Forest bioenergy harvesting changes carbon balance and risks biodiversity in boreal forest

Abstract


Introduction
Climate solutions relying on forest bioenergy may be in conflict with carbon sequestration and storage by forests as well as conservation of biodiversity.Intensifying biomass harvests for bioenergy production may cause significant net losses of carbon from forests which may partly or entirely offset the emission savings from replacing fossil fuels with bioenergy (Schulze et al. 2012).In addition, the intensification of forest biomass harvests to meet the climate goals has raised concerns on adverse effects on forest biodiversity (EASAC 2017).
Logging residues, such as branches, treetops, and stumps, are an increasingly important source of bioenergy in northern temperate and boreal forests.While biomass from other side streams and waste from forest industry are already used for energy production (Szabó et al. 2011), many studies have identified a large, unused bioenergy potential of logging residues (e.g. de Wit and Faaij 2010).As a result of policies promoting bioenergy and concerns about climate change, extraction rates of these previously unharvested residues are expected to further increase in the future (Mantau et al. 2010).This development can already be seen in Sweden (de Jong and Dahlberg 2017) and in Finland (Peltola 2014).In Finland, bioenergy production from forest chips made mainly from logging residues has multiplied by eight times since the year 2000, and placed forest chips as the most important solid wood fuel in Finland (Peltola 2014).Logging residues are an attractive source of bioenergy because the use of residues does not involve a change in land use nor cause a direct competition for land with food production.Hence, unlike Page 3 of 41 Can.J. For.Res.Downloaded from www.nrcresearchpress.com by JYVASKYLAN YLIOPISTO on 05/28/20 For personal use only.This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition.It may differ from the final official version of record.
the use of agricultural biomass, the use of logging residues is not capped in the EU Renewable Energy directive and use of logging residues will likely expand (COM 2016).
Large-scale logging residue harvesting may pose a conflict with conservation efforts to protect deadwood-dependent species (Bouget et al. 2012, Ranius et al. 2018).In Fennoscandia approximately one quarter of forest species depend on deadwood (Siitonen 2001).
Approximately half of the threatened forests species in Finland require forests rich in decaying wood (Hyvärinen et al. 2019).While national initiatives have been set to preserve and increase the amount of deadwood in forests to improve the state of biodiversity (Hjältén et al. 2010), simultaneous large-scale forest residue harvesting may result in further losses in already scarce deadwood resources and additional detrimental effects on deadwood-dependent species ( e.g.Ranius et al. 2014, Johansson et al. 2016).
Previous studies indicate that logging residue harvesting for bioenergy causes habitat reduction and destruction and changes in the temporal availability of habitats of deadwood-depended species, but the long-term impacts remain uninvestigated (Ranius et al. 2018).Previous empirical studies investigating the effects of forest bioenergy on biodiversity have focused mainly on short-term effects at the stand-level (de Jong & Dahlberg, 2017) or provided a snapshot field data at landscape level (Hiron et al. 2018).Modeling approaches have been introduced to investigate longer term impacts of logging residue extraction on deadwood (Verkerk et al. 2014;Hof et al. 2018a) or deadwood-dependent species (Geijer et al. 2014;Ranius et al. 2014;Johansson et al. 2016;Snäll et al. 2017).These modelling studies have either considered single forest stands or theoretical or real landscapes, and as response variables they have used the amount of dead wood, the population development of real or theoretical model species, or the habitat amount for a wide range of saproxylic organisms based on their specialization to deadwood types.Studies investigating deadwood dynamics or changes in different deadwood types over time on larger landscapes with variation in initial, measured conditions are scarce.The information on deadwood types is important when assessing long-term impacts of logging residue harvesting, because both volume and diversity of deadwood affect the long-term sustainability of populations of deadwood-dependent species (Tikkanen et al. 2007).Hence, despite the large body of empirical research, little is still known about how forest biodiversity would respond to a large-scale removal of logging residues in the long term, and at the landscape scale.The lack of long-term studies limit our ability to predict the effects of harvesting logging residues on deadwood in two ways.First, only long-term studies are sufficient to capture the relatively slow process of wood decay dynamics.Second, year to year fluctuations in the deadwood diversity and abundance are common, and these fluctuations may mask the effects in short-term field experiments (Riffell et al. 2011).
The question whether bioenergy from logging residues truly mitigates climate change has been substantially debated both in public (Kangas et al. 2018) and in scientific forums (Agostini et al. 2013).An increasing number of studies show that logging residue extraction decreases the carbon stock and the sink capacity of forests, which reduces the climate benefits of replacing fossil fuels with bioenergy for years and decades (Agostini et al. 2013).However studies based on single stands may not reflect carbon dynamics in larger landscapes.While the carbon loss due to logging residue removal on individual stands is acknowledged, it may have not been considered problematic because other stands in the landscape could act as carbon sinks compensating the carbon loss (Lamers and Junginger 2013).The initial age structure of the forest stands affects the carbon balance of forest landscape (e.g.Routa et al. 2012).Hence, analyses on larger landscapes are needed to investigate the effects of large-scale logging residue harvesting on carbon dynamics.
A central question in the discussion on the effects of logging residue extraction is if landscape level processes could counteract the adverse effects of residue extraction on carbon balance or deadwood.Hence, even in landscapes where large-scale logging residue harvesting takes place, different types of deadwood could be abundant or increasing in other parts of the landscape where logging residue harvesting is not practiced (Lamers and Junginger 2013).This implies that logging residue harvesting would not decrease deadwood availability for saproxylic species at the landscape level.Similar arguments have been presented for forest carbon balance (Lamers and Junginger 2013).Yet, studies quantifying both changes in forest carbon balance resulting from logging harvesting and subsequent effects on the habitat availability of deadwooddependent species on larger landscapes are missing.In addition, it has not been investigated to what extent forest stands that have a high bioenergy potential also have a high conservation potential for red-listed, deadwood-dependent species.Evaluating this potential overlap in forest values is important from the forest planning perspective.
The objectives of this study were to i) investigate how large-scale stump and slash harvesting for bioenergy affects forest carbon balance and the quality and quantity of deadwood in a landscape, ii) study how the changes in the deadwood affect the amount of suitable habitats of deadwooddependent species and iii) explore the overlap between areas of high bioenergy potential and high conservation potential for selected deadwood-depended species.To address these questions, Page 6 of 41 Can.J. For.Res.Downloaded from www.nrcresearchpress.com by JYVASKYLAN YLIOPISTO on 05/28/20 For personal use only.This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition.It may differ from the final official version of record.
we modeled the forest carbon dynamics and the availability of different types of deadwood in a landscape over 100 years.

Approach
To study the effect of large-scale forest residue harvesting on forest carbon balance and the habitat availability on deadwood-dependent species we simulated the development of forest stands in central Finland.Our study landscape consisted of four watersheds that differed in their productivity and age distribution (Table 1).The modelled landscape encompassed of ca.11 000 hectares and was derived from 6 800 forest stands on mineral soils (Table 1).Measured forest data was provided by the Finnish Forest Center.Our simulation combined modelling of forest carbon budget, deadwood quality and quantity with information of species habitat associations.Species habitat associations modelled with habitat suitability indices were based on expert judgements (Tikkanen et al. 2006(Tikkanen et al. , 2007)).Combining expert judgements with forest models has been used also in previous studies, but on individual stands (Dahlberg et al. 2011;Ranius et al. 2014).In addition, instead of a snapshot of time, our approach provided dynamic estimates of bioenergy resource, forest carbon balance, and the formation and the decay of different types of deadwood in larger forest landscapes Page of 41 Can.J. For.Res.Downloaded from www.nrcresearchpress.com by JYVASKYLAN YLIOPISTO on 05/28/20 For personal use only.This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition.It may differ from the final official version of record.
We simulated three scenarios i) with (BIO) and ii) without (BAU) forest residue harvesting for bioenergy, and iii) set aside (SA), which was used to study the conservation potential of these landscapes without management.In the BIO and BAU scenarios forest stands were managed according to the current forest management recommendations in Finland where forests were clear-cut at the age of 70-90 years (Äijälä et al. 2014).After final felling stands were artificially regenerated by planting or seeding.Stands were tended and thinned two to three times before the final felling.In the BIO scenario branches and treetops were harvested from thinnings and final fellings.In addition, stumps were extracted from all clear-cuts.We assumed that 70% of all available residues were harvested, which was consistent with the extraction percentage reported in field studies (Dahlberg et al. 2011), and with good practice guidance for forest residue harvesting in Finland (Koistinen et al. 2016).We assumed that logging residue harvesting increased the destruction and removal of snags and downed logs by 30% based on the observations of field studies (Hautala et al. 2004;Rudolphi and Gustafsson 2005;Rabinowitsch-Jokinen and Vanha-Majamaa 2010).To define the increase in destruction and removal in the modelling framework we used an iterative approach where we first estimated the baseline reduction in the annual deadwood input to be 60% in the BAU scenario.The rate of destruction and removal resulted deadwood pool values consistent with the measured values in the National Forest Inventory (Peltola 2014).High rate of deadwood destruction and removal are partly explained collection of firewood by forest owners.Foliage was assumed to be left in stands to prevent nutrient loss and corrosion in power plants (Alakangas et al. 2016).We assumed no effect on the growth of the next tree generation.
Page 8 of 41 Can.J. For.Res.Downloaded from www.nrcresearchpress.com by JYVASKYLAN YLIOPISTO on 05/28/20 For personal use only.This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition.It may differ from the final official version of record.
We predicted the forest growth and yield for each stand with a SIMO forest simulator for 100 years.The SIMO modelling framework consists of several models for describing natural processes, such as growth and mortality, and forestry operations.These models are documented in detail in the scientific literature (Kangas and Rasinmäki 2008).The growth and yield predictions were made using growth models by Hynynen et al. (2002), which were developed based on extensive data of field measurements in the National Forest Inventory.Therefore, the models based on this data cover all main tree species and forest site types in Finland (Hynynen et al. 2002).To predict the development of total biomass, simulated stem wood volumes were converted to estimates of total aboveground and belowground biomass through biomass models by Repola et al. (2007) within the SIMO framework.The SIMO model has been shown to produce equally good estimates of forest growth as another widely used Finnish forest simulator MOTTI (Mäkinen et al. 2008).
To estimate changes in litter and soil carbon stocks resulting from forest harvest residue extraction, the litter input from SIMO was used as input to the soil carbon model Yasso07 (Tuomi et al. 2011).Litter input to soil included input from thinnings and final fellings (Kangas and Rasinmäki 2008), natural mortality (Hynynen et al. 2002), and litter input from living trees (Liski et al. 2002).In addition, litter input from understory vegetation to soil was accounted for (Muukkonen and Mäkipää 2006).In the Yasso07 model the decomposition of organic matter depends on climate, litter type and litter diameter (Tuomi et al. 2011).We applied 2 cm diameter for all fine woody litter and an average diameter of 10 cm for coarse woody litter (Raumonen et al. 2011).The Yasso07 model calculates carbon stocks separately for each input type and the sum of these carbon stocks is the total litter and soil carbon stock.We used the carbon stock, calculated from the coarse woody litter without the humus fraction, as a proxy for coarse woody debris.
The initial litter and soil carbon stocks were calculated by running the Yasso07 model to a steady state with an average litter input of current recommended rotation periods (Äijälä et al. 2014).To account for the different stand ages in the landscape in the initialization, the simulated steadystate carbon stocks were adjusted to match the site type and initial age.For this, soil carbon pools were estimated for different stand ages using the information about annual litter input of recommended rotation period.These values were assigned for stands based on stand age, site type and dominant species.
We estimated the potential of the forest landscape consisting of four watersheds to provide net CO 2 emissions reductions with bioenergy while taking into account the changes in forest carbon balance.Bioenergy potentials was calculated by applying net calorific values of 19.3 MJ kg dry biomass -1 for stumps and 19.6 MJ kg dry biomass -1 for branches and treetops (Alakangas et al. 2016).The net caloric values were averages of different deciduous and coniferous species.The effect of bioenergy harvesting on the forest carbon balance was the difference between BAU and BIO scenarios.To show the magnitude of possible net emission reductions with bioenergy from forest harvest residues we applied the emission factor of 93 kg CO 2 GJ -1 to coal (Statistics

Modelling the changes in the available habitat for red-listed deadwood-dependent species
To simulate changes in the amount and the quality of snags and downed logs we linked predictions of the natural mortality (Hynynen et al. 2002) and decomposition through decay stages (1-5) of deadwood (Mäkinen et al. 2006).These models predict 1) initial wood density of the dead tree, 2) probability of dead tree being a snag or a downed log and 3) the change to successive decay class.Initial values for snags and downed deadwood were set to be consistent with the Finnish National Forest inventory data, which provided average volumes per hectare of snags and downed deadwood for tree species for Southern Finland (Peltola 2014).The initial values for snags and downed logs together ranged from 3.1 to 4.5 m 3 ha -1 , and the values were assigned as initial values based on the dominant tree species on studied stands (Cajander 1949).
To study the effects of the scenarios on the availability of the habitats of selected red-listed deadwood dependent species we applied models for habitat suitability indices (HSIs) to calculate species conservation capacity (SCC).This resource-focused approach was chosen as the reduction of deadwood along with reduction of old-growth forests and decreasing number of large trees is the primary cause of threat to threatened forest species, and the second most import cause of regional extinctions in Finland (Hyvärinen et al. 2019).We applied HSI models that connect forest stand characteristics with habitat requirements of red-listed species.The HSI models developed by Tikkanen et al. (2006Tikkanen et al. ( , 2007) ) have been created using expert opinion assessment in which Finnish experts in various taxa assessed the habitat preferences of the redlisted forest species in relation to predefined structures of forest stands and trees within stands.
Then a total of 98 species were grouped to 27 groups according to their shared habitat requirements, which can be connected to stand characteristics modelled with a forest simulator Page 11 of Can.J. For.Res.Downloaded from www.nrcresearchpress.com by JYVASKYLAN YLIOPISTO on 05/28/20 For personal use only.This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition.It may differ from the final official version of record.(Tikkanen et al. 2007).Each group is represented by one red-listed species that typifies the group (Tikkanen et al. 2007).The type species cover 11 fungi, 15 insect and one lichen species.The stand specific HSI of a type species is a product of deadwood resources, microclimate and temporal continuity of the resource (see Tikkanen et al. 2007 for equations).We estimated the combined SCC in different scenarios for each stand s for different HSIs describing stand quality for type species k (eq 1, Pakkala et al. 2002).The combined SCC is the weighted average of HSIs, in which the HSIs give the weights.The combined SCC was used as a proxy for a conservation capacity of a stand for all 27 type species for each year.Then the average combined SCC across landscape for each year was calculated for BAU, BIO and SA to compare scenarios.
In addition, to study the effect of residue harvesting on HSIs of different type species, we calculated type species-specific vulnerability to residue harvesting with the methodology introduced by Mazziotta et al. (2016).Type species-specific SCC were calculated to form a proxy for the conservation capacity of all stands for each type species and year in BIO and BAU scenarios.Vulnerabilities were calculated as the differences in the species-specific SCC across stands between BIO and BAU scenarios for each year.These vulnerabilities were summarized across landscape and time for different combinations of species requirements for deadwood resource and microclimate. (1)

Relationships between areas of high bioenergy potential and high potential conservation capacity
Page 12 of 41 Can.J. For.Res.Downloaded from www.nrcresearchpress.com by JYVASKYLAN YLIOPISTO on 05/28/20 For personal use only.This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition.It may differ from the final official version of record.

13
We identified stands with high bioenergy potential and high conservation capacity by classifying stands by combined SCC values in the set aside scenario, and by bioenergy potential in the bioenergy scenario to top 10 % quantiles.This threshold was chosen because the 10% quantiles have been used to identify important sites for conservation planning and ecosystem service hotspots (Schröter et al. 2017).The stands were classified to the highest class if the value was over the highest-class threshold at any point of the simulation period.

Changes in forest carbon stocks and balance
Forest residue harvesting for bioenergy reduced litter and soil carbon stocks on average 3.5 tC ha -1 the beginning of the simulation period and up to 12 tC ha -1 after 100 years.This carbon loss corresponded to a 3-9% decrease in the size of litter and soil carbon stock.Logging residue harvesting reduced the average amount of coarse woody debris 15-31% per hectare.
Development of total carbon stocks varied across watersheds in response to initial age distribution and site productivity.Nevertheless, the reduction in the litter and soil carbon stocks resulting from residue harvesting was of the same order of magnitude in all watersheds, hence the results were presented for all four watersheds together.
Producing bioenergy from forest residues reduced fossil fuel emissions, but also reduced the forest carbon sink compared to scenario where residues were left in place (Figure 1).The annual average bioenergy potential to replace fossil fuels in the landscape of four watersheds ranged from 323 to 1100 TJ year -1 depending on timber harvests in different years.Harvesting stumps, branches and treetops for bioenergy reduced the carbon sink of the landscape (Figure 1, white bars).Although bioenergy reduced fossil fuel emissions (Figure 1 grey bars), the decrease in the carbon sink reduced the net emission savings.Since the logging residues would release CO 2 even if left to decompose in the forest, the negative effect on the forest sink decreased over time.
Hence, after ten years the decrease in the carbon sink reduced the net emission savings 53%, 12% after 50 years and 2% after 100 years (Figure 1, grey and black bars).

Changes in deadwood and the habitat availability for deadwood-depended species
The destruction and removal of snags and logs resulting from logging residue extraction reduced the average amount of deadwood in the landscape 1 -3.7 m 3 ha -1 during the simulation period, and changed the quality of deadwood, especially reducing the amount of recently killed logs or snags of coniferous species (Figure 2) compared to management without residue harvesting.The average relative reduction in the BIO scenario compared to BAU was the largest in recently died spruce logs and snags (-43, and -41%), recently died deciduous logs (-38%) and the smallest in almost totally decayed pine logs (-19%) during the study period.Residue harvesting caused bottlenecks in the availability of at least one deadwood type (volume < 0.001 m 3 ha -1 ) on 185-2724 forest stands, which corresponds to 0.07 -1.7% increase in bottlenecks compared to the BAU scenario.Deadwood was generally a scarce resource in the simulated landscapes with the average total amount of deadwood ranging from 2.2 to 10.1 m3 ha -1 in the BAU scenario.In the set aside scenario, average total amount of deadwood in the landscape was 6 -73 m 3 ha -1 .Hence, the set aside landscape had 2.5 -33 times more deadwood per hectare than the same landscapes managed according to the current recommendations.Logging residue harvesting further increased this difference.Logging residue extraction reduced the combined species conservation capacity of the landscape for the selected deadwood-depended type species on average by 15 -21% compared to BAU and 32-54% compared to the set aside scenario (Figure 3).The species-specific vulnerabilities to residue harvesting were up to 8% during the simulation period.No large differences in the vulnerability to residue harvesting were observed between different type species (Figure 4).
Consequently, irrespective of the specific resource or microclimatic associations of the species, bioenergy harvesting reduced the ability of the landscape to provide habitats for the studied redlisted species.

Relationships between high bioenergy and conservation potential
In the data 17% of stands had both the greatest conservation capacity and the greatest potential for bioenergy exploitation when threshold for the highest potentials were set to top 10% quantile (Figure 5).However, 31% of the forest stands had high bioenergy potential but low conservation potential.Herb-rich heath forests, with Norway spruce as the dominant species, had both the highest bioenergy and the highest conservation potential.Alternatively these characteristics applied also to stands with the highest bioenergy and the lowest conservation potential (Figure 5).Hence, simple forest characteristics such as site type or stand age did not identify stands with high bioenergy potential and high/low conservation potential.

Discussion
In  (Repo et al. 2015a) showing that bioenergy from logging residues is not carbon or climate neutral, and that emission savings with bioenergy come with a delay.In our study after the first decade of residue harvesting changes in carbon balance reduced the net emissions savings by half compared to fossil carbon savings.Reductions of similar magnitude in emission savings were reported by Forsius et al. (2016) in their study conducted in the same region.During the simulation period, the forest carbon stock of the landscape increased in all scenarios because of initial conditions of the studied landscape.
Although initial age distribution and site productivity affected the forest carbon dynamics (Routa et al. 2012), accounting for these factors did not compensate for the reduction in the carbon stocks due to bioenergy harvesting.We assumed that logging residue harvesting does not affect the growth of the next tree generation.Field experiments on whole-tree harvesting report no effect and negative effects of tree growth after final felling (Thiffault et al. 2011) and growth losses after repeated whole-tree harvesting (Kaarakka et al. 2014).If forest residue harvesting caused growth loss, the effects on carbon balance would be more pronounced.
Logging residue harvesting, and the associated destruction of snags and downed logs, halved the amount of some decay classes of logs and snags, and reduced the amount of coarse woody debris for up to one third.Field studies have reported an additional damage, or an immediate loss of deadwood, due to logging residue harvesting ranging from 25 to 88% depending on the definition of deadwood (Eräjää et al. 2010;Rabinowitsch-Jokinen and Vanha-Majamaa 2010).
Some field studies have highlighted the destruction of moderately and well-decayed coarse woody debris (Rabinowitsch-Jokinen and Vanha-Majamaa 2010, Work et al. 2014), while in the current study the largest relative reduction was in the fresh deadwood.In general, in our study the average reduction in the coarse woody debris was lower than reported in field studies.The use of different definitions for deadwood and decay classes make direct comparisons to other studies challenging.
Logging residue harvesting and the associated destruction of snags and downed deadwood reduced the combined species conservation capacity of the landscape for deadwood-dependent red-listed species by one fifth compared to the general effect of forest management.The result may be an under-or overestimation because of limited empirical data on the on additional effect of residue harvesting on destruction and removal of large-diameter deadwood.However, our results highlight the importance of considering this potential loss of large-diameter deadwood in bioenergy harvesting operations.In Finland and Sweden, the majority of declining and red-listed deadwood-dependent species are associated with coarse woody debris rather than with fine woody debris (Siitonen 2001).As none of the red-listed species in Sweden is primarily associated with branches and tree tops of Norway spruce, a review study by de Jong and Dahlberg (2017) suggested that harvesting logging residues of conifers has small to negligible additional impacts on species of conservation interests in Sweden with current extraction levels.
However, our study suggests that the long-term impacts of extensive residue harvesting, and the associated removal and destruction of large-diameter deadwood are important.Since the investigated species are largely red-listed because of the reduction in deadwood (Hyvärinen et al. 2019), a further reduction of suitable habitats because of bioenergy harvesting operations may have detrimental effects on these species.
Our study indicates that combating climate change with bioenergy makes halting biodiversity loss increasingly difficult.In a study by Snäll et al. (2017), stump harvesting for bioenergy caused a rapid decline in lichen metapopulations rendering many currently common lichen species red-listed within two to three decades in Sweden.However, since the lichen metapopulations stabilized at lower equilibrium levels of the stump resource after few decades and climate benefits of stump bioenergy increased with time, Snäll et al. (2017) argued that the trade-off between biodiversity conservation and climate change mitigation with stump bioenergy is transient.This conclusion may result from the assuming a sudden increase in stump harvesting from zero to 50% (Snäll et al. 2017).A slower increase in harvest levels might result in a different conclusion.However, concluding that the trade-off is transient would mean accepting permanently lower population levels for common species.Even if somewhat lower population size does not cause high risk to the common species themselves, it may potentially cause losing some of the specialist species that are dependent on the presence of the common ones.Many Page 18 of 41 Can.J. For.Res.Downloaded from www.nrcresearchpress.com by JYVASKYLAN YLIOPISTO on 05/28/20 For personal use only.This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition.It may differ from the final official version of record.
saproxylic species are, for example, dependent on the so-called priority effects, where a preceding species changes the physiochemical conditions of the deadwood suitable for successor species (Weslien et al. 2011).It has been shown that the successor suffers more of the reduced availability of the resources (Abrego et al. 2017).
Moreover, permanently lower resource availability may be detrimental for species that are already red-listed or depend on specific deadwood types.Our study indicates that logging residue harvesting, and the associated additional deadwood destruction may increase temporal breaks in the availability of the deadwood resources at the landscape scale.These breaks may make the effects of bioenergy harvesting more severe because deadwood continuity has been shown to affect deadwood-dependent biodiversity (Nordén et al. 2014).If such a break happens on a large enough area, some of the specialist species may be totally lost.Johansson et al. (2016) showed that harvesting stumps on 20% of the clear-cuts of the landscape increases extinction risk of rare specialist species to 50%, and further increasing stump extraction to 30% of the clear-cuts negatively affects common species.Thus, extensive logging residue harvesting not only makes halting biodiversity loss less likely, but also increases the likelihood of extinctions of currently threatened species and may put currently common species under risk of becoming threatened.
We assessed the reliability of our results by comparing estimates of different modelling steps to independent data.The changes in the carbon balance due to logging residue harvesting result from changes in the litter and soil carbon pool, and the size of the soil carbon pool is determined by litter input and the decay rate of organic matter decomposition.The annual litter input estimates ranged from 1.4 -3.9 tC ha -1 year -1 , which is consistent with other estimates for Southern Finland based on net primary production (Härkönen et al. 2010), litter measurements (Ukonmaanaho et al. 2008;Ilvesniemi et al. 2009), and with other modelling frameworks (Repo et al. 2015a;Lehtonen et al. 2016).The soil carbon stock estimates were within the range 40-149 tC ha -1 reported for Southern Finland in other studies (Rantakari et al. 2012;Repo et al. 2015a;Lehtonen et al. 2016).The annual changes in the litter and soil carbon pool were modelled with a soil carbon model, which produced stock change estimates and uncertainty estimates of similar magnitude as other models or soil measurements (Rantakari et al. 2012;Ortiz et al. 2013).The average logging residue potentials per hectare were consistent with previous modelling and field studies (Eräjää et al. 2010;Repo et al. 2015b).The quantitative estimates of net emission savings that can be obtained with forest bioenergy depend on the diameters of residues harvested, time scales studied, energy content of forest chips and the fossil fuel replaced (Repo et al. 2012;Alakangas et al. 2016).However, this study shows the magnitude and time dynamics of the emission savings at landscape scale when forests are managed and residues extracted following the current recommendations.It is important to note that species conservation capacity was derived from potential availability of suitable habitats produced by model and habitat preferences based on expert judgment and our model predictions would benefit from validation in field.However, the main findings were consistent with results from a field study in Sweden (Hiron et al. 2018) Setting safe thresholds for forest residue harvesting requires answering three questions: 1) Where   For personal use only.This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition.It may differ from the final official version of record.
. For. Res.Downloaded from www.nrcresearchpress.com by JYVASKYLAN YLIOPISTO on 05/28/20For personal use only.This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition.It may differ from the final official version of record.
. For. Res.Downloaded from www.nrcresearchpress.com by JYVASKYLAN YLIOPISTO on 05/28/20For personal use only.This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition.It may differ from the final official version of record.
. For. Res.Downloaded from www.nrcresearchpress.com by JYVASKYLAN YLIOPISTO on 05/28/20For personal use only.This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition.It may differ from the final official version of record.

Finland
. For. Res.Downloaded from www.nrcresearchpress.com by JYVASKYLAN YLIOPISTO on 05/28/20For personal use only.This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition.It may differ from the final official version of record.

Figure 5 .
Figure 5.The bioenergy potential and the conservation potential (SCC) of forest stands with Can. J. For.Res.Downloaded from www.nrcresearchpress.com by JYVASKYLAN YLIOPISTO on 05/28/20For personal use only.This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition.It may differ from the final official version of record.
Can. J. For.Res.Downloaded from www.nrcresearchpress.com by JYVASKYLAN YLIOPISTO on 05/28/20For personal use only.This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition.It may differ from the final official version of record.
Johansson et al. 2016.idue harvesting reduced both forest carbon stock and carbon sink capacity of forest landscapes, and decreased deadwood compared to business as usual Can.J. For.Res.Downloaded from www.nrcresearchpress.com by JYVASKYLAN YLIOPISTO on 05/28/20For personal use only.This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition.It may differ from the final official version of record.were formed, and the same applied to litter and soil carbon stocks.Hence, large-scale logging residue harvesting increased the outflow of carbon and deadwood from the system compared to business as usual management.Our results show that logging residue harvesting decreases carbon stocks and deadwood availability for saproxylic species even when carbon dynamics are analyzed on landscape level instead of individual stands.These findings corroborate the findings of other studies on stand (e.g.Ranius et al 2011)and landscape level (e.g.Johansson et al. 2016. (Zanchi et al. 2012)ioenergy harvesting.Logging residue harvesting reduced the amount of snags and logs in different decay stages at landscape scale faster than these deadwood resources Changes in the forest carbon balance resulting from logging residue harvesting reduced the climate benefits of bioenergy and resulted in delayed emissions savings.The results support the earlier studies on different scales ranging from forest stands(Zanchi et al. 2012)to forest landscapes (McKechnie et al. 2011) and countries Can. J. For.Res.Downloaded from www.nrcresearchpress.com by JYVASKYLAN YLIOPISTO on 05/28/20For personal use only.This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition.It may differ from the final official version of record.
Can. J. For.Res.Downloaded from www.nrcresearchpress.com by JYVASKYLAN YLIOPISTO on 05/28/20For personal use only.This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition.It may differ from the final official version of record.
Can. J. For.Res.Downloaded from www.nrcresearchpress.com by JYVASKYLAN YLIOPISTO on 05/28/20For personal use only.This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition.It may differ from the final official version of record.

Table 1 .
(Cajander 1949)8bl.2014edfromwww.nrcresearchpress.com by JYVASKYLAN YLIOPISTO on 05/28/20For personal use only.This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition.It may differ from the final official version of record.al.2012).In this study, approximately every sixth stand had both high conservation potential and high bioenergy potential indicating a potential conflict between bioenergy and biodiversity objectives.However, almost one third of stands were characterized by high bioenergy but low conservation potential.Identifying these stands would be a useful way to avoid the conflict.However, our results suggest that stand characteristics, such as the main tree species, stand age and site type alone cannot separate high and low biodiversity potential.Therefore, more precise proxies will be needed for planning sustainable bioenergy harvesting.Regarding which residues to harvest, extracting fast decomposing, small diameter branches for bioenergy has shown to have smaller climate warming impact than producing bioenergy from slowly decaying stumps(Repo et al. 2012).While from the perspective of climate change mitigation focusing residue harvesting on fast decomposing branches might be beneficial, also small diameter branches have been shown to host specialist species(Juutilainen et al. 2014).Generally, de Jong et al. (2014) Can.J. For.Res.Downloaded from www.nrcresearchpress.com by JYVASKYLAN YLIOPISTO on 05/28/20 For personal use only.This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition.It may differ from the final official version of record.species extinction risk (de Jong et al. 2017), and consequently pointed out a need for landscape level planning of residue extraction.Logging residue harvesting is in conflict with the ongoing efforts to increase the amount of deadwood in forests to ensure the favorable state for biodiversity.In the short-term bioenergy from forest residues does not result in deep emission reductions, and it puts a small, but additional, pressure on species already at risk with current forest management practices.Global studies show that biodiversity is likely suffer, if cropland expansion for bioenergy is a major component of climate change mitigation(Hof et al. 2018b).Our study together with previous studies (e.g.Hiron et al. 2018), indicates that biodiversity may be at risk because of bioenergy even without land-use change when forest land stays as forest land but the harvesting intensity is increased.While guidance to good practices in energy wood harvesting along with forest Can.J. For.Res.Downloaded from www.nrcresearchpress.com by JYVASKYLAN YLIOPISTO on 05/28/20For personal use only.This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition.It may differ from the final official version of record.Characteristics of the watersheds.The high productivity site types are classified to herbrich heath forest (OMT: see(Cajander 1949)for classification) and mesic heath forests (MT), while the lower productivity watersheds composed of pine dominated sub-xeric heath forests (VT) and pine dominated xeric heath forests (CT).The development stage "Young" refers to a stand with average diameter at breast height of 8-16 cm and "Mature" similarly average diameter greater than 16 cm but ready for final felling."Ready to harvest" follows the timing of final felling according to the current forest management recommendations in Finland(Äijälä et al., 2014).Can.J. For.Res.Downloaded from www.nrcresearchpress.com by JYVASKYLAN YLIOPISTO on 05/28/20For personal use only.This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition.It may differ from the final official version of record.