Zlin Liu, Changhui Png, Davi A. MaLan, Louis D Granpr′, Jan-No¨l Canau,Danil Knshaw,＊

a School of Geographic Sciences, Hunan Normal University, Changsha, 410081, China

b Department of Biological Sciences, University of Qu′ebec at Montreal, Montreal, QC H3C 3P8, Canada

c Faculty of Forestry and Environmental Management, University of New Brunswick, Fredericton, NB E3B 5A3, Canada

d Laurentian Forestry Centre, Canadian Forest Service, Natural Resources Canada, QC G1V 4C7, Canada

e Great Lake Forestry Centre, Canadian Forest Service, Natural Resources Canada, ON P6A 2E5, Canada

Keywords:Annual defoliation Natural disturbance Forest protection Net ecosystem productivity TRIPLEX-Insect Conifers

ABSTRACT

1. Introduction

In North America, boreal forests have important ecological and economic values, including carbon (C) storage and timber supply (Chang et al., 2012a). Spruce (Picea spp.) and balsam fir (Abies balsamea (L.)Mill.),two of the most abundant tree species groups in forests of Atlantic Canada, are vulnerable to defoliation by the native insect spruce budworm (SBW; Choristoneura fumiferana (Clem.)) (Hennigar et al., 2008;Candau et al., 2018; Donovan et al., 2021). Moderate and severe SBW defoliation reduces tree growth, increases tree mortality, and converts forests from a C sink to a C source(Dymond et al.,2010;Liu et al.,2018).The 1967 to 1992 SBW outbreak in eastern Canada peaked at over 50 million hectares (Kettela, 1983), and in Qu′ebec, spruce-fir timber and wood-fiber losses were estimated at 238 million m3and led to $12.5 billion of total losses in revenue (Grondin et al., 1996). Timber harvest reductions, relative to a no-defoliation case, for the 3.0 million ha of Crown land in New Brunswick were projected to be 18% and 25% by 2052, under moderate and severe outbreak defoliation scenarios from 2012 to 2032(Hennigar et al.,2013),resulting in projected total output of the New Brunswick economy declining in present value terms by$3.3 billion(CAD)and$4.7 billion,respectively(Chang et al.,2012b).In the recent SBW outbreak, more than 21,000 km2of defoliated forests in Qu′ebec were converted from C sinks to sources between 2007 and 2017(Liu et al.,2019b).The outbreak started in 2006 in Qu′ebec.Defoliation covered 13.5 million ha in this province alone in 2020 (https://mffp.gouv.qc.ca/documents/forets/fimaq/RA_TBE_2020.pdf, accessed on 22 Nov. 2021) and has expanded to the adjacent provinces of New Brunswick,Newfoundland and Ontario.

During SBW outbreaks, foliage protection (FP) through the aerial spraying of pesticides has been used to reduce timber supply loss and tree mortality in eastern Canada. This approach is based on the use of biological insecticides (i.e. Bacillus thuringiensis kurstaki (Btk) or in some jurisdictions a molting regulator,tebufenozide)to keep host species alive during moderate and severe SBW outbreaks(MacLean et al.,2002;Carter and Lavigne,1993).In order to protect 50%of current foliage retained in Qu′ebec and 60% in New Brunswick, FP is usually applied every 2 or 3 years in spruce-fir stands(Fuentealba et al.,2019). From 1970 to 1983,FP effectively prevented extensive tree mortality over 1.5 million ha of coniferous forest in New Brunswick(Miller and Kettela,1975;Irving and Webb,1981).

Two years of moderate-severe defoliation are usually allowed to occur before application of this treatment,because it typically takes four or more years of defoliation to kill trees (MacLean, 1980; Cooke and R′egni`ere, 1999; Houndode et al., 2021). As a result of this time lag between population increases and treatment, foliage protection does not prevent the spread of SBW outbreaks despite its efficacy in keeping trees alive. As an alternative to FP, since 2014 an early intervention strategy(EIS) is being tested to control SBW populations in New Brunswick(Johns et al., 2019; MacLean et al., 2019). This approach focuses on controlling SBW at low population densities before they increase and before noticeable defoliation occurs. To be successful, EIS must detect‘hot spots’ of low but rising SBW populations, and treat these with insecticides to suppress SBW populations before they can rise (Liu et al.,2019a; Johns et al., 2019). MacLean et al. (2019) reported that second instar SBW larvae (L2) populations were reduced by 90% following 5 years of EIS in New Brunswick. Both FP and EIS can be useful for pest management,but we expect that there will be variations in the efficacy of these strategies on forest C dynamics. In theory, effective EIS would result in no mortality or carbon loss to the SBW and thus be similar to a no outbreak scenario. This has indeed been the case for 7 years of EIS trials in New Brunswick(MacLean et al.,2019).However,EIS funding is only in place for 8 years and is not guaranteed thereafter(MacLean et al.,2019).What will happen to SBW populations if EIS treatments are ceased in the 8th year but the surrounding outbreak is not over? Or what happens if the SBW populations overwhelm EIS after it is applied 8 years(assuming no further funding renewal) or 12 years (assuming an additional 4 years of funding, but no more after that). How results of successful EIS conducted for only part of an outbreak compared to FP,is not known.

Simulation modeling is an approach that can help better understand and quantify the effects of insect disturbance on forest C dynamics and compare C budgets between managed and unmanaged forests (Maleki et al.,2021).In this study,we used the TRIPLEX-Insect model(Liu et al.,2018) along with a geospatial forest database and 19 defoliation and management scenarios to perform a long-term evaluation of the impact of SBW defoliation on C dynamics in spruce-fir forests in Atlantic Canada.We determined which forests (i.e. tree species and locations) should be the most targeted for different forest pest management strategies(i.e.EIS and FP) to optimize C sequestration over various outbreak severities(moderate and severe). Objectives of this study were to use TRIPLEX-Insect simulated model outputs to:1)test the effects of EIS,FP,and combined EIS-FP, under moderate and severe SBW outbreak scenarios, on above ground biomass (AGB), net ecosystem productivity(NEP), soil organic carbon (SOC) and cumulative tree mortality from 2020 to 2039 in Atlantic Canada;2)evaluate the effects of only applying EIS for part of an outbreak,specifically only 8 or 12 years out of a 20-year SBW outbreak simulation; 3) compare the impact of EIS and FP on C dynamics(i.e.NEP and volume)for three different host species(i.e.black and red spruce (these two hybridizing species were treated as a single species),white spruce,and balsam fir);and 4)quantify effects on carbon budgets(i.e.annual volume increment,NEP and SOC)for different forest age classes.

2. Methods

2.1. Study area

The study area encompasses all of the SBW prone forests(i.e.balsam fir,white spruce,black/red spruce forests)in Atlantic Canada.It includes four provinces: New Brunswick (NB), Nova Scotia (NS), Prince Edward Island(PEI),and the island of Newfoundland(NL)(Fig.1).

2.2. Modelling

The TRIPLEX-Insect model (Liu et al., 2018), which was developed based on the process-based TRIPLEX 1.0 model(Peng et al.,2002), was parameterized for the target forest regions across the study area. It was improved by accounting for the effects of trees species mixes and age structure on defoliation severity(Hennigar et al.,2008)to simulate both short-and long-term C dynamics using a monthly time step in forests that undergo SBW disturbances(Liu et al.,2018,2019b).TRIPLEX 1.0 model runs begin with simulated tree seedlings from the year of germination.TRIPLEX-Insect estimated gross primary productivity(GPP)using forest age, air temperature, the percentage of frost-free days in a month, leaf area index, received photosynthetically active radiation, and soil water availability(Peng et al.,2002).Forest net ecosystem productivity(NEP)is estimated as:

where ER is ecosystem respiration, which is the sum of autotrophic respiration (Ra) and heterotrophic respiration (Rh). Rais calculated as a function of component C pools (e.g., foliage, branches, wood, coarse roots,and fine roots),nitrogen and air temperature.Rhis estimated as the difference between soil respiration and root respiration, which is expressed as the exponential function of temperature and Q10(a temperature sensitivity factor).Net primary productivity(NPP)is estimated by subtracting Rafrom GPP(Zhou et al.,2008).

Volume growth is based on wood C density and stem wood increment(Bossel,1996).Soil C and nitrogen are simulated based on CENTURY 4.0(Parton et al.,1993).The model uses tree mortality and litterfall rates to describe the transfer of dead biomass (e.g. 40% of total C from defoliation) to the litterfall pools. Litterfall and soil C pools (i.e. slow, passive,and active pools)and decomposition rates of soil organic C are estimated as a function of maximum decomposition rates, effects of soil moisture,and effects of soil temperature(Wang et al.,2011).

The impacts of annual defoliation(AD)in the TRIPLEX-Insect model are calculated as both leaf biomass loss due to defoliation and defoliation-caused mortality (Liu et al., 2018). Defoliation-caused mortality is estimated from defoliation stress, tree species and tree age indices.These are explained in Liu et al.(2018)but essentially consider that mature hosts are more vulnerable than stands of immature hosts,and that balsam fir is more vulnerable than white spruce which is more vulnerable than red spruce and black spruce (the values for the vulnerability rankings are taken from Hennigar et al.,2008). This variation in vulnerability applied to the stand conditions across the Atlantic Provinces is one of the main drivers of the model. Natural mortality and competition-caused mortality, based primarily on stand age by species,are also considered in the TRIPLEX-Insect model(Liu et al.,2018).In this study, because SBW defoliation results in little to no seed production during an outbreak,we did not consider natural regeneration(Schooley,1978; MacLean, 2016). In addition, the set of key parameters for the model are derived from previous studies (Peng et al., 2002; Zhou et al.,2006; Wang et al., 2011; Liu et al., 2018, 2019b). The model has been successfully calibrated and validated for defoliation of different intensity and duration in stands, host species, and management/treatment strategies under SBW outbreaks in Qu′ebec's boreal forests (Liu et al., 2018,2019b, 2020). More details on the TRIPLEX-Insect are described in Liu et al.(2018).

Fig.1. Study areas in Atlantic Canada,including 182,563 km2 of the three host species forests(i.e.,balsam fir,white spruce,black/red spruce)for which total density(expressed as percentage of total live above ground biomass) was more than 20%(pink color). (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

2.3. Data and scenarios

Forest geospatial databases used to parameterize the model were obtained from Canada's National Forest Inventory(NFI)(Beaudoin et al.,2014). Based on this database, we stratified the study area into 2,885 standard 10 km×10 km grid cells,which included 182,563 km2of the three host species (i.e., balsam fir, white spruce, black/red spruce combined because the two species hybridize).In this study,we excluded area in which total density(expressed as percentage of total live above ground biomass) of host species was less than 20% and non-forest lands. The initialization forest data of all scenario simulations are based on this data.We also used BioSIM(R′egni`ere,1996;R′egni`ere and St-Amant,2007)to obtain the monthly climate data from 1989 to 2019 for each grid cell,and then standard climatological normals were used to estimate the future climate(i.e.2020–2039)and parameterize the model.Total soil organic C data were obtained from the database of the Canadian Soil Information Service (CanSIS) (http://sis.agr.gc.ca/cansis/interpretations/carbon/in dex.html; Tarnocai and Lacelle, 1996) and projected onto our study area to estimate initialization of total organic soil C for each grid cell(Zhang et al.,2005).

Fig. 2. Comparison of observed (from geospatial forest databases) and simulated (using the TRIPLEX-Insect model) values of average volume (m3·ha-1) in 2010 in Atlantic Canada.Each data point represents the average volume per grid cell. The solid line is the 1:1 line. The dotted line is the linear regression line.

We used Landsat species biomass maps in 2010,which exported from geospatial forest databases, to calibrate (40% of the data, in 1,154 grid cells) and validate (60% of the data, in 1,731 grid cells) the TRIPLEXInsect model in Atlantic Canada (Fig. 2). We also used a Monte Carlo method to estimate uncertainty of parameters at the regional scale for the TRIPLEX-Insect model simulations(Meyer et al.,2018;Liu et al.,2019b).The resulting distributions of the TRIPLEX-Insect outputs were summarized as 95th percentiles of C dynamics (i.e. NEP, aboveground tree biomass,and soil C).

The outbreak patterns and timing were prescribed.This is a standard approach to evaluate uncertainty by testing a range of scenarios.Because SBW defoliation differs among host species(Bauce et al.,2004;Hennigar et al., 2008), we modified the moderate and severe defoliation patterns from Hennigar et al. (2007) and Hennigar and MacLean (2010) for defoliation of balsam fir, white spruce and black/red spruce with and without FP. The moderate outbreak scenario, based on the historical pattern of defoliation in New Brunswick(Hennigar et al.,2013;Hennigar and MacLean, 2010), was ＜25% defoliation for 4 years, rapidly increasing to near 100% defoliation, and then gradually declining,resulting overall in 8 years of defoliation ＞30% (Fig. 3a). Patterns of defoliation with FP, based on New Brunswick's current protection efficacy target and forest stands considered severely defoliated in New Brunswick aerial surveys (Hennigar et al., 2007), were estimated to reduce defoliation to 40% of current-year foliage in years when it was predicted to be ＞70%with no protection.The annual defoliation values of each scenario were separately input into the model according to the distribution of host species from 2020 to 2039.The purpose of our study is to compare the relative protection efficiency of different strategies.Therefore,we hope to compare them in the case of consistent defoliation.In the same scenario, we used the same defoliation patterns for each species.Because the species proportion of each grid is different,there are some spatial differences in defoliation.

Fig. 3. Estimation of annual defoliation for balsam fir, white spruce and red/black spruce under moderate and severe spruce budworm outbreak scenarios without early intervention strategy (EIS) (a, b) (modified from Hennigar and MacLean, 2010), with 8 years of EIS (b, e) and with 12 years of EIS (c, f) in the study area.Scenarios assume that the spruce budworm outbreak was merely delayed after EIS ceased,in a 20-year period.FP means foliage protection strategy.(For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

We also assumed that EIS,if completely effective,would result in no outbreak and tested a no SBW scenario(i.e.20 years of EIS)as a control to evaluate how much carbon was maintained vs.outbreak scenarios.In other words,EIS that works and is used for the duration of the outbreak(i.e.from 2020 to 2039)will be the same as our no outbreak scenario(N)in Table 1. If EIS is not completely effective, it could lead to a delayed outbreak which may also be shorter (e.g. 8 years of effective EIS could shorten a 20 years outbreak to 12 years) (Johns et al., 2019; MacLean et al.,2019).Forests would also continue to grow during the period that the EIS was effective.As the current EIS treatment has been effective for 7 years(MacLean et al.,2019),we tested scenarios where EIS was applied for 8 and 12 years,based on EIS funding ending after the current funding agreement(8 years)or renewal for only one additional 4-year period(12 years). Funding for testing EIS has been provided by the government of Canada and the four Atlantic provinces and forest industry, but is somewhat tenuous in that treatments precede the most severe defoliation(MacLean et al., 2019). Fig. 3 shows the annual defoliation dynamics under different scenarios for all pixels. Each scenario includes 3 defoliation patterns for different species. These scenarios were simulated separately in our model.

In order to compare the impact of different SBW management strategies on C dynamics of the above ground forest and the soil, we established 19 different scenarios(Table 1).The scenarios tested included no outbreak,moderate and severe outbreaks with 1)no protection,2)full FP(i.e.foliage protection of all areas with moderate or severe defoliation),3) FP10% (i.e. foliage protection applied to 10% of moderately or severely defoliated areas with the greatest estimated losses),4)8 and 12 years of EIS followed by no protection,5)8 and 12 years of EIS followed by full FP,and 6)8 and 12 years of EIS followed by FP10%(Table 1).We designed scenarios with full FP, EIS and their combination to evaluate effects on C sequestration under scenario with differing hypothetical pest intensity,but similar climate and forest throughout the study areas(i.e.2,855 grid cells). However, as applying FP to all of the outbreak area is financially and logistically unviable (it costs approximately $70 ha-1·year-1) (Forest Protection Limited,personal communication, June 2021),we also estimated scenarios in which FP is only applied to 10%of the grid cells (i.e. 285 grid cells) in Atlantic Canada with the greatest projected C losses(reduction in NEP)from SBW defoliation.The greatest C loss or efficiency of FP was estimated as:

Table 1 Description of the 19 different forest management strategies scenarios under moderate or severe defoliation by SBW with or without two levels of foliage protection (FP) – full protection everywhere or protection of 10% of defoliated areas,and with 8 years of EIS (E8)or 12 years of EIS (E12).

where NEPFP2020and NEPFP2039were the simulation of NEP with full FP scenarios in 2020 and 2039, respectively. EIS, on the other hand, treats small local SBW populations before population levels increase above an Allee threshold value(see Johns et al.,2019;MacLean et al.,2019),and so an efficiency target is not applicable for EIS. Instead all stands with rising SBW populations are treated,these are,however,concentrated on the spreading front of the outbreak rather than being found throughout the region.

3. Results

3.1. Mortality

Simulations from the TRIPLEX-Insect model indicated that average cumulative mortality for the period 2020–2039 in Atlantic Canada for the no defoliation(or 100%effective EIS)scenario was 10.3%.Across our study area,moderate and severe SBW outbreak scenarios were projected to increase cumulative mortality by 2–4 times to 26.9%–37.1% from 2020 to 2039.The E8 and E12 scenarios on average reduced mortality by 4.4%–6.7% and 15.2%–23.5%, respectively. Full FP would reduce mortality by 7.6%–16.9%, or more than E8 but less than E12 over the 2020–2039 period, but when FP was only applied to the top 10% efficiency areas,it had little effect on cumulative mortality(i.e.1.7%–2.4%reduction). Compared with EIS alone, EIS followed by FP10% did not significantly reduce the cumulative mortality (both E8FP10% and E12FP10%).The combination of E12 and full FP had the greatest effect in reducing tree mortality by 16.3%–26.4%following moderate and severe SBW outbreaks from 2020 to 2039. In addition, our model simulations showed that percent cumulative mortality of fir-spruce forests in NS and NB was almost twice that of PEI for all defoliated scenarios (Table S1).

3.2. Carbon fluxes

Compared to simulations without defoliation, the average NEP with defoliation from 2020 to 2039 decreased by 0.74–1.65 t C·ha-1·year-1under a moderate–severe outbreak in Atlantic Canada, respectively.More than 8,920–21,960 (moderate–severe, respectively) km2of defoliated areas(i.e.30.91%–76.12%of the whole study area)were predicted to be converted from C sinks to sources by 2039 if no measures are taken(Fig. 4). In our simulations, SBW had the greatest impact on C sequestration in the province of NS where the average cumulative NEP decreased by 33.0 (moderate) to 47.6 (severe) t C·ha-1from 2020 to 2039(Fig.S1).We also found that the combination of 8 or 12 years of EIS followed by full FP in areas disturbed by the SBW could maintain average cumulative NEP at levels similar to no defoliation in every province.Compared with full FP, a targeted FP10% could increase (compared to the no treatment scenario) about 42.6%–44.1% (moderate–severe,respectively) of NEP in NS, 57.4%–80.5% of NEP in NB (moderate–severe, respectively) and 48.9%–87.9% of NEP in PEI(moderate–severe, respectively) (Table S2), respectively. However,FP10%had no effect on the average cumulative NEP in NL.

Fig. 4. Simulated average net ecosystem productivity (NEP) (t C·ha-1·year-1)in 2039 under (a) no SBW outbreak, (b) a moderate SBW outbreak, and (c) a severe SBW outbreak across Atlantic Canada. The SBW scenarios applied everywhere are shown in Table 1. Level 1 means NEP ＞1.25 t C·ha-1·year-1;Level 2 NEP 0.1–1.25 t C·ha-1·year-1; Level 3 NEP –0.1–0.1 t C·ha-1·year-1;Level 4 NEP–1.25–(–0.1)t C·ha-1·year-1;Level 5 NEP ＜-1.25 t C·ha-1·year-1.

Fig.5. The efficiency of foliage protection(FP)on forest C in 2039 under(a)a moderate SBW outbreak and(b)a severe SBW outbreak across Atlantic Canada.Level 1 means FP increased NEP ＞3 t C·ha-1·year-1 than no protection;Level 2 means FP increased NEP 1–3 t C·ha-1·year-1 more than no protection; Level 3 means FP increased NEP 0–1 t C·ha-1·year-1 more than no protection; Level 4 means FP did not increase NEP more than no protection.

Based on the comparison of simulated results between no protection and the FP scenarios, we divided the efficiency of FP into four levels(Fig.5):Level 1 means FP increased NEP ＞3 t C·ha-1·year-1compared to no protection; Level 2 means FP increased NEP 1–3 t C·ha-1·year-1compared no protection; Level 3 means FP increased NEP 0–1 t C·ha-1·year-1more than no protection; Level 4 means FP did not increase NEP compared to no protection.Under a moderate SBW outbreak,the benefits of FP application to C maintenance in most areas were low for 102,078 km2of fir-spruce forests at Level 3 or nil for 76,934 km2at Level 4(Fig.5a).However,in a severe SBW outbreak,there were 50,787 km2of fir-spruce forests at Level 1 of FP efficiency and 80,948 km2of firspruce forests at Level 2 (Fig. 5b). Based on the severity of a SBW outbreak and financial support, priority could be given to using an FP strategy in Level 1–2 areas.Most of the area for the top 10%efficiency of FP was located in the province of NB for the FP10% and E8FP10% scenarios under a moderate or severe SBW outbreak (Table S2).

Fig. 6. Simulation of average net ecosystem productivity (NEP; t C·ha-1·year-1) with moderate (a) and severe (d) defoliation, soil organic carbon (SOC; t C·ha-1) with moderate (b) and severe (e) defoliation, and aboveground tree biomass (AGB; t C·ha-1)with moderate (c) and severe (f) defoliation scenarios in the study area. Shaded areas show the standard deviation with uncertainty estimates. Two foliage protection (FP)scenarios are illustrated: full FP illustrates a scenario where all areas defoliated for more than two years are sprayed with biological insecticide each year and top 10% efficiency foliage protection (FP10%) represents a scenario where only 10% of the defoliated area was protected as usually only a small proportion of affected areas can be protected.E8 means 8 years of early intervention strategy.E12 means 12 years of early intervention strategy.

Our simulated results also showed that the E8,E12,full FP and FP10%scenarios were predicted to increase average NEP by 0.2–0.7, 0.7–1.3,0.6–1.4 and 0.1–0.3 t C·ha-1·year-1in Atlantic Canada from 2020 to 2039,respectively,compared to the no treatment scenarios for moderate to severe defoliation(Fig.6a and d).Given the same period of simulated SBW outbreak, the combination of 8 years of EIS followed by full FP would maintain the most C(i.e.0.9 t C·ha-1·year-1under moderate and 1.8 t C·ha-1·year-1under severe defoliation scenarios) in study areas.However,full FP is hypothetical only,given the cost(at$70 ha-1·year-1,full FP would cost$1.3 billion per year).Only a very small of negligible gain in NEP occurred when FP10%followed EIS.The average cumulative NEP values under E8FP10%and E12FP10%were similar to E8 and E12,respectively(Table S2).

The average NEP simulated by TRIPLEX-Insect with the combination of EIS and full FP was positive for all species in defoliated areas(Fig.7a,b and c).Although balsam fir forests sustained the greatest losses of C when no protection measures were in place (i.e. from -1.6 to -1.8 t C·ha-1·year-1), the C maintained after the treatment (i.e. both EIS and full FP) was also the greatest for this forest type(Fig.7a).

As stand age increased,average NEP gradually decreased under most scenarios except the combination of EIS and FP for 60–80 years old forests,which was predicted to be 2.3 t C·ha-1·year-1for E8FP and 2.4 t C·ha-1·year-1for E12FP(Table S3).We also found that average NEP was positive for all scenarios when forests were less than 60 years old(Table S3).Under severe defoliation,E12 or FP could maintain forests of all ages as C sinks during the period 2020–2039.However,when EIS was applied for only 8 years (E8) and not combined with subsequent FP,forests over 100 years old did not remain C sinks under severe SBW disturbances during the 2020–2039 period(Table S3).When forests were over 60 years of age,E12 maintained more C than full FP under moderate scenarios but maintained less C under severe scenarios. Both E8FP and E12FP maintained more C than only full FP for all scenarios,while E8FP and E12FP maintained almost the same amount of C(Table S3).

Fig. 7. Simulated average net ecosystem productivity (t C·ha-1·year-1) from 2020 to 2039 for (a) balsam fir, (b) white spruce, and (C) red/black spruce. The blue lines indicate a carbon sink (NEP ＞0); the red lines a carbon source (NEP ＜0), under moderate (M) or severe (S) SBW defoliation with or without full foliage protection(FP),and under 8 years of early intervention strategy(E8)or 12 years of early intervention strategy(E12)scenarios.(For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

3.3. Carbon stocks

At the end of the SBW outbreak, neither the E8 nor even a full FP strategy prevented the reduction of average aboveground biomass(AGB).Compared with no outbreak, AGB decreased by about 17.4%–30.3%(moderate–severe) for E8 and 4.1%–9.2% (moderate–severe) under full FP,respectively,over the period(2020–2039)(Fig.6c and f).In contrast,E12 and the combined EIS (i.e. both 8 and 12 years of EIS) and full FP strategy had the same or higher AGB than no outbreak in Atlantic Canada(Fig. 6c and f). AGB increased by about 3.2% for E12 than in the no outbreak scenarios under moderate defoliation (Fig. 6c) and decreased by 5.6%under severe defoliation (Fig.6f)from 2020 to 2039.

Our results indicate that moderate and severe defoliation resulted in negative average annual volume increment (AVI) for most scenarios when forests were older than 60 years. The minimum AVI occurred in 80–100 years old forests without FP or EIS treatments,in which the AVI was -2.9 m3·ha-1·year-1under moderate defoliation (Table S3) and-4.0 m3·ha-1·year-1under severe defoliation (Table S3). Compared to the no treatment scenarios, all SBW management strategies increased AVI,but the combination of E12 and full FP maintained a positive AVI.In addition, the effect of E12 on live tree volume was significantly greater than a full FP for balsam fir and white spruce tree species (Fig. 8). We found that there was little difference between E8 and E12 for tree volume of red and black spruce species during either a moderate or a severe outbreak. This is also consistent with the combined EIS and full FP scenarios(i.e.E8FP and E12FP).

As the SBW outbreak progressed,simulations of average SOC indicate that it will first rise and then fall for no treatment, E8, full FP, and the combination of E8 and full FP scenarios in defoliated areas(Fig.6b and e).For E12 and the combination of E12 and full FP,there were no obvious changes in average SOC during the period 2020–2039 (Fig. 6b and e).SOC in simulations with SBW management strategies compared to those with no treatment showed that average SOC decreases for most scenarios except under severe defoliation when forests were older than 100 years(Table S3). Under any given strategy, we also found that SOC increased under severe outbreak scenarios more than under moderate scenarios when forests were less than 100 years old (Table S3). In addition,compared with E8 or E12, there were no significant changes in carbon stocks (both AGB and SOC) under E8FP10% and E12FP10% during the period 2020–2039(Fig.6b,c, e and f).

4. Discussion

Fig.8. Simulated average volume(m3·ha-1)in 2039 for balsam fir,white spruce,and red/black spruce under no(N),moderate(M)or severe(S)SBW defoliation with or without full foliage protection(FP),and under 8 years of EIS(E8)or 12 years of EIS(E12)scenarios in the study areas.(For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

In our study, balsam fir forests lost more C than spruce forests (i.e.white, black and red spruce) in unprotected forest areas. Consequently,management treatments (EIS or FP) also led to greater reductions in C losses in balsam fir than in spruce.This is due to the fact that balsam fir has a higher vulnerability to the SBW than spruce(Hennigar et al.,2008),but fir also has a higher growth rate than spruce with or without SBW disturbances (Alexander, 1987; Lemay et al., 2018). In addition, we believe that forest age,which Fuentealba et al.(2015)did not consider,is another important factor affecting the efficacy of a treatment. It is clear that younger forests grow more quickly than older forests,so they have a stronger capacity for carbon sequestration(Tang et al.,2014)while older forests are not only less productive but are also more vulnerable to the SBW (Houndode et al., 2021). In this study, most balsam fir forests, for which forest density is greater than 50%,are younger than 60 years old.For spruce forests (i.e. white, black and red spruce), the average age exceeds 80 years.This may partly explain the higher changes of C fluxes and stocks(i.e.NEP and volume)with and without treatment estimated by our TRIPLEX-Insect model for balsam fir than for any of the spruces under moderate and severe SBW outbreaks in our study areas.Fuentealba et al. (2019) also show that Btk foliage protection operations maintain foliage and keep white spruce and balsam fir alive whereas black spruce did well even in the absence of protection.Hennigar et al.(2008)showed that SBW population levels that produced 100%current year defoliation in balsam fir only resulted in an average of 28% defoliation of black spruce. Our results indicated that FP can keep cumulative mortality to 18.9%–25.4%for white spruce and to 10.0%–10.3%for black/red spruce following moderate to severe defoliation over the 2020–2039 period.In the absence of SBW defoliation,the simulated cumulative mortality was 9.8%±0.3%for white spruce and 9.9%±0.3%for black and red spruce from 2020 to 2039 in Atlantic Canada.

In this study, our TRIPLEX-Insect model simulations demonstrated that full FP could reduce forest C losses more than E8 but less than E12 over a long-term SBW outbreak period (i.e. from 2020 to 2039). Thus,unless EIS is carried out successfully for long enough, by slowing down the onset and potentially reducing the length of an outbreak, or FP is carried out at very high rates that are not currently economically viable,these treatments will still result in substantial C losses. This result thus adds a nuance to the findings of Liu et al.(2019a),as they indicated that EIS can reduce timber harvest losses and economic costs more than foliage protection from 2015 to 2026 but they assumed that EIS continues to perform with 100%efficacy over the 11-year SBW outbreak period.It remains to be tested whether such a hypothesis is realistic(Johns et al.,2019)nonetheless their results are similar to the no outbreak scenario in our study. Our results are consistent with Liu et al. (2019a), if we consider that a perfect and permanent implementation of EIS would result in a no outbreak scenario.In fact,after EIS stops,another possible scenario is that the SBW outbreak will continue for 20 years (i.e. the outbreak is not shortened). In this case, it only postpones the outbreak while allowing the forests to age. In other words, immature forests will become mature forests and thus more vulnerable to SBW induced mortality and carbon loss(Kneeshaw et al.,2021).

There is evidence that EIS has worked for the duration of the period that it has been tested, which is now 2014–2020 (Liu et al., 2019a;MacLean et al.,2019;Johns et al.,2019).This is shown by an empirical comparison of defoliation levels in QC and NB, and aerial defoliation survey results from QC and NB showing substantial differences in defoliation ending at the northern NB border. In contrast with the western area of the outbreak (i.e. the Qu′ebec-Ontario border) where spread has been slow despite a largely hands-off approach, in the eastern maritime region, which is the central part of the insect's range, the outbreak has rapidly expanded through south eastern Qu′ebec towards NB and across the Gulf of St. Lawrence to western Newfoundland. EIS seems to have created a fairly hard border between the Quebec and NB.However,there is still no evidence that EIS will perform perfectly and keep zero or very low SBW defoliation for long periods (i.e. more than 12 years) given possible long-distance migration of SBW (Boulanger et al., 2017).Therefore, in this study, we tested and simulated a range of possible durations of successful EIS treatment.

Results from this study indicate that both partially effective EIS followed by full or partial FP can reduce and mitigate SBW disturbance impacts on regional forest C sequestration. These strategies were tested for different SBW outbreak severities (i.e. moderate vs. severe defoliation), an approach of scenario testing that is complementary to previously published studies on the SBW-Decision Support System(SBW-DSS)(Hennigar et al.,2013).Thus,combining our model results with different defoliation scenarios can help forest managers to optimize forest C protection.Full FP could reduce forest C loss more effectively if the effect of continuous use of EIS weakens over time. However, as full FP is financially unrealistic our work thus provides support for EIS and fodder to convince decision makers to invest in EIS when political appetite is low as this is the most efficacious solution to reducing C loss(Kneeshaw et al.,2021). For example, even if EIS does not last, our model simulations suggest that a combination of E8 and full FP could significantly reduce tree mortality and prevent forest C losses more than only E8 or full FP,which is consistent with our research hypothesis that the combination of EIS and full FP will maintain greater forest C sequestration capacity under future moderate or severe SBW outbreaks.However,if EIS can be implemented continuously and effectively for more than 12 years,FP will no longer be required in a 20-year outbreak cycle.Thus,a commitment to continue EIS will pay greater dividends in terms of C maintained than a reliance on FP which even with the unrealistic full protection cannot achieve the same C savings.

Fig.4 illustrates spatial variability in impact of outbreaks on NEP.In the same scenario, we enter the same defoliation patterns (no random outbreak initiation).The main reasons for this spatial heterogeneity are the distribution of host species and different tree ages. As mentioned above defoliation differs by species and this is what drives the variation in response.Our goal is thus to evaluate how current forest composition and structure influence carbon dynamics under different possible SBW outcomes. Meanwhile the purpose of simulating NEP is to compare the relative protection efficiencies of different strategies. Therefore,comparing the simulation value of NEP with other studies alone would be helpful but is not within the scope of this study but could be tried in future study. No other papers have tested the same suite of scenarios.Dymond et al. (2010) did compare their NEP results with those from TRIPLEX 1.0 noting that although TRIPLEX provided slightly higher estimations than her model that they were comparable with her estimates and FLUX tower estimates.

Although our model simulations present interesting findings, there are still some limitations and uncertainties in this study. For complex boreal and Acadian forest ecosystems, multiple processes interact and impact forest C dynamics at different time and spatial scales.Factors that were not possible to capture in our model framework include: 1)TRIPLEX-Insect did not consider insect respiration processes and decomposition of dead larvae, which influence natural ecosystem C budgets at a large regional scale(MacLean,2016);2)Lowland or wetland soil conditions,which have lower growth rates(Chen et al.,2017),were not considered by the TRIPLEX-Insect model in this study(this may lead to overestimation of NEP and AVI in Atlantic Canada);3)Climate change may significantly impact forest C dynamics and SBW population dynamics (R′egni`ere et al., 2012). Although the current TRIPLEX-Insect model can simulate the impacts of different climate change scenarios,it does not yet consider the impacts of future climate change on SBW population dynamics; 4) Regeneration dynamics were also not considered and thus different recruitment between species could lead to long-term divergences in overall C response; 5) The accuracy of forest inventory data and use of 10 km ×10 km grid cells of forest may bring some uncertainty,and a follow-up study could be based on more accurate provincial inventory data. Another factor omitted here is the effect of mixed deciduous-host species stands and landscapes in reducing the defoliation levels of host species (Su et al., 1996; Zhang et al., 2018,2020; Houndode et al., 2021). Finally, our results also consider C emissions only from the selected forests and not from the treatment operations themselves.

In the future,we should consider and include these possible additional factors on forest C simulations using the TRIPLEX-Insect model. In addition,we simulated all forests in the study areas that might be infected with SBW, however, our model simulation results represent only the what-if future scenario in Atlantic Canada. Although it is impossible to accurately predict population dynamics of the pest itself under changing future environment conditions, simulating future SBW outbreak and EIS or FP treatment scenarios can help forest managers and policy-makers to make scientifically sound pest risk management decisions.

5. Conclusions

This study reports on TRIPLEX-Insect model simulations based on different forest management strategies (i.e. no treatment, EIS, foliage protection, and their combination) and 19 different scenarios to investigate the effect of moderate and severe SBW outbreaks on forest carbon dynamics and tree mortality across Atlantic Canada. Our model simulations demonstrate that 8,920–21,960 km2of forested area, affected by moderate to severe SBW outbreaks without any management treatments,could convert from C sinks to sources by 2039 in Atlantic Canada. EIS(even if only effective for a limited number of years),foliage protection(everywhere or only the top 10% of efficiency areas) and their combination can maintain higher NEP of boreal forests than no treatment during SBW disturbances. Over the 20-year period of evaluation, the combination of EIS and foliage protection had the most effective benefits on carbon sequestration. According to our scenarios, if an uncontrolled SBW outbreak follows EIS treatments,a jurisdiction would have to invest in full foliage protection of all vulnerable forests to maintain more carbon than 8 years of EIS but even full FP would not be as good as 12 years of successful EIS in maintaining forest C. We also found that black/red spruce forests or forests younger than 60 years old have minimal changes in C dynamics under different management strategies. Therefore, stand composition,age and their interactions are important factors regulating C dynamics and the efficacy of management strategies during moderate or severe SBW outbreaks and commitments to effective EIS would provide better carbon maintenance than even the most aggressive FP strategies.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgements

This study was a part of an Early Intervention Strategy research project funded by Natural Resources Canada and the Healthy Forest Partnership.The original model development was financed by the Fonds de Recherche du Qu′ebec (FQRNT) program and a Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery Grant. We wish to thank Dr.Qiuyu Liu of the University of Qu′ebec at Montreal for supplying the climate data used in these simulations.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://do i.org/10.1016/j.fecs.2022.100052.