Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Life cycle assessment of eucalyptus short rotation coppices for bioenergy production in southern France

Life cycle assessment of eucalyptus short rotation coppices for bioenergy production in southern... Introduction The recent European Directive on renewable energy set ambitious targets for all Member States, in order for the EU to reach a 20% share of energy from renewable sources by 2020 (European Commission, ). Among renewable energy sources, the biggest contribution (63%) may come from biomass, as suggested by a foresight analysis in Europe (European Commission, ). At present, biomass already contributes about 4% of the total EU energy supply, predominantly as heat, and combined heat and power applications to a lesser extent. The production of liquid biofuels for transport from biomass increased several fold in the last decade, and is currently a major issue. Among various sources of biomass (organic waste, forestry products, cereal straw, etc.), dedicated crops such as short rotation coppices (SRC) are currently being investigated. These systems involve the cultivation of a fast growing ligneous species with short to very short harvesting cycles. Species with a capacity to sprout after cutting are particularly interesting as they make it possible to harvest the same plantation several times over the lifetime of the trees. Eucalyptus ( Eucalyptus sp.) is one of the most widely known species used for biomass‐oriented SRC, particularly for pulp and paper industries (Iglesias‐Trabado & Wilstermann, ). Poplar ( Populus sp.) and willow ( Salix sp.) have been used more recently for energy purpose for example in northern Europe (Lindroth & Båth, ; Wilkinson et al ., ) or in Italy (Manzone et al ., ). In France, SRC were developed with poplar and eucalyptus in the mid 1980s on the initiative of pulp companies. Nowadays, some 2000 ha of pulp SRC are still present although only eucalyptus is still being used in the south‐western part of France with an average rate of 100–200 ha planted every year (Nguyen The et al ., 2004b). The typical plantation scheme is based on 10‐year rotations with a stand density of 1250 stems ha −1 (on a 4 × 2 m grid). Three harvests in 30 years are expected with an average productivity of 10 oven‐dry metric tons (ODT) ha −1 yr −1 with the currently used species: E. gundal , an hybrid between Eucalyptus gunnii and Eucalyptus dalrympleana (Cauvin & Melun, ). The recent drive for renewable energy sources and concerns with the sustainability of biomass production (Robertson et al ., ; Scharlemann & Laurance, ) have sparked interest for SRC given its presumed low environmental impacts as it requires less inputs than agricultural crops (WWI, ). The traditional pulp and paper SRC scheme may be directly transposed to biomass production for energy purposes. As SRC is expected to be mainly grown on former cropland, silvicultural schemes with shorter cycles than the traditional 10‐year pulp rotation are being investigated to be closer to arable farming systems. Growing cycles may be shortened to 7 years with the same productivity as long as stand density is kept within a 2000–2500 stems ha −1 range, as was already tested with poplar (Berthelot et al ., ). Similarly, SRC with shorter rotations with 3 year harvesting cycles are being tested and developed. This scheme was illustrated with willow (Dimitriou & Aronsson, ), and requires far higher stand densities, between 10 000 stems ha −1 and 15 000 stems ha −1 . Such systems are currently being trialled in France with eucalyptus and poplar. Several issues were raised regarding the environmental performance of energy crops, in particular their actual GHG benefits and impacts on water resources or biodiversity (Robertson et al ., ; Monti et al ., ). Here, we chose the LCA methodology to address these issues for eucalyptus SRC, because it is widely used for bioenergy assessment and is a multicriteria, holistic method (Blottnitz & Curran, ; Cherubini, ). No such assessments have been reported for eucalyptus SRC, to the best of our knowledge, although they exist for traditional eucalyptus forests (Lopes et al ., ; Jawjit et al ., ). There is also a growing literature on the LCA of other lignocellulosic feedstock, whether annual arable crops (Kim & Dale, ), perennial grasses such as miscanthus and switchgrass (Monti et al ., ; Shurpali et al ., ), or other types of SRC such as willow and poplar (Gasol et al ., ; Goglio & Owende, ; Djomo et al ., ), whose performance may be compared with eucalyptus. The objectives of this study were twofold: (i) to apply LCA to eucalyptus SRC in southern France, based on the currently existing pulp scheme, and extended to 3 year rotation coppices, and (ii) to investigate the possibility of including the temporary storage of atmospheric CO 2 in ecosystem carbon pools in the GHG balance of heat provision from eucalyptus SRC, following the approach suggested by Moura‐Costa & Wilson ( ) for forest products. Eucalyptus biomass was used to generate heat, and compared to equivalent fossil energy sources. Materials and methods The eucalyptus pulp SRC system was chosen as a basis for the study. This species and its silvicultural scheme have been studied in France for almost 30 years and many technical references already exist (Cauvin & Melun, ). This SRC was designed for pulp production, but may easily be extended to bioenergy production. Scope, functional unit and system boundaries for the LCA The function studied here is heat production from the combustion of SRC wood chips in a boiler. The functional unit selected was therefore 1 GJ of final heat, which means that life cycle impact indicators were calculated relatively to the production of 1 GJ of heat. The system studied is depicted on Fig. , and comprises five main stages: The production of cuttings from selected eucalyptus clones, which corresponds to current practices. It includes the production of mother trees in a biotechnology facility and transportation to a nursery. In the inventory, we used data pertaining to a research laboratory, therefore not designed nor optimized an industrial‐scale production of cuttings. Plantation establishment and removal, including site preparation, fertilization, plantation and weed control during the first 2 years, as well as stump removal at the end of the project. Harvest, including felling, forwarding and chipping for SRC and silage harvester for 3 year rotation schemes. This stage also includes the transportation of harvesting machines to the tree parcel. Transportation of wood chips from the collection site to the boiler. We used a distance of 80 km corresponding to the actual average distance between eucalyptus plantations and the pulp mill of Saint‐Gaudens (south‐western France). Handling and combustion of wood chips in a boiler. Calculation of the cumulative amounts of carbon stored in eucalyptus biomass over time in the 10‐year interval between two cuts. Management scenarios Management scenarios are detailed in Table . The reference scenario was the pulp SRC scheme based on three 10‐year harvest cycles (i.e. a total duration of 30 years), with a stand density of 1250 stems ha −1 . From this baseline, we designed a scenario dedicated to biomass production for energy by doubling the stem density (2500 stems ha −1 ) with three harvests every 7 years for a total duration of 21 years. A variant with 3 year harvesting cycles was designed with a density of 5000 stems ha −1 , which represents in the present context the maximum possible density considering the costs of eucalyptus cuttings. This scenario plans seven successive harvests over the same 21‐year time interval. Selected characteristics of the eucalyptus management scenarios Scenario name Characteristics Productivity (ODT 1 ha −1 yr −1 ) Fertilizer inputs (kg ha −1 yr −1 ) Duration Pulp SRC 1250 stems ha −1 S1 Chainsaw operator – Log harvest 11.7 N: 10P 2 O 5 : 8.7K 2 O: 14.8 3 × 10 years S2 Felling machine – Log harvest 11.7 N: 6.4P 2 O 5 : 7.8K 2 O: 10.1 3 × 10 years Energy SRC, low density 2500 stems ha −1 S3 Felling machine – Log harvest 11.7 N: 6.4P 2 O 5 : 8.3K 2 O: 10.1 3 × 7 years S4 Felling machine – Full stem harvest 14.0 N : 23.4P 2 O 5 : 11.2K 2 O: 25.2 3 × 7 years Energy 5000 stems ha −1 S5 Harvester – Full stem harvest 10 N: 40.0P 2 O 5 : 18.8K 2 O: 49.8 7 × 3 years ODT, oven‐dry metric ton. A set of technological variants technical aspects likely to influence LCA results were also considered: Harvest mechanization: approximately 50% of pulp SRC are currently harvested with felling machines rather than manual felling with chainsaws. Felling machines have a better productivity and make mechanical debarking possible in the field, which results in higher rates of nutrient returns to soils. On the other hand, felling machines consume more fuel and emit more GHGs. The 3 year old stands are usually harvested with adapted agricultural machines. Productivity: for SRC, a yield of 10 oven‐dry metric tons (ODT) ha −1 yr −1 considered as a robust average value taking into account the mortality of trees and their partial ground cover. It corresponds to a final cut at a diameter of 7 cm (commercial cut). The full stem harvest leads to an extra 20% of biomass, including leaves (Nguyen The & Deleuze, 2004a). For the second and third harvest, a 25% gain in biomass production is usually observed due to a faster growth (D. Lambrecq, Fibre excellence, Saint‐Gaudens, personal communication). For lack of consistent findings on the effects of rotation length on productivity, we assumed the same yield for all rotations (Berthelot & Gavaland, ). Fertilizer inputs: pulp SRC are currently not fertilized in France because it is not considered as a relevant operation for the sustainability of biomass production. Nevertheless, this is a very critical point, especially for the shortest rotations whose nutrient exports are expected to be significantly higher. Therefore, we assumed in all scenarios fertilizer input rates corresponding to the estimated exports of nutrients at harvest. The differences between scenarios were particularly acute across harvesting techniques, whether including debarking (with the mechanical harvest) or harvesting full stems or logs. Eucalyptus being an evergreen species, harvesting full stems rather than wood logs would lead to far larger nutrient exports because of the high nutrient contents of the leaves. The amount of N, P and K applied were calculated using state‐of‐the‐art knowledge and data on nutrient exports of eucalyptus SRC in France (Nguyen The et al ., , ) and atmospheric deposition rates (Croisé et al ., ). As a result of the above variants, a total set of five scenarios was implemented, whose characteristics are summarized in Table . LCA methodology The cut‐off threshold for neglecting system components was set at 3.6 × 10 −6 %. The production of laboratory equipment was excluded because cuttings production was only a marginal part in the use of this equipment over its total life cycle. The transportation of pesticides and fertilizers (N, P, K and Mg fertilizers in the nursery, herbicides for site preparation and plantation maintenance, field fertilization) were not taken into account due to a lack of accurate information. Chemical inputs in the nursery were exclusively attributed to the production of cuttings, except for fungicides and hormones, which were neglected due to the very low dosages used. Nursery propagators were also excluded due to the lack of information on this material (jiffy pellets made from peat). Neither waste nor coproducts are produced during the life cycle of SRC, which alleviated the need for allocations. As usually assumed in the LCA of bioenergy systems, the global warming potential of the CO 2 emitted during the combustion of biomass was considered nil (Cherubini, ). LCA calculations were done by using the TEAM 4.0 software package (Ecobilan‐PWC, Paris, France) with the EcoInvent 2000 database (V2.01, St‐Gallen, Switzerland). Field emissions related to the input of fertilizer N and P were calculated using the methods proposed in the Ecoinvent report (Nemecek et al ., ). However, the model proposed for nitrate leaching was found unsuitable for eucalyptus, and this flux was thus neglected. The leaching risk was low because fertilizers are usually applied in spring after the winter drainage, and taken up before the onset of drainage in autumn. In addition, nitrate leaching under forests is generally minimal (Galloway et al ., ). Impacts were characterized using the CML 2001 method, as described in Guinée et al . ( ), and the following categories considered: non‐renewable energy consumption, global warming (with a 100 year timeframe), acidification, eutrophication and photochemical ozone creation potential (POCP). Accounting for ecosystem C dynamics and land‐use changes In a first variant relative to our baseline LCA calculations, we investigated the possibility of accounting for the temporary storage of atmospheric CO 2 in the biomass of eucalyptus stands. The principle is to derive an equivalence factor with permanently stored CO 2 based on the cumulative radiative forcing of atmospheric CO 2 over time. Moura‐Costa & Wilson ( ) derived such as factor from the number of years over which the reduction in radiative forcing would be identical between the temporary and permanent storages. They estimated the duration for break‐even to approximately 55 years, yielding an equivalence factor of 1/55 or 0.0182. However, other factors are presently under discussion in relation to carbon trading. Two other factors were thus tested here: a coefficient of 1/26 proposed by the French Ministry for Agriculture (MAP, ), corresponding to an economical calculation involving an annual discount rate of 4%, and the 1/100 factor proposed by PAS (Bsi, ) for consistency with the IPCC time horizon in the climate change scenarios (2100). Note that some studies implicitly use an equivalence factor of 1, considering net annual CO 2 gains or losses from ecosystems as their direct contribution to global warming (Palm et al ., ; Shurpali et al ., ). Following the above approach, the temporary effect of C storage may be calculated as follows: Mitigating effect ( in t CO 2 eq . ) = Qc × T × EF where Qc is the amount of C stored in tree biomass (t CO 2 ha −1 ), T is the duration of storage (years), and EF the equivalence factor (unitless). For eucalyptus SRC, the carbon taken up by the stand during the first year of the rotation is stored until harvest, i.e. during 9 years; the C taken up during the second year is stored for 8 years, etc. The Qc × T component of the equation actually corresponds to the cumulative sum of C stored through time, except for the last year when the stand is harvested (Fig. ). System boundaries and steps of the life cycle. The C sequestration of eucalyptus SRC should be compared to a baseline scenario in terms of land‐use. Here, we chose abandoned agricultural land (referred to as wildland in the following), which typically occurs after vineyard removal in southern France. Eucalyptus SRC would therefore be established on former vineyards in our scenario, which excludes indirect land‐use change effects. The global C storage was therefore calculated by subtracting the C stock of SRC by that of wildland. The carbon stored in the aboveground biomass (AGB) of the eucalyptus stands was calculated from the C content of harvested wood, considering a C content of 47% (dry weight basis; Paixão et al ., ; Tanabe et al ., ). Belowground biomass (BGB) was estimated with an allometric relationship as a fixed proportion of AGB, set to 30% (Tanabe et al ., ). Note that BGB was kept during the 30‐year cycle of the eucalyptus plantation in the trees' rooting systems, whereas AGB dropped to zero after each cutting. For the baseline land‐use, the wildland's ABG was considered constant at 0.9 t C ha −1 yr −1 , which is the peak value for grasslands in warm temperate, dry climates given in the IPCC guidelines for GHG inventories (Tanabe et al ., ). It is in the lower end of the 0.8–3.2 t C ha −1 yr −1 range reported in Europe for former arable fields up to 3 years after abandonment (Hedlund et al ., ), i.e. in the early years of fallow regeneration. Belowground biomass was fixed at 2.0 t C ha −1 yr −1 (Tanabe et al ., ), which is slightly lower than the 2.5–3.5 t C ha −1 yr −1 range in annual returns to soils estimated in the classical Rothamsted (UK) long‐term wilderness experiments, where arable fields were allowed to undergo natural woodland regeneration in the 1880s (Jenkinson et al ., ). In the beginning of the transition from arable to wildland, only herbaceous species are involved and their net annual biomass production is entirely returned to soils as litter. Further on during the 30‐year life cycle of the eucalyptus plantation, it is likely that some woody species may also appear in the wildland and start accumulating biomass from 1 year to the next, although the exact dynamics of that transition has not been documented to the best of our knowledge. Over a longer time‐frame, observations in the ‘Geescroft wilderness’ experiment in Rothamsted (UK), an arable field allowed to undergo natural woodland regeneration in 1885, may give us some insight into this process and provide an upper limit for this component. In this plot, the accumulation of AGB was estimated at 0.6 t C ha −1 yr −1 over the first 100 years of the transition (Grogan & Matthews, ), which we considered as the upper limit of what would happen in the first 30 years of wildland growth after abandonment (the lower limit being no accumulation at all). A below to AGB ratio of 1 : 3 was assumed for the wildland (Grogan & Matthews, ), which is similar to the value used for eucalyptus trees. It is likely that the differences in soil organic carbon (SOC) will appear between the SRC eucalyptus and the baseline land‐use over time, due to differences in litter and belowground inputs (Grogan & Matthews, ). However, as eucalyptus SRC systems are relatively recent, there are no long‐term experiments documenting the dynamics of SOC after conversion to eucalyptus, let alone comparing them with other land‐uses such as arable farming or wildlands. We therefore elected to exclude differences in SOC between eucalyptus SRC and wildland in our analysis. The effect of this hypothesis is addressed in the section. Results LCA results Life cycle consumption of non‐renewable energy ranged from 77.0 to 92.7 MJ GJ −1 heat output from eucalyptus biomass (Table ). It was lowest for the S1 scenario with lower stem density and manual harvest, and highest for the shortest rotation (S5). In all scenarios, wood chips transport represented the main energy consumption hotspot with a share of 46–55%. Harvesting operations came second with 30–36% of total energy consumption, except for the very short rotation scenario, where their share was only 3.3%. This is due to the use of an adapted silage harvesting machine instead of heavy, fuel‐consuming forestry machines. In scenario S5, the most important steps were fertilization and plant production. Fertilizer inputs were larger than with the SRC schemes because the harvest of whole stems including leaves lead to higher nutrient export rates and enhanced fertilizer requirements. Stem density is also twice higher in the S5 scenario compared to the energy SRC (S3 and S4). This had a significant impact on energy consumption because the production of cuttings takes place in an energy‐intensive biotechnology laboratory. The shorter rotations and higher stem densities of S5 further enhanced this trend, making this scenario the most energy‐intensive. Its energy ratio (ratio of heat output to fossil energy inputs) was also the lowest of all scenarios, at 10.8. This ratio increased with decreasing harvesting frequency, leading to the pulp scheme achieving the highest value (13). Non‐renewable energy consumption per life cycle stage of the various SRC systems (MJ GJ −1 ), and ratio of energy delivered to primary energy consumption Scenario Cutting production Site prep. Fertilization Harvest Transport Boiler Total Energy ratio S1 1.84 2.96 5.69 23.52 42.67 0.29 77.0 13.0 S2 1.84 2.96 3.99 27.30 42.67 0.29 79.0 12.7 S3 5.24 4.23 4.00 27.30 42.67 0.29 83.7 11.9 S4 4.34 3.51 9.47 27.28 42.67 0.29 87.6 11.4 S5 12.85 5.18 28.63 3.08 42.67 0.29 92.7 10.8 Life cycle GHG emissions (excluding ecosystem C pools) varied in a narrow range for the four SRC scenarios, from 8.2 (S1) to 8.5 (S4) kg CO 2 ‐eq. GJ −1 (Fig. ). They were 50% higher for the S5 scenario, due to its requiring two to three times more NPK fertilizer inputs than the other schemes, altogether with a 20–30% lower productivity (Table ). The relative importance of the various steps of the life cycle followed a similar pattern for all scenarios except S5, with an important contribution of fertilization (38–44% of total), transport (32–33%) and harvest (18–22%), in agreement with the review of Djomo et al . ( ) for willow and poplar SRC, and of Rowe et al . ( ) for woody crops. The 3 year rotation scenario (S5) had lower emissions than the longer rotation scenarios in the harvest step due to the use of agricultural harvesters, and plant production and site preparation emitted more GHGs than harvest. Fertilization accounted for 68% of the total emissions for this scenario. Life cycle assessment results for global warming, eutrophication acidification, photochemical ozone creation potential, per GJ of heat delivered. Indicators for the eutrophication impact ranged from 48 (S2) to 152 (S5) g PO 4 2− ‐eq. GJ −1 (Fig. ), and were dominated by the fertilization phase. The losses of P from the plantation by run‐off and erosion made up 90% of the impact related to fertilization, whereas ammonia volatilization contributed the remainder, the impacts of NO emissions from soils being negligible. Because of its larger fertilizer requirements, the very short rotation system had nearly threefold higher eutrophication impacts than short rotation ones. Although the latter also received varying rates of fertilizer inputs (Table ), differences in productivities compensated for these variations and all short rotation schemes had a similar eutrophication impact within a 5% relative range. Interestingly, the best scenario was the one with the highest biomass productivity (S4) and not that with the least fertilizer inputs per ha (S2), which only achieved a mid‐range performance. The acidification indicator ranged from 39 (S1) to 110 (S5) g SO 2 ‐eq. GJ −1 , following a pattern similar to eutrophication (Fig. ). The very short rotation scenario had again a threefold larger impact than the other scenarios, and for the same reason: its higher fertilizer inputs, which translated in higher field emissions of ammonia and nitric oxide, and indirect emissions due to fertilizers' manufacturing. However, the harvest and transport steps played a more important role than for eutrophication, and the breakdown differed between the scenarios. The share of harvest ranged from 20% to 30% for the SRC, although it was nearly negligible (at 2%) for S5. This stems from the major advantage of this scheme, namely the use of agricultural machines in lieu of forestry ones, which are far more resource intensive ones. However, the associated savings did not compensate for the large requirements of synthetic fertilizer inputs compared to scenarios with longer rotations. The POCP indicator ranged from 2.4 (S5) to 6.8 (S1) g C 2 H 2 ‐eq. GJ −1 , with harvest operations and wood chips transport contributing the most (Fig. ). The much higher emissions of photo‐oxidants occurring with the scenario S1 is explained by the chainsaws used for manual felling. The chainsaws used in France are seldom equipped with catalytic exhaust pipes and release volatile organic compounds, which have a high potential for ozone formation. These emissions also occur to a lesser extent with the mechanized felling option (in scenarios 2–4) because chainsaw operators are necessary for the second and third harvest to thin the coppice before felling machines can be used. The distance between the plantation and the boiler was set at 80 km in the baseline calculations. Table illustrates the influence of this distance on the five LCA impact categories for scenario S1, from a minimum value of 10 km to the baseline value of 80 km. Energy consumption was the most sensitive indicator: it dropped by 28% when halving the transport distance, whereas GHG emissions and acidification impacts were only reduced by 16%, photochemical ozone formation by 10% and eutrophication by 3%. The energy ratio increased from 13.0 to 18.0 when the transportation distance decreased from 80 to 40 km, and reached 25.2 with a 10 km distance (Fig. ). The other indicators were less sensitive to this parameter, Energy ratio as a function of the transportation distance from the eucalyptus plantation to the boiler for scenario S1. Influence of woodchips transportation distance from plantation to boilers on life cycle assessment indicators for scenario S1, per GJ of heat Transportation distance (km) 80 40 20 10 Non‐renewable energy consumption (MJ) 77.0 55.6 45.0 39.6 Acidification (g eq. SO 2 ‐eq.) 41.7 35.0 31.7 30.0 Eutrophication (g PO 4 ‐eq.) 52.0 50.5 49.8 49.4 Photochemical ozone formation (g C 2 H 2 ‐eq.) 6.8 6.1 5.7 5.5 Global warming (kg CO 2 ‐eq.) 8.16 6.80 6.11 5.77 A comparison with fossil energy sources was carried out to assess the environmental advantages and drawbacks of using SRC biomass as a substitute to coal, fuel oil and natural gas (Fig. ). In all scenarios, the provision of heat from SRC biomass consumed 90% less fossil energy than when using fossil energy sources. Similarly, GHG emissions were reduced by more than 80% with the SRC biomass. However, the patterns with the local to regional‐range impacts (acidification, eutrophication and photochemical ozone formation) were less clear‐cut. Biomass‐derived heat had generally much lower acidification and photochemical ozone formation impacts than fossil‐based heat except with natural gas, which out‐performed the S5 scenario for eutrophication and scenario S1 (pulp SRC with manual felling) for ozone formation. Natural gas had two to 30 times lower impacts than the other fossil sources, especially coal. Conversely, the eutrophication impacts were in the 50–135 g PO 4 3− ‐eq. GJ −1 range for the eucalyptus scenarios, and in the 5–40 g PO 4 3− ‐eq. GJ −1 range for the fossils, pointing to a weakness of the biomass‐based chain. The twofold higher eutrophication impacts of the S5 compared to the other scenarios were clearly due to the larger fertilizer inputs required by the former. Life cycle assessment indicators weigthed by the average impact of an European inhabitant and compared to fossil energy sources. Inclusion of ecosystem C dynamics Figure depicts the dynamics of aboveground and belowground biomass in the eucalyptus plantation (scenario 1) and the baseline wildland representing the baseline alternative land‐use. Over the 30‐year period of the eucalyptus life cycle, the average biomass of eucalyptus plantations was several fold larger than that of the wildland, even when considering the appearance of ligneous species in the latter. This hypothesis had a significant impact as it leads to a eightfold higher estimate of total biomass after 30 years compared to a wildland solely composed of annual species. The larger biomass accumulation in the eucalyptus SRC was due to a higher net primary production and an important storage in the belowground compartment, which kept increasing through the cuts. When averaged over the 30 years of the SRC rotation, the differences between SRC and wildland range from 16 to 24 t C ha −1 for the AGB, and from 26 to 37 t C ha −1 for the total biomass (Fig. ). These gaps represent the net ecosystem CO 2 gains incurred when substituting wildland with SRC, for instance after the abandonment of a vineyard. They are related to land‐use effects and may be included in the life cycle GHG emissions of eucalyptus biomass production by using equivalence factors to account the temporal value of C sequestration in the biomass. This led to savings of 0.57–5.16 t CO 2 ‐eq. ha −1 yr −1 (Table ), depending on the equivalence factors and the carbon pools taken into account. Including these CO 2 savings, the LCA of eucalyptus‐derived heat offset GHG emissions by 70–400% (Fig. ), and therefore had a large impact on the global warming indicators. With the most favourable equivalence factors (1/26 and 1/55), the C stored in eucalyptus biomass resulted in heat provision being a net GHG sink. Dynamics of aboveground (top) and above and belowground (bottom) C storage by pulp eucalyptus SRC (solid line) and wild land with (dotted line) or without (dashed line) consideration of C accumulation in woody species, in the years following conversion to SRC . Greenhouse gas emissions (g CO 2 eq. GJ −1 heat) due to sowing and harvesting operations, fertilization and transport of chips, and CO 2 savings from CO 2 sequestration in ecosystem biomass using various equivalence factors and the lower and upper estimates. Carbon storage in the eucalyptus SRC stands (management scenario 1), relative to the baseline wildland, as averaged over the 30‐year duration of the project, in the aboveground and above and belowground biomass pools (t CO 2 ha −1 ). The lower end of the range corresponds to the emergence of woody species in the wildlands, which is ignored for the upper‐end value. C stored in biomass pools are transformed into CO 2 sequestration rates using the three possible equivalence factors detailed in the text Ecosystem pools Equivalence factors 1/26 1/55 1/100 Aboveground biomass 2.21–3.43 1.05–1.62 0.57–0.89 Aboveground and belowground biomass 3.67–5.16 1.73–2.44 0.95–1.34 Discussion Benefits and drawbacks of eucalyptus SRC Substituting fossil sources with biomass from eucalyptus SRC leads to an 80–90% abatement of life cycle GHG emissions and fossil energy consumption per MJ of heat supply, for all SRC management scenarios. These figures confirm the strong benefits of bioenergy chains and are consistent with other LCAs of heat from biomass. For instance, Reinhardt ( ) reported a 95% abatement in GHG emissions and energy consumption when displacing oil or natural gas with short rotation willow for district heating in several European countries. In addition, inclusion of the temporary storage of CO 2 in the plant biomass, which was ignored in previous literature, more than doubled the GHG savings compared to those in fossil sources. The relevance of this hypothesis is discussed in the subsection . Conversely, the benefits of SRC were far from obvious for the other impact categories, especially when displacing natural gas, which had threefold to fourfold lower impacts per functional unit than the other fossil sources. This trade‐off between global impacts (global warming and fossil energy consumption) and local impacts has often been reported for bioenergy chains (Reinhardt, ; Gabrielle & Gagnaire, ), and is almost inevitable because of the gaseous and leaching losses of nutrient occurring upon the feedstock production phase. Despite the relatively low fertilizer N requirements of eucalyptus stands compared to arable crops, none of the management scenarios achieved lower eutrophication impacts than the fossil‐based alternatives. Furthermore, the impact estimates were conservative because some losses of nutrients were neglected, as discussed in the subsection . In terms of management scenarios, the very short rotation scenario (S5) was outperformed by the other scenarios for all impact categories except ozone formation, by a factor of 50–250%. As the economics of this system is also unfavourable (Nguyen The et al ., ), S5 does not emerge as a good candidate compared to short rotation scenarios. Thus, the benefits from a quicker biomass growth and simplified harvesting made possible by the 3 year growing cycle of S5 were outweighed by their larger fertilizer input and stem density requirements. The only advantage of S5 appeared in the POCP, in which harvesting operations were predominant. Note that the productivity estimate of S5 was rather conservative, being 17–40% lower than that for the longer rotation scenarios, for lack of consistent field data (Berthelot & Gavaland, ). However, assuming a similar or even slightly higher productivity for S1 would not change the overall ranking because the relative differences between S1 and the other scenarios were larger than 50% for most impact categories. To our knowledge, no LCAs have been carried out so far on eucalyptus SRC, whether for energy or pulp and paper. Our results may still be compared with those pertaining to traditional eucalyptus plantations published by Jawjit et al . ( ) in Thailand. Their study used system boundaries and characterization factors similar to those of ours, but found much lower impact values in general. Plant‐gate life cycle GHG emissions were estimated at only 3.1 kg CO 2 ‐eq. GJ −1 , compared to the 8–12 kg CO 2 ‐eq. GJ −1 range we obtained here. The acidification impact was 22 g SO 2 ‐eq. GJ −1 in the Thailand study compared to our 40–110 g SO 2 ‐eq. GJ −1 range, whereas the photochemical ozone formation potential amounted to 1.6 g C 2 H 2 ‐eq. GJ −1 in Thailand compared to our 2.5–7.0 g C 2 H 2 ‐eq. GJ −1 range. Eutrophication was an exception with similar impacts between Thailand and France, at 41 g PO 4 2− ‐eq. GJ −1 and an average of 50 g PO 4 3− ‐eq. GJ −1 for the SRC systems respectively. Some of these discrepancies are explained by the higher yields of 17.4 ODT ha − 1 yr −1 achieved by eucalyptus under the tropical conditions of Thailand, compared to the 9.5–14 ODT ha −1 yr −1 range assumed here. The eutrophication impact was relatively higher because 35–20% of the fertilizer N and P applied was supposed to leach to water bodies in this Thailand study, whereas those losses were neglected here, as they were in other LCAs on herbaceous and tree species for lack of specific references (Gasol et al ., ; Monti et al ., ). Our results on eucalyptus SRC may be more broadly compared to other lignocellulosic feedstock: willow in France (Reinhardt, ), Italy (Goglio & Owende, ) and Europe (Djomo et al ., ), poplar SRC in Italy (Gasol et al ., ) and Europe (Djomo et al ., ), reed‐canary grass in Finland (Shurpali et al ., ) and four perennial grasses in Italy (Monti et al ., ). All these studies used similar system boundaries with the exception of the combustion step, and relied on the same set of characterization coefficients (from Guinée et al ., ). Most of them also used the EcoInvent database for the life cycle inventory. Compared to the poplar SRC system assessed by Gasol et al . ( ) in Italy, the production and harvest of eucalyptus biomass consumed 1.8–2.5 more primary energy, essentially because the harvest was threefold less energy‐intensive per ton of biomass than eucalyptus (for scenarios 1–4) or because poplars required fourfold less fertilizers (for scenario 5). Also, the data on fuel consumption by farm machinery were adapted from the EcoInvent database based on local records, but the exact corrections were not given by the authors. When including the transportation of wood chips, albeit with a shorter distance than our nominal hypothesis (25 vs. 40 kms), the GHG emissions of poplar totalled 1.93 kg·CO 2 ‐eq. GJ −1 which is four to six times less than our 8–12 kg CO 2 ‐eq. GJ −1 range for eucalyptus. The gap was even wider for the other impact categories: the eutrophication impact of poplar was estimated at 3.4 g PO 4 3− ‐eq. GJ −1 vs. 40–135 g PO 4 3− ‐eq. GJ −1 for eucalyptus; acidification amounted to 15.7 g SO 2 ‐eq. GJ −1 vs. 40–110 g SO 2 ‐eq. GJ −1 for eucalyptus; and POCP totalled 0.3 g C 2 H 2 ‐eq. GJ −1 for poplar compared to 2.4–7 C 2 H 2 ‐eq. GJ −1 for eucalyptus. Besides differences in management and inventory data for farm machinery, these large discrepancies arise because direct field emissions contributed only a minor share of the impacts in the Gasol et al . study, whereas they predominated in our LCA. There are reasons to believe that some of these emissions were somehow underestimated: for instance, N 2 O emissions from Gasol et al . were similar to our estimates on a ha basis, whereas NO emissions were twofold lower. This contradicts current literature, which indicates that NO and N 2 O emissions fall within a similar range (Stehfest & Bouwman, ). Our estimate of N 2 O emissions also included background emissions (i.e. nonanthropogenic) and the contribution of eucalyptus residues. Our LCA results for eucalyptus are overall closer to those reported by Reinhardt ( ) and Goglio & Owende ( ) for short rotation willow in Germany and Ireland respectively. These authors reported energy consumptions of 33 MJ GJ −1 heat and 56.4 MJ GJ −1 , respectively, compared to our 55.6 MJ GJ −1 figure for scenario 1 (S1) with a similar transportation distance (40 km). The lower figure from Reinhardt ( ) was due to a less energy‐intensive harvest for willow, whereas the Goglio & Owende ( ) study involved a drying phase prior to combustion. Our values also fell in the middle of the 12.6–76.9 MJ GJ −1 range reported by Djomo et al . ( ) across Europe for willow and poplar SRC. GHG emissions were very similar, at 7.13 kg CO 2 ‐eq. GJ −1 for willow in Germany vs. 6.80 kg CO 2 ‐eq. GJ −1 for the S1 eucalyptus system here, whereas the eutrophication impact for willow was 94 g PO 4 3− ‐eq. GJ −1 , well within the 40–135 g PO 4 3− ‐eq. GJ −1 range reported here for our systems, although it should be noted that the estimation of nitrate and phosphate losses was not explicitly described in the willow study. Finally, the acidification emissions of willow in Germany totalled 174 g SO 2 ‐eq. GJ −1 , compared to a 40–110 g SO 2 ‐eq. GJ −1 range for eucalyptus SRC. This is probably due to higher combustion emissions of acidifying compounds in the Reinhardt ( ) study than listed in the EcoInvent database, which pertains to more recent technologies. For the same reason, POCP impacts were also larger with willow, at 18 C 2 H 2 ‐eq. GJ −1 in comparison to 6.1 C 2 H 2 ‐eq. GJ −1 g for the S1 system. Lastly, eucalyptus SRC may be compared to the range of perennial grasses assessed by Monti et al . ( ), involving miscanthus, switchgrass, cynara and giant reed, with a cradle to farm‐gate system boundary. Energy consumption ranges from 33 to 142 MJ GJ −1 biomass energy content, compared to approximately 35 MJ GJ −1 for eucalyptus SRC (Table ), putting the latter on par with the best performers – giant reed and miscanthus. However, their GHG emissions were significantly lower, at 1.75 kg CO 2 ‐eq. GJ −1 compared to 5.5–9.4 for kg CO 2 ‐eq. GJ −1 eucalyptus. The same applied to eutrophication impacts, ranging from 4 to 20 g PO 4 3− ‐eq. GJ −1 for grasses and from 45 to 132 g PO 4 3− ‐eq. GJ −1 for eucalyptus, and also to acidification impacts, which are 2–2.5 lower for the grasses than eucalyptus. As with the Gasol et al . ( ) study, it may be that field emissions were undervalued, because fertilizer N input rates were rather higher than the eucalyptus SRC systems (at 80 kg N ha −1 yr −1 compared to a 6–40 kg N ha −1 yr −1 range for eucalyptus). The Monti et al . ( ) article does not mention direct emissions of nitrate or P in the field. Because of differences in local contexts, in the sources of life cycle inventory data and estimation methods for field emissions, it is not possible to directly compare the eucalyptus systems tested here with other coppices or herbaceous plants as these differences are likely to overrule the differences between feedstock per se . With the exception of the Gasol et al . ( ) study, the LCA indicators of eucalyptus were within the range of impacts reported for other lignocellulosic feedstock, but no robust patterns emerged in terms of ranking with other species. Uncertainties in the life cycle inventories Field emissions are particularly difficult to correctly address in the LCA of agricultural or forestry systems as they depend to a large extent on local conditions (soil properties, climate) and on their interactions with management practices, which govern the fate of chemical or organic inputs. As very little data on field emissions have been published for eucalyptus SRC in temperate zones, we used estimation methods developed for other species, or assumed that some emissions were negligible. Such was the case for nitrate leaching and P losses, which may have lead to an underestimation of eutrophication impacts. Lopes et al . ( ) found these emissions negligible in their LCA of eucalyptus‐derived paper, and so did Jawjit et al . ( ) although their estimates of nitrate and phosphate emissions from eucalyptus plantations were rather large: they assumed that 35% and 20% of fertilizer N and P inputs were leached to water bodies, respectively, according to the 1997 IPCC guidelines for GHG inventories. The 35% emission factor for nitrate (which was revised to 30% in the 2006 IPCC guidelines – Tanabe et al ., ) should in principle apply to managed forests, but no reference specific to forest or energy plantation is given in the literature base that served to determine this value. Further research is therefore warranted to provide a more accurate estimate of nitrate leaching for eucalyptus SRC. The same applies to P losses, and also to gaseous emissions of N 2 O, NH 3 and NO. The latter were calculated according to the 2006 IPCC guidelines for managed ecosystems, using default emission factors which are characterized by a large uncertainty range (Stehfest & Bouwman, ). Unfortunately, no literature data were found for eucalyptus SRC or forests in Europe to refine those estimates. Relevance of including ecosystem C dynamics Accounting for variations in ecosystem C stocks, compared to the alternative land‐use (wildland in our case) had a drastic effect on the GHG balance of eucalyptus‐derived heat, whose magnitude depended on the factor chosen for the equivalence between C stored in ecosystem pools and atmospheric CO 2 . Even when using the most conservative value of 1 : 100 (i.e. that least favourable to eucalyptus), ecosystem C pools offset GHG emissions by 50–70%, depending on the inclusion of belowground biomass. This made net eucalyptus a nearly carbon‐neutral source of heat, and stresses the influence of ecosystem C dynamics in relation to land‐use changes (LUC) in LCAs, already noted by Ndong et al . ( ) for biodiesel from jatropha in West Africa, and Shurpali et al . ( ) for reed‐canary grass in Finland. Note that the latter authors effectively used an equivalence factor of 1 : 1, as they used measurements of net ecosystem exchanges of CO 2 over reed‐canary grass, as cumulated over 1 year, as a measure of the C sink strength of the field where this crop was grown. Such hypothesis was also implicit in the GHG budgets of farmland and woodland management computed by Palm et al . ( ) in two villages in Africa, or by Ceschia et al . ( ) for cropping systems across Europe. In both references, ecosystem C fixation was put on a par with CO 2 emissions from fossil sources or N 2 O emissions from soils. This may be justified on a short‐term basis, but is misleading in the long‐run as most of the C taken up by ecosystems on a given year will be released back to the atmosphere after a few years as it enters fresh organic matter pools with rapid turnover (Jenkinson, ). From a life cycle perspective, whereby one attempts at estimating the cumulated past and future effects of substituting one product by another, using such an hypothesis would have overemphasized the sink capacity of SRC stands compared to wildland, and given wrong results on the actual GHG benefits of eucalyptus biomass. The use of equivalence factors, which are up to two orders of magnitude lower, is thus fully justified. Of course the magnitude and direction of this effect strongly depends on the LUC hypotheses made in the LCA. Adverse effects were conversely noted for biofuels when including indirect land‐use change effects whereby the displacement of food crops for biofuels in the United States entailed the conversion of natural ecosystems to arable farming in other parts of the world (Fargione et al ., ). Our scenarios for eucalyptus growth did not involve such effects as they considered the farming of eucalyptus SRC as an opportunity to value former arable land or vineyards that had been abandoned because of a drop in the market prices of wine. Assuming the resulting wildland to be reverted back to cropland, indirect LUC would apply. Using a value derived from a recent meta‐analysis on such effects for biofuels (De Cara et al ., ), we estimated this would cause an extra emission of 6.8 kg CO 2 ‐eq. GJ −1 heat, or 80% of the attributional life cycle emissions of the chain. This effect is similar (although opposed) to the inclusion of ecosystem pools and deserves further attention. Soil organic matter (SOM) pools were not included in the ecosystem pools for lack of robust estimates of SOM variations under both eucalyptus SRC and wildland. This pool was actually responsible for most of the land‐use offset of GHG emissions in the LCA of Jatropha by Ndong et al . ( ). Similarly, given the differences in net primary production between the SRC stands and the wildland, it is likely that the former have a higher SOM content than the latter, and therefore further accrue their GHG benefits. Grogan & Matthews ( ) thus argued from a very preliminary modelling study that ‘short‐rotation coppice systems have the capacity to sequester substantial amounts of carbon, comparable to, or even greater than, an undisturbed naturally regenerating woodland’. This results from C inputs from SRC being higher than from the regenerated woodland, which is comparable to our wildland system here. Field samplings were carried out in our study area to estimate SOM contents under vineyards, eucalyptus SRC of various ages, wildlands and arable land. Although the comparison was confounded by soil clay content, SOM was clearly lowest under the vineyards and comparable between wildlands and SRC. Conversion shortly after vineyard abandonment would therefore maximize the benefits of eucalyptus SRC in terms of SOM gains from land‐use change. Grogan & Matthews ( ) estimated that willow SRC sequestered 0.11 t C ha −1 yr −1 compared to abandoned cropland, which would translate for eucalyptus SRC as an additional offset of 2.5 kg CO 2 ‐eq. GJ −1 heat, or 30% of the GHG emissions of the chain. Further work is nevertheless required to provide more robust estimates of these potential gains. Acknowledgements These results were obtained in the framework of the CULIEXA project funded by the ENERBIO fund of the TUCK foundation (Rueil‐Malmaison, France). http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png GCB Bioenergy Wiley

Life cycle assessment of eucalyptus short rotation coppices for bioenergy production in southern France

Loading next page...
 
/lp/wiley/life-cycle-assessment-of-eucalyptus-short-rotation-coppices-for-HgmseRdvG3

References (63)

Publisher
Wiley
Copyright
"Copyright © 2013 Blackwell Publishing Ltd"
ISSN
1757-1693
eISSN
1757-1707
DOI
10.1111/gcbb.12008
Publisher site
See Article on Publisher Site

Abstract

Introduction The recent European Directive on renewable energy set ambitious targets for all Member States, in order for the EU to reach a 20% share of energy from renewable sources by 2020 (European Commission, ). Among renewable energy sources, the biggest contribution (63%) may come from biomass, as suggested by a foresight analysis in Europe (European Commission, ). At present, biomass already contributes about 4% of the total EU energy supply, predominantly as heat, and combined heat and power applications to a lesser extent. The production of liquid biofuels for transport from biomass increased several fold in the last decade, and is currently a major issue. Among various sources of biomass (organic waste, forestry products, cereal straw, etc.), dedicated crops such as short rotation coppices (SRC) are currently being investigated. These systems involve the cultivation of a fast growing ligneous species with short to very short harvesting cycles. Species with a capacity to sprout after cutting are particularly interesting as they make it possible to harvest the same plantation several times over the lifetime of the trees. Eucalyptus ( Eucalyptus sp.) is one of the most widely known species used for biomass‐oriented SRC, particularly for pulp and paper industries (Iglesias‐Trabado & Wilstermann, ). Poplar ( Populus sp.) and willow ( Salix sp.) have been used more recently for energy purpose for example in northern Europe (Lindroth & Båth, ; Wilkinson et al ., ) or in Italy (Manzone et al ., ). In France, SRC were developed with poplar and eucalyptus in the mid 1980s on the initiative of pulp companies. Nowadays, some 2000 ha of pulp SRC are still present although only eucalyptus is still being used in the south‐western part of France with an average rate of 100–200 ha planted every year (Nguyen The et al ., 2004b). The typical plantation scheme is based on 10‐year rotations with a stand density of 1250 stems ha −1 (on a 4 × 2 m grid). Three harvests in 30 years are expected with an average productivity of 10 oven‐dry metric tons (ODT) ha −1 yr −1 with the currently used species: E. gundal , an hybrid between Eucalyptus gunnii and Eucalyptus dalrympleana (Cauvin & Melun, ). The recent drive for renewable energy sources and concerns with the sustainability of biomass production (Robertson et al ., ; Scharlemann & Laurance, ) have sparked interest for SRC given its presumed low environmental impacts as it requires less inputs than agricultural crops (WWI, ). The traditional pulp and paper SRC scheme may be directly transposed to biomass production for energy purposes. As SRC is expected to be mainly grown on former cropland, silvicultural schemes with shorter cycles than the traditional 10‐year pulp rotation are being investigated to be closer to arable farming systems. Growing cycles may be shortened to 7 years with the same productivity as long as stand density is kept within a 2000–2500 stems ha −1 range, as was already tested with poplar (Berthelot et al ., ). Similarly, SRC with shorter rotations with 3 year harvesting cycles are being tested and developed. This scheme was illustrated with willow (Dimitriou & Aronsson, ), and requires far higher stand densities, between 10 000 stems ha −1 and 15 000 stems ha −1 . Such systems are currently being trialled in France with eucalyptus and poplar. Several issues were raised regarding the environmental performance of energy crops, in particular their actual GHG benefits and impacts on water resources or biodiversity (Robertson et al ., ; Monti et al ., ). Here, we chose the LCA methodology to address these issues for eucalyptus SRC, because it is widely used for bioenergy assessment and is a multicriteria, holistic method (Blottnitz & Curran, ; Cherubini, ). No such assessments have been reported for eucalyptus SRC, to the best of our knowledge, although they exist for traditional eucalyptus forests (Lopes et al ., ; Jawjit et al ., ). There is also a growing literature on the LCA of other lignocellulosic feedstock, whether annual arable crops (Kim & Dale, ), perennial grasses such as miscanthus and switchgrass (Monti et al ., ; Shurpali et al ., ), or other types of SRC such as willow and poplar (Gasol et al ., ; Goglio & Owende, ; Djomo et al ., ), whose performance may be compared with eucalyptus. The objectives of this study were twofold: (i) to apply LCA to eucalyptus SRC in southern France, based on the currently existing pulp scheme, and extended to 3 year rotation coppices, and (ii) to investigate the possibility of including the temporary storage of atmospheric CO 2 in ecosystem carbon pools in the GHG balance of heat provision from eucalyptus SRC, following the approach suggested by Moura‐Costa & Wilson ( ) for forest products. Eucalyptus biomass was used to generate heat, and compared to equivalent fossil energy sources. Materials and methods The eucalyptus pulp SRC system was chosen as a basis for the study. This species and its silvicultural scheme have been studied in France for almost 30 years and many technical references already exist (Cauvin & Melun, ). This SRC was designed for pulp production, but may easily be extended to bioenergy production. Scope, functional unit and system boundaries for the LCA The function studied here is heat production from the combustion of SRC wood chips in a boiler. The functional unit selected was therefore 1 GJ of final heat, which means that life cycle impact indicators were calculated relatively to the production of 1 GJ of heat. The system studied is depicted on Fig. , and comprises five main stages: The production of cuttings from selected eucalyptus clones, which corresponds to current practices. It includes the production of mother trees in a biotechnology facility and transportation to a nursery. In the inventory, we used data pertaining to a research laboratory, therefore not designed nor optimized an industrial‐scale production of cuttings. Plantation establishment and removal, including site preparation, fertilization, plantation and weed control during the first 2 years, as well as stump removal at the end of the project. Harvest, including felling, forwarding and chipping for SRC and silage harvester for 3 year rotation schemes. This stage also includes the transportation of harvesting machines to the tree parcel. Transportation of wood chips from the collection site to the boiler. We used a distance of 80 km corresponding to the actual average distance between eucalyptus plantations and the pulp mill of Saint‐Gaudens (south‐western France). Handling and combustion of wood chips in a boiler. Calculation of the cumulative amounts of carbon stored in eucalyptus biomass over time in the 10‐year interval between two cuts. Management scenarios Management scenarios are detailed in Table . The reference scenario was the pulp SRC scheme based on three 10‐year harvest cycles (i.e. a total duration of 30 years), with a stand density of 1250 stems ha −1 . From this baseline, we designed a scenario dedicated to biomass production for energy by doubling the stem density (2500 stems ha −1 ) with three harvests every 7 years for a total duration of 21 years. A variant with 3 year harvesting cycles was designed with a density of 5000 stems ha −1 , which represents in the present context the maximum possible density considering the costs of eucalyptus cuttings. This scenario plans seven successive harvests over the same 21‐year time interval. Selected characteristics of the eucalyptus management scenarios Scenario name Characteristics Productivity (ODT 1 ha −1 yr −1 ) Fertilizer inputs (kg ha −1 yr −1 ) Duration Pulp SRC 1250 stems ha −1 S1 Chainsaw operator – Log harvest 11.7 N: 10P 2 O 5 : 8.7K 2 O: 14.8 3 × 10 years S2 Felling machine – Log harvest 11.7 N: 6.4P 2 O 5 : 7.8K 2 O: 10.1 3 × 10 years Energy SRC, low density 2500 stems ha −1 S3 Felling machine – Log harvest 11.7 N: 6.4P 2 O 5 : 8.3K 2 O: 10.1 3 × 7 years S4 Felling machine – Full stem harvest 14.0 N : 23.4P 2 O 5 : 11.2K 2 O: 25.2 3 × 7 years Energy 5000 stems ha −1 S5 Harvester – Full stem harvest 10 N: 40.0P 2 O 5 : 18.8K 2 O: 49.8 7 × 3 years ODT, oven‐dry metric ton. A set of technological variants technical aspects likely to influence LCA results were also considered: Harvest mechanization: approximately 50% of pulp SRC are currently harvested with felling machines rather than manual felling with chainsaws. Felling machines have a better productivity and make mechanical debarking possible in the field, which results in higher rates of nutrient returns to soils. On the other hand, felling machines consume more fuel and emit more GHGs. The 3 year old stands are usually harvested with adapted agricultural machines. Productivity: for SRC, a yield of 10 oven‐dry metric tons (ODT) ha −1 yr −1 considered as a robust average value taking into account the mortality of trees and their partial ground cover. It corresponds to a final cut at a diameter of 7 cm (commercial cut). The full stem harvest leads to an extra 20% of biomass, including leaves (Nguyen The & Deleuze, 2004a). For the second and third harvest, a 25% gain in biomass production is usually observed due to a faster growth (D. Lambrecq, Fibre excellence, Saint‐Gaudens, personal communication). For lack of consistent findings on the effects of rotation length on productivity, we assumed the same yield for all rotations (Berthelot & Gavaland, ). Fertilizer inputs: pulp SRC are currently not fertilized in France because it is not considered as a relevant operation for the sustainability of biomass production. Nevertheless, this is a very critical point, especially for the shortest rotations whose nutrient exports are expected to be significantly higher. Therefore, we assumed in all scenarios fertilizer input rates corresponding to the estimated exports of nutrients at harvest. The differences between scenarios were particularly acute across harvesting techniques, whether including debarking (with the mechanical harvest) or harvesting full stems or logs. Eucalyptus being an evergreen species, harvesting full stems rather than wood logs would lead to far larger nutrient exports because of the high nutrient contents of the leaves. The amount of N, P and K applied were calculated using state‐of‐the‐art knowledge and data on nutrient exports of eucalyptus SRC in France (Nguyen The et al ., , ) and atmospheric deposition rates (Croisé et al ., ). As a result of the above variants, a total set of five scenarios was implemented, whose characteristics are summarized in Table . LCA methodology The cut‐off threshold for neglecting system components was set at 3.6 × 10 −6 %. The production of laboratory equipment was excluded because cuttings production was only a marginal part in the use of this equipment over its total life cycle. The transportation of pesticides and fertilizers (N, P, K and Mg fertilizers in the nursery, herbicides for site preparation and plantation maintenance, field fertilization) were not taken into account due to a lack of accurate information. Chemical inputs in the nursery were exclusively attributed to the production of cuttings, except for fungicides and hormones, which were neglected due to the very low dosages used. Nursery propagators were also excluded due to the lack of information on this material (jiffy pellets made from peat). Neither waste nor coproducts are produced during the life cycle of SRC, which alleviated the need for allocations. As usually assumed in the LCA of bioenergy systems, the global warming potential of the CO 2 emitted during the combustion of biomass was considered nil (Cherubini, ). LCA calculations were done by using the TEAM 4.0 software package (Ecobilan‐PWC, Paris, France) with the EcoInvent 2000 database (V2.01, St‐Gallen, Switzerland). Field emissions related to the input of fertilizer N and P were calculated using the methods proposed in the Ecoinvent report (Nemecek et al ., ). However, the model proposed for nitrate leaching was found unsuitable for eucalyptus, and this flux was thus neglected. The leaching risk was low because fertilizers are usually applied in spring after the winter drainage, and taken up before the onset of drainage in autumn. In addition, nitrate leaching under forests is generally minimal (Galloway et al ., ). Impacts were characterized using the CML 2001 method, as described in Guinée et al . ( ), and the following categories considered: non‐renewable energy consumption, global warming (with a 100 year timeframe), acidification, eutrophication and photochemical ozone creation potential (POCP). Accounting for ecosystem C dynamics and land‐use changes In a first variant relative to our baseline LCA calculations, we investigated the possibility of accounting for the temporary storage of atmospheric CO 2 in the biomass of eucalyptus stands. The principle is to derive an equivalence factor with permanently stored CO 2 based on the cumulative radiative forcing of atmospheric CO 2 over time. Moura‐Costa & Wilson ( ) derived such as factor from the number of years over which the reduction in radiative forcing would be identical between the temporary and permanent storages. They estimated the duration for break‐even to approximately 55 years, yielding an equivalence factor of 1/55 or 0.0182. However, other factors are presently under discussion in relation to carbon trading. Two other factors were thus tested here: a coefficient of 1/26 proposed by the French Ministry for Agriculture (MAP, ), corresponding to an economical calculation involving an annual discount rate of 4%, and the 1/100 factor proposed by PAS (Bsi, ) for consistency with the IPCC time horizon in the climate change scenarios (2100). Note that some studies implicitly use an equivalence factor of 1, considering net annual CO 2 gains or losses from ecosystems as their direct contribution to global warming (Palm et al ., ; Shurpali et al ., ). Following the above approach, the temporary effect of C storage may be calculated as follows: Mitigating effect ( in t CO 2 eq . ) = Qc × T × EF where Qc is the amount of C stored in tree biomass (t CO 2 ha −1 ), T is the duration of storage (years), and EF the equivalence factor (unitless). For eucalyptus SRC, the carbon taken up by the stand during the first year of the rotation is stored until harvest, i.e. during 9 years; the C taken up during the second year is stored for 8 years, etc. The Qc × T component of the equation actually corresponds to the cumulative sum of C stored through time, except for the last year when the stand is harvested (Fig. ). System boundaries and steps of the life cycle. The C sequestration of eucalyptus SRC should be compared to a baseline scenario in terms of land‐use. Here, we chose abandoned agricultural land (referred to as wildland in the following), which typically occurs after vineyard removal in southern France. Eucalyptus SRC would therefore be established on former vineyards in our scenario, which excludes indirect land‐use change effects. The global C storage was therefore calculated by subtracting the C stock of SRC by that of wildland. The carbon stored in the aboveground biomass (AGB) of the eucalyptus stands was calculated from the C content of harvested wood, considering a C content of 47% (dry weight basis; Paixão et al ., ; Tanabe et al ., ). Belowground biomass (BGB) was estimated with an allometric relationship as a fixed proportion of AGB, set to 30% (Tanabe et al ., ). Note that BGB was kept during the 30‐year cycle of the eucalyptus plantation in the trees' rooting systems, whereas AGB dropped to zero after each cutting. For the baseline land‐use, the wildland's ABG was considered constant at 0.9 t C ha −1 yr −1 , which is the peak value for grasslands in warm temperate, dry climates given in the IPCC guidelines for GHG inventories (Tanabe et al ., ). It is in the lower end of the 0.8–3.2 t C ha −1 yr −1 range reported in Europe for former arable fields up to 3 years after abandonment (Hedlund et al ., ), i.e. in the early years of fallow regeneration. Belowground biomass was fixed at 2.0 t C ha −1 yr −1 (Tanabe et al ., ), which is slightly lower than the 2.5–3.5 t C ha −1 yr −1 range in annual returns to soils estimated in the classical Rothamsted (UK) long‐term wilderness experiments, where arable fields were allowed to undergo natural woodland regeneration in the 1880s (Jenkinson et al ., ). In the beginning of the transition from arable to wildland, only herbaceous species are involved and their net annual biomass production is entirely returned to soils as litter. Further on during the 30‐year life cycle of the eucalyptus plantation, it is likely that some woody species may also appear in the wildland and start accumulating biomass from 1 year to the next, although the exact dynamics of that transition has not been documented to the best of our knowledge. Over a longer time‐frame, observations in the ‘Geescroft wilderness’ experiment in Rothamsted (UK), an arable field allowed to undergo natural woodland regeneration in 1885, may give us some insight into this process and provide an upper limit for this component. In this plot, the accumulation of AGB was estimated at 0.6 t C ha −1 yr −1 over the first 100 years of the transition (Grogan & Matthews, ), which we considered as the upper limit of what would happen in the first 30 years of wildland growth after abandonment (the lower limit being no accumulation at all). A below to AGB ratio of 1 : 3 was assumed for the wildland (Grogan & Matthews, ), which is similar to the value used for eucalyptus trees. It is likely that the differences in soil organic carbon (SOC) will appear between the SRC eucalyptus and the baseline land‐use over time, due to differences in litter and belowground inputs (Grogan & Matthews, ). However, as eucalyptus SRC systems are relatively recent, there are no long‐term experiments documenting the dynamics of SOC after conversion to eucalyptus, let alone comparing them with other land‐uses such as arable farming or wildlands. We therefore elected to exclude differences in SOC between eucalyptus SRC and wildland in our analysis. The effect of this hypothesis is addressed in the section. Results LCA results Life cycle consumption of non‐renewable energy ranged from 77.0 to 92.7 MJ GJ −1 heat output from eucalyptus biomass (Table ). It was lowest for the S1 scenario with lower stem density and manual harvest, and highest for the shortest rotation (S5). In all scenarios, wood chips transport represented the main energy consumption hotspot with a share of 46–55%. Harvesting operations came second with 30–36% of total energy consumption, except for the very short rotation scenario, where their share was only 3.3%. This is due to the use of an adapted silage harvesting machine instead of heavy, fuel‐consuming forestry machines. In scenario S5, the most important steps were fertilization and plant production. Fertilizer inputs were larger than with the SRC schemes because the harvest of whole stems including leaves lead to higher nutrient export rates and enhanced fertilizer requirements. Stem density is also twice higher in the S5 scenario compared to the energy SRC (S3 and S4). This had a significant impact on energy consumption because the production of cuttings takes place in an energy‐intensive biotechnology laboratory. The shorter rotations and higher stem densities of S5 further enhanced this trend, making this scenario the most energy‐intensive. Its energy ratio (ratio of heat output to fossil energy inputs) was also the lowest of all scenarios, at 10.8. This ratio increased with decreasing harvesting frequency, leading to the pulp scheme achieving the highest value (13). Non‐renewable energy consumption per life cycle stage of the various SRC systems (MJ GJ −1 ), and ratio of energy delivered to primary energy consumption Scenario Cutting production Site prep. Fertilization Harvest Transport Boiler Total Energy ratio S1 1.84 2.96 5.69 23.52 42.67 0.29 77.0 13.0 S2 1.84 2.96 3.99 27.30 42.67 0.29 79.0 12.7 S3 5.24 4.23 4.00 27.30 42.67 0.29 83.7 11.9 S4 4.34 3.51 9.47 27.28 42.67 0.29 87.6 11.4 S5 12.85 5.18 28.63 3.08 42.67 0.29 92.7 10.8 Life cycle GHG emissions (excluding ecosystem C pools) varied in a narrow range for the four SRC scenarios, from 8.2 (S1) to 8.5 (S4) kg CO 2 ‐eq. GJ −1 (Fig. ). They were 50% higher for the S5 scenario, due to its requiring two to three times more NPK fertilizer inputs than the other schemes, altogether with a 20–30% lower productivity (Table ). The relative importance of the various steps of the life cycle followed a similar pattern for all scenarios except S5, with an important contribution of fertilization (38–44% of total), transport (32–33%) and harvest (18–22%), in agreement with the review of Djomo et al . ( ) for willow and poplar SRC, and of Rowe et al . ( ) for woody crops. The 3 year rotation scenario (S5) had lower emissions than the longer rotation scenarios in the harvest step due to the use of agricultural harvesters, and plant production and site preparation emitted more GHGs than harvest. Fertilization accounted for 68% of the total emissions for this scenario. Life cycle assessment results for global warming, eutrophication acidification, photochemical ozone creation potential, per GJ of heat delivered. Indicators for the eutrophication impact ranged from 48 (S2) to 152 (S5) g PO 4 2− ‐eq. GJ −1 (Fig. ), and were dominated by the fertilization phase. The losses of P from the plantation by run‐off and erosion made up 90% of the impact related to fertilization, whereas ammonia volatilization contributed the remainder, the impacts of NO emissions from soils being negligible. Because of its larger fertilizer requirements, the very short rotation system had nearly threefold higher eutrophication impacts than short rotation ones. Although the latter also received varying rates of fertilizer inputs (Table ), differences in productivities compensated for these variations and all short rotation schemes had a similar eutrophication impact within a 5% relative range. Interestingly, the best scenario was the one with the highest biomass productivity (S4) and not that with the least fertilizer inputs per ha (S2), which only achieved a mid‐range performance. The acidification indicator ranged from 39 (S1) to 110 (S5) g SO 2 ‐eq. GJ −1 , following a pattern similar to eutrophication (Fig. ). The very short rotation scenario had again a threefold larger impact than the other scenarios, and for the same reason: its higher fertilizer inputs, which translated in higher field emissions of ammonia and nitric oxide, and indirect emissions due to fertilizers' manufacturing. However, the harvest and transport steps played a more important role than for eutrophication, and the breakdown differed between the scenarios. The share of harvest ranged from 20% to 30% for the SRC, although it was nearly negligible (at 2%) for S5. This stems from the major advantage of this scheme, namely the use of agricultural machines in lieu of forestry ones, which are far more resource intensive ones. However, the associated savings did not compensate for the large requirements of synthetic fertilizer inputs compared to scenarios with longer rotations. The POCP indicator ranged from 2.4 (S5) to 6.8 (S1) g C 2 H 2 ‐eq. GJ −1 , with harvest operations and wood chips transport contributing the most (Fig. ). The much higher emissions of photo‐oxidants occurring with the scenario S1 is explained by the chainsaws used for manual felling. The chainsaws used in France are seldom equipped with catalytic exhaust pipes and release volatile organic compounds, which have a high potential for ozone formation. These emissions also occur to a lesser extent with the mechanized felling option (in scenarios 2–4) because chainsaw operators are necessary for the second and third harvest to thin the coppice before felling machines can be used. The distance between the plantation and the boiler was set at 80 km in the baseline calculations. Table illustrates the influence of this distance on the five LCA impact categories for scenario S1, from a minimum value of 10 km to the baseline value of 80 km. Energy consumption was the most sensitive indicator: it dropped by 28% when halving the transport distance, whereas GHG emissions and acidification impacts were only reduced by 16%, photochemical ozone formation by 10% and eutrophication by 3%. The energy ratio increased from 13.0 to 18.0 when the transportation distance decreased from 80 to 40 km, and reached 25.2 with a 10 km distance (Fig. ). The other indicators were less sensitive to this parameter, Energy ratio as a function of the transportation distance from the eucalyptus plantation to the boiler for scenario S1. Influence of woodchips transportation distance from plantation to boilers on life cycle assessment indicators for scenario S1, per GJ of heat Transportation distance (km) 80 40 20 10 Non‐renewable energy consumption (MJ) 77.0 55.6 45.0 39.6 Acidification (g eq. SO 2 ‐eq.) 41.7 35.0 31.7 30.0 Eutrophication (g PO 4 ‐eq.) 52.0 50.5 49.8 49.4 Photochemical ozone formation (g C 2 H 2 ‐eq.) 6.8 6.1 5.7 5.5 Global warming (kg CO 2 ‐eq.) 8.16 6.80 6.11 5.77 A comparison with fossil energy sources was carried out to assess the environmental advantages and drawbacks of using SRC biomass as a substitute to coal, fuel oil and natural gas (Fig. ). In all scenarios, the provision of heat from SRC biomass consumed 90% less fossil energy than when using fossil energy sources. Similarly, GHG emissions were reduced by more than 80% with the SRC biomass. However, the patterns with the local to regional‐range impacts (acidification, eutrophication and photochemical ozone formation) were less clear‐cut. Biomass‐derived heat had generally much lower acidification and photochemical ozone formation impacts than fossil‐based heat except with natural gas, which out‐performed the S5 scenario for eutrophication and scenario S1 (pulp SRC with manual felling) for ozone formation. Natural gas had two to 30 times lower impacts than the other fossil sources, especially coal. Conversely, the eutrophication impacts were in the 50–135 g PO 4 3− ‐eq. GJ −1 range for the eucalyptus scenarios, and in the 5–40 g PO 4 3− ‐eq. GJ −1 range for the fossils, pointing to a weakness of the biomass‐based chain. The twofold higher eutrophication impacts of the S5 compared to the other scenarios were clearly due to the larger fertilizer inputs required by the former. Life cycle assessment indicators weigthed by the average impact of an European inhabitant and compared to fossil energy sources. Inclusion of ecosystem C dynamics Figure depicts the dynamics of aboveground and belowground biomass in the eucalyptus plantation (scenario 1) and the baseline wildland representing the baseline alternative land‐use. Over the 30‐year period of the eucalyptus life cycle, the average biomass of eucalyptus plantations was several fold larger than that of the wildland, even when considering the appearance of ligneous species in the latter. This hypothesis had a significant impact as it leads to a eightfold higher estimate of total biomass after 30 years compared to a wildland solely composed of annual species. The larger biomass accumulation in the eucalyptus SRC was due to a higher net primary production and an important storage in the belowground compartment, which kept increasing through the cuts. When averaged over the 30 years of the SRC rotation, the differences between SRC and wildland range from 16 to 24 t C ha −1 for the AGB, and from 26 to 37 t C ha −1 for the total biomass (Fig. ). These gaps represent the net ecosystem CO 2 gains incurred when substituting wildland with SRC, for instance after the abandonment of a vineyard. They are related to land‐use effects and may be included in the life cycle GHG emissions of eucalyptus biomass production by using equivalence factors to account the temporal value of C sequestration in the biomass. This led to savings of 0.57–5.16 t CO 2 ‐eq. ha −1 yr −1 (Table ), depending on the equivalence factors and the carbon pools taken into account. Including these CO 2 savings, the LCA of eucalyptus‐derived heat offset GHG emissions by 70–400% (Fig. ), and therefore had a large impact on the global warming indicators. With the most favourable equivalence factors (1/26 and 1/55), the C stored in eucalyptus biomass resulted in heat provision being a net GHG sink. Dynamics of aboveground (top) and above and belowground (bottom) C storage by pulp eucalyptus SRC (solid line) and wild land with (dotted line) or without (dashed line) consideration of C accumulation in woody species, in the years following conversion to SRC . Greenhouse gas emissions (g CO 2 eq. GJ −1 heat) due to sowing and harvesting operations, fertilization and transport of chips, and CO 2 savings from CO 2 sequestration in ecosystem biomass using various equivalence factors and the lower and upper estimates. Carbon storage in the eucalyptus SRC stands (management scenario 1), relative to the baseline wildland, as averaged over the 30‐year duration of the project, in the aboveground and above and belowground biomass pools (t CO 2 ha −1 ). The lower end of the range corresponds to the emergence of woody species in the wildlands, which is ignored for the upper‐end value. C stored in biomass pools are transformed into CO 2 sequestration rates using the three possible equivalence factors detailed in the text Ecosystem pools Equivalence factors 1/26 1/55 1/100 Aboveground biomass 2.21–3.43 1.05–1.62 0.57–0.89 Aboveground and belowground biomass 3.67–5.16 1.73–2.44 0.95–1.34 Discussion Benefits and drawbacks of eucalyptus SRC Substituting fossil sources with biomass from eucalyptus SRC leads to an 80–90% abatement of life cycle GHG emissions and fossil energy consumption per MJ of heat supply, for all SRC management scenarios. These figures confirm the strong benefits of bioenergy chains and are consistent with other LCAs of heat from biomass. For instance, Reinhardt ( ) reported a 95% abatement in GHG emissions and energy consumption when displacing oil or natural gas with short rotation willow for district heating in several European countries. In addition, inclusion of the temporary storage of CO 2 in the plant biomass, which was ignored in previous literature, more than doubled the GHG savings compared to those in fossil sources. The relevance of this hypothesis is discussed in the subsection . Conversely, the benefits of SRC were far from obvious for the other impact categories, especially when displacing natural gas, which had threefold to fourfold lower impacts per functional unit than the other fossil sources. This trade‐off between global impacts (global warming and fossil energy consumption) and local impacts has often been reported for bioenergy chains (Reinhardt, ; Gabrielle & Gagnaire, ), and is almost inevitable because of the gaseous and leaching losses of nutrient occurring upon the feedstock production phase. Despite the relatively low fertilizer N requirements of eucalyptus stands compared to arable crops, none of the management scenarios achieved lower eutrophication impacts than the fossil‐based alternatives. Furthermore, the impact estimates were conservative because some losses of nutrients were neglected, as discussed in the subsection . In terms of management scenarios, the very short rotation scenario (S5) was outperformed by the other scenarios for all impact categories except ozone formation, by a factor of 50–250%. As the economics of this system is also unfavourable (Nguyen The et al ., ), S5 does not emerge as a good candidate compared to short rotation scenarios. Thus, the benefits from a quicker biomass growth and simplified harvesting made possible by the 3 year growing cycle of S5 were outweighed by their larger fertilizer input and stem density requirements. The only advantage of S5 appeared in the POCP, in which harvesting operations were predominant. Note that the productivity estimate of S5 was rather conservative, being 17–40% lower than that for the longer rotation scenarios, for lack of consistent field data (Berthelot & Gavaland, ). However, assuming a similar or even slightly higher productivity for S1 would not change the overall ranking because the relative differences between S1 and the other scenarios were larger than 50% for most impact categories. To our knowledge, no LCAs have been carried out so far on eucalyptus SRC, whether for energy or pulp and paper. Our results may still be compared with those pertaining to traditional eucalyptus plantations published by Jawjit et al . ( ) in Thailand. Their study used system boundaries and characterization factors similar to those of ours, but found much lower impact values in general. Plant‐gate life cycle GHG emissions were estimated at only 3.1 kg CO 2 ‐eq. GJ −1 , compared to the 8–12 kg CO 2 ‐eq. GJ −1 range we obtained here. The acidification impact was 22 g SO 2 ‐eq. GJ −1 in the Thailand study compared to our 40–110 g SO 2 ‐eq. GJ −1 range, whereas the photochemical ozone formation potential amounted to 1.6 g C 2 H 2 ‐eq. GJ −1 in Thailand compared to our 2.5–7.0 g C 2 H 2 ‐eq. GJ −1 range. Eutrophication was an exception with similar impacts between Thailand and France, at 41 g PO 4 2− ‐eq. GJ −1 and an average of 50 g PO 4 3− ‐eq. GJ −1 for the SRC systems respectively. Some of these discrepancies are explained by the higher yields of 17.4 ODT ha − 1 yr −1 achieved by eucalyptus under the tropical conditions of Thailand, compared to the 9.5–14 ODT ha −1 yr −1 range assumed here. The eutrophication impact was relatively higher because 35–20% of the fertilizer N and P applied was supposed to leach to water bodies in this Thailand study, whereas those losses were neglected here, as they were in other LCAs on herbaceous and tree species for lack of specific references (Gasol et al ., ; Monti et al ., ). Our results on eucalyptus SRC may be more broadly compared to other lignocellulosic feedstock: willow in France (Reinhardt, ), Italy (Goglio & Owende, ) and Europe (Djomo et al ., ), poplar SRC in Italy (Gasol et al ., ) and Europe (Djomo et al ., ), reed‐canary grass in Finland (Shurpali et al ., ) and four perennial grasses in Italy (Monti et al ., ). All these studies used similar system boundaries with the exception of the combustion step, and relied on the same set of characterization coefficients (from Guinée et al ., ). Most of them also used the EcoInvent database for the life cycle inventory. Compared to the poplar SRC system assessed by Gasol et al . ( ) in Italy, the production and harvest of eucalyptus biomass consumed 1.8–2.5 more primary energy, essentially because the harvest was threefold less energy‐intensive per ton of biomass than eucalyptus (for scenarios 1–4) or because poplars required fourfold less fertilizers (for scenario 5). Also, the data on fuel consumption by farm machinery were adapted from the EcoInvent database based on local records, but the exact corrections were not given by the authors. When including the transportation of wood chips, albeit with a shorter distance than our nominal hypothesis (25 vs. 40 kms), the GHG emissions of poplar totalled 1.93 kg·CO 2 ‐eq. GJ −1 which is four to six times less than our 8–12 kg CO 2 ‐eq. GJ −1 range for eucalyptus. The gap was even wider for the other impact categories: the eutrophication impact of poplar was estimated at 3.4 g PO 4 3− ‐eq. GJ −1 vs. 40–135 g PO 4 3− ‐eq. GJ −1 for eucalyptus; acidification amounted to 15.7 g SO 2 ‐eq. GJ −1 vs. 40–110 g SO 2 ‐eq. GJ −1 for eucalyptus; and POCP totalled 0.3 g C 2 H 2 ‐eq. GJ −1 for poplar compared to 2.4–7 C 2 H 2 ‐eq. GJ −1 for eucalyptus. Besides differences in management and inventory data for farm machinery, these large discrepancies arise because direct field emissions contributed only a minor share of the impacts in the Gasol et al . study, whereas they predominated in our LCA. There are reasons to believe that some of these emissions were somehow underestimated: for instance, N 2 O emissions from Gasol et al . were similar to our estimates on a ha basis, whereas NO emissions were twofold lower. This contradicts current literature, which indicates that NO and N 2 O emissions fall within a similar range (Stehfest & Bouwman, ). Our estimate of N 2 O emissions also included background emissions (i.e. nonanthropogenic) and the contribution of eucalyptus residues. Our LCA results for eucalyptus are overall closer to those reported by Reinhardt ( ) and Goglio & Owende ( ) for short rotation willow in Germany and Ireland respectively. These authors reported energy consumptions of 33 MJ GJ −1 heat and 56.4 MJ GJ −1 , respectively, compared to our 55.6 MJ GJ −1 figure for scenario 1 (S1) with a similar transportation distance (40 km). The lower figure from Reinhardt ( ) was due to a less energy‐intensive harvest for willow, whereas the Goglio & Owende ( ) study involved a drying phase prior to combustion. Our values also fell in the middle of the 12.6–76.9 MJ GJ −1 range reported by Djomo et al . ( ) across Europe for willow and poplar SRC. GHG emissions were very similar, at 7.13 kg CO 2 ‐eq. GJ −1 for willow in Germany vs. 6.80 kg CO 2 ‐eq. GJ −1 for the S1 eucalyptus system here, whereas the eutrophication impact for willow was 94 g PO 4 3− ‐eq. GJ −1 , well within the 40–135 g PO 4 3− ‐eq. GJ −1 range reported here for our systems, although it should be noted that the estimation of nitrate and phosphate losses was not explicitly described in the willow study. Finally, the acidification emissions of willow in Germany totalled 174 g SO 2 ‐eq. GJ −1 , compared to a 40–110 g SO 2 ‐eq. GJ −1 range for eucalyptus SRC. This is probably due to higher combustion emissions of acidifying compounds in the Reinhardt ( ) study than listed in the EcoInvent database, which pertains to more recent technologies. For the same reason, POCP impacts were also larger with willow, at 18 C 2 H 2 ‐eq. GJ −1 in comparison to 6.1 C 2 H 2 ‐eq. GJ −1 g for the S1 system. Lastly, eucalyptus SRC may be compared to the range of perennial grasses assessed by Monti et al . ( ), involving miscanthus, switchgrass, cynara and giant reed, with a cradle to farm‐gate system boundary. Energy consumption ranges from 33 to 142 MJ GJ −1 biomass energy content, compared to approximately 35 MJ GJ −1 for eucalyptus SRC (Table ), putting the latter on par with the best performers – giant reed and miscanthus. However, their GHG emissions were significantly lower, at 1.75 kg CO 2 ‐eq. GJ −1 compared to 5.5–9.4 for kg CO 2 ‐eq. GJ −1 eucalyptus. The same applied to eutrophication impacts, ranging from 4 to 20 g PO 4 3− ‐eq. GJ −1 for grasses and from 45 to 132 g PO 4 3− ‐eq. GJ −1 for eucalyptus, and also to acidification impacts, which are 2–2.5 lower for the grasses than eucalyptus. As with the Gasol et al . ( ) study, it may be that field emissions were undervalued, because fertilizer N input rates were rather higher than the eucalyptus SRC systems (at 80 kg N ha −1 yr −1 compared to a 6–40 kg N ha −1 yr −1 range for eucalyptus). The Monti et al . ( ) article does not mention direct emissions of nitrate or P in the field. Because of differences in local contexts, in the sources of life cycle inventory data and estimation methods for field emissions, it is not possible to directly compare the eucalyptus systems tested here with other coppices or herbaceous plants as these differences are likely to overrule the differences between feedstock per se . With the exception of the Gasol et al . ( ) study, the LCA indicators of eucalyptus were within the range of impacts reported for other lignocellulosic feedstock, but no robust patterns emerged in terms of ranking with other species. Uncertainties in the life cycle inventories Field emissions are particularly difficult to correctly address in the LCA of agricultural or forestry systems as they depend to a large extent on local conditions (soil properties, climate) and on their interactions with management practices, which govern the fate of chemical or organic inputs. As very little data on field emissions have been published for eucalyptus SRC in temperate zones, we used estimation methods developed for other species, or assumed that some emissions were negligible. Such was the case for nitrate leaching and P losses, which may have lead to an underestimation of eutrophication impacts. Lopes et al . ( ) found these emissions negligible in their LCA of eucalyptus‐derived paper, and so did Jawjit et al . ( ) although their estimates of nitrate and phosphate emissions from eucalyptus plantations were rather large: they assumed that 35% and 20% of fertilizer N and P inputs were leached to water bodies, respectively, according to the 1997 IPCC guidelines for GHG inventories. The 35% emission factor for nitrate (which was revised to 30% in the 2006 IPCC guidelines – Tanabe et al ., ) should in principle apply to managed forests, but no reference specific to forest or energy plantation is given in the literature base that served to determine this value. Further research is therefore warranted to provide a more accurate estimate of nitrate leaching for eucalyptus SRC. The same applies to P losses, and also to gaseous emissions of N 2 O, NH 3 and NO. The latter were calculated according to the 2006 IPCC guidelines for managed ecosystems, using default emission factors which are characterized by a large uncertainty range (Stehfest & Bouwman, ). Unfortunately, no literature data were found for eucalyptus SRC or forests in Europe to refine those estimates. Relevance of including ecosystem C dynamics Accounting for variations in ecosystem C stocks, compared to the alternative land‐use (wildland in our case) had a drastic effect on the GHG balance of eucalyptus‐derived heat, whose magnitude depended on the factor chosen for the equivalence between C stored in ecosystem pools and atmospheric CO 2 . Even when using the most conservative value of 1 : 100 (i.e. that least favourable to eucalyptus), ecosystem C pools offset GHG emissions by 50–70%, depending on the inclusion of belowground biomass. This made net eucalyptus a nearly carbon‐neutral source of heat, and stresses the influence of ecosystem C dynamics in relation to land‐use changes (LUC) in LCAs, already noted by Ndong et al . ( ) for biodiesel from jatropha in West Africa, and Shurpali et al . ( ) for reed‐canary grass in Finland. Note that the latter authors effectively used an equivalence factor of 1 : 1, as they used measurements of net ecosystem exchanges of CO 2 over reed‐canary grass, as cumulated over 1 year, as a measure of the C sink strength of the field where this crop was grown. Such hypothesis was also implicit in the GHG budgets of farmland and woodland management computed by Palm et al . ( ) in two villages in Africa, or by Ceschia et al . ( ) for cropping systems across Europe. In both references, ecosystem C fixation was put on a par with CO 2 emissions from fossil sources or N 2 O emissions from soils. This may be justified on a short‐term basis, but is misleading in the long‐run as most of the C taken up by ecosystems on a given year will be released back to the atmosphere after a few years as it enters fresh organic matter pools with rapid turnover (Jenkinson, ). From a life cycle perspective, whereby one attempts at estimating the cumulated past and future effects of substituting one product by another, using such an hypothesis would have overemphasized the sink capacity of SRC stands compared to wildland, and given wrong results on the actual GHG benefits of eucalyptus biomass. The use of equivalence factors, which are up to two orders of magnitude lower, is thus fully justified. Of course the magnitude and direction of this effect strongly depends on the LUC hypotheses made in the LCA. Adverse effects were conversely noted for biofuels when including indirect land‐use change effects whereby the displacement of food crops for biofuels in the United States entailed the conversion of natural ecosystems to arable farming in other parts of the world (Fargione et al ., ). Our scenarios for eucalyptus growth did not involve such effects as they considered the farming of eucalyptus SRC as an opportunity to value former arable land or vineyards that had been abandoned because of a drop in the market prices of wine. Assuming the resulting wildland to be reverted back to cropland, indirect LUC would apply. Using a value derived from a recent meta‐analysis on such effects for biofuels (De Cara et al ., ), we estimated this would cause an extra emission of 6.8 kg CO 2 ‐eq. GJ −1 heat, or 80% of the attributional life cycle emissions of the chain. This effect is similar (although opposed) to the inclusion of ecosystem pools and deserves further attention. Soil organic matter (SOM) pools were not included in the ecosystem pools for lack of robust estimates of SOM variations under both eucalyptus SRC and wildland. This pool was actually responsible for most of the land‐use offset of GHG emissions in the LCA of Jatropha by Ndong et al . ( ). Similarly, given the differences in net primary production between the SRC stands and the wildland, it is likely that the former have a higher SOM content than the latter, and therefore further accrue their GHG benefits. Grogan & Matthews ( ) thus argued from a very preliminary modelling study that ‘short‐rotation coppice systems have the capacity to sequester substantial amounts of carbon, comparable to, or even greater than, an undisturbed naturally regenerating woodland’. This results from C inputs from SRC being higher than from the regenerated woodland, which is comparable to our wildland system here. Field samplings were carried out in our study area to estimate SOM contents under vineyards, eucalyptus SRC of various ages, wildlands and arable land. Although the comparison was confounded by soil clay content, SOM was clearly lowest under the vineyards and comparable between wildlands and SRC. Conversion shortly after vineyard abandonment would therefore maximize the benefits of eucalyptus SRC in terms of SOM gains from land‐use change. Grogan & Matthews ( ) estimated that willow SRC sequestered 0.11 t C ha −1 yr −1 compared to abandoned cropland, which would translate for eucalyptus SRC as an additional offset of 2.5 kg CO 2 ‐eq. GJ −1 heat, or 30% of the GHG emissions of the chain. Further work is nevertheless required to provide more robust estimates of these potential gains. Acknowledgements These results were obtained in the framework of the CULIEXA project funded by the ENERBIO fund of the TUCK foundation (Rueil‐Malmaison, France).

Journal

GCB BioenergyWiley

Published: Jan 1, 2013

Keywords: ; ; ; ; ;

There are no references for this article.