Lus Moreu, Evelyne Thiffult, Domini Cyr, Yn Boulnger, Roert Beuregrd

a D′epartement des sciences du bois et de la for^et, Universit′e Laval, Qu′ebec, Canada

b Science and Technology Branch, Environment and Climate Change Canada, Gatineau, Canada

c Laurentian Forestry Centre, Natural Resources Canada, Qu′ebec, Canada

Keywords:Climate change Carbon Forest sector Forest management Boreal landscapes Northern temperate landscape Mitigation potential

ABSTRACT Background: Forest based climate mitigation emerged as a key component of the Paris Agreement, and thus requires robust science to reduce uncertainties related to such strategies. The aim of this study was to assess and compare the cumulative effects on carbon dynamics of forest management and climate change on boreal and northern temperate forest sector in eastern Canada for the 2020–2100 period.Methods: We used the spatially explicit forest landscape model LANDIS-II and its extension Forest Carbon Succession,in conjunction with the Carbon Budget Model for Harvested Wood Products framework.We simulated the dynamics of forest composition and carbon flows from forest ecosystems to wood products and their substitution effect on markets under increasing climate forcing, according to a tonne-year approach. Simulations were conducted for a series of forest management scenarios based on realistic practices principally by clearcut in the boreal territory and continuous-cover forestry in the northern temperate one.These scenarios included:i)a business-asusual scenario (BaU), representing the current management strategy, ii) increased harvesting by 6.3% to 13.9%,iii) increased conservation (i.e. reduced harvesting by 11.1% to 49.8%), iiii) and a scenario representing the natural evolution of the forest landscape (i.e. without any management activity).Results:Our study revealed that increasing harvesting levels had contrasting effects on the mitigation potential in northern temperate(enhance net sequestration)and boreal forest sector(enhance net emissions)in comparison to the BaU from 2040 onwards, regardless of the future climate. Carbon storage in wood products and the substitution effect were not sufficient to offset carbon emissions from ecosystems. Moreover, climate change had a strong impact on the capacity of both landscapes to act as carbon sinks.Northern temperate landscapes became a net source of carbon over time due to their greater vulnerability to climate change than boreal landscapes.Conclusions: Our study highlights the need to consider the initial landscape characteristics in simulations to maximize the mitigation potential of alternative forest management strategies.The optimal management solution can be very different according to the characteristics of forest ecosystems.This opens the possibility of optimizing management for specific forest stands, with the objective of maximizing the mitigation potential of a given landscape.

1. Background

Forest-based climate mitigation has emerged as a key component of the Paris Agreement,providing up to a quarter of planned net reductions in emissions (Friedlingstein et al., 2019; Grassi et al., 2017). Many jurisdictions of the temperate and boreal biomes have been managing their forests and have well-implemented forest industries around which communities are organized. The connection between forest ecosystems and industrial networks allow for a flow of wood products to be used to meet material and energy needs of societies,hence representing a climate change mitigation potential through substitution of GHG-intensive, fossil-based products and an increased accumulation of carbon stock in these wood products.Decisions related to the role of the forest sector for climate change mitigation therefore need to take into account carbon fluxes of forest ecosystems, but also those of wood products that are sourced from forests and their use on markets(Smyth et al.,2014).

Several studies have projected important effects of climate change on forest ecosystems over the next centuries (Boulanger et al., 2017; Boulanger and Pascual Puigdevall,2021;Whitman et al.,2019).Changes in forest productivity and composition as well as in the natural disturbance regimes (fire, insect outbreaks, climate extremes) are expected, with potential consequences on harvested wood volume(Brecka et al.,2020)and on the ability of forests to act as a carbon sink(Valade et al.,2017).

The province of Quebec, in eastern Canada, has significant forest resources distributed along a climatic gradient from temperate hardwood forests in the south to boreal coniferous forests in the north.Latitudinal changes in climate create a gradient of growing season length across Quebec forest landscapes. This allows for transitional zones at the temperate-boreal biome interface due to interspecific competitive limitations.In these northern temperate landscapes,species living at the edge of their range coexist in highly unstable and competitive ecosystems in which small variations in climatic or microclimatic conditions can result in changes in forest structure and composition (Goldblum and Rigg,2010). Northern temperate landscapes are therefore affected by small-gap dynamics that creates irregular structures within stands. In contrast, Quebec forest landscapes located further north in the boreal biome are more stable in terms of composition and structure despite the large and severe natural disturbances occurring in the territory(Gauthier et al.,2015). As a result of these different dynamics, along with the impacts of historical land use and tenure,forest management strategies used in both regions vary markedly: long rotation and large clear-cuts are more common in the north, and recurring small-gap partial logging are more common within the northern temperate landscapes.

Anthropogenic disturbances such as harvesting may disrupt the natural carbon cycling of an ecosystem and thus the forest carbon budget,i.e. the balance between carbon sequestration and emission processes(Malhi et al., 1999; Goldblum and Rigg, 2010). Moreover, the compounding effects of wood harvesting in a changing climate can lead to significant variations in the forest landscape characteristics that may ultimately alter ecosystem capacity to store carbon (Steenberg et al.,2013;Brice et al.,2019).For the main ecozone of Quebec(Boreal Shield East), a reduction in harvesting level, coupled with an increased proportion of wood harvesting going towards long-lived wood products,was predicted to provide the highest mitigation potential until 2050;for other Canadian ecozones, an increase in harvest levels might be more beneficial(Smyth et al.,2014).However,the finer carbon dynamics and fluxes of smaller forest areas with their specific ecosystems and management regimes,and the compounded effects of wood harvesting and a changing climate(for which the impacts are projected to be region-specific),have yet to be investigated. Therefore, it is crucial to analyse how the forest sector at a regional level can sustain or increase its carbon sequestration and emission mitigation potential in a changing environment.

The aim of this study was to assess and compare the cumulative effects of forest management and climate change on carbon fluxes in the boreal and northern temperate forest sectors of Quebec. The objective was to compare the effects on forest productivity,forest composition and the resulting temporal carbon balance of alternative forest management strategies based on local practices under projected scenarios of climate forcing for the 2020–2100 period. Two case studies with contrasting characteristics were selected, located respectively in the boreal (Montmorency Forest) and northern temperate (Hereford Forest) vegetation zones. We used the spatially explicit forest landscape model LANDIS-II and its Forest Carbon Succession (ForCS) extension, in conjunction with the Carbon Budget Model for Harvested Wood Products (CBMFHWP) framework, to simulate forest dynamics and carbon flows from ecosystems to products and markets in a tonne-year approach.

2. Methods

2.1. Case study areas

Two areas were selected for this study: the Montmorency Forest(37,050 ha) and the Hereford Forest (5,668.75 ha). The Montmorency Forest is located in the balsam fir (Abies balsamea) - white birch (Betula papyrifera)bioclimatic domain of Quebec(Saucier et al.,1998),which is part of the boreal vegetation zone.It has a mean annual temperature of 0.5°C and a mean annual rainfall of 1,585 mm (Saucier et al., 1998).Forest species include balsam fir, white birch, white and black spruces(Picea glauca and Picea mariana) and trembling aspen (Populus tremuloides). The Hereford Forest is located in the sugar maple(Acer saccharum)-yellow birch(Betula alleghaniensis)bioclimatic domain and is part of the northern temperate vegetation zone, with a mean annual temperature of 4°C and a mean annual rainfall of 938 mm (Saucier et al.,1998). It comprises a diverse species composition that is typical of the northern hardwood forests (Acer saccharum, Betula alleghaniensis, Abies balsamea, Picea sp., Populus sp., Fagus grandifolia, Quercus sp.) (Saucier et al.,1998).

Both forest areas have been under forest management for several decades, and have experienced strategies that were broadly representative of common practices in their respective area.It should be noted that forest management in the southern regions of Quebec, including the Hereford Forest, has been characterized by high-grading selection cutting.This is why most of these hardwood ecosystems are now considered with low productivity (B′edard and Majcen, 2003; Kenefic et al., 2014;Pr′evost and Charette, 2019). Multiple silvicultural scenarios are experimented for the rehabilitation of these northern hardwood stands(B′edard et al.,2014).Boreal forests are more stable in composition and structure despite both the large and severe natural disturbances occurring in the territory(Gauthier et al.,2015)and the century-old forest management.Sustainable forestry principles are now implemented in the province in accordance with the Sustainable Forest Development Act (Qu′ebec,2013). Ecosystem-based forest management is at the heart of this act:it aims to reduce differences between natural and managed landscapes,and provides regulations for the protection of soil, water and biodiversity(Qu′ebec, 2013).

2.2. Forest management framework

For each of the two territories, several management scenarios were defined based on inputs from local forest managers, reflecting realistic actions for their respective territories.Scenarios representing a gradient of intensity of forest management were applied over the 2020–2100 period. These scenarios included: i) a business-as-usual scenario (BaU),representing the current management strategy, ii) increased harvesting by 6.3%to 13.9%,iii)increased conservation(i.e.reduced harvesting by 11.1% to 49.8%), iiii) and a scenario without any harvest or other management activity, thus representing the natural evolution of the forest landscape. Management scenarios were modulated across the territories according to local characteristics and associated silvicultural practices (Table 1 and Table 2). For For^et Montmorency, intensity of forest management and harvest level varied both by harvested volume(increased or decreased harvested volume)and by type of cut(proportion of clearcut vs. partial cut). For the Hereford forest, the intensity of management and harvest level differed between scenarios according to the rotation time between two partial cuts(either shorter or longer)and by the absence or presence of a conservation area(Table 2).Harvesting of logging residues was not considered in any of our scenarios, as this practice is still marginal in Qu′ebec. Moreover, no site preparation was simulated.

Based on the modelling framework used in this study, there were certain limitations in terms of the ability to simulate partial cuts,which had to be simplified for the purposes of this exercise. Moreover, harvesting was simulated to be carried out to the extent that the stand types planned for harvesting were available in the model. As such, no optimization of timber supply or assessment of annual allowable cut was performed; harvesting levels became emerging variables from the simulations and not fixed parameters. However, the volume targeted and harvested followed an increasing gradient from the natural evolution scenario with no harvest to the intensification scenario with the maximum harvest level.

2.3. Climate framework

Climate forcing scenarios were also projected over the simulation period (2020–2100), and used in combination with the forest management scenarios. Climate scenarios were based on three different global warming trajectories: a reference scenario without climate change,which corresponds to a projection of current (1981–2010) climate conditions without any change, and two climates forcing scenarios (RCP -Representative Concentration Pathways), i.e.RCP 4.5 and RCP 8.5(van Vuuren et al.,2011).The future climate values of the RCP scenarios were calculated from Canadian Earth System Model version 2 simulations(Arora and Boer, 2010), downloaded from the Climate Model Intercomparison Project Phase 5 of the World Climate Research Program and Ouranos (2021). A correction has been made to the Canadian Earth System Model data for the 1981–2010 period to incorporate data from McKenney et al. (2013), particularly for temperature values and precipitation ratios. RCP scenarios were produced using data from climate station records(McKenney et al.,2013).Further details can be found in Boulanger and Pascual Puigdevall(2021).

3. Modelling

3.1. Forest ecosystems

Scenarios were simulated using the LANDIS-II (LANdscape, DIsturbance and Succession) model in conjunction with the Forest Carbon Succession (ForCS) extension. LANDIS-II is designed to simulate the spatial dynamics of ecosystems and the interaction of processes governing them (Scheller et al., 2007). It integrates different ecological processes such as natural disturbances, succession, seed dispersal and the effect of climate change.Succession is simulated as a function of species traits, growth and mortality parameters, which determine intra- and inter-specific interactions (Fundation, 2018). As described in Dymond et al. (2016, 2021) the ForCS dead organic matter (DOM) and soil dynamics are built from the CBM-CFS3 model (Kurz et al., 2009). TheCBM-CFS3 has been investigated for parameter sensitivity (e.g. White et al., 2008) and uncertainty (Metsaranta et al., 2017), compared with empirical observations of carbon stocks(Shaw et al.,2014;Heffner et al.,2021)and with other estimates of NEP(Wang et al.,2011).It is also the core ecosystem model of Canada's National Forest Carbon Monitoring,Accounting and Reporting System (NFCMARS; Kurz et al., 2009; Environment and Climate Change Canada,2021), which produces estimates of GHG emissions for Canada's annual submission to the United Nations Framework Convention on Climate Change (UNFCCC). The CBM-CFS3 adopts an age-based yield curve approach to simulate stand development,which is well suited for large portions of the Canadian boreal forest that develop after stand initiating disturbances such as fire or clearcutting, but which may not adequately represents the uneven-aged stand dynamics typical of temperate forest of Eastern Canada,and also that of large portions of the boreal forests that have not been affected by stand-initiating disturbances for a long time. Consequently, it may not capture the effects of partial harvesting, an essential component of the management systems and simulated treatments of our case study areas.By combining CBM-CFS3′DOM and soil dynamics with LANDIS-II Biomass Succession extension's capability to simulate multi-cohort stand dynamics,ForCS allows for a more comprehensive representation of the carbon dynamics in such forests.

Table 1 Forest management scenarios for the Montmorency Forest.

Table 2 Forest management scenarios for the Hereford Forest.

Forest stands are modelled as a collection of cohorts of any number of species, interacting with each other within a cell (in our case, 250 m ×250 m resolution).At year 0,each cell was initialized with a set of species X age cohorts, which then evolved over time during the simulation.Forest composition and structure in each cell were initialized using provincial ecoforestry maps and cohort data from provincial permanent and temporary forest inventory plots. Cells with less than 50% forest cover were excluded from the simulations.The simulations were carried out annually and over a 80-year horizon, i.e. from 2020 to 2100. More details about initialization procedure can be found in Boulanger and Pascual Puigdevall(2021).LANDIS-II integrates extensions that manage different processes at the stand and landscape level (Tremblay et al.,2018). The following extensions were used in the simulations to represent natural processes and anthropogenic disturbances: a modified Base Biological Disturbance Agent v3.0(Sturtevant et al.,2004),Base Harvest v4.0 (Gustafson et al., 2000) and Base Wind v.3 (Mladenoff and He,1999).

Gap model PICUS v1.5 (Lexer and H¨onninger, 2001) was used to set LANDIS-II growth and establishment parameters for all tree species as a function of climate conditions and soil characteristics. The PICUS model simulates the dynamics of individual trees on a grid of small patches (10 m × 10 m) located across a 1-ha forest stand. The model accounts for spatially-explicit interactions among patches via a 3D light module, simulates seed dispersal, and climate and soil properties effects on tree population dynamics (Lexer and H¨onninger, 2001). In our study, we used PICUS model to simulate mono-specific stands for each of the tree species simulated by LANDIS-II. A factorial simulation design was used to simulate all mono-specific stands for each landtype under climate conditions for specific periods (baseline, 2011–2040,2041–2070, 2071–2100) and radiative forcing scenarios (baseline, RCP 4.5, RCP 8.5). Each of these stands were simulated for 300 years,starting from bare-ground, and using the local soil and climate time-series data for the specific landtype. Species establishment probability (SEP), maximum aboveground NPP (maxANPP) and maximum aboveground biomass (maxAGB) used in LANDIS were then derived from these simulations (Tremblay et al., 2018). The validation of dynamic growth parameters, as well as static growth- and mortality curve shape parameters, were assessed under baseline climate conditions and provincial sample plot data for eastern Canada (Taylor et al., 2017).Growth and establishment parameters derived from PICUS were updated in 2040 and 2070 to account for climate change. Climate dynamic parameters were calibrated using 30-yr climate data which corresponds to a standard climate normals time period. As such, the growth parameters for the first time period (2020–2040) were actually calculated using the 2011–2040 climate data in order to be consistent all along the simulation, i.e., having growth parameters assessed on 30-yr time periods. Hence, these 30-yr climate normals time periods begin in 2011, 2041 and 2071. Although values do not change during these 30 yr periods(21 years for 2020–2040),they were assessed using spatially downscaled CanESM2 monthly time in PICUS. As such, the 30-yr dynamic parameters are representative of the climate variability for this time period.

The Biological Disturbance Agent extension was parameterized to simulate spruce budworm outbreaks at year 2030,2062 and 2094;each outbreak was set to last for one year, and was based on age and tree species vulnerability, the older cohorts being most vulnerable and the most vulnerable species being, in decreasing order: balsam fir, white spruce, red spruce (Picea rubens) and black spruce (Hennigar et al.,2008).Outbreaks were simulated as probabilistic events at the cell level with probabilities being a function of the site and neighborhood resource dominance(e.g,host species occurrence within a 1 km radius)as well as regional outbreak status.Outbreak-related tree mortality was contingent on these probabilities as well as on host species- and age-specific susceptibility.The 32-year periodicity was based on a dendrochronological reconstruction of SBW outbreaks in southern Quebec for the last 400 years(Boulanger et al.,2012).SBW outbreak parameters were calibrated and validated using various studies conducted within the boreal and mixedwood forests (Boulanger and Pascual Puigdevall, 2021). In the context of this study,wildfires were excluded from the simulations,since such events are rare in both case study regions(Boulanger et al.,2014).Indeed,if a wildfire would have occurred in a given simulation,it would have had a disproportionate influence on the overall results and would have masked the management effect.

Forest Carbon Succession (ForCS) v2.1 (Dymond et al., 2021) was used with LANDIS-II core extensions to simulate the evolution of forest stands and carbon dynamics.ForCS monitors the evolution of above-and below-ground carbon stocks and fluxes in the main compartments of ecosystems (living biomass, dead organic matter, soils) and transfers to wood products. This extension uses the carbon pool definitions of the Canadian Forest Sector Carbon Budget Model(CBM-CFS3)as well as the same carbon transfer processes between pools(Kurz et al.,2009).ForCS distributes biomass from turnover and age-related mortality and adds it to the appropriate DOM pool.The litterfall and turnover are derived from the expected annual mortality and the expected Aboveground Net Primary Production (ANPP). Root turnover is determined based on a user-defined proportion that is applied to root stocks, and is calculated every year. Root mortality is determined only when the aboveground biomass in the cohort decreases(Dymond et al.,2021).

To assign the appropriate living and dead biomass values estimated for each site to each cell,ForCS performs the following spin-up process.The extension iterates the number of time steps equal to the maximum cohort age for each site.Beginning at time(t-oldest cohort age),cohorts are added at each time step corresponding to the time when the existing cohorts were established. Each cohort undergoes growth and mortality for the number of years equal to its current age, and its initial biomass value reflects competition among cohorts. Then for DOM initial stock,ForCS relies on the soil pool initialization process (spin up), where the model operates the above-mentioned biomass spin-up process multiple times,each time assuming that all cohorts present at the end have been killed, and then will regrow exactly as before. This process is repeated until the slow soil pools have stabilized.The last cycle of the initialization procedure starts with a stand replacing disturbance and then simulates growth and decay dynamics until all cohorts reach the age in the initial communities file (Dymond et al., 2021). Harvesting is simulated as a transfer of carbon from living biomass to harvested wood products. A proportion of the aboveground living biomass classified as merchantable(which includes merchantable wood and bark) is removed from the ecosystem and added to the wood product pool; this proportion varies according to species and the age of cohort,using the regionally specific data typically used with CBM-CFS3 (Kurz et al., 2009) to produce Canada's GHG report(Environment and Climate Change Canada,2021).The remaining aboveground tree biomass is transferred to DOM, along with all belowground tree biomass.

For each combination of climate and management scenario, five replicates were simulated, to take into account the effects of stochastic parameters. All the results were based on the average of these five replicates. Results for the following variables were compiled: ecosystem carbon stocks, ecosystems carbon fluxes (GHG fluxes) and forest composition (expressed as species proportion of aboveground living biomass).Carbon fluxes of the ecosystems were compiled and expressed in tonnes of CO2-equivalent per year; positive values for these fluxes represented emissions to the atmosphere (i.e. ecosystems are carbon sources) and negative values represented sequestration (i.e. ecosystems are carbon sinks).

3.2. Wood products and markets

Harvest data emerging from the Forest Carbon Succession extension,and expressed as the amount of carbon transferred from forest biomass to harvested wood products, were then processed into the Carbon Budget Model - Forest Harvested Wood Products (CBM-FHWP) (Environment and Climate Change Canada, 2020). First, the amount of harvested carbon was divided into specific baskets of wood products based on local wood sales data from each study area. Wood products included: sawnwood, pulp and paper, panels and bioenergy (used only for internal heating of sawmills) (see Appendix A.1 and Appendix A.2 for further details).At the beginning of the simulations,reservoirs of wood products were considered empty;only the products created during the simulations were thus considered. CBM-FHWP makes it possible to track the fate of the carbon within wood products according to their respective half-life time. The values of half-life times for sawnwood, panels (other wood)and pulp and paper were respectively 35,25 and 2 years,i.e.the default values provided by the Intergovernmental Panel on Climate Change(IPCC) (IPCC, 2014). The retirement of wood products was simulated following the same parameters as in Smyth et al.(2014).After retirement of solid wood products(i.e.,sawnwood and other wood products such as panels),a large part(99%)was assumed to be sent to landfills,while the rest was assumed to be incinerated(without capture for energy).For pulp and paper,the proportions were 93%to landfills and 7%to incineration.In landfills, 77% of solid wood products were assumed to not be degradable, while the other 23% was degradable with a half-life of 29 years(IPCC,2006).For pulp and paper products,44%was assumed to be not degradable while the remaining 56%was assumed to be degradable with a half-life of 14.5 years (IPCC, 2006). Landfills emissions were assumed to be comprised of 50% CO2and 50% CH4with no capture or use for bioenergy,as such a practice is still very marginal in Quebec.

To take into account the impact of additional wood products generated by forest management entering the markets and replacing GHGintensive, fossil-based products for material used to meet societal needs,a substitution effect for wood products was estimated and applied to alternative scenarios. This effect was calculated by multiplying local displacement factors (DF) with the amount of carbon within each category of wood product. A DF corresponds to the efficiency index with which the use of wood products reduce net GHG emissions by replacing a functionally comparable non-wood material:it quantifies the amount of GHG emissions (expressed as equivalent of carbon emissions) avoided per unit of carbon within the wood product(Sathre and O'Connor,2010).The following DF were used in our study, based on average provincial values for Quebec (Beauregard et al., 2019): 0.91 tonne of carbon emissions avoided per tonne of carbon in sawnwood and 0.77 for panels and other solid sawmill products(hereinafter named“other wood”).No substitution effect was assumed for pulp and paper products and internal consumption of bioenergy.The substitution effect of wood products was calculated only for the additional harvested wood in alternative scenarios relative to the BaU.The substitution effect was assumed to occur during the year the wood was harvested,and was recorded as a negative carbon emission expressed in tonnes of CO2-equivalent. Similarly, to take into account the impact of a reduction of the amount of wood products available on markets resulting from a management scenario that decreases harvesting levels relative to the BaU, we made the assumption that wood products would then be replaced by functionally equivalent,more GHG-intensive non-wood products,causing a negative substitution potential. Such an effect was calculated using the same DF mentioned above; the negative substitution potential was computed as a positive emission to the atmosphere(in tonnes of CO2-equivalent).

Emissions from extraction, transport and manufacturing (ETM) processes were also considered in this study according to each product family,i.e.,sawnwood,pulp and paper and other wood products.These emissions were accounted for in the “Products” emissions. Products under substitution were not included in this sum to avoid any double counting of ETM emissions,since they are already included in DF values.We use a value of 0.3532,0.1915,and 0.6725 tCO2e per tonne of carbon in the processed product for ETM emissions for other wood,sawnwood,and pulp and paper,respectively(Athena Sustainable Materials Institute,2018a,b,c;Skog,2008;Sun et al.,2018).

3.3. Carbon budgets and climate change mitigation potential

The simple decay approach (Rüter et al., 2019) was used as the carbon accounting perimeter,to which we added the substitution effect of wood products on markets when relevant. This approach considers both emission and sequestration fluxes between the ecosystem and the atmosphere, as well as emissions from wood products sourced from the case study area, whether they are used domestically or exported to other countries (Rüter et al., 2019). CH4and N2O emissions from wood product decay in landfills were also accounted for. For both of our case studies, we compiled carbon budgets for each combination of climate and management scenario by summing annual GHG fluxes (i.e.tonne-year approach) from forest ecosystems, decay of wood products and their substitution effect. We defined the climate change mitigation potential of the forest sector (including forests, products and substitution) for each territory as the difference between the carbon budget of the BaU and each alternative scenario as defined by the local forest managers (see above for details of scenarios) (Smyth et al., 2014; Xu et al., 2018).

4. Results

4.1. Evolution of forest landscapes

4.1.1. Forest composition

4.1.1.1. Boreal landscape-Montmorency. General trends in the evolution of forest composition over the simulated period were directly correlated with the type of forest management (Appendix A.3). An increase in harvesting was associated with an increase in the proportion of intolerant broadleaf species (Acer rubra, Betula papyrifera, Populus sp.) at the expense of balsam fir (Abies balsamea) and Picea sp. Results also suggested that for a given forest management, changes in anthropogenic climate forcing had little influence on forest composition over the time period considered.

4.1.1.2. Northern temperate landscape - Hereford. For the northern temperate landscape, the overall simulated trend for forest composition was an increase in the proportion of American beech(Fagus grandifolia)at the expense of all other species except spruces (Picea sp.), which remained stable across the forest landscape (Appendix A.4). American beech appeared to colonize and grow more vigorously with increasing climate forcing and management intensification. Only the Natural evolution scenario showed a stabilization of beech biomass accumulation and a decrease in spruces to the benefit of other hardwood species.Our results also suggested that an increase in climate forcing could have a negative impact on spruce species for this area. Spruce growth rate and regeneration capacity declined dramatically with time and increasing climate forcing. Sudden changes in 2041 and 2071 were due to the updating of parameters in PICUS (maximum Aboveground Net Primary Production (maxANPP) and maximum Aboveground biomass(maxAGB)).

4.1.2. Carbon stocks

4.1.2.1. Boreal landscape - Montmorency. Our simulations results suggested a general decrease of carbon stocks in forest biomass (including above-and belowground)for both the Baseline and the RCP 8.5 climate scenarios,irrespective of the forest management scenario(Fig.1).For the baseline scenario,this overall decrease was slight but constant,whereas under the RCP 8.5,results suggested a relatively constant situation in the 2020–2060 simulation time period and a steady decrease afterward.However,under the RCP 4.5 climate scenario,carbon stocks in biomass was projected to increase slightly throughout the simulation period(Fig. 1). For any given climate scenario, the increased harvesting associated with an intensification of forest management relative to the BaU reduced biomass carbon stocks; conversely, increased conservation caused an increase of these stocks.In addition,major and sudden‘drops’in carbon stocks of biomass for all management scenarios were strongly associated with spruce budworm outbreaks that occurred in the year 2030, 2062 and 2094. Outbreaks all led to a periodic but substantial decrease in biomass carbon stocks. However, simulations suggested a significant biomass re-growth in the years following the disturbance,regardless of the climate pathway(Fig.1).

Regarding carbon stocks in the dead organic matter pool (DOM),simulations suggested that they were higher than biomass stocks; total ecosystem carbon stocks were thus largely influenced by DOM variations(Fig.1). Spruce budworm outbreaks also appeared to be responsible for temporary but important increases of DOM carbon stocks(Fig.1).Apart from these events,there was a constant increase of DOM stocks over time regardless of the climate projection or forest management scenario.This increase was more important with the least intensive forest management strategies.The lowest DOM carbon stock was associated with the highest climate forcing(RCP 8.5).

4.1.2.2. Northern temperate landscape - Hereford. For this territory,trends for carbon stocks in biomass were directly linked to the vegetation growth potential and forest management strategies.

Contrary to Montmorency Forest,spruce budworm outbreaks only led to a slight decrease in biomass stocks that were rapidly recovered(Fig.1).Simulations also suggested a general increase in biomass stocks under both the baseline and the RCP 4.5 climate scenarios. Under RCP 8.5,stable biomass carbon stocks were simulated between 2030 and 2070,followed by a decrease over the following decades.Moreover,for a given climate scenario, increasing harvesting levels associated with the intensification scenario led to overall lower biomass stocks.

As in Montmorency, total ecosystem carbon stocks were largely influenced by DOM variations(Fig.1).The evolution of carbon stocks in DOM appeared to be influenced by two somewhat synergetic drivers.

First, higher harvesting levels caused higher DOM quantities. Second,high climate forcing was correlated with lower DOM quantities; the climate-associated reductions in DOM were also more evident with increasing harvesting levels.Thus intensification of management led to a higher total ecosystem carbon stock(despite lower living biomass).

4.1.3. Carbon dynamics and fluxes

4.1.3.1. Boreal landscape - Montmorency. The simulated net biomass growth was higher for forest management scenarios with higher harvesting levels (i.e Extensive III and Intensification) (Fig. 2). Conversely,scenarios with lower harvesting levels(i.e.Intensive I),or no harvest at all (i.e. Natural evolution) appeared to store larger amounts of dead organic matter(DOM)due to a much higher turnover(the rate at which DOM is depleted and replaced), resulting in higher Net Primary Production (NPP). As a result, the simulated Net Ecosystem Productivity(NEP), combination of the Net Primary Production (NPP) and heterotrophic respiration (RH), increased with the reduction of management intensity (i.e. NEP was highest with the Natural evolution scenario).Nevertheless, climate forcing appeared to be the main driver of net growth and was the primary factor responsible for important and sudden NEP decreases in 2041 and 2071(i.e.the years for which the growth and establishment parameters were updated to account for climate change).For their part, spruce budworm outbreaks did not appear to be a determining factor in NEP changes, compared with the effect of climate forcing,although outbreaks slightly stimulated net growth(Fig.2).

4.1.3.2. Northern temperate landscape-Hereford. The productivity of this area was mainly stimulated by forest harvesting,with the most intensive management scenarios showing the highest net growth, NPP and NEP.Spruce budworm outbreaks also caused temporary increases in NEP.Conversely,Natural evolution appeared to lead to a negative net growth and NEP for most of the simulation period.Nevertheless,climate forcing projections also played an important role under all management scenarios.Higher climate forcing led to a decrease in NEP,which seemed to be due to a decrease in net growth combined with stabilization or even a slight increase in heterotrophic respiration (Fig. 2). We predicted a negative NEP for the Natural evolution scenario regardless of the RCP climate simulated. However, similarly to what was observed for Montmorency, the significant drops simulated in the years 2041 and 2071,especially under RCP 8.5,were directly related to the parameter updates in species growth curves in PICUS(Fig.2).

4.2. Carbon budgets

4.2.1. Budget overview

4.2.1.1. Boreal landscape - Montmorency. Overall, the simulated boreal landscape yielded a negative cumulative carbon budget, i.e. a carbon sequestration,regardless of the combination of management and climate scenarios. Based on our methodological assumptions and simulation tools, this study area would remain a carbon sink for the entire 2020–2100 period regardless of the climate and the management scenario(Fig.3).

Nevertheless, annual carbon flows showed important variations between climate scenarios. Under the RCP 8.5, the sudden growth curve reduction in 2070 significantly affected the forest sink, causing the cumulative emissions to climb. The time during which annual flows remained emitters as well as the magnitude of these emissions was strongly linked to the forest management strategy, with higher harvest levels causing higher emissions. Moreover, we also forecast important variations between forest management scenarios regardless of the climatic forcing simulated. Indeed, lower levels of harvest produced a larger sequestration than more intensive management scenarios(Fig.3).

Fig. 1. Annual carbon stocks in ecosystem pools according to forest management and climate forcing scenarios for the two case study areas.Note:NPP: Net primary productivity; RH: Heterotrophic respiration; NEP: Net Ecosystem Productivity; Biomass: Above- and Belowground; DOM: Dead organic matter.

Fig. 2. Annual ecosystem carbon fluxes according to forest management and climate forcing scenarios for the two case study areas.Note: NPP: Net Primary Production; RH: Heterotrophic respiration; NEP: Net Ecosystem Production.

4.2.1.2. Northern temperate landscape - Hereford. For the northern temperate area, simulations forecast a shift in the capacity of the landscape to be a sink over time.After a period of sequestration(2020–2030),all management scenarios would see stabilization or an increase in their annual emissions, regardless of the RCP considered (Fig. 3). Emissions appeared to be driven by two compounding factors:an increasing climate forcing and a reduction of forest harvest levels.Both caused emissions to increase, with the scale of the impact of harvest level on the carbon budget depending heavily on the climate projection (Fig. 3). Under all climate forcing projections, the Natural evolution scenario would be a cumulative carbon source by the end of the period.The source was more pronounced with higher climate forcing;however,the increase in emissions was even more severe for other management scenarios.Thus,under a RCP 8.5,all forest management strategies would cause the territory to become a net carbon source by 2100(Fig.3).

Fig. 3. Cumulative carbon budgets according to forest management and climate forcing scenarios. Budgets include net emissions from forests, products and substitution.

4.2.2. Mitigation potential

4.2.2.1. Boreal landscape - Montmorency. The mitigation potential was assessed as the difference in carbon fluxes between the BaU and alternative forest management scenarios(i.e.,net emissions relative to BaU).Even though simulations suggested that different climate forcing projections had distinct impacts on the boreal landscape(Fig.3), no significant interaction between climate and forest management strategies was forecast for this area.Indeed,the mitigation potential of a given scenario remained approximately the same irrespective of climate projections(Fig.4).

Fig.4. Cumulative mitigation potential according to forest management strategy and climate forcing scenarios.Potentials include net emissions from forests,products and substitution relative to the Business-as-usual scenario.

For the first decades of the simulation period, no clear distinction between forest management scenarios was revealed by simulations, as their net emissions relative to the BaU were all hovering near 0(Fig.4).However,from the year 2040 and beyond,strategies with lower harvest levels created a clear cumulative mitigation potential, with the Natural evolution scenario (i.e. no harvest) causing the highest net carbon sequestration at the end of the simulation period and thus providing the highest mitigation potential. Conversely, an increase in harvest levels relative to BaU caused a net cumulative carbon source, and would therefore not provide any mitigation benefit.

When looking at the breakdown of the cumulative mitigation potential according to the components Forests-Products-Substitution(Fig. 5), it can be seen that the increase of mitigation potential associated with lower harvesting levels was primarily due to an increased net sequestration in the forest ecosystem pool relative to BaU. To a lesser extent, it was also due to a reduction of decay emissions from wood products relative to the BaU (Fig. 5). Lower harvest levels were associated with a negative substitution potential,displayed as emissions to the atmosphere caused by the use of carbon-intensive products instead of wood products. However, these additional emissions were marginal relative to other carbon fluxes.Conversely,an increase of harvest levels negatively affected the forest ecosystem sink while increasing decay emissions from products;increased net emissions from forests and wood products could not be offset by the substitution effect on markets(Fig.5).

Fig.5. Breakdown of the cumulative mitigation potential by forest sector pool.Illustration for the RCP 4.5.Potentials include net emissions from forests,products and substitution relative to the Business-as-usual scenario.

4.2.2.2. Northern temperate landscape - Hereford. In terms of mitigation potential, results for the northern temperate landscape again differed from the boreal landscape. Significant interactions were observed between the simulated forest management strategies and climate forcing projections (Fig. 4). First, decreasing harvest levels relative to the BaU resulted in increased net emissions,and therefore no mitigation benefit.However, an increase in the climate forcing tended to reduce the gap between the more conservation-oriented management scenarios and the BaU(Fig.4),although the former still remained higher emitters than the BaU. Conversely, increased climate forcing was predicted to reduce the mitigation potential associated with increased harvest levels relative to the BaU.

The breakdown of the cumulative mitigation potential according to the components Forests-Products-Substitution (Fig. 5) suggested that reducing harvest levels negatively affected the forest sink relative to BaU,although this effect was moderated by decreased decay emissions from wood products (Fig. 5). The negative substitution potential associated with increased forest conservation only had a marginal effect on overall mitigation potential. Conversely, simulations suggested that increasing harvest levels relative to BaU provided mitigation benefits (Fig. 5), by increasing the capacity for carbon sequestration and storage of the forest ecosystems,especially in the dead organic matter pool(Fig.4).This effect was nevertheless lessened by the decay emissions from the degradation of wood products, with the substitution playing a minimal effect on the total carbon budget(Fig.5).

5. Discussion

Our study revealed that forest harvesting have contrasting effects on carbon balance in the northern temperate and boreal landscapes studied here,regardless of the future climate pathway;increased harvesting level improves the mitigation potential in the former, whereas it reduces this potential in the latter. These variations between landscapes are largely induced by the response of forest ecosystems to their respective management and silvicultural practices. That being said, the intensity of radiative forcing was a strong determinant of the capacity of forest ecosystems to act as a carbon sink,and interacted with the characteristics and dynamics of the vegetation. Moreover, for the two territories, the simulated carbon storage in wood products and substitution effect of wood on markets are not sufficient to offset emissions generated by the alternative forest management and climate change in ecosystems.

Our main observation for boreal landscapes, i.e., that reducing harvesting levels contribute to climate change mitigation, is in line with some of the previous boreal studies(Smyth et al.,2014;Heinonen et al.,2017;Pukkala,2018;Chen et al.,2018a;Sepp¨al¨a et al.,2019;Skytt et al.,2021; Ter-Mikaelian et al., 2021; Landry et al., 2021). However, the opposite conclusion was drawn for other boreal studies (Poudel et al.,2012;Lundmark et al.,2014;Gustavsson et al.,2017;Chen et al.,2018b;Smyth et al.,2018).One of the reasons for this difference is the role that substitution on markets can play for offsetting emissions from forest ecosystems under intensive management.We used the average values of displacement factors calculated for Canada,thus reflecting conditions of our case study areas; these values were significantly lower than values from global meta-analyses that are used elsewhere. Nevertheless, it contributes to underline the importance of jointly considering carbon fluxes from forests, products and markets when assessing mitigation potential.

Two determining factors can explain our contrasting results between boreal and northern temperate ecosystems: the type of harvest and the initial ecosystem characteristics. Firstly, the increase of harvest in the northern temperate area was assumed to be performed solely by partial cutting,for which a portion of the stand biomass is periodically removed,and part of the canopy cover is preserved;on the other hand,most of the harvesting activities in the boreal forest was simulated as clearcutting(with some proportion performed as partial cutting depending on the scenarios). This difference in harvesting practices may explain the contrasting responses of the simulated areas to the increased harvest levels.Our simulations were not designed to discriminate the effect of increasing harvest levels versus a variation in the type of harvest.Further research will be needed to understand the importance of each factor and whether synergies emerge.Clearcutting induces a period of net emissions to the atmosphere in the years following harvest,due to emissions from decomposition being substantially higher than sequestration by photosynthesis (for at least 13 years following harvest in balsam fir stands of Montmorency Forest) (Paradis et al., 2019). Conversely,continuous-cover forestry is associated with less variation in NPP; it allows the maintenance of stocks of dead organic matter (Paradis et al.,2019) while stimulating growth of remaining trees (Laiho et al., 2011)and would thus have a more favourable carbon balance. While a systematic comparison of partial vs.clearcut harvesting would be needed to confirm this hypothesis in the context of our study areas,continuous-cover forestry may allow for a continuum in the accumulation of carbon in the ecosystem, while maintaining a level of wood production and product substitution on markets (Pukkala, 2014; Lundmark et al.,2016).

Secondly, differences in initial forest landscape characteristics, in particular the initial forest composition and age structure induced by past climate and disturbances, may also partly explain the contrasting response of the northern temperate and boreal landscapes. It should be noted that initial carbon stocks in 2020 were low in both territories compared to other studies.Kurz et al.(2013)estimated a mean value of 193 tC per ha for the boreal managed forest and Payne et al. (2019) a mean value of 213 tC per ha for the mature boreal mixedwood forest.Our results forecast a global stock of carbon around 120 tC per ha and 150 MgC per ha at the beginning of the simulation for the boreal and the northern temperate territories respectively (160 and 220 tC in 2100 under the baseline climate). Our results forecast a unique trend in the northern temperate territory as the biomass(above and below-ground)is expected to largely increase in the first 20 years of the simulation(regardless of the forest management), while the total biomass stock in the boreal area is expected to remain fairly constant during the whole simulated period.Thus,our model considered that ecosystems were not saturated in living biomass at the beginning of the simulations in the northern temperate ecosystems. Temperate forest landscapes of eastern Canada often display the legacy of unsustainable past management practices that caused a reduction of the forest productivity (B′edard and Majcen, 2003; Kenefic et al., 2014; Pr′evost and Charette, 2019) and created forest areas with age and species structure that are not optimal for carbon sequestration (Kurz et al., 2013). In such instances, forest landscapes can benefit from natural and anthropogenic disturbances(including harvesting)that remove parts of the vegetation overstory and thus release stand growth potential by allowing the regeneration of new vegetation cohorts (Pugh et al., 2019). Stands of the boreal biome may have fewer constraints on net growth,and would not necessarily benefit from increased disturbances.Because average net primary productivity is higher in the northern temperate biome than in boreal one,the effect of a potential growth release is even more significant.

Regarding the effect of climate change, our study suggests that increased climate forcing may have a greater impact on the carbon balance of northern temperate landscapes than on boreal ones.On the one hand, an increased climate forcing could lead to a higher ecosystem productivity on the short-term (Denman et al., 2007; Gonsamo et al.,2017), as was observed in our simulations for the boreal landscape.

On the other hand, the ecological functioning of landscapes dominated by species that are poised to become ill-suited to the emerging climate conditions is more likely to be disrupted by climate change, as observed in our northern temperate landscape (Boulanger et al., 2017;Brecka et al., 2020). For example, spruces (Picea sp.) are predicted to have a severely reduced growth in northern temperate landscapes under high climate forcing,with potential impacts on ecosystem carbon fluxes,likely beyond the end of our simulation period (Brecka et al., 2020).Ultimately, climate change can limit net vegetation growth even in the boreal area, and may also enhance respiration by increasing ground temperature (Bond-Lamberty and Thomson, 2010; Krishna and Mohan,2017;Peng et al.,2008),thus decreasing sequestration by growth while increasing emissions by degradation. Landscapes primarily managed by high-frequency operations such as partial cut can also be even more impacted by increased climate forcing.Indeed,the frequent disturbances may contribute to accelerate the changes in ecosystem dynamics that are induced by climate change (Brice et al., 2019). Nevertheless, while the possible interaction of forest management and climate change on ecosystem functioning is a vector of uncertainty, there still can be a significant role for forest management in adapting the land to its future characteristics (Messier et al., 2019; Mina et al., 2021). For example,harvest operations could target stands that are most susceptible to insect outbreaks,fire or productivity decline,thus reducing the impacts of such events while maintaining the harvest level (Hennigar and MacLean,2010;MacLean et al.,2007).Foresters could also actively manage species composition (through operational selection of harvested trees during partial cuts) and the connectivity of stands and landscapes to increase their resilience and sustainability in the face of a changing climate.Modelling artifacts, mainly updating climate change impacts on stand dynamics at specific time period (i.e., 2041 and 2071), likely do not interfere with our conclusions, as all scenarios are equally affected by those and as results are almost all interpreted as relative to a BAU scenario.

While increasing harvesting levels might somewhat contribute to maintain or even increase the ecosystem carbon sink as observed in the northern temperate areas, the overall balance of harvested wood products and substitution is not necessarily a carbon sink.Indeed,our results underline the magnitude of the impact of the basket of wood products sourced from forest landscapes on carbon budgets (Chen et al., 2018b;Dugan et al.,2018;Gustavsson et al.,2017;Poudel et al.,2012;Pukkala,2014; Smyth et al., 2014). In our study, emissions associated with the procurement and manufacturing of wood products and their decay during their life-time heavily influenced the carbon balance,and were more important than any substitution effect that these products had on markets.This was compounded by the fact that the decline of species(such as Picea sp.)that are currently valued by the timber industry(Brecka et al.,2018)was accompanied by an increase of other species that are currently mostly dedicated to pulp and paper (Fagus grandifolia, Populus sp.), i.e.products with a short lifespan and virtually no substitution effect on markets. Increasing the material use of wood as long-lived, durable products and encouraging their recycling and cascading at their end-of-life, both for material and energy uses would reduce annual product emissions while increasing the substitution effect,with a strong and direct impact on the forestry sector's carbon budget (Chen et al.,2018b;Dugan et al.,2018;Smith et al.,2014).Reducing GHG emissions associated with wood processing and product manufacturing, and targeting the displacement of the most GHG-intensive fossil-based products would also lead to higher displacement factors. This would have the potential to promote an increase in harvesting at the expense of conservation(Chen et al.,2018b;Smyth et al.,2017,2018;Smith et al.,2014;Xu et al., 2018; Sepp¨al¨a et al., 2019). Alternatively, any future improvements in the GHG footprint of industries that produce carbon-intensive products (e.g. cement, steel) would reduce the climate change mitigation potential of the forest sector, due to an ever-lower substitution effect of wood products (Harmon, 2019). Several assumptions used in the DF determination are only partially supported by the literature,such as assumptions related to market dynamics,i.e.,pricing,leakages and rebound effects (Harmon, 2019; Howard et al., 2021).Brunet-Navarro et al. (2021) even indicated that the displacement of GHG-intensive products with wood should be reduced between 2030(33%lower)and 2100(96%lower)due to the gradual roll-out of climate policies around the world. Therefore the DF impact must be taken with caution in forestry studies.

Our simulations did not include wildfires: the scale of the territories did not allow any meaningful insight on the potential impact of this natural disturbance in the context of climate change,and our study areas are also naturally minimally exposed to this disturbance.Simulations on larger areas have nevertheless suggested that modifications of wildfire regimes can lead to profound transformations in the productivity and regeneration capacity of forest ecosystems (Boulanger et al., 2017) and jeopardize the sustainability of forest management strategies and wood production. Similar consequences can also be expected with insect outbreaks not simulated in our study (Boucher et al., 2018; Brecka et al.,2018).

In our study,apart from internal heating,residues from sawmills were not simulated to be used for external bioenergy production on markets.Moreover,none of the scenarios took into consideration the possibility of harvesting logging residues for energy production, even though a mitigation potential may exist through this practice(Kilpel¨ainen et al.,2016;Lagani‵ere et al., 2017; Pingoud et al., 2016). However, any increased removal of organic material from the ecosystem would need to take into account possible impacts on the DOM pool: it represented the largest ecosystem carbon pool of our simulations,and was particularly sensitive to changes in forest management practices and climate forcing.For both territories, the DOM stock was constantly increasing during the simulated period. For the northern temperate area, the intensification of management caused both a faster stand growth and a greater accumulation of DOM (the latter being a result of the proportion of biomass assumed to be transferred from living biomass to DOM at the moment of harvest).However,it should be noted that while inputs to the DOM pool and decomposition processes can be predicted reasonably well,the initial size of the DOM pool was one of the biggest sources of uncertainty in simulations of the ecosystem carbon cycling,a feature shared with other studies using the CBM-CFS3 framework(Metsaranta et al.,2017),as it is the case when using the ForCS extension of Landis-II.This issue is limited in our study as we put our focus on mitigation potential relative to a BaU.Further research should therefore look at the finer dynamics of DOM under the compounding effects of management and climate. Similarly,any end-of-life alternative for wood products will have the potential to reduce carbon emissions relative to the BaU(Head et al.,2020)and may therefore be included in the global response to climate change.The fact that almost all retired wood products are sent to landfills (i.e., the most common practice in Quebec at the moment) represents a missed opportunity for increasing the mitigation potential of wood production. In Scandinavia, up to 70% of solid wood products and 30% of paper products at their end-of-life are used for bioenergy production(Zubizarreta-Gerendiain et al., 2016), suggesting that there is no major technological barrier for this alternative. In our study, solid wood products that are sent to landfills correspond to a cumulative amount of 109,000 and 34,450 tonnes of carbon of solid wood products in 2100 in the business-as-usual scenario for the boreal and northern temperate territories,respectively. If these products were recovered for bioenergy,and assuming a displacement factor of 0.76 for converting heat production from a fossil source to bioenergy in Quebec (Beauregard et al.,2019), this could create an additional substitution effect of 82,840 and 26,182 tonnes of C for Montmorency and Hereford forests,respectively.This underlines the importance of exploring cascading use of wood products and end-of-life alternatives,which may play an important role in the mitigation potential of forestry.

Alternative strategies can also reduce or accelerate climate change by affecting biophysical processes such as albedo and evapotranspiration(Smith et al.,2014).Such processes were not considered in our study as the lack of information for our territories was too important.Assessments that consider these processes along with carbon might provide new insights for identifying the most efficient climate change mitigation strategies.

Our study defined the mitigation potential of the forest sector using its net carbon balance, i.e. according to the tonne-year approach. However, such a definition has certain limitations related to the temporality of emissions and their real effect on the climate(Head et al.,2019,2020,2021).Accounting for the cumulative radiative forcing caused by carbon emissions in the atmosphere as a function of time(i.e.over a given time horizon, accounting for the fact that earlier emissions create higher cumulative radiative forcing than emissions occurring later)can provide a way to compare scenarios on the basis of their capacity to mitigate climate change(Levasseur et al., 2010, 2012a, 2012b). Nevertheless, in our study,there was a clear and linear distinction between scenarios over the simulation period with the tonne-year approach, and no major differences in the temporality of emissions. The ranking of forest management strategies in terms of mitigation potential should therefore be similar whether the accounting is based on the balance of carbon emissions or on the resulting radiative forcing.

6. Conclusion

Our results suggest that blanket statements should be avoided with regard to the respective benefits of conservation and intensification of forest management from the perspective of the carbon balance and with regard to the ecosystem capacity to act as carbon sink.In our simulation experiment, the northern temperate and boreal landscapes reacted management intensity gradient in contrasting manner, suggesting that climate change mitigation strategies need to be tailored to the ecosystem dynamics and initial characteristics. Ecosystem-level carbon dynamics were the main determinant of the climate change mitigation potential of the forest sectors. Net growth ecosystem stimulation by forest management that results in an increase in sequestration should be coupled with the limitation of carbon loss to the atmosphere to result in higher and stable carbon sequestration in the forest sector. Living biomass, dead biomass and wood products stocks are therefore levers of action to ensure a steady carbon storage in the forest sector and enable a strong mitigation benefit following an alternative forest management. As our study suggests, the optimal management solution can be very different according to the initial characteristics of forest ecosystems. This opens the possibility of optimizing management and wood products basket for specific forest stands,with the objective of maximizing the mitigation potential of a given landscape.Within a landscape,conservation of forest areas could be used in conjunction with harvesting of specific stands(e.g.stands high productivity potential or a significant level of stagnation),with the aim of improving overall ecosystem quality and carbon sequestration while producing wood products for substitution purposes. Our study shows above all the importance of disturbance regimes in forest carbon management, especially the need to consider uneven-aged dynamics in carbon budget simulations,as the high degree of complexity of the forestry sector cannot avoid simulations of the same scope in order to develop management alternatives based on the most exhaustive information possible.

Funding

This study was funded by the Quebec Ministry of Forests, Wildlife,and Parks(contrats de service de recherche foresti‵er 142332156-2018-A and 142332174-E: PI: E. Thiffault), and by the Natural Science and Engineering Research Council through a Discovery Grant to E. Thiffault(grant number RGPIN-2018-05755).

Availability of data and materials

The datasets generated and/or analyzed during the current study are available:https://doi.org/10.6084/m9.figshare.16874710.v1.

Ethics approval and consent to participate

Not applicable.

Competing interests

The authors declare that they have no competing interests.

Consent for publication

Not applicable.

Authors' contributions

LM analyzed and interpreted the data and was a major contributor in writing the manuscript. ET secured the funding for the study, helped to interpret the data and was a major contributor in writing the manuscript.DC have made substantial contributions to the conception and acquisition of data, and substantively revised the manuscript. YB have made substantial contributions to the conception and acquisition of data, and substantively revised the manuscript. RB substantively revised the manuscript. All authors read and approved the final manuscript.

Authors’ information

Not applicable.

Acknowledgements

Authors would like to thank Julie Bouliane and Dany Senay,respectively from For^et Montmorency and For^et Hereford, for their help in defining forest management scenarios.

Appendix

Fig. A.1. Basket of wood products for the Montmorency Forest

Fig. A.2. Basket of wood products for the Hereford Forest

Fig. A.3. Forest Composition Montmorency

Fig. A.4. Forest Composition Hereford