methods (e.g. the choice between the fully mature forest or the just beginning to grow forest) can therefore be debated.

Since the definition of a reference state or point of reference is necessary for the land use and carbon cycle methods to obtain assessment results, the most applicable approach for a practitioner, for now, is the explicit statement of his/her choice. This promotes transparency and enables discussions about results with stakeholders, a process which eventually might lead to the development of conventions.

Related to the choice of the point of reference for the carbon cycle methods is the debate whether to use a stand or a landscape approach. All three tested methods use a stand approach, i.e. what is being considered is re-growth on the same plot as where biomass was harvested (see also the description of the GWP_bio method in Sect. 2.2). An alternative approach for managed forests would be to use a landscape approach, which could be done in two ways:

Recently, Cintas et al. (2017) demonstrated that the results of assessment at the landscape level depend on how the spatial system boundaries are defined, that is, whether these boundaries are constant or are expanding during an accounting period. Cintas et al. (2017) argue that using constant spatial boundaries is preferred because in that case all carbon flows in a landscape are accounted for and thus is a true landscape approach. This is contrary to Cherubini et al. (2013) (see (2) above) who assume that the spatial system boundaries expand during the accounting period (by offsetting the carbon emissions of a newly harvested plot) and thus account for all carbon flows in the landscape only at the end of an accounting period when all stands that comprise the landscape are included in the assessment.

Another methodological difference found between the carbon cycle and land use methods is the way time is considered. The carbon cycle methods are time-dependent. Particularly for the calculation of the GWP_bio and the GWP_netbio characterization factors, the consideration of the actual timing of carbon release and uptake is at the core of the method. The WF method is less time-dependent, since it considers the carbon release as occurring at the time of harvest, instead of during the rest of the life cycle.

In comparison, the land use methods have a static approach, using time as a means to distribute emissions due to transformation over several consecutive harvests through amortization, without any actual consideration of timing. For example, the GWP_soil as described by Brandão et al. (2011) amortizes emissions from land transformation over 100 years of production, a choice which appears as value laden as any other amortization period currently presented in the literature (see e.g. Cederberg et al. (2011)). It should be noted that the GWP_soil method differs from the UNEP-SETAC guideline (Koellner et al. 2013) in that it accounts for the carbon flows during the time of occupation, instead of accounting for delayed relaxation. Moreover, the CRP method uses a time horizon of 500 years to determine the climate impact of CO₂ emissions (i.e. the GWP₅₀₀), which seems rather arbitrary, in particular when considering the argument provided for this choice, namely that shorter time horizons would make land use look worse. The CRP method also does not explicitly treat the issue of how to amortize or allocate the impact of land transformation to several consecutive years of occupation, which is needed to operationalize the method. Milà et al. (2013), who used the CRP method in a case study, applied an amortization period of 20 years, which is also recommended by the UNEP-SETAC guideline (Koellner et al. 2013).

The issue of which carbon pools to consider in the assessment is another methodological challenge, at least for the carbon cycle methods. There is no clear description on which carbon pools to consider for these methods, and which ones are considered seems to be an issue of data availability. For example, Cherubini et al. (2012) consider both the below and above ground carbon, including the carbon in growing, as well as decaying biomass (e.g. forest residues left on site). However, this type of data might not be available for every case.

The land use methods tested give more clear directions. The CRP method takes above as well as below ground carbon pools into consideration. The GWP_soil method only considers soil carbon, however without any claim that this would fully account for the carbon balances of cultivation. Rather, Brandão et al. (2011) use the results of this method as a complement to a standard life cycle carbon balance in order to obtain a more complete picture of the carbon flows occurring in the system under study. It should be noted that the computational framework described in the UNEP-SETAC guideline assumes that land quality, which can be measured with the soil organic carbon content, stays constant during the time of occupation. However, the guideline acknowledges that land quality is not always constant during occupation (Koellner et al. 2013, p.1198, and indicated in Fig. 2 in this paper).

An issue particular for the carbon cycle methods is the consideration of temporary carbon storage in products, which is handled differently among the methods. For example, the WF method does not account for storage in products but models an immediate oxidation of the product (in our case study, the incineration of PE), co-occurring with the harvest. This implies that the life span of the product is ignored (or is set to 0). This makes the WF method less suitable for the assessment of, e.g. forest products used in construction which typically have a life span of many years. In contrast, the GWP_netbio and the GWP_bio methods consider when in time the product oxidises and cut-off emissions occur past the time horizon under assessment. The latter resembles the approach of the Lashof method, which is one of the methods used to assess temporary carbon storage in LCA (Levasseur et al. 2012). Including the timing of product oxidation arguably increases the adequacy of the assessment because a more realistic description of the life cycle of the product is used. When cutting off emissions at the time horizon of the assessment, this delay in the emission, or temporary storage, of carbon in a product leads to a lower impact during the accounting period.

A practical challenge encountered for all methods was data availability. Especially, for the application of the GWP_netbio method to the sugarcane PE route, data that would represent the re-evolution of a sugarcane field back to natural vegetation in a continuous manner could not be found, and a simplified estimate needed to be made. This is potentially not only an issue for sugarcane, but for short rotation crops in general. Therefore, although outside the scope of this study, further and more extended searches are recommended to explore potential data sources. Another challenge connected to data is their quality and consistency. For example, considering above- as well as belowground carbon could have a considerable influence on assessment results (Cherubini et al. 2012). Therefore, consistent and qualitatively comparable data are a necessary prerequisite to ensure the comparability of studies.

Considerable computational effort and related time needs were an issue for implementing the GWP_bio and the GWP_netbio methods. On the one hand, this might not be justifiable for everyday use where time is limited and pre-calculated characterization factors are an essential part of the LCA analyst’s toolbox. On the other hand, as pointed out by Helin et al. (2013), the use of already provided factors might not be suitable for every purpose, as conditions might be very different. For this reason, over the medium term, the calculation and provision of characterization factors for a broad spectrum of conditions might be worthwhile in order to foster the impact assessment of biogenic CO₂.

In addition, the acceptance of the assessment methods is a challenge related to application as well. We presented LCA results based on the tested methods to both industrial stakeholders and researchers developing biomass-based technology. They were rather sceptical towards the methods and questioned the results. Especially the carbon cycle methods were debated, in particular the single stand approach and use of the mature forest as a reference point. Both choices are clearly value laden and methodology developers should be open to discussion about them with other stakeholders. Moreover, there is the risk that if assessment methods are not accepted as legitimate or scientifically sound, they are not likely to be acted upon. Nevertheless, the need for discussion should be no excuse for not using debated methods, since they can help identifying new, potential environmental hotspots, which otherwise might develop to substantial threats.