Life cycle assessment (LCA) is a tool designed to quantify the environmental impact potential of products and services [1
]. Recent advances in the field of LCA, such as the inclusion of temporal dynamism [2
] and the coupling of LCA to urban metabolism [3
] increase the applicability of the LCA methodology. Dynamism in LCA allows for the quantification of impacts while taking into consideration changing background and foreground systems, e.g., amounts of renewable and fossil energy sources in the electrical energy mix of a specific location in the background, and improvement to processing technologies in the foreground. On the other hand, coupling urban metabolism to LCA allows for large-scale assessments that better predict large-scale consequences of implementing a change at regional scale. These advances are an especially important input that can help guide the transition into a sustainable bioeconomy, as they allow for prospective studies. LCA of production systems/technologies, such as various agricultural productions, e.g., wine, cereal, and meat, can benefit from applying some of the new developments, since the large inputs and outputs to these systems, most likely, will have great environmental implications when changes to the production are implemented.
By applying the TM-LCA framework, as used in this study, it is possible to assess said systems in the specific context of the region, i.e., taking into consideration the region’s infrastructure, feedstock availability and accessibility, and the technical feasibility of technology implementation. Assessing large systems, as mentioned above, can be approached by defining the geographical boundaries in terms of a “producer territory” [4
] so that the LCA can be applied for assessment of a delimited “territory”, e.g., wine-producing areas, within a broadly defined region, e.g., Southern France. The producer territory is thus defined as the area of interaction between the aggregated producers and other systems within the region. The TM-LCA framework reduces data demand by aggregating individual areas of the production of, for example, a specific product, supply chain or waste treatment technology, while ignoring unchanging background systems, i.e., only changes to the region interacting with the producer territory are assessed. At the same time, representativeness is increased by merging local inventory data from individual producers with regional and nation-wide data in order to fill in data gaps. In this way, an environmental performance improvement in the territory, due to, e.g., the implementation of a new technology or new management technique, can be quantified in the non-contiguous production area and is reflected in the results for the region. When combined with dynamic and prospective LCA [2
], this approach offers a comprehensive assessment that gives temporally and geographically resolved results. Moreover, it has the added utility of providing prospective insights that can more accurately support decision makers, production owners, and technology developers [4
A point of departure for many LCAs is a static product system, where, for example, technology A might be assessed against technology B for the making of a product. The static nature of LCA is problematic when applied to products or systems with long service lives [5
], due to inconsistencies in time horizons and changes in background systems [6
]. Previous work has demonstrated the importance of incorporating various types of dynamism into LCA, as this can significantly affect the results of the study [6
]. In this regard, it is possible to add dynamism to the various stages of the LCA in a consistent, systematic, and transparent manner, as outlined in [2
] and shown in various other publications [7
]. Following the TM-LCA framework, dynamism can be added in a consistent manner from the start, which provides added information regarding the sensitivity of the system to background changes. Real production systems are rarely static, and results based on static systems can sometimes exhibit rank reversal when compared to dynamic results [10
]. Thus, basing future decisions on static LCAs can result in building significant error into the models and associated results. Adding dynamic aspects to LCAs can increase the analytical accuracy of results [11
The added layers of information to the TM-LCA mean that the interpretation phase becomes more resource demanding. This can be eased by the use of extra tools, such as multi-criteria decision assessment (MCDA). Midpoint results for 18 different impact categories of an LCA are often difficult and time consuming to synthesize into clear and readily applicable decision support. When adding dynamism, this translates into temporally specific results for, e.g., each year of the time horizon, for each of the 18 impact categories. Out of the many MCDA methods that exist, one that has shown great capability in dealing with LCA results is Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) [12
]. The output from TOPSIS is given in the form of a single score performance index, which is used to derive preference between the scenarios being assessed. By checking a multiple criteria decision support tool used with equal weightings for all midpoint impact categories, it is easy to realize and visualize burden shifting amongst the midpoint impact categories, when used in conjunction with a visual inspection of internally normalized results. The MCDA approach is considered preferable, as using carbon footprint alone has been shown to give potentially misleading results [14
The present study’s goal is to implement an assessment based on the TM-LCA approach [4
] in order to provide a comparison of potential biorefinery choices for the treatment of agricultural residues. For the demonstration of TM-LCA, a biogas production scenario is compared to a scenario of combined biogas and Polyhydroxyalkanoates (PHA) production, which is currently being developed at pilot scale. Polyhydroxyalkanoates are naturally occurring polymers produced by a consortium of bacteria, which can feed on the volatile fatty acid (VFA) stream generated by the acidogenic phase of anaerobic digestion (AD) [15
]. PHA, which is also found as polyhydroxybutyrate (PHB), can be used to produce biodegradable plastic products. In this case, PHB production substitutes the production of polyethylene terephthalate (PET) or polylactide (PLA). The two biorefinery scenarios are modeled with dynamics built into both foreground and background systems. In the foreground system, dynamics are included as a yearly decrease in the amount of energy consumption needed to produce PHA. In the background system, the electrical energy mix, hereafter referred to as energy mix or energy grid, of both locations is varied yearly for a period of 20 years with four possible provision mixes for Oregon, and five possible choices of provision for the energy mix futures of France. The scenarios are then tested at a territorial scale as described above, i.e., processing all the feedstock in the region in the two geographically dissimilar production territories, to observe the effects of regional differences on territorial performance. Since the use of global warming potential (GWP) as a single indicator has been shown to provide potentially misleading results [14
], MCDA is applied in the interpretation phase to help ease the interpretation of results.
Results showed that the PHA scenarios outperformed the biogas-only scenarios in almost every impact category with a few exceptions (Figure 4
). Exceptions included the French energy scenarios for the Ionizing Radiation (IR) impact category and almost all scenarios for Land Use (LU), except in one instance, the Oregon Static scenario, where PHA-biogas performed better than biogas-only in terms of LU.
It is worth noting that in some of the impact categories the difference between the two scenarios is so small that, keeping in mind the considerable uncertainty of LCA results in general, it is fair to say that both PHA-biogas and biogas-only are essentially equal in terms of environmental impact. This is true for the Particulate Matter (PM), Fresh Water Ecotoxicity (FWE), Land Use (LU), Marine Ecotoxicity (MEtox), Marine Eutrophication (ME), Mineral Resource Scarcity (MRC), both Ozone Formation categories, Terrestrial Acidification (TA), and Stratospheric Ozone Depletion (SOD) impact categories. The remaining impact categories show a greater degree of difference, where it is clear that the PHA scenarios are generally preferable. Midpoint impact category results are presented as percent reduction in environmental impact from the implementation of PHA production in relation to biogas-only scenarios, for all energy provision scenarios. These are shown both for scenarios replacing PET with a ca. 93% RR and a 30% RR, to show the influence of RR in impact results (Figure 5
The first model run shown in Figure 4
has PET as the conventional polymer to be replaced by PHB. The model was checked to see if a different polymer substitution material would alter the results. It was found that a change to PLA as the polymer substitution material did not change the general ranking, but the magnitude of the difference between PHA-biogas and biogas-only, i.e., the advantage that PHA-biogas has over biogas-only, decreased. Figures and tables for the PHA-biogas results for PLA are shown in the SI (Supplementary Figure S4 and Table S7)
shows the difference between the PHA-biogas and biogas-only scenarios, i.e., PHA-biogas CO2
biogas-only, in CO2
-eq. For all 20 years, the PHA-biogas scenario induces greater savings than the biogas-only scenarios, which is why the results are always negative. Furthermore, the general negative slope of all scenario lines shows that as time progresses PHA-biogas becomes more attractive, inducing higher savings in comparison to biogas-only. More interestingly, it is possible to observe the difference between plans for energy grid development in the two locations. Hence, Oregon scenarios show a steeper slope, i.e., a drastic pull back from the use of fossil fuels and, more specifically, the use of coal. In contrast, the French slopes are less pronounced, as improvements to the grid are subtler because there is already a large share of non-fossil-based energy production in use in France. The difference between the two regions is larger at the beginning of the period, getting smaller in time as the grids progressively increase their share of renewable energy.
3.1. Sensitivity Results
The robustness of model results was checked by varying different parameters, as described in the methodology, Section 2.7
. After each change, indicators were checked with the TOPSIS and GWP single indicators, but for the most part, there was no change to the preference ranking of the scenarios, and combined PHA-biogas production continued to perform better. Thus, it can be said that the model results are robust in regards to the most influential parameters analyzed.
In more detail, changes to the replacement ratio (RR), i.e., the PHB: PET mass ratio that is allowed by different material properties, as discussed in Section 2.7.2
, was shown to be a moderately sensitive parameter. A 5% change in the replacement ratio lead to a 3–4% change in results for PHA-biogas with PET (Figure 7
), and a 2.5–4% change in results for PHA-biogas with PLA. Thus, it can be said that a general trend is observed of lower savings with lower RR (or higher savings with higher RR), while the effect of the change is nearly proportional to the change seen in the results.
The sensitivity to efficiency improvements for PHA-producing technology was also tested and it is shown in the SI, Figure S3
. This parameter was showed to have very little effect on overall model results, with GWP changing in the range of 0.1–1.5%.
Sensitivity of N2O Emission Factor
Cumulative global warming impacts switch from a savings inducing status to a burden inducing status when N2
O emission factors for the field application of digestate from [22
] are applied (Supplementary Figure S4
). However, the ranking between PHA-biogas and biogas-only stays the same, with combined PHA-biogas scenarios continuing to perform better than biogas-only scenarios. The results show that N2
O emissions play an important role, and considering the strong dependency on local conditions, they should as much as possible be spatially differentiated. The variability of N2
O emissions for the EFs employed can be seen in Figure 8
3.2. Territorial Scale Application
Application of the biorefinery alternatives at a territorial scale would lead to potential reductions in regional environmental impact. In order to give a measure of scale to the potential savings induced by the implementation of maximum (limited by feedstock availability) PHA-biogas production relative to biogas-only, the GWP impacts were normalized using planetary boundary carrying capacity-based normalization factors [41
]. Assuming a 985 kg CO2
eq. per person year (PY) carrying capacity (C.Cap) [41
], and assuming that PHA replaces PET with a 93% RR and that the PHA process improves in terms of energy efficiency at 1% annually, the production of PHA induces an average reduction in GWP impacts relative to biogas-only equating to nearly 1400 PY of C.Cap. When broken down by region, the French scenarios indicate an average relative maximum potential GWP saving of over 2400 PY of C.Cap, with Oregon exhibiting just over 80 PY of C.Cap in average relative maximum potential GWP savings. Using the same assumptions, except exchanging the replacement polymer with PLA production at a 64% RR, then the maximum implementation in France and Oregon of the PHA-biogas scenario induces an average annual potential relative GWP impact reduction of 493 PY of C.Cap when compared to production of biogas-only, with 871 and 21 PY of C.Cap in France and Oregon, respectively, see Table 3
Sensitivity Analysis of Transport at Territorial Scale
The importance of transport was tested via sensitivity analysis of different theoretical grape marc transport distances for both the biogas-only and PHA-biogas scenarios (Table 4
). For all scenarios, a 500 km transport distance results in overall elimination of environmental benefits, and at 200 km, transport of grape marc reduces average impact savings from the various biorefinery-region scenarios by 42.5% for all midpoint indicators. In terms of GWP, a 200 km transport distance induces impacts of a maximum of appx. 284% and a minimum of 68% of the magnitude of GWP savings without transport. At 50 km, all scenarios show reductions in GWP. At 100 km, all PHA production scenarios and France biogas-only scenarios induce GWP savings, while the Oregon biogas-only production scenarios eliminate the GWP benefit of implementing the biorefinery. Furthermore, if the introduction of centralized PHA-Biogas biorefineries were to induce transport of grape marc, relative to existing decentralized biogas production, then GWP savings are overwhelmed by the induced impact from transportation at any distance greater than appx. 125 km.
Overall, the model results obtained were robust and indicate that implementing PHA production technology is preferable to conventional anaerobic digestion, when the functional unit (FU) equals 1 ton of feedstock treated. Combined PHA-biogas scenarios, whether with PET or PLA as the replaced polymer, performed better across almost every impact category. This is largely due to the added benefit of replacing conventional polymers, which are associated with significant impacts. As evidenced by the replacement ratio (RR) sensitivity analysis, decreasing or increasing the amount of PHB needed to equate the function of PET or PLA resulted in an almost proportional effect in the outcome. RR of PET would have to decrease by around 80% and be as low as 20% before there is rank reversal between the two options in some of the impact categories. This was confirmed by both single score indicators, which prefer combined PHA-biogas scenarios until reaching values close to 20% RR (Table 5
). However, the GWP single indicator still preferred PHA-biogas, even at a 20% RR, except for the OR-Static Scenario. On the contrary, the TOPSIS single indicator, which is equally weighted between impact categories, starts preferring biogas-only scenarios earlier, with a 35% RR. In this regard, there was less operating space for the GWP indicator, when PLA is the replacement polymer, which starts signaling biogas-only as the preferred choice already at 30% RR. On the contrary, TOPSIS selects biogas-only at low RR of 9–16%. Thus, there is disagreement between the GWP and TOPSIS single indicators, which is, furthermore, replacement polymer-dependent. This points to two issues to consider: (1) choosing GWP as the only impact category for the assessment can potentially result in burden shifting to other environmental impact categories and (2) the choice of polymer substitution affects impact categories other than GWP, here exemplified by the difference in the TOPSIS results when choosing PET or PLA as polymer replacement. To elaborate, the difference lies in PET’s production being more burdensome for impact categories other than GWP in comparison to PLA’s production. However, the single score indicators employed generally indicated a similar scenario prioritization, i.e., combined PHA-biogas production being the preferred choice across all future energy scenarios, as long as RRs were higher than 20% for PET and 30% for PLA. It is worth noting that such a low replacement ratio is considered unrealistic, as the material properties of PHB allow for various applications [40
Much like with polymer replacement ratios, TOPSIS and GWP do not always agree when the limits of process energy consumption are tested. If process energy consumption reaches 134 kWh per FU of added energy demand for PHA production, then TOPSIS (unlike GWP) indicates preference for biogas-only, for all energy scenarios, which indicates there is a potential for burden shifting if GWP is chosen as the only indicator. However, unlike the replacement ratio, improvements in process energy consumption for the production of PHA lead to very small changes in results. If there is no improvement in process energy consumption, meaning production of PHA consumes 7 kWh more per FU than the biogas-only scenario, results still stay the same. The break-even point of energy consumption for PHA production is high, i.e., it takes 12 times this value, 85 kWh of added process energy consumption of PHA per ton feedstock, before the TOPSIS-derived single indicator shows preference for biogas-only over combined PHA-biogas production for several of the French energy scenarios and one Oregon scenario. Moreover, it takes 16 times this value, or 113 kWh/FU more, before it is possible to observe prioritization change for the GWP single indicator for one Oregon scenario, the OR-Static Scenario, and 32 times the initial value, 226kWh/FU, before all Oregon energy scenarios show a preference for biogas-only. As for France, it is not until PHA production consumes 55 times this value, 389 kWh/FU, before there is a change in the GWP single indicator in preference of one of the energy future scenarios; the FR-Static Scenario. Thus, it is possible to conclude that there is large leeway in process energy consumption for PHA production before the decision support will change, in terms of GWP. As exemplified here, this is also dependent on the share of renewable energy sources in the future energy grid, which is why results are more robust for France in terms of GWP, i.e., requiring 55 times, 7 kWh/FU, more energy consumption before seeing a change in GWP impact category. The energy prediction mix is thereby an important factor when deriving the impacts of the system, which are heavily affected by energy mix usage.
In this regard, using dynamic energy grids for the background is a powerful tool. Many nuances are highlighted and originate from the predicted/expected changes in the share of renewable energy for the different locations. The most obvious of these subtleties can be observed in the Ionizing Radiation category (Figure 4
), where it is evident that there is a higher share of nuclear energy in the French background system than in that of Oregon. As seen in Figure 6
, the evolution of the energy grid reveals a sharp decrease for Oregon, while France’s energy grid remains somewhat unaltered. This is due to legal requirements in Oregon, which are intended to increase the share of renewables from 15% to 50% by 2040 [28
]. Greening of the energy grids increases the difference between biogas-only and PHA-biogas in the future, as is exhibited by the negative slopes of the lines in Figure 6
. Despite the increasing environmental importance of plastic replacement as opposed to electricity replacement, it is worth restating that PHA-biogas is consistently preferable in terms of GWP, i.e., negative values throughout the assessment period. One major area discussion regarding the dynamic inventory is the use of local energy mix scenarios in commodity replacement. It is likely that the increased production of PHA would have no direct effect on the production of PET or PLA in Oregon or France. However, by using a local instead of global process, it is possible to develop processes that are treated equally, in terms of system dynamism, for their inventory development. Furthermore, this is seen as a cautious choice, as the localized dynamic processes for the replaced polymers exhibit lower impacts than the global average. Thus, it is possible that this inclusion slightly under-represents the potential impact reduction gains from increased PHA production and is hence considered unlikely to over-state impact reduction gains.
As shown in the sensitivity analyses, biogas-only scenarios are preferred only in extreme cases where polymer replacement ratio or consumption of energy during PHA production are set to extreme values, i.e., very low RR and very high process energy consumption for PHA. Another area of uncertainty is N2
O emissions after digestate application, which have also been shown to be highly uncertain in several LCAs [42
O emissions were shown to have the potential to induce impacts for all scenarios, though the ranking of PHA-biogas in relation to biogas-only was not affected. Due to the closeness in results from the field application of digestates generated from the model for biogas and PHA scenarios, it can be concluded that both digestates act more or less in the same way during field application. Results were also tested without the field emissions, leading to the same technology prioritization. Nevertheless, it is important to highlight the large impact that N2
O emissions have in assessing agricultural product systems, and the necessity to improve inventories of these emissions in LCA assessments. Incidentally, the TM-LCA framework advocates for the use of local inventory data as much as possible.
One area that is made evident by including the territorial assessment, where there is potential for inducing impacts that would eliminate the environmental benefits of the system, is transport. Due to the relatively low energy and chemical value density in grape marc, increases in present transport of grape marc greater than 200 km cause induced impacts in all biogas-only scenarios. When transporting grape marc 250 km, both PHA-biogas with PET replacement and biogas-only induce impacts, except for the PHA-biogas scenario with static energy grid in Oregon, i.e., a dirtier energy mix than impacts from transport. Furthermore, if the PHA-biogas scenario induces transport relative to the biogas-only scenario (no added transport for biogas-only), then 150 km of grape marc transport eliminates the GWP benefit of the PHA-biogas scenario. While the PHA production scenario remains clearly preferable to biogas-only in all transport scenarios, this result does underline the need to assess potential re-routing of the feedstock if a new biorefinery technology were to be implemented.
It is also notable that the present use of feedstock, omitted in the results of this study as the impacts would be equal in both the PHA-biogas and the biogas-only scenarios, varies significantly between the two assessed territories. In France, there is a well-established market for distillation of wine residues, and in Oregon the wine residues are often used as compost. This said, it is also important to highlight that the feedstock mix used in this assessment can also be changed, as the PHA-producing technology is compatible with all types of organic waste, e.g., the organic fraction of household waste, waste-water treatment sludge, other animal slurries, other crop residues etc. The option to change the feedstock mix was not investigated in this study, as it would change the functional unit and was thus omitted from the present work. However, it is quite possible that there is further exploitable feedstock in both assessed regions. A good indication of feasibility is if there is an industrial sized biogas plant already in operation in the region; this would indicate that there is already feedstock enough to run PHA production. However, it is important to keep in mind that the use of crops has not been investigated in this report and so this study’s conclusions do not apply if the feedstock is food crops.