• Open Access

Quantifying the global warming potential of CO2 emissions from wood fuels

Authors


Abstract

Recent studies have introduced the metric GWPbio, an indicator of the potential global warming impact of CO2 emissions from biofuels. When a time horizon of 100 years was applied, the studies found the GWPbio of bioenergy from slow-growing forests to be significantly lower than the traditionally calculated GWP of CO2 from fossil fuels. This result means that bioenergy is an attractive energy source from a climate mitigation perspective. The present paper provides an improved method for quantifying GWPbio. The method is based on a model of a forest stand that includes basic dynamics and interactions of the forest's multiple carbon pools, including harvest residues, other dead organic matter, and soil carbon. Moreover, the baseline scenario (with no harvest) takes into account that a mature stand will usually continue to capture carbon if not harvested. With these methodological adjustments, the resulting GWPbio estimates are found to be two to three times as high as the estimates of GWPbio found in other studies, and also significantly higher than the GWP of fossil CO2, when a 100-year time horizon is applied. Hence, the climate impact per unit of CO2 emitted seems to be even higher for the combustion of slow-growing biomass than for the combustion of fossil carbon in a 100-year time frame.

Introduction

Global warming potential (GWP) is a frequently used metric when the climate impacts of greenhouse gases (GHGs) need to be compared. GWP quantifies the cumulative potential warming effect of a pulse of GHGs over a specified time frame, taking its absorption of infrared radiation and atmospheric lifetime into account. GWP is a relative measure and the GWP of CO2 is given the ratio 1.

Traditionally, bioenergy has been considered to be carbon neutral because the released carbon is absorbed by the harvested crops' regrowth. Thus, CO2 released from the combustion of bioenergy has been given a GWP of zero in most LCA analyses. For the same reason, no country imposes taxes on CO2 emissions from the combustion of bioenergy, and firms included in the European Union emissions trading market are not committed to acquiring and surrendering allowances for CO2 emissions from the combustion of bioenergy.

There is now increasing agreement that biofuels from forests should not be considered to be carbon neutral because there is a significant time lag between harvesting and regrowth; see for example Chum et al. (2012), Haberl (2013), Haberl et al. (2012a, b), Holtsmark (2013a, b), Hudiburg et al. (2011), Schulze et al. (2012), Searchinger et al. (2009). An article by Fargione et al. (2008) triggered different studies that estimated the length of the carbon debt payback period of biofuels from slow-growing forests (McKechnie et al., 2011; Zanchi et al., 2011; Holtsmark, 2012; Bernier & Paré, 2013; Dehue, 2013; Jonker et al., 2013; Lamers & Junginger, 2013).

At the same time, Cherubini et al. (2011a) introduced a new concept labeled GWPbio, which was proposed as an indicator of the net potential warming impact of CO2 released by the combustion of biomass when it is taken into account how the regrowth of harvested trees recaptures the amount of CO2 that was released by the combustion of the harvest. Taking the time profile of this regrowth into account, the calculated lifetime of a pulse of CO2 from bioenergy was found to be shorter than the lifetime of a pulse of CO2 from fossil fuels. Consequently, it seems reasonable that the potential warming impact of CO2 from bioenergy is smaller than the potential warming impact of CO2 from fossil fuels. With a time horizon of 100 years, Cherubini et al. (2011a) found GWPbio to be 0.43 when they considered a stand of a slow-growing forest that was harvested at an age of 100 years. With shorter rotations, they found lower estimates of GWPbio.

Later Cherubini et al. (2011b, 2012), Bright et al. (2012), Guest et al. (2012), and Pingoud et al. (2011) presented estimates of GWPbio in the interval 0.34–0.62 when slow-growing forest stands were considered. The fact that these estimates are significantly below 1.0 indicates that bioenergy from slow-growing forests ‘becomes an attractive climate change mitigation option’ (Cherubini et al., 2011b, p 65).

However, an examination of the above-mentioned papers reveals some serious weaknesses with regard to their method for modeling and calculating GWPbio. Bright et al. (2012), Cherubini et al. (2011a, b, 2012) and Pingoud et al. (2011) applied models of a forest stand that did not include the harvest's effects on the dynamics of important carbon pools such as harvest residues, natural deadwood, and soil carbon. Moreover, only Pingoud et al. (2011) included a realistic baseline scenario. The other studies made the simplifying assumption that, if not harvested, there is no further growth and accumulation of carbon in a mature stand. Guest et al. (2012) did include harvest residues in their analysis, but they did not include natural deadwood or effects on soil carbon, and they did not construct a realistic baseline scenario.

The main purpose of this paper is to present an improved method for modeling and quantifying the potential warming impact of CO2 from the combustion of biomass from slow-growing forests (GWPbio). The methodological improvements are related to the model of the considered forest stand and the construction of a realistic baseline scenario (Helin et al. 2012, Holtsmark, 2013a, b).

Firstly, the proposed method applies a no-harvest baseline scenario that takes into account that stands are usually considered mature and therefore harvested before growth has culminated (Faustmann, 1849; Samuelson, 1976; Scorgie & Kennedy, 1996; Holtsmark et al., 2012). At this point, note that when this paper uses the expression ‘a mature stand’, it refers to a stand that is considered ready for harvesting. The Faustmann rule states that a stand should be harvested before the point in time when marginal growth drops below average growth; see Holtsmark, 2012, 2013b for further details.

Secondly, the proposed method includes modeling the dynamics of the forest stand's main carbon pools, including harvest residues, the pool of natural deadwood as well as all parts of growing trees such as branches, tops, stumps, and roots in addition to the stems. The effects of harvesting on the pool of soil carbon are also modeled (Buchholz et al., 2013). Including the forest stand's different carbon pools in the model is important as the dynamics of these pools, which are influenced by harvesting, determine the path of the net carbon flux between the considered forest stand and the atmosphere. The resulting GWPbio ratios will be misleading if only the carbon flux generated by the regrowth of the harvested trees is taken into account, while not taking into account how other carbon fluxes between the considered stand and the atmosphere are altered.

To show the importance of the methodological improvements, the paper also presents some numerical examples. When a 100-year time horizon was applied to a forest stand of age 100 years, the resulting GWPbio estimate was found to be 1.54, i.e., two to three times as high as the estimates of GWPbio found in the above-mentioned studies, and significantly higher than 1. This result is not very sensitive to the considered stand's age. If harvest instead takes place when the stand's age is for example 70 years or 200 years, the GWPbio was found to be 1.79 and 1.38, respectively. Hence, to the extent that GWPbio is a useful index, bioenergy from slow-growing forests is not as attractive from a climate perspective as concluded in some of the above-mentioned studies.

The outline of the paper is as follows. The next section presents a model of a forest stand and all parameter values. Thereafter, the applied model for the accumulation of carbon in the atmosphere is described before the proposed method for calculating GWPbio is presented. The results section consists of three parts. First, the basic results are presented. A number of sensitivity analyses then follow, and a set of model simulations are presented in a subsection to show the effects of the different methodological simplifications made in the studies by Cherubini et al. (2011b), Guest et al. (2012), and Pingoud et al. (2011). It is shown that, with corresponding simplifications of the model applied in this paper, their results are reproduced. Finally, there is a section discussing the results and concluding.

Materials and methods

The model of the forest stand

Figure 1 provides an overview of the basic properties of the model of the considered forest stand. The basis for the estimation of GWPbio is a comparison of the time profile of the forest stand's total carbon stock in the harvest scenario (Fig. 1a) and in the no-harvest scenario (Fig. 1b) and the corresponding net flux of CO2 between the stand and the atmosphere. As a starting point, it was assumed that the stand's age at time of harvest (t = 0) is 100 years, with a total carbon stock of 162 tC (before harvesting). In the harvest scenario, all stems of living trees are removed from the stand at time t = 0 with subsequent combustion giving rise to a pulse of CO2 corresponding to the amount of carbon contained in the stems (39 tC). Hence, after harvesting, the stand stores 123 tC; see Fig. 1a. A case including the use of harvest residues was also considered, giving rise to a correspondingly higher emission pulse at time t = 0 and correspondingly smaller subsequent emissions from the decomposition of residues. After harvesting, new trees start growing; see the hatched and cross-hatched areas in Fig. 1a. Residues left on the forest floor decompose; see the black area. Moreover, natural deadwood (NDOM) that was present in the stand at the time of harvesting also gradually decomposes, while new naturally dead biomass is generated; see the dotted area in Fig. 1a.

Figure 1.

Development of the carbon pools of a single stand in a case where no residues were harvested. (a) The harvest scenario. (b) The no-harvest scenario.

With regard to the dynamics of the soil's carbon pool, it was assumed that harvesting results in some years with a net release of carbon from the soil. Thereafter, the soil's carbon pool gradually returns to its original state; see Fig. 1a.

The development of the stand's carbon stock in the no-harvest baseline scenario is shown in Fig. 1b. The starting point is that the stand's age is 100 years at t = 0. Hence, at time t = 0 in the no-harvest scenario, the sizes of the carbon pools are the same as at time t = 100 in the harvest scenario, cf. Fig. 1a and b. Moreover, in the no-harvest scenario, there is continued forest growth after t = 0 with a corresponding continued accumulation of natural deadwood. In the no-harvest scenario, the soil's carbon pool is assumed to be constant over time.

There is great uncertainty about the likely development of the carbon stock of an old stand (Helin et al. 2012). However, in accordance with, e.g., Luyssaert et al. (2008), I assumed continued accumulation of carbon even in old stands. As this is an uncertain part of the scenario, a sensitivity analysis is carried out with a significantly smaller accumulation of carbon in older stands.

In addition to the case where the stand age is 100 years at time of harvest, the section with sensitivity analyses presents results of simulations where the stand's age is 70 and 200 years at time of harvest, with corresponding adjustments of the dynamics of the carbon pools. For example, if harvest takes place when the stand's age is 200 years, the growth and accumulation of biomass in the baseline no-harvest case is almost negligible.

A detailed description of the construction of the numerical model follows below. The basic building block in the model is the growth function for tree trunks:

display math(1)

where G(τ) is the timber volume of a forest stand of 1 ha (measured in tons of carbon per ha, tC ha−1), whereas τ is the stand's age and v1, v2, and v3 are parameters. This is a functional form frequently used in the literature; see for example Asante & Armstrong (2012), and Asante et al. (2011). The value of the parameters v2 and v3 are based on Cherubini et al. (2011b), and Holtsmark (2013ab). The scale parameter v1 is calibrated, so that, at stand age 100 years, the stand's volume of trunks is 194 m3 ha−1. This is in agreement with results of simulations using the Norwegian forest model AVVIRK-2000 (see Eid & Hobbelstad, 2000), which indicate that harvesting in a typical Norway spruce forest would yield an average harvest of 194 m3 ha−1.

For all calculations, the starting point is that, at time = 0, the stand age is τh and the stand is considered to be mature. In the harvest case, the harvesting takes place at time t = 0, and regrowth restarts along the path described by G(t) whereas, in the baseline scenario, there is no harvest and the forest growth continues along the path described by G(τh + t) as defined in (1).

On average, trunks are assumed to constitute a proportion θ = 0.48 of total living biomass (Løken et al., 2012). Hence, the total living biomass B (τ) in a stand is:

display math(2)

Next, consider the dynamics of the pool of harvest residues. At the time of harvesting, the stock G(τh) of stems is removed from the stand. In addition, a proportion σ of the residues is harvested. Hence, the total harvest is

display math(3)

It will be assumed here that the harvested biomass is used as energy immediately after harvesting. Hence, in the harvest scenario, there will at time t = 0 be a pulse emission equal to E(τh, σ).

In the harvest case, an amount of residues, math formula, is generated at time t = 0, while there are no harvest residues in the baseline scenario. Hence, only in the harvest scenario is there an amount of harvest residues on the forest floor as described by the function:

display math(4)

where ω is the annual decomposition rate for dead organic matter. Based on the results and the discussion in Liski et al. (2005), this parameter was set to 0.04. As it is known that decomposition rates differ greatly between different components of the trees, it would have improved the model to let the speed and time profile of decomposition depend on the type of residues and NDOM components (Repo et al., 2011). However, sensitivity simulations were carried out that showed that the results are relatively insensitive to the size of this parameter; see the results section. Hence, although the results' sensitivity with respect to different decomposition rates for different residue components has not been tested for, this indicates that the assumed speed of decomposition in general is not very important to the results.

Let subscript H refer to the harvest scenario, whereas subscript 0 refers to the reference scenario without harvesting. Consider the pool of natural deadwood, DNi(t), H, 0, which also decomposes at the rate ω. Define the parameters δ0 = 1 and δH = 0. The NDOM pool develops as follows:

display math(5)

where β is a positive parameter and the term math formula represents litterfall, whereas ωDNi(t) represents decomposition. This means that the amount of NDOM generated at time k that is left at time t is math formula Hence, the time profile of the stock of NDOM is as follows:

display math(6)

where D0 represents the amount of DOM in the stand at time t = 0. Thus, the first term on the right-hand side represents the amount of DOM that remains from the previous rotations, and the second term on the right-hand side represents NDOM generated after time t = 0. Based on Asante & Armstrong (2012) and Asante et al. (2011), the parameter β was set to 0.01357.

Next, consider the dynamics of soil carbon. An important question is the extent to which harvesting triggers the release of carbon from soil. As emphasized by Fontaine et al. (2007), Friedland & Gillingham (2010), Jonker et al. (2013), Kjønaas et al. (2000), and Nilsen et al. (2008), the accumulation and possible release of carbon from the soil are complicated processes and there is a high degree of uncertainty here. However, according to field experiments reported by Olsson et al. (1996), the loss of carbon after clear-cutting in a spruce forest could be substantial. Olsson et al. (1996) found that, 15 years after clear-cutting, the net loss of soil carbon from a spruce site is within the range 9–15 tC ha−1. They found that in mature forests most of the soil carbon has been recaptured.

The following model of soil carbon was therefore constructed:

display math

where S0 is the constant amount of soil carbon in the stand in the no-harvest case, whereas s1, s2, and s3 are parameters. The parameter values are given in Table 1. They were calibrated to give a maximum soil carbon loss of 12 tC ha−1 15 years after harvesting. After 15 years, the stand's soil carbon pool was assumed to gradually increase back to its original state, see Fig. 1. Although not important for this analysis, the fixed baseline stock of soil carbon, S0, was set to 60 tC ha−1. This corresponds to a mean of the estimates of the amount of carbon contained in the organic part of the soil found by de Wit & Kvindesland (1999).

Table 1. Parameter values
y 0 0.217
y 1 0.259
y 2 0.338
y 3 0.186
α 1 172.9
α 2 18.51
α 3 1.186
β 0.01357
ω 0.04
v 1 103.067
v 2 0.0245
v 3 2.6925
δ H 0
δ NH 1
s 0.48
s 1 −113.5
s 2 −0.09
s 3 3.003

It should be noted here that it was assumed that forest residue removal does not amplify the loss of soil carbon after harvest and does not reduce future growth. This is probably somewhat optimistic (Johnson & Curtis, 2001).

The stand's total carbon stock, labeled math formula includes the carbon pool of all living biomass B(t), the pool of harvest residues DR (t), the NDOM pool math formula and soil carbon Si(t):

display math(7)

To sum up, in the harvest scenario, there will be a pulse emission math formula at time t = 0, followed by a phase of regrowth and carbon capture, leading to a net flux from the stand to the atmosphere following the path of math formula math formula In the baseline no-harvest scenario, there will be no pulse emission at t = 0, but continued growth will lead to a negative net flux following the path of math formula math formula All parameter values are listed in Table 1. math formula represents the time derivative of math formula which is the net carbon flux from the atmosphere to the stand due to the stands's growth as well as the release of soil carbon and the release of CO2 from the decomposition of harvest residues and NDOM.

Accumulation of carbon in the atmosphere

With regard to the fraction of an initial pulse of CO2 at time t = 0, that remains in the atmosphere at time t, labeled y(t), the following function is applied:

display math(8)

where αi and yi are parameters. This decay function is based on Joos & Bruno (1996), Joos et al. (1996), and Joos et al. (2001), labeled the Bern 2.5CC carbon cycle model. It takes into account how a pulse of CO2 leads to increased absorption of CO2 by the terrestrial biosphere and the sea. For example, the profile of the solid single-lined curve in Fig. 2 describes the remaining proportion at time t of the CO2 pulse generated at time t = 0 in the harvest scenario. However, the Bern 2.5CC model is also applied to fluxes of CO2 generated by the stand's growth, as well as the release of CO2 due to decomposition of NDOM and harvest residues left on the forest floor; see further details below.

Figure 2.

Development of atmospheric carbon released from the forest stand's main carbon pools.

Let AH (t) be the amount of atmospheric carbon at time t that is caused by the harvest with subsequent combustion of the biomass and the stand's regrowth, while A0 (t) is the amount of atmospheric carbon in the no-harvest case, i.e., taking continued growth into account. We then have:

display math(9)
display math(10)

where E(τhσ) represents the pulse emission at time t = 0, see (3). As all variables are measured with regard to their carbon content, this pulse is equal to the harvested biomass E(τh, σ), which depends on the stand age and the proportion of residues harvested.

The terms on the right-hand side of (9) and (10) are illustrated in Fig. 2 in a case without any residues harvested. The solid single-lined curve depicts the first term on the right-hand side of (9), i.e., the remaining share at time t of the pulse emission from combustion of the harvest at time t = 0.

The dashed curve in Fig. 2 represents the second term on the right-hand side of (9), i.e., the accumulated net effect on atmospheric carbon of decomposition of harvest residues left on the forest floor and NDOM (release of carbon) in addition to the stand's regrowth (carbon capture) and net release of soil carbon. In the first phase after harvesting, the decomposition of residues and NDOM dominates the effect of regrowth. Hence, the dashed curve is upward sloping, corresponding to the net release of CO2 from the stand. Later on, when a large proportion of the residues have decomposed, regrowth dominates and the dashed curve becomes downward sloping, which means that there is net accumulation of carbon in the stand.

The dotted curve in Fig. 2 represents the right-hand side of Eqn (10), i.e., the accumulated effect on atmospheric carbon of the continued growth and carbon capture in the forest in the no-harvest scenario.

The net effect on atmospheric carbon of harvesting compared with the baseline scenario without harvesting is:

display math(11)

The double-lined curve in Fig. 2 depicts the profile of A(t) in a case where no residues were harvested. It is found by vertically adding the dashed and solid lines in Fig. 2 and then vertically subtracting the dotted line in the same diagram. The double-lined curve is above the x-axis during the first 115 years after harvesting. This means that the harvest would result in a higher atmospheric CO2 concentration in this period compared to a no-harvest case. At t = 115, the double-lined curve crosses the x-axis. This means that harvesting would result in lower atmospheric CO2 when t ≥ 115, using the no-harvest scenario as the reference point.

Global warming potentials

The concept of global warming potential (GWP) was introduced as a relative measure of how much heat a greenhouse gas traps in the atmosphere compared with the amount of heat trapped by a similar mass of carbon dioxide. Hence, the GWP factor of CO2 is 1. GWP is commonly calculated over time horizons of 20, 100 or 500 years.The absolute global warming potential, math formula, of a CO2 pulse math formula is usually calculated as follows:

display math(12)

where math formula is the radiative forcing effect of CO2 at time t, T is the applied time horizon, whereas y(t) is the proportion of the emission pulse that is still in the atmosphere at time t, see Eqn (8).

More recently, Cherubini et al. (2011a) introduced the concept math formula which is intended to measure the absolute warming potential of a pulse of CO2 caused by the combustion of biomass when it is taken into account that harvesting is followed by regrowth of the trees in the forest stand and other dynamic processes triggered by the harvesting. Using the model of a forest stand described above, the appropriate definition of math formula is then:

display math(13)

where A(t), defined in (11), represents the net effect of harvesting on the atmospheric carbon stock, compared with the baseline scenario without harvesting. To measure the relative global warming effect of biomass combustion, Cherubini et al. (2011a) next defined the math formula factor

display math(14)

The radiative forcing effect of CO2, math formula, is expected to decrease over time as the concentration of CO2 increases. For the sake of simplicity, I will make the approximation that math formula is constant over time (see Caldeira & Kasting, 1993). It follows that (14) could be simplified to:

display math(15)

where Y(t) and A(t) are defined above. For easier interpretation of the results presented in the next section, recall that the profile of A(t) is described by the double-lined curve in Fig. 2, while the profile of math formula is described by the black solid curve in Fig. 2.

Results

Consider the estimates of GWPbio provided in Table 2. Two cases are displayed, one with no residues harvested and one with collection of 25 percent of the residues. A proportion of 25 percent was chosen because that could represent a case in which most of the branches and tops were harvested, while stumps and below ground residues are left in the stand. This study did not consider cases where more than 25 percent of the residues were harvested, because knowledge about the consequences for forest productivity and soil carbon of such harvesting is limited (Helmisaari et al., 2011).

Table 2. Estimates of GWPbio ratios for 20, 100, and 500-year time horizons for different proportions of harvest residues
 TH = 20TH = 100TH = 500
No residues harvested1.921.540.31
25% of residues harvested1.651.250.25

Assuming the stand's age at time of harvest to be 100 years, the GWPbio factor is here estimated to be 1.54 when no residues are harvested (100 years time horizon). If 25 percent of the residues are harvested, GWPbio drops to 1.25.

As mentioned in the introduction, the estimates of GWPbio presented here exceed 1.0, while estimates of GWPbio in earlier studies are significantly below 1.0. For example, when they considered a forest stand that was mature and harvested at a stand age of 100 years, Cherubini et al. (2011a, b) found a GWPbio factor of 0.44 and 0.43, respectively. Moreover, when using a time horizon of 100 years, Guest et al. (2012) found the GWPbio factor to be 0.62 when all harvest residues were left on the forest floor. And finally, Pingoud et al. (2011) estimated the GWPbio factor to be 0.60 when the stand age at felling was 100 years. The discussion section provides a detailed explanation for these differences. It will be shown that implementing different sets of restrictions/simplifications of the model used in this paper means that the model becomes comparable to the models used in the aforementioned studies, and their results are reproduced.

There are various reasons why the estimates of GWPbio found here exceed 1.0, when a time horizon of 100 years was applied. First, the release of CO2 from the decomposition of the residues left on the forest floor is significant and it comes in addition to the pulse emission generated by the combustion of the harvested stems. Secondly, the dynamics of the pool of carbon stored in natural deadwood are important, and especially the different dynamics of this carbon pool in the harvest scenario compared to the baseline no-harvest scenario. More generally, the harvest scenario is evaluated against a no-harvest baseline scenario. In the baseline scenario, there is continued forest growth although at a declining rate, and there is continued accumulation of dead organic matter. Finally, the release of carbon from the soil after harvesting plays a role, although not a major one.

The results discussed so far relate to a time perspective of 100 years. If a time perspective of 500 years is found to be more relevant than the standard 100 years discussed above, the results become significantly more in favor of wood fuels. For example, in the case with no residues harvested, and with a 500-year time horizon, GWPbio was found to be 0.31; see Table 2. See also Fig. 3, which shows how the GWPbio estimates vary depending on the time horizon. With no residues harvested, GWPbio exceeds 1.0 if the time horizon is 166 years or less. If 25 percent of the residues are harvested, GWPbio exceeds 1.0 if the time horizon is 133 years or less.

Figure 3.

GWPbio with and without collection of harvest residues for different time horizons.

As regards the lower GWPbio ratio when residues are harvested, it should be emphasized that the GWPbio factor is a relative (unit-based) measure of warming, see (15). In the case with residues included, more biomass is harvested and combusted, and this generates a larger CO2 emission pulse at the time of harvesting. Recall that the emission pulse math formula is increasing in σ. Hence, even though the relative warming potential of the harvest is lower in the case with residues than in the case where only the stems were harvested, the absolute potential warming impact is higher in the case in which harvest residues are collected. In the case where 25 percent of residues were harvested, the absolute warming potential, as defined in (13), was three percent higher than in the case where no residues were harvested.

Sensitivity analysis

The numerical model used in this paper relies on a number of uncertain factors. At the same time, different assumptions might be of significant importance (Lamers & Junginger, 2013). Forest growth after the age of maturity is probably most important. The accumulation of natural deadwood and the loss of soil carbon after harvesting are other uncertain parts of the model. A sensitivity analysis is therefore presented below. Here, it was assumed that the release of soil carbon to the atmosphere after harvesting and the accumulation of natural deadwood are reduced by 50 percent compared to the reference case. Moreover, it was assumed that, after the stand has reached the age of maturity, the growth of the living biomass is reduced by 50 percent compared to the reference case. Fig. 4 provides an overview of how the revised assumptions change the paths of the carbon pools. While Fig. 4a shows the harvest case, Fig. 4b shows the no-harvest case.

Figure 4.

Sensitivity analysis. Development of the carbon pools of a single stand, in a case with reduced growth in mature stands, reduced accumulation of natural deadwood and smaller effect of harvesting on soil carbon. (a) The harvest scenario. (b) The no-harvest scenario.

The results of the sensitivity analysis are shown in Table 3. In the case without any residues harvested, GWPbio is 1.13 with a time horizon of 100 years. If 25 percent of the residues are harvested, GWPbio drops to 0.94.

Table 3. Sensitivity analysis. Estimates of GWP ratios for 20, 100, and 500-year time horizons for fossil fuels and for wood fuels for different proportions of harvest residues. Case with reduced growth on mature stands, less accumulation of natural deadwood and less loss of soil carbon after harvesting
 TH = 20TH = 100TH = 500
No residues harvested1.601.130.21
25% of residues harvested1.400.940.17

In addition to these sensitivity analyses, the sensitivity of the decomposition rate for dead organic matter (ω) has been checked. In the base case, ω = 0.04. If this parameter was reduced to 0.02 or increased to 0.06, GWPbio changed to 1.60 and 1.49, respectively.

And, finally, it was checked how the harvesting age influences the results. If the harvesting age was reduced to 70 years, GWPbio was increased to 1.79. If the harvesting age was increased to a stand age of 200 years, GWPbio dropped to 1.38. Note that when the stand's age has reached 200 years, its carbon stock is almost in equilibrium.

Comparison of method and results in three other studies

As mentioned above, previous studies estimated the GWPbio ratio to be significantly lower than found in this study, see Table 4. Explanations for these differences are presented in the following, using the papers by Cherubini et al. (2011b), Guest et al. (2012), and Pingoud et al. (2011) as examples. It will be shown that results very close to the results in these three papers were achieved by placing appropriate restrictions on the model applied in this paper.

Table 4. Estimates of GWPbio for 20, 100, and 500-year time horizons when different restrictions are put on the model parameters, and the results of three corresponding studies
 TH = 20TH = 100TH = 500
β = 0, θ = 1, s1 = 0 and use of the function in expression (16)0.960.430.08
Cherubini et al. (2011b)0.970.440.08
β = 0, θ = 0.48, s1 = 0 and use of the function in expression (16)1.160.580.10
Guest et al. (2012)1.300.620.09
β = 0, θ = 1, s1 = 0 and use of the function in expression (1)1.020.610.12
Pingoud et al. (2011)1.000.60na

First, the paper by Cherubini et al. (2011b) is considered. Fig. 5 describes the basic structure of their model, which did not include any residues left on the forest floor or any pools of natural deadwood. Neither was soil carbon included in their model. In the harvest scenario, all biomass from the forest stand was removed at time t = 0 and immediately followed by an emission pulse corresponding to the release of all the carbon stored in the stand. This was followed by regrowth of the stand. With regard to regrowth, Cherubini et al. (2011b) assumed that it follows a typical stand's growth path until the biomass stock has reached the level it had at the time of harvesting. At that point in time, the stand growth is assumed to stop abruptly, as described in Fig. 5a. In the baseline (no-harvest) scenario, the forest stand's carbon stock is constant; see Fig. 5b. Hence, there is no growth and carbon capture in the no-harvest scenario.

Figure 5.

Illustration of the development of carbon stored in the considered forest stand as modeled by Cherubini et al. (2011b). (a) The harvest scenario. (b) The no-harvest scenario.

Certain adjustments and simplifications of the model applied in this paper lead to a model very close to the model used by Cherubini et al. (2011b). With regard to the abrupt cessation of growth in the harvest scenario and the constant carbon stock in the no-harvest case, this could be formulated as follows:

display math(16)

where math formula now is the timber volume of a forest stand of one ha. The function G(τ) was defined in Eqn (1). Moreover, if it is assumed that θ = 1, then harvesting would mean that all living biomass in the stand is removed at the time of harvesting. Assuming that β = 0 means that no dead organic matter is generated, while s1 = 0 means that harvesting has no effects on soil carbon. Using these parameter values and the growth function described by (16), the GWP factor was estimated to be 0.44, which is very close to the result of 0.43 found by Cherubini et al. (2011b). Note that their estimates of GWPbio for the 20 and 500-year time horizons were also reproduced; see the first two rows of results in Table 4.

Next, the study by Guest et al. (2012) is considered. In comparison with Cherubini et al. (2011b), an improved approach was applied by Guest et al. (2012) as they took into account that stems constitute approximately half of the carbon stock of a typical forest stand. It follows that harvest residues then become an issue. Moreover, they considered different scenarios for the extraction of harvest residues. Fig. 6a illustrates their model in the case where all residues were left on the forest floor. In that case, there is an emission pulse at time t = 0 corresponding to the amount of carbon contained in the stock of stems in the stand at the time of harvesting. However, we again observe that, at the point in time when the biomass of the stand has reached the level it had at the time of harvesting, forest growth stops abruptly, see Fig. 6a. Note that Fig. 6b shows their baseline (no-harvest) scenario. In that case, the stand's biomass is fixed.

Figure 6.

Illustration of the development of carbon stored in the considered forest stand as modeled by Guest et al. (2012). (a) The harvest scenario. (b) The no-harvest scenario.

It follows that the model applied in this paper becomes similar to the model applied by Guest et al. (2012) if β = 0 (no naturally dead organic matter is generated) and s1 = 0 (the harvest has no effects on the stock of soil carbon). And, finally, the function described by (16) should be applied instead of (1). With these simplifications of the model, the GWPbio ratio was found to be 0.58, relatively close to the ratio of 0.62 found by Guest et al. (2012); see the third and fourth rows of results in Table 4.

Finally, Pingoud et al. (2011) should be considered. Fig. 7 describes the basic structure of their model, which is similar to the model applied by Cherubini et al. (2011b). They did not include any residues left on the forest floor or any pools of natural deadwood. Neither was soil carbon included in their model. In the harvest scenario, all biomass from the forest stand was removed at time t = 0, and this was immediately followed by an emission pulse corresponding to the release of all the carbon stored in the stand. With regard to regrowth, however, Pingoud et al. (2011) did not assume an abrupt stop at the time of maturity, and a baseline with continued growth was adopted.

Figure 7.

Illustration of the development of carbon stored in the considered forest stand as modeled by Pingoud et al. (2011). (a) The harvest scenario. (b) The no-harvest scenario.

It follows that the model applied in this paper becomes similar to the model applied by Pingoud et al. (2011) if it is assumed that θ = 1 (harvesting would then mean that all living biomass on the stand is removed at the time of harvesting), β = 0 (no naturally dead organic matter is generated) and s1 = 0 (the harvest has no effects on the stock of soil carbon). With these adjustments of the model, the GWPbio ratio was found to be 0.61, relatively close to the ratio of 0.60 found by Pingoud et al. (2011); see the two last rows in Table 4.

The calculations in this section give an indication of the importance of the different assumptions. Some readers might be looking for a more precise quantification of how large a proportion each of the different assumptions contributed to the deviation in results. However, such an exercise might yield limited value added because the interactions between the different assumptions are of crucial importance. Nevertheless, it is clear that the inclusion of harvest residues in the calculations is the most important factor. The inclusion of soil carbon, on the other hand, is of minor importance. However, as emphasized, the release of soil carbon after harvesting is not well understood and might be larger than assumed in this paper.

Discussion

Some comments are warranted on the limitations of the scope of this study. The purpose is not to paint a complete picture of all the environmental pros and cons of wood fuels. For example, this paper only considered CO2. It is left to future studies to include the non-CO2 climate-forcing effects of forestry, for example albedo, the effects of aerosols, etc. (Spracklen et al., 2008; Bright et al., 2012). Moreover, the combustion of both fossil fuels and wood fuels generates, to varying degrees, different substances harmful to health as well as greenhouse gases other than CO2. For example, the combustion of wood fuels in open fireplaces and stoves leads to the release of substantial amounts of methane (CH4); see Haakonsen & Kvingedal (2001) and the IPCC-guidelines for energy, Eggleston et al. (2006). The harmful emissions from wood burning in open fireplaces and stoves are substantial (Haakonsen & Kvingedal, 2001).

It should also be emphasized that the estimated effects of the harvesting and combustion of forest biomass can only be directly applied by decision makers if the harvest from the studied stand is used for bioenergy purposes. The aggregated approach applied does not consider cases in which different forest biomass components originating from harvests are used for other purposes. In conventional forest management, a significant proportion of stemwood from final felling is often used, for example, in construction materials, fiber-products, etc., while only some of the stems are used for energy purposes together with the harvested residues. This supports an approach that identifies the differences in the climate impacts of energy use of specific fractions of the harvest (Repo et al., 2012).

It should also be noted that the approach taken in the present paper results in a description of the climate impacts of current, unchanged forestry practice. The study does not describe the climate impacts of a change in forest management from, e.g., increased harvest levels to meet increased bioenergy targets (described in, e.g., Holtsmark, 2012) or the climate impacts of specific forest biomass fractions, e.g., stems, branches or stumps (as in, e.g., Repo et al., 2011, 2012).

Moreover, it should be noted that knowledge is limited about the extent to which the harvesting of residues will trigger an increased release of soil carbon after the harvest. That is not accounted for in the present calculations. Hence, the estimated GWPbio factors when the harvesting of residues was included might be too optimistic; see the discussion in Repo et al. (2011). Moreover, the collection of forest residues might not just have an impact on soil carbon, but also influence forest growth as well (Helmisaari et al., 2011; Lamers et al., 2013).

With regard to the numerical model applied in this paper, there is a considerable potential for improvement. A simple geometric model for the decomposition of forest residues was applied. Although sensitivity analyses (not presented here) show that the results are relatively insensitive to the speed and profile of the decomposition rate, improvements on this point could easily be implemented, for example, based on the Yasso model (Tuomi et al., 2009, 2011). Moreover, along the same lines, the time profiles for the accumulation of natural deadwood and more general accumulation of dead and living biomass in old forests should be studied further (Carey et al., 2001; Luyssaert et al., 2008). Helin et al. (2012) emphasized that the development of the carbon stock of mature forests is uncertain and that different scenarios in that respect are important. The sensitivity analysis presented points in the same direction. This study is also limited to an analysis of a forest stand that was harvested at a stand age of 100 years. An interesting extension would be to consider stands that grow both faster and slower along with different harvesting ages, as observed in Cherubini et al. (2011a,b).

Despite these limitations, the study still presents an improved method for estimating GWPbio. The proposed method includes modeling the dynamics of the forests' multiple carbon pools, how these pools are impacted by harvesting, and comparing the harvest scenario with a realistic baseline without harvesting. Based on the proposed method, the paper re-estimated the GWPbio ratio. The numerical examples demonstrate that the proposed method results in estimates of GWPbio that are two to three times as high as the estimates of GWPbio found in earlier studies. While earlier studies estimated GWPbio to be significantly below 1, this study estimated GWPbio to be significantly above 1 when a slow-growing forest stand was considered.

An important question is how the estimates of the GWPbio ratio should be interpreted. For example, Cherubini et al. (2011b) found GWPbio estimates significantly below 1 when a 100-year time horizon was applied, and they concluded on that basis that ‘bioenergy becomes an attractive climate mitigation option […] which cools the climate when particular forest management practices are applied’ (Cherubini et al., 2011b, p 65). It should be noted here that any positive GWP values smaller than 1 signify that the climate impact of a mass unit of the greenhouse gas considered is lower than the warming impact of a mass unit of fossil CO2, but it still warms the climate. Only negative GWP values mean that the emissions considered cool the climate in absolute terms. Thus, the conclusions in the quoted text are potentially misleading for the reader.

The results of this paper provide a basis for the following conclusions. First, the climate impact of the harvesting and combustion of slow-growing forest biomass seems to be higher than previous assessments have concluded. Second, the climate impact per unit of CO2 emitted seems to be even higher for the combustion of slow-growing biomass than for the combustion of fossil carbon in a 100-year time frame.

Acknowledgements

The author gratefully acknowledges valuable comments and suggestions from Sebastiaan Luyssaert and Ernst-Detlef Schulze as well as two anonymous referees. While carrying out this research, the author was affiliated to the Oslo Centre for Research on Environmentally friendly Energy (CREE). CREE is supported by the Research Council of Norway. The work has also been supported by the Research Council of Norway through the project ‘Biodiversity and Nature Index: understanding, adaptive planning, and economic policy means for management of open lowlands and forests’ (Project number 204348).

Ancillary