- Email: [email protected]
Carbon debt and payback time – Lost in the forest?
Renewable and Sustainable Energy Reviews 73 (2017) 1211–1217
Contents lists available at ScienceDirect
Renewable and Sustainable Energy Reviews journal homepage: www.elsevier.com/locate/rser
Carbon debt and payback time – Lost in the forest?
MARK
Niclas Scott Bentsen University of Copenhagen, Faculty of Science, Department of Geosciences and Natural Resource Management, Rolighedsvej 23, DK-1958 Frederiksberg C, Denmark
A R T I C L E I N F O
A BS T RAC T
Keywords: Forest bioenergy Carbon debt Payback time Machine learning Meta-analysis
In later years the potential contribution of forest bioenergy to mitigate climate change has been increasingly questioned due to temporal displacement between CO2 emissions when forest biomass is used for energy and subsequent sequestration of carbon in new biomass. Also disturbance of natural decay of dead biomass when used for energy affect the carbon dynamics of forest ecosystems. These perturbations of forest ecosystems are summarized under the concept of carbon debt and its payback time. Narrative reviews demonstrate that the payback time of apparently comparable forest bioenergy supply scenarios vary by up to 200 years allowing amble room for confusion and dispute about the climate benefits of forest bioenergy. This meta-analysis confirm that the outcome of carbon debt studies lie in the assumptions and find that methodological rather than ecosystem and management related assumptions determine the findings. The study implies that at the current development of carbon debt methodologies and their lack of consensus the concept in it-self is inadequate for informing and guiding policy development. At the management level the carbon debt concept may provide valuable information directing management principles in a more climate benign directions.
1. Introduction In later years the contribution of forest bioenergy to potentially mitigate global warming has been increasingly questioned [1] due to the temporal displacement between CO2 emissions when forest biomass is used for energy and subsequent sequestration of carbon in new biomass. Disturbance of natural decay of dead biomass and growth of living biomass when used for energy affect the carbon dynamics of forest ecosystems. These perturbations of forest ecosystems are summarized under the concept of carbon debt and its payback time. A number of recent narrative reviews discussed the implications of carbon dynamics and carbon debt of forest bioenergy with reference to climate impact and policy development [2–4]. Lamers and Junginger [3] demonstrated that the carbon payback time of apparently comparable forest bioenergy scenarios vary by up to 200 years allowing amble room for confusion and dispute about the potential climate benefit of forest bioenergy. The birth of the carbon debt concept is often attributed a paper in Science in 2008 [5], which did not treat forest bioenergy but potential forest clearing as a consequence of agricultural expansion driven by increased demand for biofuels. The underlying mechanisms describing how forest carbon dynamics may be influenced by increased demand for bioenergy was however treated much earlier [6,7]. An analysis from 1996 by Leemans et al. [8] developed to support the second assessment report of the IPCC [9] describe the now well-known pattern of a
transition period, where increased deployment of bioenergy increases CO2 emissions to the atmosphere followed by an extended period with reduced emissions. More recent papers describe the same pattern conceptually; see e.g. Mitchel et al. [10]. While few if any argue against the existence of a potential carbon debt, quantification of same remains controversial. Naudts et al. [11] reported a carbon debt of Europe's forest of 3.1 Pg C since 1750 because of forest management compared against an untouched forest baseline. Nabuurs et al. [12] contested the relevance of an untouched forest baseline assumption and find no carbon debt in the outlooks for Europe's forest for the same reason, forest management. Carbon debt and payback time studies aim to inform scientists, policy makers, forest managers, the utility sector and other stakeholders on the climate consequences of extracting more biomass from forests to meet an increased demand for non-fossil energy. In the vast body of literature one can find support for almost any view on the climate impact of forest bioenergy, from being instantly beneficial to analyses showing that it will not in the next 10,000 years contribute to global warming mitigation. The objectives of this review are a) to identify patterns and commonalities in assumptions and outcomes across the current scientific literature on forest bioenergy, carbon dynamics and global warming mitigation potential; b) to identify factors influencing carbon debt and payback times of energy production based on forest biomass; and c) to provide guidance to policy and decision makers on how to understand and treat the carbon debt
E-mail address: [email protected]. http://dx.doi.org/10.1016/j.rser.2017.02.004 Received 8 April 2016; Received in revised form 21 December 2016; Accepted 1 February 2017 1364-0321/ © 2017 Elsevier Ltd. All rights reserved.
Renewable and Sustainable Energy Reviews 73 (2017) 1211–1217
N.S. Bentsen
2.2. Scale and data
concept with reference to forest management, energy resource procurement and energy policy development. Buchholz et al. [13] demonstrated recently that reported carbon debt payback times of forest bioenergy are particularly influenced by inclusion of wildfire dynamics in the analyses and models. This review follows in the line of Buchholz et al. [13], but considerably more scenarios or cases are included (245 here vs. 123 in Buchholz et al.) and emphasis is put on exploring how applied methodology and assumptions influence the outcome of carbon debt studies.
Lamers and Junginger [3] distinguished between three different scales when analyzing carbon debts. Same distinction is used here. Stand scale indicate that carbon dynamics is modelled for a uniform even-aged mono-culture. When harvested, the entire stand is cut and the wood is used for materials and/or energy. Analyses applying a fixed landscape scale attempts to counteract the obvious simplification of stand scale studies by assuming a hypothetical landscape of uniform even-aged compartments of monoculture, where each compartment is displaced in time but otherwise modelled as described for stand scale studies. In the fixed landscape the compartments in the forest landscape describe a so-called normal forest where all successional stages from regeneration to final harvest are equally represented [47]. The dynamic landscape representation is based on a true representation of an actual forest landscape encompassing diversity in e.g. species, ages, and rotation lengths. Data underlying the studies are based on either hypothetical data or spatially explicit data, usually from forest inventories.
2. Materials and methods Carbon debt is caused by a number of factors. With respect to forest biomass factors of particular relevance are: Temporal displacement between CO2 emission from biomass conversion to energy and subsequent CO2 sequestration in new biomass. The long rotations in forestry increase the importance of this factor [14,15]. Additional harvest of biomass perturbate the forest ecosystems and may change growth and decay rates of living and dead biomass. Upstream fossil CO2 emissions from resource production and extraction are different than those associated with the extraction of fossil resources displaced by biomass. Differences in efficiency between fossil and biomass conversion technologies. Heat and to some extent electricity can be generated as efficiently with forest biomass as with fossil resources. Biomass with high content of chlorine (Cl), potassium (K2O) and sodium (Na2O) causes corrosion, slagging and fouling of boilers, heat exchangers and super heaters [16,17]. Problems increase with increasing temperatures, why biomass boilers often operate at lower steam temperatures. This is particular true for straw fired boilers, but a high share of bark, leaves and twigs in forest biomass can also reduce attainable operating temperatures. The oxygen to carbon ratio is higher in biomass than in fossil hydrocarbons [16]. Fossil material (old biomass) is in a more reduced state than living or recently dead plant tissues. Consequently CO2 emission per energy unit from combustion is higher from biomass than from coal, oil and natural gas irrespective of conversion efficiencies. Carbon debt is comparable to financial debt in that it can be paid back over a period of time. Increasing harvest of biomass from forests may for a shorter or longer period of time reduce the amount of carbon stored in the forest, either in living or dead biomass. If increased harvest change the hydrology of forest ecosystems due to reduced evapotranspiration increased emissions of methane and nitrous oxide may be observed [18]. When increased harvest of forest biomass is done with a purpose of displacing other resources, GHG emissions from extraction and use of these are avoided. The payback time of carbon debt is modelled as the number of years it takes to reach parity between the cumulated additional emissions from biomass harvest and use, and avoided emissions from extraction and use of displaced resources [10]. This review builds on the scientific literature published in the last 20+ years reporting payback times of using forest biomass to displace fossil resources for energy generation. A total of 245 scenarios are included and characterized relative to a number of descriptive variable presented in Table 1. Data are extracted from [6,7,10,19–46].
2.3. Model A range of models are used to model the carbon dynamics of increased biomass harvest. In this study 13 different named models were found as well as a number of un-named models. Only named model were included, but the methodology accounts for missing values by attributing arbitrary values to the un-named models. 2.4. Biome and geography Here I distinguish between three different biomes from boreal to temperate. Sub-tropic and tropic biomes are not represented specifically in the literature on carbon debt repayment. A few scenarios included in this review have a global scope and include the sub-tropics and tropics indirectly. Particularly scenarios from North America and Europe dominate the review (Fig. 1), reflecting well that these are the regions that dominate use, production or trade of solid biofuels [48]. 2.5. Species class and land use history The studies included here model a wide range of tree species and forest types, some as monocultures and others as mixed species forests or stands. In this review the modelled stands or landscapes are characterized as either coniferous, broadleaves or mixed species. The history of land use prior to the stand or landscape being harvested for energy purposes may have an influence on the payback time. Particularly if former land use was agriculture, a large mass of carbon have been sequestered from the atmosphere and stored in living and dead biomass. Natural forests store large amounts of carbon, but may not contribute much to further sequestration of carbon [49,50]. Here I distinguish between three different types of land use history; agriculture, natural forest or plantation forest. 2.6. Counterfactuals Carbon debt studies analyze the impact of bioenergy scenarios as an alternative to other energy supply scenarios. In the studies included here the bioenergy is an alternative to fossil energy. Setting up such alternatives requires a number of ‘what if’ assumptions. What if the forest was not managed for bioenergy production; how would the forest then have been managed? What if biomass (living or dead) was not harvested for energy; what would have happened to it? These ‘what if’ assumptions are usually termed counterfactuals. Mitchel et al. [10] distinguished between carbon debt repayment and carbon offset parity. Carbon debt repayment represents the time it takes a forest bioenergy system to offset temporary
2.1. Payback time In this review I included payback times reported in reviewed literature in the form of tabulated data, clearly legible graphs or payback times described or discussed in-text. Consequently a few publications were disregarded and not all scenarios in all publication are included. E.g. Mitchel et al. [10] analyze 1764 different scenarios for increased harvest in Ontario, 14 of these are included here as they are specifically discussed in the paper. 1212
Renewable and Sustainable Energy Reviews 73 (2017) 1211–1217
N.S. Bentsen
Table 1 Variables included in the analysis of patterns in carbon debt payback times. † The role of individual variables is characterized as D=dependent/response or I=independent. ‡ Variables are either Con=continuous or Cat=categorical. Variable
Role†
Type‡
Levels
Values
Payback time Scale Data Model
D I I I
Con Cat Cat Cat
– 3 2 13
Forest baseline Biome Geography Species class Land use history Counterfactual forest management Counterfactual biomass destiny Fossils displaced Energy service Biomass resource
I I I I I I I I I I
Cat Cat Cat Cat Cat Cat Cat Cat Cat Cat
2 3 7 3 3 2 3 3 4 5
0–1000 Stand, fixed landscape, dynamic landscape Hypothetical, spatially explicit CBM-CFS3, FSM, FORCARB-ON, FRAMES, FVS, GFLUM, GORCAM, LANDCARB, Q-model, SIMA, SRTS, STANDCARB, Yasso07 Reference point, dynamic Boreal, sub-boreal, temperate Australia, Canada, Central Europe, Europe, Globe, North Europe, USA Coniferous, broadleaves, mixed species Agriculture, natural forest, plantation Continued timber harvest, protection Roadside burning, natural decay, continued growth Coal, oil, natural gas Combined heat and power, electricity, heat, liquid fuel Whole tree, round wood, residue, stump, mixed
2.7. Biomass resource
reductions in carbon storage resulting from bioenergy production. Carbon offset parity represents the time it takes for the same system to equal the amount of carbon that would be stored had the forest been left unharvested. Other studies make similar distinctions, but definitions and nomenclature are not used consequently and unambiguously [3,13]. Here studies are characterized as assuming a reference point or a dynamic forest baseline. A reference point forest baseline defines a static carbon reservoir in the forest and compares future carbon exchanges between the bioenergy system and the atmosphere against this static reference state. A dynamic forest baseline includes a modelled trend in carbon pools in dead and living biomass. Relating to, but not necessarily identical to assumptions on forest baselines are assumptions on future forest management practices. Two assumptions are identified; forests left unmanaged in the absence of bioenergy or forests continue under some sort of management including harvest of wood. The last counterfactual included in the analysis characterizes the fate of the potential bioenergy resources in the absence of a bioenergy program or tries to answer the question: What would have happened to the biomass in the absence of a forest bioenergy program. The counterfactual biomass fate is characterized as roadside burning, natural decay or continued growth.
The studies included in this review model bioenergy based on a variety of different biomass resources or fractions. Five different biomass fractions were identified in the studies used for classification i.e. whole tree harvest, round wood, forest residues, stumps or mixed sources. 2.8. Fossils displaced and energy service Harvest or additional extraction of forest biomass has the purpose to displace fossil resources and it could be expected that the type of fossil resource displaced would have an impact on the time it takes to bayback the carbon debt. Fossil resources displaced are characterized as coal, oil, natural gas, or mixed resources that include different fossil resources in various proportions. Forest biomass can be used to meet different energy services. In the studies included her I identified four different energy services; combined heat and power production, electricity production, heat production or the production of liquid transportation fuels. 3. Theory/calculation To search for patterns and commonalities across the different studies binary recursive partitioning was used. Partitioning is a
Fig. 1. Overview of the origin of scenarios and references included in the pattern search analysis. Country name indicates the country on which the carbon debt analysis is performed. The first number indicates the number of references reviewed; the second number indicates the number of scenarios or cases from that country.
1213
Renewable and Sustainable Energy Reviews 73 (2017) 1211–1217
N.S. Bentsen
methodology within the area of machine learning or supervised learning. Partitioning is a multivariate non-parametric procedure that recursively partitions data to find groupings in predictor variables that best predicts a response variable. Partitioning does not require prior knowledge of distributions, models and interactions, but it might be computationally demanding. Recursive partitioning has shown to be a powerful tool for meta-analyses on forest management and carbon debt studies [13], but also in other areas of ecological research [51,52]. The modelling was performed with the JMP12Pro software [53]. 3.1. Splitting criterion Splitting is based on the LogWorth statistic, which is calculated as: −log10(p-value). The highest LogWorth value determines the next split. The p values are based on the null hypothesis that there is no difference in the target mean between two groups [53]. 3.2. Model validation
Fig. 3. Relative contribution of model parameters to the partitioning model.
There is a trade-off between the model's descriptive power and its predictive power. Increasing the number of splits increase the descriptive power at the risk of overfitting the model and thus reducing the model's ability to predict the response of a new dataset. To validate the model I used k-fold validation with k=5. The model is trained on the entire data set and the validation procedure splits the same data randomly into 5 portions to test the models ability to predict each of the five subsets. K-fold validation was preferred over hold-out validation because of the relatively small data set (N=245) with relatively many variables (N=13) analyzed here. To avoid overfitting and build a model as simple as possible the optimal number of splits was determined with the so-called one minus SE rule [54]. This stopping rule selects the smallest regression tree within the range of one standard error of the minimum SSE value of the validation set. The stopping rule selected an optimal model with 15 splits (Fig. 2).
The choice of model comes out as the most influential parameter in determining carbon debt pay back times. The parameter contributes 66% to the model description. Two of the 13 namedmodel, i.e. the STANDCARB and the LANDCARB model, tend to return comparably long payback times. Both of these models are developed at Oregon State University. The mean payback time found in scenarios built on the STANDCARB model [29] is 234 years and 392 years for the LANDCARB model [10]. For the other 11 named models the mean payback time varies between 2 and 52 years (Fig. 4). Scenarios built on the STANDCARB or the LANDCARB model all focus on roundwood harvest for energy from the temperate forests of the Pacific North West of USA. Buchholz et al. [13] showed a comparable influence. In their meta-analysis the variable ‘Author clusters’ came out as highly influential on payback times. Studies from three authors, author groups or institutions (Bernier & Pare [24], Holtsmark [31,55] or Joanneum Research in Graz, Austria) returned a mean payback time of 112 years, while the mean of other studies were 26 years. The strong model dependence calls for caution in policy making, which must acknowledge the uncertainty in carbon debt and payback time projections [56]. The second most influential parameter is fossils displaced (10%). When forest biomass is used to displace coal the payback time tends to be shorter (mean=31 years) than when oil or natural gas is displaced
4. Results and discussion The purpose of the study was to identify parameters particularly influential in determining the time it takes a forest bioenergy system to contribute to global warming mitigation. The characterizing parameters included in the analysis (Table 1) are ranked according to their contribution to the model (Fig. 3).
Fig. 2. a: Relative error (SSE) of model validation set as a function on number of splits. The dotted line indicate the one minus SE range for determining the optimal number of splits. b: R2 for training and validation sets. Error bars indicate one standard error of the mean based on three repicates.
1214
Renewable and Sustainable Energy Reviews 73 (2017) 1211–1217
N.S. Bentsen
Fig. 4. Mean and range of carbon payback times in years across the five most influential independent variables (see Table 1 and Fig. 3).
analysis Central Europe is represented by Austria and contributes with 15 scenarios or cases from one study performed at Joanneum Research in Graz. Buchholz et al. [13] found that research by researchers affiliated with Joanneum Research found longer payback times than other author groups. Biomass resource contributes 5% to the model description. Fig. 4 shows that the use of residues and stumps for energy generation return shorter payback times (mean values of 18 and 14 years respectively) than roundwood and whole trees (102 and 74 years respectively). Stump harvest and use for energy is included in only five scenarios, while forest residues appear in 33 scenarios. A number of factors were considered and included in the analysis, but had no or marginal influence on reported payback times. Scale contributed 1% to the model's explanation of data. Studies applying a stand scale scope often model a stand in a certain state prior to harvest. Bioenergy is harvested in a single intervention, where after the stand regrows to the same state as prior to harvest (see. e.g. Cherubini et al. [15]). Under such assumptions the bioenergy harvest event releases a large fraction of the total carbon stored in the ecosystem to the atmosphere. Assuming a static landscape scale evens out annual fluctuations as harvest events in any year comprise smaller parts of the total forest landscape [47] even though the fundamental assumption, that the forest regrows to the same state as prior to harvest, is the same. Nabuurs et al. [12] argue that using a stand scale approach combined with comparisons against a hypothetical forest development baseline assuming that managed forests grow into untouched forests set up a unrealistic counterfactual; at least for European forests. The conceptual difference between a reference point and a dynamic forest baseline is illustrated in e.g. Mitchell et al. [10] or Buchholz et al. [13]. Here the forest baseline contributed 0.6% to the model and returned mean values of 67 and 99 years respectively for reference point and dynamic baselines. In Buchholz et al. [13] baseline assumption is the least influential parameter included in their model. The modest influence of forest baseline may be attributable to the fact that a dynamic baseline not necessarily leads to a larger carbon reservoir in forests absent of bioenergy. Sedjo and Tian [61] argue that an expected demand for forest bioenergy would stimulate forest owners to tend their forests more intensively and lead to increased stocking in forests managed for bioenergy. Similar is shown by Graudal et al. [62], that if the forest owner responds to market signals of increased demand for bioenergy he or she could change management to increase stocking as well as harvest potential. It is widely debated to what age or development stage forests continue to act as sinks. Luyssaert et al. [49] show
(133 and 105 years respectively). This trend does not only hold in this analysis but also when one look at individual studies that included coal and one or more fossils resources in the same modelling framework. Cherubini et al. [14] found that forest biomass displacing coal for heat production had a mean carbon payback time of 41 years, while displacing natural gas returned a payback time of 101 years. Repo et al. [21] showed that forest biomass displacing coal for electricity had a mean payback time of 0 years. Corresponding values for oil and natural gas were 8.5 and 13.5 years respectively. Similar findings, but with other values are reported by e.g. Cintas et al. [44], Schlamadinger et al.[7], Walker et al. [19] and Zanchi et al. [20]. Hektor et al. [57] argue that the most relevant comparator to forest bioenergy should be coal combustion as this is the most common fuel replaced. It is, however, demonstrated that the true marginal technology often is a composite of different technologies and resources influenced by the hour to hour dynamics of the specific energy supply and demand [58], and the competition between energy producing units [59]. York [60] showed more generally that renewable energy does not necessarily displace fossil energy in a one-to-one ratio even when looking at the amount of energy produced because of the influence of the energy market. The studies included in this meta-analysis assume the same amount of energy is produced in fossil and alternative scenarios but differences in conversion efficiency leads to different energy resource consumption. Such an assumption is, following the findings of York [60] and Mathiesen [58], a simplification of the dynamics of the energy system and energy market. Land use history contributes 9% to the model description. Scenarios assuming conversion of natural forests to bioenergy production tend to return much longer payback times than the conversion of agricultural lands or plantation forests. Natural forests tend to store more carbon in living and dead biomass than plantation forests and when that carbon is used for energy and released to the atmosphere a very large carbon debt is instigated and consequently long time is required to build up a similar amount of carbon in new forests. Mean payback time in the studies included here is 136 years. Plantation forests are under some sort of management and a bioenergy program most commonly will increase extraction rates and bioenergy will be part of a multiple product system, which reduces the payback time (mean=25 years). Furthermore forest plantations can be expected grown on more productive locations. Geography tends to play a role in determining payback times (6%). The studies from Central Europe returned a longer payback time (mean=145 years) than other (mean from 29 to 87 years). In this 1215
Renewable and Sustainable Energy Reviews 73 (2017) 1211–1217
N.S. Bentsen
render the concept of evidence-based policy making on bioenergy issues essentially meaningless [79]. At the supply chain management level, the carbon debt concept does provide valuable guidance directing management principles in more climate benign directions. This review indicates that residue biomass should be prioritized over roundwood and whole tree biomass; that wood from forest plantations should be prioritized over wood from natural forests; and that forest biomass preferably should be used to displace coal in the utility sector. Similar recommendations were reported by the Joint Research Center of the European Union [80].
that although net assimilation rates drop with increased stand age, forests continue to act as carbon sinks past several hundred years of age. Similar pattern is found for forests in Canada [63]. For the forests in Scandinavia it is shown that they continue as carbon sinks well beyond normal rotation age [64]. On a regional scale Europe's forests has acted as carbon sinks for decades [65], however Nabuurs et al. [66] found signs of carbon sink saturation. Contrasting views are expressed by e.g. Lippke et al. [50] arguing that past ~100 years forests in the US Douglas regions more or less stop acting as carbon sinks. Similar results are shown for Scandinavia based on simulations with three different models [67]. The carbon dynamics of forests left untouched is context specific and as such different modelling approaches, tools and assumptions may influence carbon payback times in both directions. The year of publication could be expected to be influential taking the development of more advanced models and incorporation of a more comprehensive understanding of forest ecosystem processes into account. This review shows that the number of studies or reported cases has increased dramatically since 2010, but nothing indicates that it has influenced reported payback times. In this review only studies reporting payback times based on cumulative net GHG emissions was considered as the majority of studies published in the last +20 years used this metric. In later years a number of studies have applied payback metrics that expand on cumulative GHG emission. Most frequent is the use of the metric cumulative radiative forcing (CRF), see e.g. [45,68–72]. CRF is calculated by multiplying an impulse response function (IRF) to the annual net GHG emission, and by that modelling the fate of greenhouse gases in the atmosphere and their radiative forcing. Carbon dioxide and other greenhouse gasses emitted to the atmosphere do not stay in the atmosphere permanently. Methane (CH4) and nitrous oxide (N2O) have a relatively short life in the atmosphere, while carbon dioxide is believed to remain in the atmosphere for considerably longer time. As the long term fate of CO2 is not well known a variety of IRFs for CO2 can be used [73] but generally they find that 20–30% of CO2 emitted to the atmosphere remain there for centuries. Further analytical expansions can include calculating temperature impacts and sea level rise caused by increased radiative forcing again caused by net GHG emissions to the atmosphere [74]. These expansions increase the policy relevance of such studies but at the same time increase the scientific uncertainty [45]. Payback times based on cumulative radiative forcing or temperature change are not directly comparable to those based on cumulative GHG emissions; however the ranking of different bioenergy supply chains or management options does not change significantly [28,31,45].
Acknowledgements The research presented here was supported by the ENERWOODS project (www.enerwoods.dk) financed by Nordic Energy Research, Nordic Council of Ministers, grant number 0037m, and the BIORESOURCE project (www.bioresource.dk) financed by the Danish Council for Strategic Research, grant number 11-116725. References [1] Schulze E-D, Körner C, Law BE, Haberl H, Luyssaert S. Large-scale bioenergy from additional harvest of forest biomass is neither sustainable nor greenhouse gas neutral. GCB Bioenergy 2012;4:611–6. [2] Ter-Mikaelian MT, Colombo SJ, Chen J. The burning question: does forest bioenergy reduce carbon emissions? A review of common misconceptions about forest carbon Accounting. J For 2015;113:57–68. [3] Lamers P, Junginger M. The ‘debt’ is in the detail: a synthesis of recent temporal forest carbon analyses on woody biomass for energy. Biofuels, Bioprod Bioref 2013;7:373–85. [4] Dehue B. Implications of a ‘carbon debt' on bioenergy's potential to mitigate climate change. Biofuels, Bioprod Bioref 2013;7:228–34. [5] Fargione J, Hill J, Tilman D, Polasky S, Hawthorne P. Land Clearing and the Biofuel Carbon Debt. Science 2008;319:1235–8. [6] Marland G, Schlamadinger B. Biomass fuels and forest-management strategies: how do we calculate the greenhouse-gas emissions benefits?. Energy 1995;20:1131–40. [7] Schlamadinger B, Marland G. The role of forest and bioenergy strategies in the global carbon cycle. Biomass- Bioenergy 1996;10:275–300. [8] Leemans R, van Amstel A, Battjes C, Kreileman E, Toet S. The land cover and carbon cycle consequences of large-scale utilizations of biomass as an energy source. Glob Environ Change 1996;6:335–57. [9] Watson RT, Zinyowera MC, Moss RH, editors. Climate change 1995-Impacts, adaptions and mitigation of climate change: scientific-technical analyses, Contribution of Working Group II to the Second Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge, UK: Cambridge University Press; 1996. [10] Mitchell SR, Harmon ME, O'Connell KEB. Carbon debt and carbon sequestration parity in forest bioenergy production. GCB Bioenergy 2012;4:818–27. [11] Naudts K, Chen Y, McGrath MJ, Ryder J, Valade A, Otto J, et al. Europe's forest management did not mitigate climate warming. Science 2016;351:597–600. [12] Nabuurs G-J, EJMM Arets, Schelhaas M-J. European forests show no carbon debt, only a long parity effect. Forest Policy and Economics; 2016. [13] Buchholz T, Hurteau MD, Gunn J, Saah D. A global meta-analysis of forest bioenergy greenhouse gas emission accounting studies. GCB Bioenergy 2016;8:281–9. [14] Cherubini F, Bright RM, Stromman AH. Site-specific global warming potentials of biogenic CO2 for bioenergy: contributions from carbon fluxes and albedo dynamics. Environ Res Lett 2012:7. [15] Cherubini F, Peters GP, Berntsen T, Strømman AH, Hertwich E. CO2 emissions from biomass combustion for bioenergy: atmospheric decay and contribution to global warming. GCB Bioenergy 2011;3:413–26. [16] Demirbas A. Combustion characteristics of different biomass fuels. Prog Energy Combust Sci 2004;30:219–30. [17] Capareda S. Introduction to biomass energy conversions. Boca Raton, FL, USA: CRC Press; 2013. [18] Christiansen JR, Gundersen P, Frederiksen P, Vesterdal L. Influence of hydromorphic soil conditions on greenhouse gas emissions and soil carbon stocks in a Danish temperate forest. For Ecol Manag 2012;284:185–95. [19] Walker T, Cardellichio P, Gunn JS, Saah DS, Hagan JM. Carbon accounting for woody biomass from Massachusetts (USA) managed forests: a framework for determining the temporal impacts of wood biomass energy on atmospheric greenhouse gas levels. J Sustain For 2012;32:130–58. [20] Zanchi G, Pena N, Bird N. Is woody bioenergy carbon neutral? A comparative assessment of emissions from consumption of woody bioenergy and fossil fuel. GCB Bioenergy 2012;4:761–72. [21] Repo A, Tuomi M, Liski J. Indirect carbon dioxide emissions from producing bioenergy from forest harvest residues. GCB Bioenergy 2011;3:107–15. [22] Lamers P, Junginger M, Dymond CC, Faaij A. Damaged forests provide an
5. Conclusions This meta-analysis corroborates earlier findings that the outcome of carbon debt studies to a high degree lie in the assumptions [13,75]. Methodological approaches and applied tools determine the findings rather than issues related to ecosystem and supply chain management. On one hand the carbon debt and repayment concept informs the discussion of climate change benefits of bioenergy by challenging the carbon neutrality assumption. On the other hand this review implies that at the current level of development of carbon debt methodologies and their lack of consensus, the concept is inadequate in itself for informing and guiding concrete policy development; too much of the outcomes and conclusions rely on methodology and tools. Future policy making must acknowledge the uncertainty in carbon debt and payback time projections [56] and apply time perspectives relevant not only to the policies themselves [76] but also to forest management [12] and the inertia of the climate system. The strong model dependence calls for scientific attention towards harmonizing projections as also seen for estimates on bioenergy potentials [77], and for a deeper understanding of model difference's influence on outcomes [78]. Lack of scientific consensus on the potential climate benefits of forest bioenergy may 1216
Renewable and Sustainable Energy Reviews 73 (2017) 1211–1217
N.S. Bentsen opportunity to mitigate climate change. GCB Bioenergy 2014;6:44–60. [23] McKechnie J, Colombo S, Chen J, Mabee W, MacLean HL. forest bioenergy or forest carbon? Assessing trade-offs in greenhouse gas mitigation with wood-based fuels. Environ Sci Technol 2010;45:789–95. [24] Bernier P, Paré D. Using ecosystem CO2 measurements to estimate the timing and magnitude of greenhouse gas mitigation potential of forest bioenergy. GCB Bioenergy 2013;5:67–72. [25] Jonker JGG, Junginger M, Faaij A. Carbon payback period and carbon offset parity point of wood pellet production in the South-eastern United States. GCB Bioenergy 2014;6:371–89. [26] Colnes A, Doshi K, Emick H, Evans A, Perschel R, Robards T, et al. Biomass supply and carbon accounting for southeastern forests. Montpelier, VT, USA: Biomass Energy Resource Center; 2012. p. 123. [27] Mika AM, Keeton WS. Net carbon fluxes at stand and landscape scales from wood bioenergy harvests in the US Northeast. GCB Bioenergy 2015;7:438–54. [28] Withers MR, Malina R, Barrett SRH. Carbon, climate, and economic breakeven times for biofuel from woody biomass from managed forests. Ecol Econ 2015;112:45–52. [29] Mitchell SR, Harmon ME, O'Connell KEB. Forest fuel reduction alters fire severity and long-term carbon storage in three Pacific Northwest ecosystems. Ecol Appl 2009;19:643, [55]. [30] Ter-Mikaelian MT, McKechnie J, Colombo SJ, Chen J, MacLean HL. The carbon neutrality assumption for forest bioenergy: a case study for northwestern Ontario. For Chron 2011:87. [31] Holtsmark B. Harvesting in boreal forests and the biofuel carbon debt. Clim Change 2012;112:415, [28]. [32] Baral A, Guha GS. Trees for carbon sequestration or fossil fuel substitution: the issue of cost vs. carbon benefit. Biomass- Bioenergy 2004;27:41–55. [33] Winford EM, Gaither JC. Jr, Carbon outcomes from fuels treatment and bioenergy production in a Sierra Nevada forest. For Ecol Manag 2012;282:1–9. [34] Oliver CD, Nassar NT, Lippke BR, McCarter JB. Carbon, Fossil Fuel, and Biodiversity Mitigation With Wood and Forests. J Sustain For 2014;33:248–75. [35] Ter-Mikaelian MT, Colombo SJ, Lovekin D, McKechnie J, Reynolds R, Titus B, et al. Carbon debt repayment or carbon sequestration parity? Lessons from a forest bioenergy case study in Ontario, Canada. GCB Bioenergy 2015;7:704–16. [36] McKechnie J, MacLean HL. Implications of emissions timing on the costeffectiveness of greenhouse gas mitigation strategies: application to forest bioenergy systems. GCB Bioenergy 2014;6:414–24. [37] Daigneault A, Sohngen B, Sedjo R. Economic approach to assess the forest carbon implications of biomass energy. Environ Sci Technol 2012;46:5664–71. [38] Galik CS, Abt RC. The effect of assessment scale and metric selection on the greenhouse gas benefits of woody biomass. Biomass- Bioenergy 2012;44:1–7. [39] Kilpeläinen A, Kellomäki S, Strandman H. Net atmospheric impacts of forest bioenergy production and utilization in Finnish boreal conditions. GCB Bioenergy 2012;4:811–7. [40] Schlamadinger B, Marland G. Full fuel cycle carbon balances of bioenergy and forestry options. Energy Convers Manag 1996;37:813–8. [41] Schlamadinger B, Spitzer J, Kohlmaier GH, Ludeke M. Carbon balance of bioenergy from logging residues. Biomass- Bioenergy 1995;8:211–34. [42] Zetterberg L, Chen D. The time aspect of bioenergy – climate impacts of solid biofuels due to carbon dynamics. GCB Bioenergy 2015;7:785–96. [43] Ximenes F d A, George BH, Cowie A, Williams J, Kelly G. Greenhouse Gas Balance of Native Forests in New South Wales, Australia. Forests 2012;3:653–83. [44] Cintas O, Berndes G, Cowie AL, Egnell G, Holmström H, Ågren GI. The climate effect of increased forest bioenergy use in Sweden: evaluation at different spatial and temporal scales. Wiley Interdiscip Rev: Energy Environ 2016;5:351–69. [45] Cherubini F, Bright RM, Strømman AH. Global climate impacts of forest bioenergy: what, when and how to measure?. Environ Res Lett 2013;8:1–12. [46] Taeroe A, Mustapha W, Stupak I, Raulund-Rasmussen K. Does forests best mitigate CO2 emissions to the atmosphere by setting them aside for maiximization of carbon storage or by management for fossil fuel substitution? Journal of Environmental Management (accepted). [47] Eliasson P, Svensson M, Olsson M, Ågren GI. Forest carbon balances at the landscape scale investigated with the Q model and the CoupModel – Responses to intensified harvests. For Ecol Manag 2013;290:67–78. [48] Lamers P, Junginger M, Hamelinck C, Faaij A. Developments in international solid biofuel trade—An analysis of volumes, policies, and market factors. Renew Sustain Energy Rev 2012;16:3176–99. [49] Luyssaert S, Schulze ED, Borner A, Knohl A, Hessenmoller D, Law BE, et al. Oldgrowth forests as global carbon sinks. Nature 2008;455:213–5. [50] Lippke B, Oneil E, Harrison R, Skog K, Gustavsson L, Sathre R. Life cycle impacts of forest management and wood utilization on carbon mitigation: knowns and unknowns. Carbon Manag 2011;2:303–33. [51] De'ath G, Fabricius KE. Classification and Regression Trees: a powerful yet simple technique for ecological data analysis. Ecology 2000;81:3178–92. [52] Thiffault E, Béchard A, Paré D, Allen D. Recovery rate of harvest residues for
[53] [54] [55] [56]
[57] [58]
[59]
[60] [61] [62]
[63]
[64]
[65] [66]
[67]
[68]
[69]
[70] [71]
[72]
[73]
[74]
[75]
[76] [77] [78]
[79] [80]
1217
bioenergy in boreal and temperate forests: a review. Wiley Interdiscip Rev: Energy Environ 2015;4:429–51. SAS Institute Inc . JMP® 11 Specialized Models. Cary, NC, USA: SAS Institute Inc.; 2014. Breiman L, Friedman J, Stone CJ, Olshen RA. Classification and Regression Trees. London, UK: Chapman & Hall/CRC; 1984. Holtsmark B. Boreal forest management and its effect on atmospheric CO2. Ecol Model 2013;248:130–4. Berndes G, Abt B, Asikainen A, Cowie A, Dale V, Egnell G, et al. Forest biomass, carbon neutrality and climate change mitigation. Joensuu, FI: European Forest Institute; 2016. Hektor B, Backéus S, Andersson K. Carbon balance for wood production from sustainably managed forests. Biomass- Bioenergy 2016;93:1–5. Mathiesen BV, Münster M, Fruergaard T. Uncertainties related to the identification of the marginal energy technology in consequential life cycle assessments. J Clean Prod 2009;17:1331–8. Lund H, Mathiesen BV, Christensen P, Schmidt JH. Energy system analysis of marginal electricity supply in consequential LCA. Int J Life Cycle Assess 2010;15:260–71. York R. Do alternative energy sources displace fossil fuels?. Nat Clim Change 2012;2:441–3. Sedjo R, Tian X. Does wood bioenergy increase carbon stocks in Forests?. J For 2012;110:304–11. Graudal L, Nielsen UB, Schou E, Thorsen BJ, Hansen JK, Bentsen NS, et al. The contribution of Danish forestry to increase wood production and offset climate change 2010–2100. IEA Bioenergy Task 2016;43. D'Amato AW, Bradford JB, Fraver S, Palik BJ. Forest management for mitigation and adaptation to climate change: insights from long-term silviculture experiments. For Ecol Manag 2011;262:803–16. Framstad E, Wit Hd, Mäkipää R, Larjavaara M, Vesterdal L, Karltun E. Biodiversity, carbon storage and dynamics of old northern forests. TemaNord. Copenhagen, DK: Nordic Council of Ministers; 2013. Forest Europe. State of the Europe’s Forests 2015. Madrid, ES; 2015. p. 312. Nabuurs G-J, Lindner M, Verkerk PJ, Gunia K, Deda P, Michalak R, et al. First signs of carbon sink saturation in European forest biomass. Nat Clim Change 2013;3:792–6. Schmid S, Thurig E, Kaufmann E, Lischke H, Bugmann H. Effect of forest management on future carbon pools and fluxes: a model comparison. For Ecol Manag 2006;237:65–82. Guest G, Cherubini F, Strømman AH. Climate impact potential of utilizing forest residues for bioenergy in Norway. Mitig Adapt Strateg Glob Change 2012;18:1089–108. Guest G, Bright RM, Cherubini F, Strømman AH. Consistent quantification of climate impacts due to biogenic carbon storage across a range of bio-product systems. Environ Impact Assess Rev 2013;43:21–30. Sathre R, Gustavsson L. Time-dependent climate benefits of using forest residues to substitute fossil fuels. Biomass- Bioenergy 2011;35:2506–16. Repo A, Känkänen R, Tuovinen J-P, Antikainen R, Tuomi M, Vanhala P, et al. Forest bioenergy climate impact can be improved by allocating forest residue removal. GCB Bioenergy 2012;4:202–12. Haus S, Gustavsson L, Sathre R. Climate mitigation comparison of woody biomass systems with the inclusion of land-use in the reference fossil system. BiomassBioenergy 2014;65:136–44. Joos F, Roth R, Fuglestvedt JS, Peters GP, Enting IG, von Bloh W, et al. Carbon dioxide and climate impulse response functions for the computation of greenhouse gas metrics: a multi-model analysis. Atmos Chem Phys 2013;13:2793–825. Kirkinen J, Sahay A, Savolainen I. Greenhouse impact of fossil, forest residues and jatropha diesel: a static and dynamic assessment. Prog Ind Ecol, Int J 2009;6:185–206. Holtsmark B. The outcome is in the assumptions: analyzing the effects on atmospheric CO2 levels of increased use of bioenergy from forest biomass. GCB Bioenergy 2013;5:467–73. Ollikainen M. Forest management, public goods and optimal policies. Annu Rev Resour Econ 2016;8:207–26. Searle S, Malins C. A reassessment of global bioenergy potential in 2050. GCB Bioenergy 2015;7:328–36. Gustavsson L, Haus S, Ortiz CA, Sathre R, Le Truong N. Climate effects of bioenergy from forest residues in comparison to fossil energy. Appl Energy 2015;138:36–50. Geden O. The Paris Agreement and the inherent inconsistency of climate policymaking. Wiley Interdiscip Rev: Clim Change 2016;7:790–7. Agostini A, Giuntoli J, Boulamanti A. Carbon accounting of forest bioenergy: conclusions and recommendations from a critical literature review. In: Marelli L, editor. Joint Research Centre. Ispra, Italy: Institute for Energy and Transport; 2013.
Contents lists available at ScienceDirect
Renewable and Sustainable Energy Reviews journal homepage: www.elsevier.com/locate/rser
Carbon debt and payback time – Lost in the forest?
MARK
Niclas Scott Bentsen University of Copenhagen, Faculty of Science, Department of Geosciences and Natural Resource Management, Rolighedsvej 23, DK-1958 Frederiksberg C, Denmark
A R T I C L E I N F O
A BS T RAC T
Keywords: Forest bioenergy Carbon debt Payback time Machine learning Meta-analysis
In later years the potential contribution of forest bioenergy to mitigate climate change has been increasingly questioned due to temporal displacement between CO2 emissions when forest biomass is used for energy and subsequent sequestration of carbon in new biomass. Also disturbance of natural decay of dead biomass when used for energy affect the carbon dynamics of forest ecosystems. These perturbations of forest ecosystems are summarized under the concept of carbon debt and its payback time. Narrative reviews demonstrate that the payback time of apparently comparable forest bioenergy supply scenarios vary by up to 200 years allowing amble room for confusion and dispute about the climate benefits of forest bioenergy. This meta-analysis confirm that the outcome of carbon debt studies lie in the assumptions and find that methodological rather than ecosystem and management related assumptions determine the findings. The study implies that at the current development of carbon debt methodologies and their lack of consensus the concept in it-self is inadequate for informing and guiding policy development. At the management level the carbon debt concept may provide valuable information directing management principles in a more climate benign directions.
1. Introduction In later years the contribution of forest bioenergy to potentially mitigate global warming has been increasingly questioned [1] due to the temporal displacement between CO2 emissions when forest biomass is used for energy and subsequent sequestration of carbon in new biomass. Disturbance of natural decay of dead biomass and growth of living biomass when used for energy affect the carbon dynamics of forest ecosystems. These perturbations of forest ecosystems are summarized under the concept of carbon debt and its payback time. A number of recent narrative reviews discussed the implications of carbon dynamics and carbon debt of forest bioenergy with reference to climate impact and policy development [2–4]. Lamers and Junginger [3] demonstrated that the carbon payback time of apparently comparable forest bioenergy scenarios vary by up to 200 years allowing amble room for confusion and dispute about the potential climate benefit of forest bioenergy. The birth of the carbon debt concept is often attributed a paper in Science in 2008 [5], which did not treat forest bioenergy but potential forest clearing as a consequence of agricultural expansion driven by increased demand for biofuels. The underlying mechanisms describing how forest carbon dynamics may be influenced by increased demand for bioenergy was however treated much earlier [6,7]. An analysis from 1996 by Leemans et al. [8] developed to support the second assessment report of the IPCC [9] describe the now well-known pattern of a
transition period, where increased deployment of bioenergy increases CO2 emissions to the atmosphere followed by an extended period with reduced emissions. More recent papers describe the same pattern conceptually; see e.g. Mitchel et al. [10]. While few if any argue against the existence of a potential carbon debt, quantification of same remains controversial. Naudts et al. [11] reported a carbon debt of Europe's forest of 3.1 Pg C since 1750 because of forest management compared against an untouched forest baseline. Nabuurs et al. [12] contested the relevance of an untouched forest baseline assumption and find no carbon debt in the outlooks for Europe's forest for the same reason, forest management. Carbon debt and payback time studies aim to inform scientists, policy makers, forest managers, the utility sector and other stakeholders on the climate consequences of extracting more biomass from forests to meet an increased demand for non-fossil energy. In the vast body of literature one can find support for almost any view on the climate impact of forest bioenergy, from being instantly beneficial to analyses showing that it will not in the next 10,000 years contribute to global warming mitigation. The objectives of this review are a) to identify patterns and commonalities in assumptions and outcomes across the current scientific literature on forest bioenergy, carbon dynamics and global warming mitigation potential; b) to identify factors influencing carbon debt and payback times of energy production based on forest biomass; and c) to provide guidance to policy and decision makers on how to understand and treat the carbon debt
E-mail address: [email protected]. http://dx.doi.org/10.1016/j.rser.2017.02.004 Received 8 April 2016; Received in revised form 21 December 2016; Accepted 1 February 2017 1364-0321/ © 2017 Elsevier Ltd. All rights reserved.
Renewable and Sustainable Energy Reviews 73 (2017) 1211–1217
N.S. Bentsen
2.2. Scale and data
concept with reference to forest management, energy resource procurement and energy policy development. Buchholz et al. [13] demonstrated recently that reported carbon debt payback times of forest bioenergy are particularly influenced by inclusion of wildfire dynamics in the analyses and models. This review follows in the line of Buchholz et al. [13], but considerably more scenarios or cases are included (245 here vs. 123 in Buchholz et al.) and emphasis is put on exploring how applied methodology and assumptions influence the outcome of carbon debt studies.
Lamers and Junginger [3] distinguished between three different scales when analyzing carbon debts. Same distinction is used here. Stand scale indicate that carbon dynamics is modelled for a uniform even-aged mono-culture. When harvested, the entire stand is cut and the wood is used for materials and/or energy. Analyses applying a fixed landscape scale attempts to counteract the obvious simplification of stand scale studies by assuming a hypothetical landscape of uniform even-aged compartments of monoculture, where each compartment is displaced in time but otherwise modelled as described for stand scale studies. In the fixed landscape the compartments in the forest landscape describe a so-called normal forest where all successional stages from regeneration to final harvest are equally represented [47]. The dynamic landscape representation is based on a true representation of an actual forest landscape encompassing diversity in e.g. species, ages, and rotation lengths. Data underlying the studies are based on either hypothetical data or spatially explicit data, usually from forest inventories.
2. Materials and methods Carbon debt is caused by a number of factors. With respect to forest biomass factors of particular relevance are: Temporal displacement between CO2 emission from biomass conversion to energy and subsequent CO2 sequestration in new biomass. The long rotations in forestry increase the importance of this factor [14,15]. Additional harvest of biomass perturbate the forest ecosystems and may change growth and decay rates of living and dead biomass. Upstream fossil CO2 emissions from resource production and extraction are different than those associated with the extraction of fossil resources displaced by biomass. Differences in efficiency between fossil and biomass conversion technologies. Heat and to some extent electricity can be generated as efficiently with forest biomass as with fossil resources. Biomass with high content of chlorine (Cl), potassium (K2O) and sodium (Na2O) causes corrosion, slagging and fouling of boilers, heat exchangers and super heaters [16,17]. Problems increase with increasing temperatures, why biomass boilers often operate at lower steam temperatures. This is particular true for straw fired boilers, but a high share of bark, leaves and twigs in forest biomass can also reduce attainable operating temperatures. The oxygen to carbon ratio is higher in biomass than in fossil hydrocarbons [16]. Fossil material (old biomass) is in a more reduced state than living or recently dead plant tissues. Consequently CO2 emission per energy unit from combustion is higher from biomass than from coal, oil and natural gas irrespective of conversion efficiencies. Carbon debt is comparable to financial debt in that it can be paid back over a period of time. Increasing harvest of biomass from forests may for a shorter or longer period of time reduce the amount of carbon stored in the forest, either in living or dead biomass. If increased harvest change the hydrology of forest ecosystems due to reduced evapotranspiration increased emissions of methane and nitrous oxide may be observed [18]. When increased harvest of forest biomass is done with a purpose of displacing other resources, GHG emissions from extraction and use of these are avoided. The payback time of carbon debt is modelled as the number of years it takes to reach parity between the cumulated additional emissions from biomass harvest and use, and avoided emissions from extraction and use of displaced resources [10]. This review builds on the scientific literature published in the last 20+ years reporting payback times of using forest biomass to displace fossil resources for energy generation. A total of 245 scenarios are included and characterized relative to a number of descriptive variable presented in Table 1. Data are extracted from [6,7,10,19–46].
2.3. Model A range of models are used to model the carbon dynamics of increased biomass harvest. In this study 13 different named models were found as well as a number of un-named models. Only named model were included, but the methodology accounts for missing values by attributing arbitrary values to the un-named models. 2.4. Biome and geography Here I distinguish between three different biomes from boreal to temperate. Sub-tropic and tropic biomes are not represented specifically in the literature on carbon debt repayment. A few scenarios included in this review have a global scope and include the sub-tropics and tropics indirectly. Particularly scenarios from North America and Europe dominate the review (Fig. 1), reflecting well that these are the regions that dominate use, production or trade of solid biofuels [48]. 2.5. Species class and land use history The studies included here model a wide range of tree species and forest types, some as monocultures and others as mixed species forests or stands. In this review the modelled stands or landscapes are characterized as either coniferous, broadleaves or mixed species. The history of land use prior to the stand or landscape being harvested for energy purposes may have an influence on the payback time. Particularly if former land use was agriculture, a large mass of carbon have been sequestered from the atmosphere and stored in living and dead biomass. Natural forests store large amounts of carbon, but may not contribute much to further sequestration of carbon [49,50]. Here I distinguish between three different types of land use history; agriculture, natural forest or plantation forest. 2.6. Counterfactuals Carbon debt studies analyze the impact of bioenergy scenarios as an alternative to other energy supply scenarios. In the studies included here the bioenergy is an alternative to fossil energy. Setting up such alternatives requires a number of ‘what if’ assumptions. What if the forest was not managed for bioenergy production; how would the forest then have been managed? What if biomass (living or dead) was not harvested for energy; what would have happened to it? These ‘what if’ assumptions are usually termed counterfactuals. Mitchel et al. [10] distinguished between carbon debt repayment and carbon offset parity. Carbon debt repayment represents the time it takes a forest bioenergy system to offset temporary
2.1. Payback time In this review I included payback times reported in reviewed literature in the form of tabulated data, clearly legible graphs or payback times described or discussed in-text. Consequently a few publications were disregarded and not all scenarios in all publication are included. E.g. Mitchel et al. [10] analyze 1764 different scenarios for increased harvest in Ontario, 14 of these are included here as they are specifically discussed in the paper. 1212
Renewable and Sustainable Energy Reviews 73 (2017) 1211–1217
N.S. Bentsen
Table 1 Variables included in the analysis of patterns in carbon debt payback times. † The role of individual variables is characterized as D=dependent/response or I=independent. ‡ Variables are either Con=continuous or Cat=categorical. Variable
Role†
Type‡
Levels
Values
Payback time Scale Data Model
D I I I
Con Cat Cat Cat
– 3 2 13
Forest baseline Biome Geography Species class Land use history Counterfactual forest management Counterfactual biomass destiny Fossils displaced Energy service Biomass resource
I I I I I I I I I I
Cat Cat Cat Cat Cat Cat Cat Cat Cat Cat
2 3 7 3 3 2 3 3 4 5
0–1000 Stand, fixed landscape, dynamic landscape Hypothetical, spatially explicit CBM-CFS3, FSM, FORCARB-ON, FRAMES, FVS, GFLUM, GORCAM, LANDCARB, Q-model, SIMA, SRTS, STANDCARB, Yasso07 Reference point, dynamic Boreal, sub-boreal, temperate Australia, Canada, Central Europe, Europe, Globe, North Europe, USA Coniferous, broadleaves, mixed species Agriculture, natural forest, plantation Continued timber harvest, protection Roadside burning, natural decay, continued growth Coal, oil, natural gas Combined heat and power, electricity, heat, liquid fuel Whole tree, round wood, residue, stump, mixed
2.7. Biomass resource
reductions in carbon storage resulting from bioenergy production. Carbon offset parity represents the time it takes for the same system to equal the amount of carbon that would be stored had the forest been left unharvested. Other studies make similar distinctions, but definitions and nomenclature are not used consequently and unambiguously [3,13]. Here studies are characterized as assuming a reference point or a dynamic forest baseline. A reference point forest baseline defines a static carbon reservoir in the forest and compares future carbon exchanges between the bioenergy system and the atmosphere against this static reference state. A dynamic forest baseline includes a modelled trend in carbon pools in dead and living biomass. Relating to, but not necessarily identical to assumptions on forest baselines are assumptions on future forest management practices. Two assumptions are identified; forests left unmanaged in the absence of bioenergy or forests continue under some sort of management including harvest of wood. The last counterfactual included in the analysis characterizes the fate of the potential bioenergy resources in the absence of a bioenergy program or tries to answer the question: What would have happened to the biomass in the absence of a forest bioenergy program. The counterfactual biomass fate is characterized as roadside burning, natural decay or continued growth.
The studies included in this review model bioenergy based on a variety of different biomass resources or fractions. Five different biomass fractions were identified in the studies used for classification i.e. whole tree harvest, round wood, forest residues, stumps or mixed sources. 2.8. Fossils displaced and energy service Harvest or additional extraction of forest biomass has the purpose to displace fossil resources and it could be expected that the type of fossil resource displaced would have an impact on the time it takes to bayback the carbon debt. Fossil resources displaced are characterized as coal, oil, natural gas, or mixed resources that include different fossil resources in various proportions. Forest biomass can be used to meet different energy services. In the studies included her I identified four different energy services; combined heat and power production, electricity production, heat production or the production of liquid transportation fuels. 3. Theory/calculation To search for patterns and commonalities across the different studies binary recursive partitioning was used. Partitioning is a
Fig. 1. Overview of the origin of scenarios and references included in the pattern search analysis. Country name indicates the country on which the carbon debt analysis is performed. The first number indicates the number of references reviewed; the second number indicates the number of scenarios or cases from that country.
1213
Renewable and Sustainable Energy Reviews 73 (2017) 1211–1217
N.S. Bentsen
methodology within the area of machine learning or supervised learning. Partitioning is a multivariate non-parametric procedure that recursively partitions data to find groupings in predictor variables that best predicts a response variable. Partitioning does not require prior knowledge of distributions, models and interactions, but it might be computationally demanding. Recursive partitioning has shown to be a powerful tool for meta-analyses on forest management and carbon debt studies [13], but also in other areas of ecological research [51,52]. The modelling was performed with the JMP12Pro software [53]. 3.1. Splitting criterion Splitting is based on the LogWorth statistic, which is calculated as: −log10(p-value). The highest LogWorth value determines the next split. The p values are based on the null hypothesis that there is no difference in the target mean between two groups [53]. 3.2. Model validation
Fig. 3. Relative contribution of model parameters to the partitioning model.
There is a trade-off between the model's descriptive power and its predictive power. Increasing the number of splits increase the descriptive power at the risk of overfitting the model and thus reducing the model's ability to predict the response of a new dataset. To validate the model I used k-fold validation with k=5. The model is trained on the entire data set and the validation procedure splits the same data randomly into 5 portions to test the models ability to predict each of the five subsets. K-fold validation was preferred over hold-out validation because of the relatively small data set (N=245) with relatively many variables (N=13) analyzed here. To avoid overfitting and build a model as simple as possible the optimal number of splits was determined with the so-called one minus SE rule [54]. This stopping rule selects the smallest regression tree within the range of one standard error of the minimum SSE value of the validation set. The stopping rule selected an optimal model with 15 splits (Fig. 2).
The choice of model comes out as the most influential parameter in determining carbon debt pay back times. The parameter contributes 66% to the model description. Two of the 13 namedmodel, i.e. the STANDCARB and the LANDCARB model, tend to return comparably long payback times. Both of these models are developed at Oregon State University. The mean payback time found in scenarios built on the STANDCARB model [29] is 234 years and 392 years for the LANDCARB model [10]. For the other 11 named models the mean payback time varies between 2 and 52 years (Fig. 4). Scenarios built on the STANDCARB or the LANDCARB model all focus on roundwood harvest for energy from the temperate forests of the Pacific North West of USA. Buchholz et al. [13] showed a comparable influence. In their meta-analysis the variable ‘Author clusters’ came out as highly influential on payback times. Studies from three authors, author groups or institutions (Bernier & Pare [24], Holtsmark [31,55] or Joanneum Research in Graz, Austria) returned a mean payback time of 112 years, while the mean of other studies were 26 years. The strong model dependence calls for caution in policy making, which must acknowledge the uncertainty in carbon debt and payback time projections [56]. The second most influential parameter is fossils displaced (10%). When forest biomass is used to displace coal the payback time tends to be shorter (mean=31 years) than when oil or natural gas is displaced
4. Results and discussion The purpose of the study was to identify parameters particularly influential in determining the time it takes a forest bioenergy system to contribute to global warming mitigation. The characterizing parameters included in the analysis (Table 1) are ranked according to their contribution to the model (Fig. 3).
Fig. 2. a: Relative error (SSE) of model validation set as a function on number of splits. The dotted line indicate the one minus SE range for determining the optimal number of splits. b: R2 for training and validation sets. Error bars indicate one standard error of the mean based on three repicates.
1214
Renewable and Sustainable Energy Reviews 73 (2017) 1211–1217
N.S. Bentsen
Fig. 4. Mean and range of carbon payback times in years across the five most influential independent variables (see Table 1 and Fig. 3).
analysis Central Europe is represented by Austria and contributes with 15 scenarios or cases from one study performed at Joanneum Research in Graz. Buchholz et al. [13] found that research by researchers affiliated with Joanneum Research found longer payback times than other author groups. Biomass resource contributes 5% to the model description. Fig. 4 shows that the use of residues and stumps for energy generation return shorter payback times (mean values of 18 and 14 years respectively) than roundwood and whole trees (102 and 74 years respectively). Stump harvest and use for energy is included in only five scenarios, while forest residues appear in 33 scenarios. A number of factors were considered and included in the analysis, but had no or marginal influence on reported payback times. Scale contributed 1% to the model's explanation of data. Studies applying a stand scale scope often model a stand in a certain state prior to harvest. Bioenergy is harvested in a single intervention, where after the stand regrows to the same state as prior to harvest (see. e.g. Cherubini et al. [15]). Under such assumptions the bioenergy harvest event releases a large fraction of the total carbon stored in the ecosystem to the atmosphere. Assuming a static landscape scale evens out annual fluctuations as harvest events in any year comprise smaller parts of the total forest landscape [47] even though the fundamental assumption, that the forest regrows to the same state as prior to harvest, is the same. Nabuurs et al. [12] argue that using a stand scale approach combined with comparisons against a hypothetical forest development baseline assuming that managed forests grow into untouched forests set up a unrealistic counterfactual; at least for European forests. The conceptual difference between a reference point and a dynamic forest baseline is illustrated in e.g. Mitchell et al. [10] or Buchholz et al. [13]. Here the forest baseline contributed 0.6% to the model and returned mean values of 67 and 99 years respectively for reference point and dynamic baselines. In Buchholz et al. [13] baseline assumption is the least influential parameter included in their model. The modest influence of forest baseline may be attributable to the fact that a dynamic baseline not necessarily leads to a larger carbon reservoir in forests absent of bioenergy. Sedjo and Tian [61] argue that an expected demand for forest bioenergy would stimulate forest owners to tend their forests more intensively and lead to increased stocking in forests managed for bioenergy. Similar is shown by Graudal et al. [62], that if the forest owner responds to market signals of increased demand for bioenergy he or she could change management to increase stocking as well as harvest potential. It is widely debated to what age or development stage forests continue to act as sinks. Luyssaert et al. [49] show
(133 and 105 years respectively). This trend does not only hold in this analysis but also when one look at individual studies that included coal and one or more fossils resources in the same modelling framework. Cherubini et al. [14] found that forest biomass displacing coal for heat production had a mean carbon payback time of 41 years, while displacing natural gas returned a payback time of 101 years. Repo et al. [21] showed that forest biomass displacing coal for electricity had a mean payback time of 0 years. Corresponding values for oil and natural gas were 8.5 and 13.5 years respectively. Similar findings, but with other values are reported by e.g. Cintas et al. [44], Schlamadinger et al.[7], Walker et al. [19] and Zanchi et al. [20]. Hektor et al. [57] argue that the most relevant comparator to forest bioenergy should be coal combustion as this is the most common fuel replaced. It is, however, demonstrated that the true marginal technology often is a composite of different technologies and resources influenced by the hour to hour dynamics of the specific energy supply and demand [58], and the competition between energy producing units [59]. York [60] showed more generally that renewable energy does not necessarily displace fossil energy in a one-to-one ratio even when looking at the amount of energy produced because of the influence of the energy market. The studies included in this meta-analysis assume the same amount of energy is produced in fossil and alternative scenarios but differences in conversion efficiency leads to different energy resource consumption. Such an assumption is, following the findings of York [60] and Mathiesen [58], a simplification of the dynamics of the energy system and energy market. Land use history contributes 9% to the model description. Scenarios assuming conversion of natural forests to bioenergy production tend to return much longer payback times than the conversion of agricultural lands or plantation forests. Natural forests tend to store more carbon in living and dead biomass than plantation forests and when that carbon is used for energy and released to the atmosphere a very large carbon debt is instigated and consequently long time is required to build up a similar amount of carbon in new forests. Mean payback time in the studies included here is 136 years. Plantation forests are under some sort of management and a bioenergy program most commonly will increase extraction rates and bioenergy will be part of a multiple product system, which reduces the payback time (mean=25 years). Furthermore forest plantations can be expected grown on more productive locations. Geography tends to play a role in determining payback times (6%). The studies from Central Europe returned a longer payback time (mean=145 years) than other (mean from 29 to 87 years). In this 1215
Renewable and Sustainable Energy Reviews 73 (2017) 1211–1217
N.S. Bentsen
render the concept of evidence-based policy making on bioenergy issues essentially meaningless [79]. At the supply chain management level, the carbon debt concept does provide valuable guidance directing management principles in more climate benign directions. This review indicates that residue biomass should be prioritized over roundwood and whole tree biomass; that wood from forest plantations should be prioritized over wood from natural forests; and that forest biomass preferably should be used to displace coal in the utility sector. Similar recommendations were reported by the Joint Research Center of the European Union [80].
that although net assimilation rates drop with increased stand age, forests continue to act as carbon sinks past several hundred years of age. Similar pattern is found for forests in Canada [63]. For the forests in Scandinavia it is shown that they continue as carbon sinks well beyond normal rotation age [64]. On a regional scale Europe's forests has acted as carbon sinks for decades [65], however Nabuurs et al. [66] found signs of carbon sink saturation. Contrasting views are expressed by e.g. Lippke et al. [50] arguing that past ~100 years forests in the US Douglas regions more or less stop acting as carbon sinks. Similar results are shown for Scandinavia based on simulations with three different models [67]. The carbon dynamics of forests left untouched is context specific and as such different modelling approaches, tools and assumptions may influence carbon payback times in both directions. The year of publication could be expected to be influential taking the development of more advanced models and incorporation of a more comprehensive understanding of forest ecosystem processes into account. This review shows that the number of studies or reported cases has increased dramatically since 2010, but nothing indicates that it has influenced reported payback times. In this review only studies reporting payback times based on cumulative net GHG emissions was considered as the majority of studies published in the last +20 years used this metric. In later years a number of studies have applied payback metrics that expand on cumulative GHG emission. Most frequent is the use of the metric cumulative radiative forcing (CRF), see e.g. [45,68–72]. CRF is calculated by multiplying an impulse response function (IRF) to the annual net GHG emission, and by that modelling the fate of greenhouse gases in the atmosphere and their radiative forcing. Carbon dioxide and other greenhouse gasses emitted to the atmosphere do not stay in the atmosphere permanently. Methane (CH4) and nitrous oxide (N2O) have a relatively short life in the atmosphere, while carbon dioxide is believed to remain in the atmosphere for considerably longer time. As the long term fate of CO2 is not well known a variety of IRFs for CO2 can be used [73] but generally they find that 20–30% of CO2 emitted to the atmosphere remain there for centuries. Further analytical expansions can include calculating temperature impacts and sea level rise caused by increased radiative forcing again caused by net GHG emissions to the atmosphere [74]. These expansions increase the policy relevance of such studies but at the same time increase the scientific uncertainty [45]. Payback times based on cumulative radiative forcing or temperature change are not directly comparable to those based on cumulative GHG emissions; however the ranking of different bioenergy supply chains or management options does not change significantly [28,31,45].
Acknowledgements The research presented here was supported by the ENERWOODS project (www.enerwoods.dk) financed by Nordic Energy Research, Nordic Council of Ministers, grant number 0037m, and the BIORESOURCE project (www.bioresource.dk) financed by the Danish Council for Strategic Research, grant number 11-116725. References [1] Schulze E-D, Körner C, Law BE, Haberl H, Luyssaert S. Large-scale bioenergy from additional harvest of forest biomass is neither sustainable nor greenhouse gas neutral. GCB Bioenergy 2012;4:611–6. [2] Ter-Mikaelian MT, Colombo SJ, Chen J. The burning question: does forest bioenergy reduce carbon emissions? A review of common misconceptions about forest carbon Accounting. J For 2015;113:57–68. [3] Lamers P, Junginger M. The ‘debt’ is in the detail: a synthesis of recent temporal forest carbon analyses on woody biomass for energy. Biofuels, Bioprod Bioref 2013;7:373–85. [4] Dehue B. Implications of a ‘carbon debt' on bioenergy's potential to mitigate climate change. Biofuels, Bioprod Bioref 2013;7:228–34. [5] Fargione J, Hill J, Tilman D, Polasky S, Hawthorne P. Land Clearing and the Biofuel Carbon Debt. Science 2008;319:1235–8. [6] Marland G, Schlamadinger B. Biomass fuels and forest-management strategies: how do we calculate the greenhouse-gas emissions benefits?. Energy 1995;20:1131–40. [7] Schlamadinger B, Marland G. The role of forest and bioenergy strategies in the global carbon cycle. Biomass- Bioenergy 1996;10:275–300. [8] Leemans R, van Amstel A, Battjes C, Kreileman E, Toet S. The land cover and carbon cycle consequences of large-scale utilizations of biomass as an energy source. Glob Environ Change 1996;6:335–57. [9] Watson RT, Zinyowera MC, Moss RH, editors. Climate change 1995-Impacts, adaptions and mitigation of climate change: scientific-technical analyses, Contribution of Working Group II to the Second Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge, UK: Cambridge University Press; 1996. [10] Mitchell SR, Harmon ME, O'Connell KEB. Carbon debt and carbon sequestration parity in forest bioenergy production. GCB Bioenergy 2012;4:818–27. [11] Naudts K, Chen Y, McGrath MJ, Ryder J, Valade A, Otto J, et al. Europe's forest management did not mitigate climate warming. Science 2016;351:597–600. [12] Nabuurs G-J, EJMM Arets, Schelhaas M-J. European forests show no carbon debt, only a long parity effect. Forest Policy and Economics; 2016. [13] Buchholz T, Hurteau MD, Gunn J, Saah D. A global meta-analysis of forest bioenergy greenhouse gas emission accounting studies. GCB Bioenergy 2016;8:281–9. [14] Cherubini F, Bright RM, Stromman AH. Site-specific global warming potentials of biogenic CO2 for bioenergy: contributions from carbon fluxes and albedo dynamics. Environ Res Lett 2012:7. [15] Cherubini F, Peters GP, Berntsen T, Strømman AH, Hertwich E. CO2 emissions from biomass combustion for bioenergy: atmospheric decay and contribution to global warming. GCB Bioenergy 2011;3:413–26. [16] Demirbas A. Combustion characteristics of different biomass fuels. Prog Energy Combust Sci 2004;30:219–30. [17] Capareda S. Introduction to biomass energy conversions. Boca Raton, FL, USA: CRC Press; 2013. [18] Christiansen JR, Gundersen P, Frederiksen P, Vesterdal L. Influence of hydromorphic soil conditions on greenhouse gas emissions and soil carbon stocks in a Danish temperate forest. For Ecol Manag 2012;284:185–95. [19] Walker T, Cardellichio P, Gunn JS, Saah DS, Hagan JM. Carbon accounting for woody biomass from Massachusetts (USA) managed forests: a framework for determining the temporal impacts of wood biomass energy on atmospheric greenhouse gas levels. J Sustain For 2012;32:130–58. [20] Zanchi G, Pena N, Bird N. Is woody bioenergy carbon neutral? A comparative assessment of emissions from consumption of woody bioenergy and fossil fuel. GCB Bioenergy 2012;4:761–72. [21] Repo A, Tuomi M, Liski J. Indirect carbon dioxide emissions from producing bioenergy from forest harvest residues. GCB Bioenergy 2011;3:107–15. [22] Lamers P, Junginger M, Dymond CC, Faaij A. Damaged forests provide an
5. Conclusions This meta-analysis corroborates earlier findings that the outcome of carbon debt studies to a high degree lie in the assumptions [13,75]. Methodological approaches and applied tools determine the findings rather than issues related to ecosystem and supply chain management. On one hand the carbon debt and repayment concept informs the discussion of climate change benefits of bioenergy by challenging the carbon neutrality assumption. On the other hand this review implies that at the current level of development of carbon debt methodologies and their lack of consensus, the concept is inadequate in itself for informing and guiding concrete policy development; too much of the outcomes and conclusions rely on methodology and tools. Future policy making must acknowledge the uncertainty in carbon debt and payback time projections [56] and apply time perspectives relevant not only to the policies themselves [76] but also to forest management [12] and the inertia of the climate system. The strong model dependence calls for scientific attention towards harmonizing projections as also seen for estimates on bioenergy potentials [77], and for a deeper understanding of model difference's influence on outcomes [78]. Lack of scientific consensus on the potential climate benefits of forest bioenergy may 1216
Renewable and Sustainable Energy Reviews 73 (2017) 1211–1217
N.S. Bentsen opportunity to mitigate climate change. GCB Bioenergy 2014;6:44–60. [23] McKechnie J, Colombo S, Chen J, Mabee W, MacLean HL. forest bioenergy or forest carbon? Assessing trade-offs in greenhouse gas mitigation with wood-based fuels. Environ Sci Technol 2010;45:789–95. [24] Bernier P, Paré D. Using ecosystem CO2 measurements to estimate the timing and magnitude of greenhouse gas mitigation potential of forest bioenergy. GCB Bioenergy 2013;5:67–72. [25] Jonker JGG, Junginger M, Faaij A. Carbon payback period and carbon offset parity point of wood pellet production in the South-eastern United States. GCB Bioenergy 2014;6:371–89. [26] Colnes A, Doshi K, Emick H, Evans A, Perschel R, Robards T, et al. Biomass supply and carbon accounting for southeastern forests. Montpelier, VT, USA: Biomass Energy Resource Center; 2012. p. 123. [27] Mika AM, Keeton WS. Net carbon fluxes at stand and landscape scales from wood bioenergy harvests in the US Northeast. GCB Bioenergy 2015;7:438–54. [28] Withers MR, Malina R, Barrett SRH. Carbon, climate, and economic breakeven times for biofuel from woody biomass from managed forests. Ecol Econ 2015;112:45–52. [29] Mitchell SR, Harmon ME, O'Connell KEB. Forest fuel reduction alters fire severity and long-term carbon storage in three Pacific Northwest ecosystems. Ecol Appl 2009;19:643, [55]. [30] Ter-Mikaelian MT, McKechnie J, Colombo SJ, Chen J, MacLean HL. The carbon neutrality assumption for forest bioenergy: a case study for northwestern Ontario. For Chron 2011:87. [31] Holtsmark B. Harvesting in boreal forests and the biofuel carbon debt. Clim Change 2012;112:415, [28]. [32] Baral A, Guha GS. Trees for carbon sequestration or fossil fuel substitution: the issue of cost vs. carbon benefit. Biomass- Bioenergy 2004;27:41–55. [33] Winford EM, Gaither JC. Jr, Carbon outcomes from fuels treatment and bioenergy production in a Sierra Nevada forest. For Ecol Manag 2012;282:1–9. [34] Oliver CD, Nassar NT, Lippke BR, McCarter JB. Carbon, Fossil Fuel, and Biodiversity Mitigation With Wood and Forests. J Sustain For 2014;33:248–75. [35] Ter-Mikaelian MT, Colombo SJ, Lovekin D, McKechnie J, Reynolds R, Titus B, et al. Carbon debt repayment or carbon sequestration parity? Lessons from a forest bioenergy case study in Ontario, Canada. GCB Bioenergy 2015;7:704–16. [36] McKechnie J, MacLean HL. Implications of emissions timing on the costeffectiveness of greenhouse gas mitigation strategies: application to forest bioenergy systems. GCB Bioenergy 2014;6:414–24. [37] Daigneault A, Sohngen B, Sedjo R. Economic approach to assess the forest carbon implications of biomass energy. Environ Sci Technol 2012;46:5664–71. [38] Galik CS, Abt RC. The effect of assessment scale and metric selection on the greenhouse gas benefits of woody biomass. Biomass- Bioenergy 2012;44:1–7. [39] Kilpeläinen A, Kellomäki S, Strandman H. Net atmospheric impacts of forest bioenergy production and utilization in Finnish boreal conditions. GCB Bioenergy 2012;4:811–7. [40] Schlamadinger B, Marland G. Full fuel cycle carbon balances of bioenergy and forestry options. Energy Convers Manag 1996;37:813–8. [41] Schlamadinger B, Spitzer J, Kohlmaier GH, Ludeke M. Carbon balance of bioenergy from logging residues. Biomass- Bioenergy 1995;8:211–34. [42] Zetterberg L, Chen D. The time aspect of bioenergy – climate impacts of solid biofuels due to carbon dynamics. GCB Bioenergy 2015;7:785–96. [43] Ximenes F d A, George BH, Cowie A, Williams J, Kelly G. Greenhouse Gas Balance of Native Forests in New South Wales, Australia. Forests 2012;3:653–83. [44] Cintas O, Berndes G, Cowie AL, Egnell G, Holmström H, Ågren GI. The climate effect of increased forest bioenergy use in Sweden: evaluation at different spatial and temporal scales. Wiley Interdiscip Rev: Energy Environ 2016;5:351–69. [45] Cherubini F, Bright RM, Strømman AH. Global climate impacts of forest bioenergy: what, when and how to measure?. Environ Res Lett 2013;8:1–12. [46] Taeroe A, Mustapha W, Stupak I, Raulund-Rasmussen K. Does forests best mitigate CO2 emissions to the atmosphere by setting them aside for maiximization of carbon storage or by management for fossil fuel substitution? Journal of Environmental Management (accepted). [47] Eliasson P, Svensson M, Olsson M, Ågren GI. Forest carbon balances at the landscape scale investigated with the Q model and the CoupModel – Responses to intensified harvests. For Ecol Manag 2013;290:67–78. [48] Lamers P, Junginger M, Hamelinck C, Faaij A. Developments in international solid biofuel trade—An analysis of volumes, policies, and market factors. Renew Sustain Energy Rev 2012;16:3176–99. [49] Luyssaert S, Schulze ED, Borner A, Knohl A, Hessenmoller D, Law BE, et al. Oldgrowth forests as global carbon sinks. Nature 2008;455:213–5. [50] Lippke B, Oneil E, Harrison R, Skog K, Gustavsson L, Sathre R. Life cycle impacts of forest management and wood utilization on carbon mitigation: knowns and unknowns. Carbon Manag 2011;2:303–33. [51] De'ath G, Fabricius KE. Classification and Regression Trees: a powerful yet simple technique for ecological data analysis. Ecology 2000;81:3178–92. [52] Thiffault E, Béchard A, Paré D, Allen D. Recovery rate of harvest residues for
[53] [54] [55] [56]
[57] [58]
[59]
[60] [61] [62]
[63]
[64]
[65] [66]
[67]
[68]
[69]
[70] [71]
[72]
[73]
[74]
[75]
[76] [77] [78]
[79] [80]
1217
bioenergy in boreal and temperate forests: a review. Wiley Interdiscip Rev: Energy Environ 2015;4:429–51. SAS Institute Inc . JMP® 11 Specialized Models. Cary, NC, USA: SAS Institute Inc.; 2014. Breiman L, Friedman J, Stone CJ, Olshen RA. Classification and Regression Trees. London, UK: Chapman & Hall/CRC; 1984. Holtsmark B. Boreal forest management and its effect on atmospheric CO2. Ecol Model 2013;248:130–4. Berndes G, Abt B, Asikainen A, Cowie A, Dale V, Egnell G, et al. Forest biomass, carbon neutrality and climate change mitigation. Joensuu, FI: European Forest Institute; 2016. Hektor B, Backéus S, Andersson K. Carbon balance for wood production from sustainably managed forests. Biomass- Bioenergy 2016;93:1–5. Mathiesen BV, Münster M, Fruergaard T. Uncertainties related to the identification of the marginal energy technology in consequential life cycle assessments. J Clean Prod 2009;17:1331–8. Lund H, Mathiesen BV, Christensen P, Schmidt JH. Energy system analysis of marginal electricity supply in consequential LCA. Int J Life Cycle Assess 2010;15:260–71. York R. Do alternative energy sources displace fossil fuels?. Nat Clim Change 2012;2:441–3. Sedjo R, Tian X. Does wood bioenergy increase carbon stocks in Forests?. J For 2012;110:304–11. Graudal L, Nielsen UB, Schou E, Thorsen BJ, Hansen JK, Bentsen NS, et al. The contribution of Danish forestry to increase wood production and offset climate change 2010–2100. IEA Bioenergy Task 2016;43. D'Amato AW, Bradford JB, Fraver S, Palik BJ. Forest management for mitigation and adaptation to climate change: insights from long-term silviculture experiments. For Ecol Manag 2011;262:803–16. Framstad E, Wit Hd, Mäkipää R, Larjavaara M, Vesterdal L, Karltun E. Biodiversity, carbon storage and dynamics of old northern forests. TemaNord. Copenhagen, DK: Nordic Council of Ministers; 2013. Forest Europe. State of the Europe’s Forests 2015. Madrid, ES; 2015. p. 312. Nabuurs G-J, Lindner M, Verkerk PJ, Gunia K, Deda P, Michalak R, et al. First signs of carbon sink saturation in European forest biomass. Nat Clim Change 2013;3:792–6. Schmid S, Thurig E, Kaufmann E, Lischke H, Bugmann H. Effect of forest management on future carbon pools and fluxes: a model comparison. For Ecol Manag 2006;237:65–82. Guest G, Cherubini F, Strømman AH. Climate impact potential of utilizing forest residues for bioenergy in Norway. Mitig Adapt Strateg Glob Change 2012;18:1089–108. Guest G, Bright RM, Cherubini F, Strømman AH. Consistent quantification of climate impacts due to biogenic carbon storage across a range of bio-product systems. Environ Impact Assess Rev 2013;43:21–30. Sathre R, Gustavsson L. Time-dependent climate benefits of using forest residues to substitute fossil fuels. Biomass- Bioenergy 2011;35:2506–16. Repo A, Känkänen R, Tuovinen J-P, Antikainen R, Tuomi M, Vanhala P, et al. Forest bioenergy climate impact can be improved by allocating forest residue removal. GCB Bioenergy 2012;4:202–12. Haus S, Gustavsson L, Sathre R. Climate mitigation comparison of woody biomass systems with the inclusion of land-use in the reference fossil system. BiomassBioenergy 2014;65:136–44. Joos F, Roth R, Fuglestvedt JS, Peters GP, Enting IG, von Bloh W, et al. Carbon dioxide and climate impulse response functions for the computation of greenhouse gas metrics: a multi-model analysis. Atmos Chem Phys 2013;13:2793–825. Kirkinen J, Sahay A, Savolainen I. Greenhouse impact of fossil, forest residues and jatropha diesel: a static and dynamic assessment. Prog Ind Ecol, Int J 2009;6:185–206. Holtsmark B. The outcome is in the assumptions: analyzing the effects on atmospheric CO2 levels of increased use of bioenergy from forest biomass. GCB Bioenergy 2013;5:467–73. Ollikainen M. Forest management, public goods and optimal policies. Annu Rev Resour Econ 2016;8:207–26. Searle S, Malins C. A reassessment of global bioenergy potential in 2050. GCB Bioenergy 2015;7:328–36. Gustavsson L, Haus S, Ortiz CA, Sathre R, Le Truong N. Climate effects of bioenergy from forest residues in comparison to fossil energy. Appl Energy 2015;138:36–50. Geden O. The Paris Agreement and the inherent inconsistency of climate policymaking. Wiley Interdiscip Rev: Clim Change 2016;7:790–7. Agostini A, Giuntoli J, Boulamanti A. Carbon accounting of forest bioenergy: conclusions and recommendations from a critical literature review. In: Marelli L, editor. Joint Research Centre. Ispra, Italy: Institute for Energy and Transport; 2013.