Salvage logging following fires can minimize boreal caribou habitat loss while maintaining forest quotas: An example of compensatory cumulative effects

Journal of Environmental Management 163 (2015) 234e245

Contents lists available at ScienceDirect

Journal of Environmental Management journal homepage: www.elsevier.com/locate/jenvman

Research article

Salvage logging following fires can minimize boreal caribou habitat loss while maintaining forest quotas: An example of compensatory cumulative effects
de
ric Raulier b, d Julien Beguin a, *, Eliot J.B. McIntire b, c, Fre D
epartment de biologie, Universit
e Laval, Qu
ebec, Qu
ebec, G1V 0A6, Canada
Centre d'Etude de la For^ et (CEF), Canada c Pacific Forestry Centre (Canadian Forest Service, Natural Resources Canada), 506 West Burnside Road, Victoria, British Columbia, V8Z 1 M5, Canada d
Departement des sciences du bois et de la for^ et, 2405, rue de la Terrasse, Universit
e Laval, Qu
ebec, Qu
ebec, G1V 0A6, Canada a

b

a r t i c l e i n f o

a b s t r a c t

Article history: Received 24 October 2014 Received in revised form 21 July 2015 Accepted 11 August 2015 Available online xxx

Protected area networks are the dominant conservation approach that is used worldwide for protecting biodiversity. Conservation planning in managed forests, however, presents challenges when endangered species use old-growth forests targeted by the forest industry for timber supply. In many ecosystems, this challenge is further complicated by the occurrence of natural disturbance events that disrupt forest attributes at multiple scales. Using spatially explicit landscape simulation experiments, we gather insights into how these large scale, multifaceted processes (fire risk, timber harvesting and the amount of protected area) influenced both the persistence of the threatened boreal caribou and the level of timber supply in the boreal forest of eastern Canada. Our result showed that failure to account explicitly and a priori for fire risk in the calculation of timber supply led to an overestimation of timber harvest volume, which in turn led to rates of cumulative disturbances that threatened both the long-term persistence of boreal caribou and the sustainability of the timber supply itself. Salvage logging, however, allowed some compensatory cumulative effects. It minimised the reductions of timber supply within a range of ~10% while reducing the negative impact of cumulative disturbances caused by fire and logging on caribou. With the global increase of the human footprint on forest ecosystems, our approach and results provide useful tools and insights for managers to resolve what often appear as loseelose situation between the persistence of species at risk and timber harvest in other forest ecosystems. These tools contribute to bridge the gap between conservation and forest management, two disciplines that remain too often disconnected in practice. © 2015 Elsevier Ltd. All rights reserved.

Keywords: Woodland caribou Rangifer Forest management Salvage Fire regime Protected areas Conservation planning Landscape simulation model Population viability Timber supply

1. Introduction Protected area networks are the dominant conservation approach that is used worldwide for protecting biodiversity (Millennium Ecosystem Assessment, 2005). Conservation planning in managed forests, however, presents many challenges when species at risk are found in or use old stands (“old-growth”) that have been targeted by the forest products industry for their timber supplies. On one hand, prohibition of timber harvesting within

* Corresponding author. E-mail addresses: [email protected] (J. Beguin), [email protected]. ca (E.J.B. McIntire), [email protected] (F. Raulier). http://dx.doi.org/10.1016/j.jenvman.2015.08.009 0301-4797/© 2015 Elsevier Ltd. All rights reserved.

protected areas may carry high economic opportunity costs, i.e., the loss of economic gain from timber harvesting when protection is chosen. On the other hand, the absence of regulations for timber harvests may result in unacceptable rates of habitat and species losses (Verkerk et al., 2014). Conservation issues that are related to changes in compositional and structural attributes of old-growth forests caused by forest management operations are recurrent in many ecosystems for a wide range of organisms, including lichens (McMullin et al., 2013), carabid beetles (Paillet et al., 2010), birds (Imbeau et al., 2001), and mammals (Di Marco et al., 2014). In the boreal forests of North America, conservation of forestdwelling boreal caribou (Rangifer tarandus caribou, hereafter “boreal caribou”), which is a wide-ranging and threatened ecotype that ranges from Labrador to the Northwest Territories in Canada

J. Beguin et al. / Journal of Environmental Management 163 (2015) 234e245

(Environment Canada, 2011), is a key challenge for conservationists and forest managers, as both boreal caribou and the forest industry value the same old-growth conifer forests. The sensitivity of boreal caribou to natural and human-induced disturbances is well documented with respect to fire (Beguin et al., 2013), logging (Courtois et al., 2008), roads (Wasser et al., 2011), and petroleum and natural gas infrastructures (Dyer et al., 2001). Furthermore, anthropogenic disturbances are the main cause of range recession in eastern North America (Schaefer, 2003; Vors et al., 2007; Festa-Bianchet et al., 2011). In this context, understanding how conservation and forest management plans are influenced by the interactions among fire, timber harvest, and protected areas is a priority. Catastrophic wildfire is a key disturbance process shaping the natural dynamics of boreal forest ecosystems (Johnson 1996). Because fires are dynamic spatial processes, they influence the outcomes of forest management and conservation strategies in terms of timber supply and habitat suitability, respectively. For instance, fires can cause timber losses for the industry in the medium-term, but also make a large volume of timber readily available over a very short temporal window through salvage logging, i.e., the logging of burned trees. Costs and benefits of salvage logging, however, may greatly depend upon the spatial characteristics of landscape attributes, such as the development of road networks and the existence of protected areas. Despite its relevance for conservation, the nature and extent to which interactions among fires, conservation areas, and forest management practices influence trade-offs between economics and conservation objectives are still little known (but see Schneider et al., 2012). Several studies have investigated these interactions using a subset of processes or using biological indicators that are not directly related to demographic parameters of focal species (Morgan et al., 2007; Rempel et al., 2007; Hauer et al., 2010), such as the probability of occurrence or habitat suitability indices. Yet it is unclear how changes in these indicators directly relate to risks to species persistence over time. Moreover, we are aware of no study that has evaluated the effect of salvage logging on trade-offs at large spatial scales between forestry and conservation of a wide-ranging species at risk. When long-term conservation objectives (e.g., persistence of species at risk) are at odds with short-term socio-economic goals (e.g., timber supply), a proactive way for improving gains in conservation efficiency would explicitly integrate economic indicators

235

into conservation planning (Polasky et al., 2005; Naidoo et al., 2006). Our main objective, therefore, was to quantify how interactions among fires, protected areas and forest management shape the population dynamics of boreal caribou (a conservation objective) and the annual timber supply for the forest industry (an economic objective) in order to identify possible trade-offs between caribou conservation and forest management. We first assessed how conservation areas affected levels of sustainable timber supply under fire risk in three forest management units of > 104 km2 each, hereafter referred to as FMUs. This assessment allowed us to quantify the relative effect of the current conservation strategy on a key economic indicator for the forest industry. Second, we assessed the separate and joint impacts of fire, (salvage) logging, and road networks on the annual population growth rate of boreal caribou under various levels of protection. This allowed us to quantify the ecological consequences of timber harvesting and fire on the probability of persistence of boreal caribou under various conservation schemes. We controlled the amount and spatial configuration of protected areas, (salvage) logging, fire, and roads using a spatially explicit landscape simulation model in which each of these processes interacts annually in space and time over a time horizon of 150 years. Based on existing empirical disturbancepopulation growth equations (Environment Canada, 2011), we used these simulated disturbance events as inputs in caribou population viability analyses (ecological indicator). Using yield tables, we also calculated the amount of timber volume that can be harvested (economic indicator) following the sustainable principles of forest management. We designed a series of experiments to uncover the influence of disturbance types and the amount of protected area on both economic and ecological indicators. 2. Material & methods 2.1. Study area The study region (38 844 km2) is located in the southeastern ^te-Nord continuous range of boreal caribou and is part of the Co (North Shore) administrative region, which is located in the eastern part of the Province of Quebec, Canada (Fig. 1A). Forested areas represent 76% of the study region and are dominated by coniferous

Fig. 1. Location of the study area, with delineation of A) the three Forest Management Units (FMUs) used in this study, and B) conservation areas (IUCN protected areas in black and caribou protection blocks in light grey). See Table A.1. in Appendix 1 for further details on initial conditions in each FMU.

236

J. Beguin et al. / Journal of Environmental Management 163 (2015) 234e245

stands in which black spruce (Picea mariana (Miller) BSP), balsam fir (Abies balsamea (L.) Miller) and jack pine (Pinus banksiana Lambert) are the most frequently encountered tree species (Table A.1. in Appendix 1). Soil and climate conditions are described in detail by Bouchard et al. (2008). The two main disturbance regimes are fire and timber harvesting. The mean fire return interval is estimated at ~270 years [95% CI: 234e310] (Bouchard et al., 2008) and the empirical fire size distribution is known for the last 200 years (Bouchard and Pothier, 2008). Industrial forestry (clear-cutting systems) commenced during the second half of the 20th century and current logging activities are organised according to FMUs (see Fig. 1 and Table A.1.). Governmental authorities are responsible for calculating the sustainable timber volume that can be harvested annually in each FMU. In all the Province of Quebec, the boreal caribou has the status of “vulnerable species” and the cornerstone of the current recovery strategy relies upon a protected area network that is aimed at main
taining a sufficient amount of old-growth conifer forest (ERCFQ, 2008, 2013). The current network in our study region is composed of two

types of protected areas. The first type corresponds to permanent protected areas belonging to categories Ia and III of IUCN (the International Union for Conservation of Nature) (Fig. 1B; see Table A.1 for more details). The second corresponds to caribou protection blocks that have been established as floating reserves in space and time. In both types of protected areas, logging and road construction are forbidden during the period of the planning horizon. 2.2. Data and simulation sub-models We used forest inventory data to map information on forested and non-forested areas, disturbance (logging and fire) history, and road networks (Table A.1.). All maps were rasterised to a 200 m
200 m pixel size prior to being used in our simulation experiment. We used the Spatially Explicit Landscape Event Simulator (SELES, Fall and Fall, 2001), a generic platform to build landscape simulation modules (or sub-models). We used SELES to build modules for forest growth and aging, fires, conventional logging, salvage logging, and the amount of protected area. We

Fig. 2. Diagram showing the main steps of our simulation model, together with a description of the main inputs, processes, outputs, and indicators. Boxes and arrows indicate how elements interrelate with one another in space and time. Inputs are composed of static spatial variables (their values do not change in time and space), dynamic spatial variables (i.e., values change in time and space with processes) and stand yield tables. The main processes involve three spatiotemporal sub-models: fire regime, logging regime, and road building (see Methods for details). Outputs can be defined as the spatially explicit results of actions caused by processes operating on input data. These outputs are spatially updated at each time-step. Indicators track the changes in output maps at each time-step that are relevant from conservation (population growth rate of boreal caribou populations: l) and forest management (the annual timber supply).

J. Beguin et al. / Journal of Environmental Management 163 (2015) 234e245

created these modules in modifying preexisting modules already developed in SELES to investigate ecological and management questions within forested landscapes (Fall et al., 2004; Didion et al., 2007). Using the SELES platform allows our dynamic modules to interact over space and time in three FMUs. We used these modules to vary the fire regime, the logging regime (with and without salvage logging), and road construction (Fig. 2). We did very little direct model calibration; rather we used empirically determined parameters and relationships for most parts of the sub-models (Table A.2. in Appendix 2). However, for fires, we did calibrate the percolation model to fit with the empirical fire distribution of our landscape, as described below. We classified forest stands into four different species groups based on their composition, growth profile, tolerance to light and successional status as: 1) shade-tolerant conifers (black spruce and balsam fir), 2) shade-intolerant conifer (jack pine), 3) mixed/deciduous species (trembling aspen and paper birch), and 4) lichen woodland (Table A.1.). For each species group, we used yield curve predictions of stand volumes of coniferous species as a function of age. These yield curves were estimated as area-based weighted averages from yield tables and were calculated by the Chief Forester's Office (Quebec) for the last timber supply analysis of 2008e2013, following the model of Pothier and Savard (1998). We assumed pixels that were undergoing a stand-replacing disturbance (fire or logging, see below) remained in their initial species group. We simulated a fire regime using a standard percolation model with a mean fire interval of 300 years (Bouchard and Pothier 2008). We modelled the annual number of fires over the study region as a Poisson process, i.e., Nfire ~ Poisson(lfire), where lfire is the mean number of fires > 10 km2 that ignite each year over the study region (Table A.2.; Fig. A.2. in Appendix 2). We calibrated the percolation model by varying the two parameters, scale and shape, of the Reverse Weibull distribution and comparing them to data about fire size distribution in our landscape (Bouchard and Pothier, 2008). We were able to obtain good approximations of the observed fire size distribution (Table A.2.; Fig. A.1. in Appendix 3). To prevent unrealistically high fire sizes, we truncated maximum fire size to 2130 km2, which corresponds to the largest fire that has been observed in our study area since 1800 (Bouchard et al., 2008). We based our logging component on a simple harvesting module developed by Didion et al. (2007) and we modified it for our landscape to allow logging rules based on cut block size, stand age, and distance from roads and mills. Specifically, we simulated cut-blocks with sizes ranging from 50 to 150 ha (Table A.2.). We distinguished logging of undisturbed stands from salvage logging. For undisturbed stands, logging in a focal pixel was determined by the following criteria: 1) stand age had to be greater than 90 years (Table A.2.); 2) older stands were harvested first; 3) stands that were located at shorter distances to roads were given priority over stands that were located further away (Table A.2.); and 4) stands that were located at a shorter distance to mills were logged in priority. These criteria were translated into probabilities and

237

combined to give a harvest probability surface at each time-step. Other rules for logging could have been added, but we limited the number of constraints to maintain a balance between complexity and generality.
bec, the percentage of salvaged fires regularly increased In Que after 1986 and between 1995 and 2007, approximately 1.8
106 m3 of wood was salvaged per year (z6% of the annual allowable cut) with the highest peaks in years of high burn rate (Nappi et al., 2011). Overall, salvage logging is limited by economic considerations (lack of access, low volumes that can be salvaged) and by the rapid degradation of burned wood (Nappi et al., 2011). For simulating salvage logging, we retained two main criteria: 1) the salvageable volume; and 2) the distance to the closest road. The maximum standing salvage volume that was available for harvest in a given year in each pixel was thirty percent of the volume prior to the fire event (J. Duval, MRNF, Pers. comm.) and a minimum of 50 m3/ha prior to disturbance was required for salvage logging to occur (Table A.2.). When a fire occurred and conditions for salvage logging were satisfied, salvage logging was given higher priority (probability) over standard logging because of the urgency of harvesting burned volumes to avoid rapid degradation of wood by insects (Nappi et al., 2004). We constrained the occurrence of salvage logging within the first three years after fire and we prevented salvage logging from occurring far from the existing road network when the salvageable volume for harvest was low. In all cases, the amount of salvage volume that was harvested substitutes for (and is not added to) the annual allowable cut. Once a burned area was salvaged, it was reclassified as a logged area and included as such in the recruitment-disturbance relationship for caribou (see below). We followed the recommendations of Environment Canada (2012) and considered stands as disturbed for caribou (either by logging or fire) as long as the time since disturbance was 40 years. We based the road construction component on the default road construction in Didion et al. (2007), again modified for our landscape. Specifically, the construction of roads closely matched the occurrence of cut-blocks. Each time a cut-block was located further than 500 m from the nearest road (maximum skidding distance), a new road was built using the Euclidean distance to connect the centroid of the new cut-block to the nearest existing road. As cutblocks are more likely to occur at a short distances from roads, this rule emulates the known pattern of progressive south-to-north logging intensity and road density. Roads did not occur within protected areas and water bodies. 2.3. Simulation experiment and definition of scenarios To gain a comprehensive view of the interactions among fire, logging, salvage logging and protected areas, and their effects on caribou population dynamics and timber supply, we designed a factorial experiment in which we simultaneously controlled for the type of disturbance regime and the amount of protected area (Table 1). We first controlled the level of disturbance by teasing

Table 1 Description of simulated scenarios of conservation and disturbance regimes used in this study. In the sub-model “gradient of protected areas,” we simulated four different levels of protection for each disturbance scenario: no protected areas (control), only caribou protection blocks, only IUCN protected areas, and both caribou protection blocks and IUCN protected areas simultaneously (see text for further details). Class of sub-models

Sub-models

Disturbances

Fire Conventional logging Salvage logging Road construction Gradient of Protected areas

Conservation

Conservation-disturbance scenarios Logging only

Logging þ fire (no salvage logging)

Logging þ fire þ salvage logging

√

√ √

√ √

√ √

√ √ √ √ √

238

J. Beguin et al. / Journal of Environmental Management 163 (2015) 234e245

apart the relative effects of fire, logging, and salvage logging. Within each FMU, we designed three disturbance scenarios: 1) logging only; 2) fire and logging without salvage logging; and 3) fire and logging with salvage logging (Table 1). The “logging only” scenario mirrors a strategy where fire suppression would be 100% effective, which is similar to the assumption made by the Government of Quebec for calculating the level of timber supply in every FMU. This scenario is also useful for comparative purposes. The second and third disturbance scenarios depict two different strategies regarding the use (or not) of salvage logging as a management tool. Second, we designed four conservation scenarios that differed in the amount of conserved area (Table 1). In each FMU, we designed a gradient of proportions of conservation areas by subsampling currently protected areas into four subclasses: 1) no protected areas (control); 2) IUCN protected areas only; 3) caribou protection blocks only; and 4) a combination of both types of protected areas. IUCN protected areas were created by governmental agencies as part of a general strategy to protect natural areas, while caribou protection blocks were designed specifically to address the needs of boreal caribou (Courtois et al., 2004). This partitioning within each of the three FMUs yielded twelve combinations for the proportions of conservation areas, which encompassed an overall gradient ranging from 0 to 25.6% of the FMU areas. This procedure allowed us to avoid the creation of arbitrary and random protection areas, which would have been unrealistic in practice. Although caribou protection blocks are aimed to be dynamic, their effective duration and spatial localisation in the future are currently unknown. We overcame this limitation by considering caribou protection blocks as permanent and we interpreted the results accordingly. The combinations of levels of conservation and of disturbance within each FMU, which are hereafter referred to as conservationdisturbance scenarios (Table 1), covered a wide assortment of conditions, which ranged from a low proportion of conservation areas with low disturbances rates to a high proportion of conservation areas with high disturbance rates. To investigate the effectiveness of the current protected area network for boreal caribou and its relationship with timber supply and fire risk, we evaluated the impact of each conservation-disturbance scenario for each FMU according to two indicators: 1) the annual population growth rate (l) of boreal caribou (biological indicator); and 2) the sustainable timber supply (m3/year) (economic indicator).

In our study area, no available empirically-based relationship currently links adult caribou survival to habitat and landscape features, so we estimated adult female survival for each year independently of the habitat attributes using random draws from a normal distribution, based on the national average following Environment Canada (2011) (Table A.2.). We derived mortality, M, as 1 e survival. In line with the current recovery strategy for boreal
caribou in Quebec (ERCFQ, 2013), we retained the delineations of the FMUs to assess demographic parameters. The rationale for this choice was as follows: 1) the area of each FMU is greater than the minimum area (104 km2) that is recommended for evaluating demographic parameters for boreal caribou (Environment Canada, 2011); and 2) previous estimates of demographic parameters around our study region were made within three main study areas (see Courtois et al., 2007). 3) Boreal caribou occur within each FMU. One of the areas (Manicouagan) fully encompassed our study region and overlapped, at more than 85%, FMU 1. To include parameter uncertainty as a source of stochasticity in our simulations, we drew normal random variables for b0, b1 , recruitment rate, and survival rate (Table A.2.). Once we estimated R and M, we calculated l at each time-step for each FMU and each conservation-disturbance scenario. To match thresholds that were established by Environment Canada (2012), we calculated, for each year and for each FMU, the probability that l > 1 (Probl>1), which is the number of replicates where l > 1 divided by 100, the total number of replicates. In all conservation-disturbance scenarios, the relationship between l and time followed an inverse S-shaped curve, indicating the existence of a transition phase between the effects of initial conditions and reaching an equilibrium stage. As our primary focus was to evaluate the long-term persistence of boreal caribou, we defined leq as the arithmetic mean value of l when the equilibrium stage was reached (see Fig. 4). Following risk analysis principles and recommendations from Environment

2.4. Indicators Following Hatter and Bergerud (1991) and assuming that immigration and emigration compensate one another, l can be expressed as a function of the annual adult mortality rate (M) and the annual recruitment rate (R), where l ¼ (1-M)/(1-R). In a Canada-wide analysis across a sample of 24 boreal caribou population ranges, Environment Canada (2011) found that 69% of the variation in caribou recruitment rate, which was expressed as the calf:cow ratio and assuming a 1:1 sex ratio of calves, was explained by the percentage of total disturbance in population ranges, according to the following linear relationship:

Recruitment rate ¼ b0 þ b1
total disturbance þ

3

(1)

where b0 and b1 were regression coefficients estimated by Environment Canada (2011) (see Table A.2.) and Ɛ, the residual model error. Here, as is the case for Environment Canada (2011), total disturbance refers to burned areas plus logged areas with a buffer of 500 m, plus roads with a buffer of 500 m. For each conservation-disturbance scenario that we simulated, all of these variables were annual outputs of the fire, logging and road models (Fig. 2).

Fig. 3. Relationship between the proportion of conservation areas (IUCN protected areas þ caribou protection blocks) and the reduction of the annual timber supply as a function of Forest Management Unit (FMU) (circles: FMU 1, diamonds: FMU 2, and triangles: FMU 3) and disturbance scenarios (red colour: Logging only; green: Logging þ Fire with salvage logging; blue: Logging þ Fire without salvage logging). The reference scenario for calculating the reduction of annual timber supply corresponds to the conservation-disturbance scenario with no fire and no conservation areas (see text for further descriptions). Grey-shaded rectangles correspond to the current amount of ^ te-Nord region, Que
bec. conservation areas in 2012 in the three FMUs of the Co

J. Beguin et al. / Journal of Environmental Management 163 (2015) 234e245

239

3. Results 3.1. The amount of conserved area vs timber supply under multiple disturbance regimes

Fig. 4. Relationship between the percentage reduction of annual timber supply that is necessary to achieve a sustainable timber yield and the probability that the rate of caribou population growth at equilibrium (leq) will be stable or increasing, Pr(leq>1), as a function of Forest Management Unit (FMU) (circles: FMU 1, diamonds: FMU 2, and triangles: FMU 3) and disturbance scenarios (red colour: Logging only; green: Logging þ Fire with salvage logging; blue: Logging þ Fire without salvage logging). The dashed lines and grey shaded area demarcate the regions of low (Pr(leq>1) < 0.4), medium (0.4  Pr(leq>1)  0.6), and high (Pr(leq>1) > 0.6) probability of persistence to boreal caribou. The figure on the right corner shows how leq is calculated from the inverse S-shape relationship between Pr(l > 1) and the year of simulation (Nsimulation ¼ 100).

Canada (2012), if Probl>1 is higher than 0.6, the risk that the population range cannot support a self-sustaining caribou population is from low to very low, if 0.4 < Probl>1 < 0.6, the risk is intermediate, and if Probl>1 < 0.4, the risk is from high to very high. For assessing the socio-economic impact of each conservationdisturbance scenario, we calculated the annual timber supply for each FMU that could be harvested according to the long-term sustainability of wood production (Bettinger et al., 2009). Under the sustainable yield paradigm, a sustainable supply of timber is the largest amount of timber volume that can be harvested within a given period of time without degrading the productive capacity of the timber stock (the capital). It is a key economic indicator for the forestry sector as it relates directly to potential socio-economic activity. Iterative simulations and maximization procedures using linear programming are the most common techniques to estimate sustainable timber supply (Bettinger et al., 2009), while the former is often used when stochastic events are taken into account (e.g. fire occurrence). As we simulated both stochastic and deterministic events, we determined the maximum timber supply in each FMU using iterative simulations with the following tolerance rule: 95% of simulations for each conservation-disturbance scenario had to achieve no shortfall in timber supply greater than 10% over the entire simulation length (Peter and Nelson, 2005). In other words, the amount of annual timber supply was adjusted by iteration until this stopping rule was not rejected. For each FMU and for each conservation-disturbance scenario, this iterative process yielded an estimate of sustainable annual timber supply (m3/year) (Table A.2.). We initialised the simulations with the most recent maps of the study region (i.e., 2010). Each scenario was run for one hundred and fifty years with an annual time-step. We conducted exploratory analyses and performed 100 independent simulations of each conservation-disturbance scenario, which provided stable estimates of indicators.

In the absence of conservation areas and fires (logging scenario only), no reduction in annual timber supply occurs in all three FMUs over the duration of the simulation (Fig. 3). This result indicates the values that we obtained for annual harvest yield (1 010 000, 386 000, and 300 000 m3/year in FMUs 1, 2, and 3, respectively) were appropriate based on sustained yield principles. Increasing conservation areas within FMUs by 1% decreases the percentage of annual timber supply by the same proportion (Fig. 3; slope coefficient ¼ 0.95 ± 0.07 (standard error - SE), see Table A.5. in Appendix 4). Interestingly, this 1:1 relationship is robust to heterogeneity both among disturbance regimes and FMU initial conditions, as we found no effect of random slopes on these variables using mixed-effect regressions (see Table A.3. in Appendix 4). This 1:1 relationship can thus be confidently applied across the study region. Much variation, however, was found in the intercept of this linear relationship, with differences among disturbance scenarios explaining three times more variation (variance of the random intercept on disturbance scenarios ¼ 207.1) than differences among FMUs (variance of the random intercept on FMUs ¼ 62.2) (see Fig. 3 and Table A.5.). High variability in the intercepts has deep implications for the identification of mitigation measures between conservation and forest management. For instance, along the gradient of conservation areas, the presence of logging and fires (without salvage logging) respectively decreases the annual timber supply by ~25, ~40 and ~50% in FMUs 1, 3, and 2, compared to the hypothetical baseline “logging only” scenario (Fig. 3). Consequently, under fire regimes and current conservation schemes (see shaded areas in Fig. 3), FMU 1 is likely to exhibit the lowest reduction in annual timber volume among the three FMUs (from 20 to 30% for FMU 1 with or without salvage logging, compared to 20e65% and 30e60% for FMUs 2 and 3, respectively) (see Fig. 3). Overall, fires have proportionally greater negative impacts on timber supply shortfalls than does increasing the proportion of conservation areas within FMUs from 0 to 25.6% (Fig. 3). 3.2. Timber supply and the probability of self-sustaining populations of boreal caribou The probability that the number of boreal caribou will be stable or would increase over the long-term (Pr(leq>1)) increases linearly in each FMU with the increasing percentage reduction in annual timber supply, according to a 1:1 ratio (Fig. 4; slope coefficient ¼ 1.11 ± 0.11 (SE), see Table A.5.). Contrary to the conservation area-timber harvest relationship, however, we detected an effect of both random intercept and slope on the variable FMUs (see Table A.4. in Appendix 4). This 1:1 relationship, therefore, is not similar across landscapes and an equal reduction in the percentage of annual timber supply will have different effects on boreal caribou persistence in each FMU (Fig. 4). If the goal is that Pr(leq > 1) be at least equal to 0.6 (see Environment Canada, 2012) under current conservation levels, an overall reduction of annual timber supply from 25 to 68% is required, but the exact magnitude of timber volume reduction depends upon the disturbance scenario and FMU (Table 2; see column “Total” for all scenarios). In the presence of fire, salvage logging, and current conservation levels, no FMU achieves Pr(leq > 1)  0.6, even after the annual timber supply is reduced by as much as the 20e28% that is necessary to achieve a sustainable timber harvest (Fig. 4, green symbols; Table 2, column “Y” for scenarios Logging þ Fire þ Salvage logging). Under such scenarios, additional reductions of 48%, 7% and 7% in annual timber supply are

240

J. Beguin et al. / Journal of Environmental Management 163 (2015) 234e245

Table 2 Percentage reduction in annual timber supply that is needed to obtain a sustainable yield (labelled “Y” for yield), caribou persistence or Pr(leq > 1) ¼ 0.6 (labelled “C” for caribou; see section 2.4), and both sustainable yield and caribou persistence simultaneously (labelled “Total”, with Total ¼ Y þ C), as a function of disturbance scenarios, forest ^te-Nord. management units (FMU), and the current percentage of conservation areas (IUCN protection areas þ caribou protection blocks), as is currently applied in the Co Results were obtained by extrapolation in Fig. 4. FMU

Percentage of conservation areas

Reduction in timber supply (%) Logging þ fire (no salvage logging)

Logging only

1 2 3

9.3 11.3 25.6

Logging þ fire þ salvage logging

Total

Y

C

Total

Y

C

Total

Y

C

61% 42% 25%

10% 22% 25%

51% 20% 0%

62% 64% 56%

32% 64% 56%

30% 0% 0%

68% 28% 35%

20% 21% 28%

48% 7% 7%

necessary to achieve a minimum of Pr(leq > 1) ¼ 0.6 in FMUs 1, 2 and 3, respectively (Table 2, column “C”). Under the same scenario, but without salvage logging, maintenance of a sustainable yield is sufficient to allow persistence of boreal caribou in both FMU 2 and 3, but not in FMU 1 (Fig. 4; Table 2, C ¼ 0). In all cases, FMU 1 requires an additional reduction in annual timber supply of at least 30% for specifically maintaining a minimal Pr(leq > 1) ¼ 0.6, which always leads to an overall reduction in timber supply  60% (Table 2). 3.3. Potential for zonation A decrease of 60% in the level of timber harvesting in FMU 1 (¼ 0.6*1 010 000 ¼ 600 000 m3/year) corresponds roughly to 85% of the sum of timber supplies in both FMU 2 and 3 (686 000 m3/year) (Table 2; Table A.2.). The southernmost FMU 1 had the lowest amount of conserved area, the highest human footprint (road density and logging areas), and stands with the highest productivity. FMU 1 also had the lowest percentage of reduction in annual timber harvest volume due to fire (i.e. high resistance to fires) and the lowest potential for supporting a self-sustaining boreal caribou population. In contrast, the two northern FMUs showed an opposite pattern, with a low resistance to fires and high potential for boreal caribou populations to be self-sustainable. 4. Discussion Using a spatially explicit analysis of stochastic risk (such as forest fires) and resource exploitation (such as forest harvesting), our results showed that there is a way via salvage harvesting of fire killed trees to minimize species at risk decline and maintain economic output. This is because the cumulative effects of disturbance and harvesting become partially compensatory. This is critical because in the case of boreal caribou in the managed boreal forest, fire risk and protected areas directly impact the amount and the spatial distribution of timber harvest. Failure to account explicitly for these impacts in the calculation of timber supply leads to an overestimation of timber harvest volume, which in turn leads to rates of cumulative effects that threaten both the long-term persistence of boreal caribou and the sustainability of the timber supply itself. Fire is the process that had the strongest effects on timber supply, causing reductions from 22% to 52% along a gradient of conservation areas ranging from 0 to 25.6% (see Fig. 3). The magnitude of these reductions reinforces the requirement that timber supply analyses should account explicitly and a priori for fire risk, if sustainability is the goal of forest management (Savage et al., 2010; Raulier et al., 2013). In addition to mean effects, the great variability in reductions among FMUs shows that the capacity of forest landscapes to provide ecosystem goods, such as timber, in the presence of fire varies across spatial scales < 104 km2. Provided that reductions in timber supply proportionally increase the probability of caribou persistence, spatial variation of forest

productive capacity is thus a key driver influencing the delicate balance between economic (e.g., timber supply) and conservation (e.g., the long-term persistence of boreal caribou) objectives. Although we did not investigate the underlying cause(s) of such heterogeneity, Boychuk and Martell (1996) have shown that the maintenance of forest productive capacity is related to the existence of a buffer stock of timber, which is itself related to initial forest age structure and to the level of disturbance. Overall, these results reveal that the effectiveness of current conservation planning for boreal caribou would greatly benefit from explicitly considering disturbance regimes (fire and logging), heterogeneity in forest productive capacity, and forest age class structure. Such integration could bridge an important gap between current conservation and forest management strategies, especially in the context of ecosystem-based forest management. One key principle of ecosystem-based management is to limit the rate of anthropogenic disturbance within the range of ecosystem natural variability (Gauthier et al., 2009). We acknowledge that under ecosystem-based management, our results must be interpreted with caution. First, we set no threshold on the proportion of old-growth forests to be protected within FMUs, as recommended by ecosystem management guidelines (e.g., 30%). Boychuk and Martell (1996) again have demonstrated that maintaining un-harvested stands beyond rotation age has a negative effect upon the level of timber supply. Second, we did not account for the predicted three-fold increase in the annual area that would burn over our study region in the next 100 years (Boulanger et al., 2013). In this context, our current estimates of timber harvest reductions that are caused by fire might be conservative. However, we did not include variation in fire suppression efforts and effectiveness. If fire suppression increases in the near future, it could relieve constraints on timber supplies, but this remains to be demonstrated as current fire suppression cannot stop large fires in years of high fire activity (Le Goff et al., 2005). Finally, we did not incorporate information about connectivity and spatial configurations of caribou habitat in our population viability analysis, as the direct influence on these variables on vital rates of boreal caribou populations remains to be quantified in our study region. Adding spatial constraints in timber harvest schedule would likely result in additional reductions of timber supply. Baskent and Keles (2005) have shown that adding spatial constraints on timber harvest can induce reductions of timber supply up to 40%. Further empirical and simulation studies, therefore, are needed to quantify the effects of these additional processes on future mitigation measures between the conservation of boreal caribou and timber harvesting. Many conservation strategies for species at risk rely upon the amount of undisturbed suitable habitat that is retained within the landscape (e.g., the amount of undisturbed mature conifers for boreal caribou; Environment Canada, 2011, 2012). Although protected areas are a convenient way of increasing the amount of mature forest across landscapes, tracts of mature forest are also found outside protected areas. To attain a suitable threshold of

J. Beguin et al. / Journal of Environmental Management 163 (2015) 234e245

undisturbed forests, conservationists and forest managers, therefore, can act directly on two non-exclusive components: increasing the amount of conserved areas and decreasing the level of timber harvesting outside protected areas. Managers can also act on the spatial connectivity between suitable patches within the forest matrix, an attribute that is known to influence space use patterns of boreal caribou (Courbin et al., 2014), while its direct influence on population dynamic needs to be formally tested and quantified. When IUCN protected areas and static caribou protection blocks are used as the main conservation strategy, our results show that the relationship between the amount of conserved areas and the sustainable level of timber supply is proportional, linear and robust, both to variation in disturbance regimes and to heterogeneity in initial conditions among FMUs. At this time, however, the spatial arrangement of caribou protection blocks in the future is unknown, so we could not answer whether the delineation of dynamic caribou blocks will alter this relationship. Nevertheless, it should be noted that if changing the location of caribou blocks over the simulation horizon will have any beneficial effect on caribou, it would impose additional spatial constraints on timber harvest (Baskent and Keles, 2005). Consequently, under a dynamic scheme of caribou protection blocks over space and time, our estimates of reductions in timber supply appear conservative and additional reductions due to spatial constraints in timber harvest should be anticipated. We found a high degree of variation among FMUs in terms of conservation potential, resistance to fire, forest productivity and the human footprint. Salvage logging allowed limited reductions in timber harvests (within a range of ~10%), while increasing the sustainable yield of timber harvests by more than 20% in the two northern FMUs. In addition, salvage logging decreased the overall cumulative amount of disturbed areas, which had a beneficial effect on the probability of caribou persistence compared to a scenario without salvage logging. Nevertheless, the benefit of salvage logging was not spatially homogeneous, as no better trade-offs were found in the FMU that contained the highest initial rate of anthropogenic disturbances (e.g., road density, cutovers). This heterogeneity in the potential of conservation and timber supply is likely common to many other ecosystems and suggests that zonation might be a promising option for identifying trade-offs with minimal losses between conservation and economic objectives. It is possible that concentrating forest management practices (including intensive silviculture) on a small part of a boreal landscape may counterbalance losses that are associated with increasing protected areas at larger scales (Tittler et al., 2012). Our results tend to support this hypothesis, but a robust evaluation of any zonation strategy will require extending the present analysis to the entire population of forest management units that are located
bec. within the range of the recovery plan for boreal caribou in Que When the delineation of future caribou protection blocks in our system becomes available, it would also be relevant to assess how dynamic reserves influence population growth rates, together with those of timber supplies. The interdependence among fires, logging, and protected areas makes the integration of fire risk into timber supply analyses a critical factor for identifying mitigation methods between caribou conservation and wood supply. Surprisingly, inclusion of fire risk is often overlooked in assessments of habitat for endangered species, which can result to overly optimistic projections. Moreover, except for the Canadian province of Ontario (Savage et al., 2010), past and current timber supply analyses in the boreal forest of Canada have not tended to account for the effects of fire risk during the process of timber supply calculations (Nelson, 2003; Gauthier et al., 2009;

241

Conrod, 2010). As highlighted by our results, this optimistic situation leads to an overestimation of timber supplies, which increases the tension between the long-term persistence of boreal caribou and forest harvesting. Consequently, if forest managers and conservationists simultaneously aim to improve landscape suitability for boreal caribou and the sustainability of timber supply, a downward review of planned supplies that explicitly takes into account the interactions among fires, timber harvesting and protected areas should be the first priority. The practical manner in which this reduction is set up on the ground will be a key element of any efficient conservation strategy. 5. Conclusion Our study demonstrates the need (1) to incorporate a priori fire risk in both timber supply analyses and caribou persistence estimates and (2) to account for landscape level variability in forest productive capacity, fire risk, and conservation potentials into the design of conservation and forest management plans. We found that salvage logging, an industrial activity confined to disturbed areas, can compensate for negative impacts caused by fire on timber supply (Peter and Nelson, 2005) and can reduce cumulative effects of disturbances on caribou by overlapping fire and logging with on another (Schneider et al., 2003). This finding suggests that the use of salvage logging might offer perspectives for identifying trade-offs with minimal losses between timber harvesting and the persistence of boreal caribou. Salvage logging, however, also impacts negatively local post-fire communities and biodiversity (Nappi et al., 2004; Lindenmayer and Noss, 2006; Schmiegelow et al., 2006). A more extensive use of salvage logging could therefore limit the availability of suitable post fire habitats and threaten a range of species and ecological communities if no mitigation measures at local scales are developed (Swanson et al., 2010; Venier et al., 2014). Balancing the benefit of salvage logging at a strategic level (regional landscapes) with the negative impacts on ecological communities at local scales (burned sites) is therefore an emerging challenge that, we feel, has not been fully appreciated. We are aware of no other study that has evaluated the usefulness of salvage logging at a strategic leveldand within a conservation context at large spatial scalesdfor a wide-ranging species at risk. Our results indicate that salvage logging may offer some relief from the antagonism that often occurs between conservation efforts and forest harvesting. Research on this topic needs to be conducted at multiple scales, including over large spatial extents where tradeoffs between conservation and timber harvest objectives are more likely to occur. Acknowledgements Funding for this research was provided by a scholarship from NSERC (Natural Sciences and Engineering Research Council of Canada) to J.B. We sincerely thank S. Heppel for giving us access to re des the caribou protection blocks and the MRNFQ (Ministe
bec) for the forest digital ressources naturelles et de la faune du Que inventory maps. We are grateful to A. Fall and M. Bouchard for sharing their codes in SELES, G. Cyr for his help with yield curves, and J. Duval for sharing information about salvage logging. We also
bec (Chief forester thank the Bureau du Forestier en Chef du Que
bec) for giving us access to yield tables and curves. office of Que Finally, we thank I.D. Thompson, M. Bouchard and P. James for their comments on an early version of this manuscript and W.F.J. Parsons for English revisions.

242

J. Beguin et al. / Journal of Environmental Management 163 (2015) 234e245

Appendix 1

Table A.1 Description of initial characteristics and land allocation associated with the three forest management units that were used in this study. Data come from forest inventory maps (MRNFQ 2003), updated to 2010. Class

Variables

FMU 1 (93-51) 2

(km ) Area of FMU Forest Spruce-fir stands Deciduous stands Jack pine stands Lichen woodland sum of forested areas Non-Forest Water Wetlands Lichen heaths Heaths without lichen Other sum of non-forested areas Disturbances Fire < 10-years-old Fire < 20-years-old Fire < 40-years-old Fire < 60-years-old Logging < 10-years-old Logging < 20-years-old Logging < 40-years-old Logging < 60-years-old Road density (km/km2) Land allocation IUCN protected areas Caribou protection blocks Non-protected areas Operable productive areas Level of timber supply (m3/year)

15 460 11 341 920 305 19 12 585 2106 234 191 281 63 2875 7 314 505 511 791 2103 3358 3740 0.92 218 1220 14 022 11 147 1 010 000

FMU 2 (93-52) (%) 100 73.4 5.9 2.0 0.1 81.4 13.6 1.5 1.2 1.8 0.4 18.6 0.0 2.0 3.3 3.3 5.1 13.6 21.7 24.2 e 1.4 7.9 90.7 72.1

2

(km )

FMU 3 (94-52) (%)

12 909 7524 369 584 163 8640 2510 597 671 454 37 4269 245 722 2487 2639 263 509 509 514 0.14 424 1036 11 450 7180 386 000

100 58.3 2.9 4.5 1.3 66.9 19.4 4.6 5.2 3.5 0.3 33.1 1.9 5.6 19.3 20.4 2.0 3.9 3.9 4.0 e 3.3 8.0 88.7 55.6

2

(km ) 10 474 7395 337 49 502 8283 1129 225 578 236 23 2191 20 20 117 151 230 741 751 757 0.27 482 2202 7791 5599 300 000

Total (%) 100 70.6 3.2 0.5 4.8 79.1 10.8 2.2 5.5 2.3 0.2 20.9 0.2 0.2 1.1 1.4 2.2 7.1 7.2 7.2 e 4.6 21.0 74.4 53.5

(km2) 38 844 26 261 1625 938 684 29 507 5745 1057 1439 971 124 9336 272 1057 3108 3301 1285 3353 4619 5011 0.48 1124 4458 33 263 23 925 1 696 000

(%) 100 67.6 4.2 2.4 1.8 76.0 14.8 2.7 3.7 2.5 0.3 24.0 0.7 2.7 8.0 8.5 3.3 8.6 11.9 12.9 e 2.9 11.5 85.6 61.6

Reference re des ressources naturelles et de la Faune du Que
bec e MRNFQ (2003) Normes de cartographie e
coforestie re du troisie me inventaire e
coforestier. Fore ^t Que
bec. Ministe Direction des inventaires forestiers. ISBN 2-551-xxxxx-x. 109 p. [in French].

Appendix 2

Table A.2 Parameters of the different sub-models used in our simulation experiment. Sub-model

Variables

Probability distribution

YIELD FIRE

Yield tables Mean number of fire per year Fire size

fixed Poisson(lfire) Reverse Weibull(shape, scale)

LOGGING

Size of cut-blocks (ha)

Uniform(min, max) 3

CARIBOU

a

Planned timber harvest volume (m /year) in FMU 1 Planned timber harvest volume (m3/year) in FMU 2 Planned timber harvest volume (m3/year) in FMU 3 Minimum stand age for logging (yrs) Maximum distance to roads (km) Minimum volume for salvage logging (m3/ha) b0 (see eqn 1 in Methods)

fixed fixed fixed fixed fixed fixed Normal(mean(0), sd(0))

b1 (see eqn 1 in Methods)

Normal(mean(1), sd(1))

Standard deviation of recruitment rate Survival rate

fixed Normal(mean(2), sd(2))

Parameter(s)

References

e

Pothier and Savard (1998) Modified from Bouchard and Pothier (2008b) Bouchard and Pothier (2008b) Bouchard and Pothier (2008b) RNI (2002) RNI (2002) By iteration (see Methods) By iteration (see Methods) By iteration (see Methods) Pothier and Savard (1998) e Resolute Forest Products (Pers. Comm.) Env. Can (2011) Env. Can (2011) Env. Can (2011) Env. Can (2011) Courtois et al. (2007, 2008) Env. Can (2011) Courtois et al. (2007, 2008)

lfire ¼ 0.86

shape ¼ 0.7357 scale ¼ 0.0386 min ¼ 50 max ¼ 150 1 010 000a 386 000a 300 000a 90 2 50 mean(0) ¼ 44.265 sd(0) ¼ 2.942 mean(1) ¼ 0.429 sd(1) ¼ 0.061 0.049 mean(2) ¼ 0.85 sd(2) ¼ 0.0525

Values of planned harvest volume are for commercial coniferous species: balsam fir, spruce, jack pine, and larch.

J. Beguin et al. / Journal of Environmental Management 163 (2015) 234e245

243

0.4 0.3

simulated

0.1

0.2

observed

0

Proportional abundance of size classes

Appendix 3. Comparisons between observed and simulated frequencies of fire size classes and examples of simulated fire patterns.

10-30

30-90

90-270

270-810 810-2100

Fire Size (km2) Fig. A.1. Comparison between observed and simulated frequencies of fire size classes. See Bouchard and Pothier (2008b) for more information on observed frequency of fire size ^ te-Nord region. classes for the Co

Fig. A.2. Three different simulated fire patterns obtained with our fire model after 150 years of simulation. Each figure represents a different simulation run, where oldest (nearing 150-years-old) simulated fires are coloured in green and most recent simulated fires are coloured in red.

References: Bouchard, M., and D. Pothier. 2008b. Simulations of the effects of changes in mean fire return intervals on balsam fir abundance, and implications for spruce budworm outbreaks. Ecological Modelling 218: 207e218. Appendix 4. Results of the model selection procedures and parameter estimates.

Table A.3 Procedure of comparison among models varying in the specification of random effects. Data are presented in Fig. 3. Variable names: PctVolumeReduction ¼ reduction in annual timber harvest volume (%); PctProtectedAreas ¼ proportion of conservation areas within FMU (%); DisturbanceScenario ¼ scenario of disturbance (logging only; Fire þ logging; Fire þ logging þ salvage logging); FMU ¼ forest management unit. Restricted maximum likelihood method (REML ¼ TRUE) was used to calculate AICc. The model that is highlighted in boldface type has the greatest support among all candidate models. ID

Model in R (lmer function in lme4 package, Bates et al., 2013)

Random effects

K

AICc

DAICc

1

PctVolumeReduction ~ PctProtectedAreas þ (1jDisturbanceScenario) þ (0 þ PctProtectedAreas j DisturbanceScenario) þ (1 j FMU:DisturbanceScenario) þ (0 þ PctProtectedAreas j FMU:DisturbanceScenario) PctVolumeReduction ~ PctProtectedAreas þ (1 j FMU:DisturbanceScenario) þ (0 þ PctProtectedAreas j FMU:DisturbanceScenario) PctVolumeReduction ~ PctProtectedAreas þ (1jDisturbanceScenario) þ (0 þ PctProtectedAreas j DisturbanceScenario)

uncorrelated slope and intercept random effects on both FMU and disturbance scenarios

6

228.4

2.6

uncorrelated slope and intercept random effects on FMU

4

229.8

4.0

uncorrelated slope and intercept random effects on disturbance scenarios

4

264.1

38.3

2

3

(continued on next page)

244

J. Beguin et al. / Journal of Environmental Management 163 (2015) 234e245

Table A.3 (continued ) ID

Model in R (lmer function in lme4 package, Bates et al., 2013)

Random effects

K

AICc

DAICc

4

PctVolumeReduction ~ PctProtectedAreas þ (0 þ PctProtectedAreas j DisturbanceScenario) þ (0 þ PctProtectedAreas j FMU:DisturbanceScenario) PctVolumeReduction ~ PctProtectedAreas þ (1 j DisturbanceScenario) þ (1 j FMU:DisturbanceScenario) PctVolumeReduction ~ PctProtectedAreas þ (0 þ PctProtectedAreas j FMU:DisturbanceScenario) PctVolumeReduction ~ PctProtectedAreas þ (1 j FMU:DisturbanceScenario) PctVolumeReduction ~ PctProtectedAreas þ (0 þ PctProtectedAreas j DisturbanceScenario) PctVolumeReduction ~ PctProtectedAreas þ (1 j DisturbanceScenario) PctVolumeReduction ~ PctProtectedAreas þ (1 j fuzzy)

slope random effects on both FMU and disturbance scenarios

4

303.8

78.0

intercept random effects on both FMU and disturbance scenarios slope random effect on FMU

4

225.8

0.0

3

304.4

78.6

intercept random effect on FMU

3

230.2

4.4

slope random effect on disturbance scenarios intercept random effect on disturbance scenarios null intercept random effect

3

302.0

76.2

3

261.6

35.8

3

309.8

84.0

5 6 7 8 9 10

Table A.4 Procedure of comparison among models varying in the specification of random effects. Data are presented in Fig. 4. Variable names: ProbLambda ¼ Pr(leq>1) (see Fig. 4 for more details); PctVolumeReduction ¼ reduction in annual timber harvest volume (%); DisturbanceScenario ¼ scenario of disturbance (logging only; Fire þ logging; Fire þ logging þ salvage logging); FMU ¼ forest management unit. Restricted maximum likelihood method (REML ¼ TRUE) was used to calculate AICc. The model highlighted in bold has the best support among all candidate models. ID

Model in R (lmer function in lme4 package, Bates et al., 2013)

Random effects

K

AICc

DAICc

1

ProbLambda ~ PctVolumeReduction þ (1jDisturbanceScenario) þ (0 þ PctVolumeReduction j DisturbanceScenario) þ (1 j FMU:DisturbanceScenario) þ (0 þ PctVolumeReduction j FMU:DisturbanceScenario) ProbLambda ~ PctVolumeReduction þ (1 j FMU:DisturbanceScenario) þ (0 þ PctVolumeReduction j FMU:DisturbanceScenario) ProbLambda ~ PctVolumeReduction þ (1jDisturbanceScenario) þ (0 þ PctVolumeReduction j DisturbanceScenario) ProbLambda ~ PctVolumeReduction þ (0 þ PctVolumeReduction j DisturbanceScenario) þ (0 þ PctVolumeReduction j FMU:DisturbanceScenario) ProbLambda ~ PctVolumeReduction þ (1 j DisturbanceScenario) þ (1 j FMU:DisturbanceScenario) ProbLambda ~ PctVolumeReduction þ (0 þ PctVolumeReduction j FMU:DisturbanceScenario) ProbLambda ~ PctVolumeReduction þ (1 j FMU:DisturbanceScenario) ProbLambda ~ PctVolumeReduction þ (0 þ PctVolumeReduction j DisturbanceScenario) ProbLambda ~ PctVolumeReduction þ (1 j DisturbanceScenario) PctVolumeReduction ~ PctVolumeReduction þ (1 j fuzzy)

uncorrelated slope and intercept random effects on both FMU and disturbance scenarios

6

220.5

5.2

uncorrelated slope and intercept random effects on FMU

4

215.3

0.0

uncorrelated slope and intercept random effects on disturbance scenarios slope random effects on both FMU and disturbance scenarios

4

282.3

67.0

4

249.5

34.2

2

3 4

5 6 7 8 9 10

intercept random effects on both FMU and disturbance scenarios slope random effect on FMU

4

227.8

12.5

3

247.0

31.7

intercept random effect on FMU slope random effect on disturbance scenarios

3 3

226.4 292.9

11.1 77.6

intercept random effect on disturbance scenarios null intercept random effect

3 3

280.0 288.7

64.7 73.4

Reference: Bates, D., M. Maechler, B. Bolker, and S. Walker. 2013. lme4: Linear mixed-effects models using Eigen and S4. R package version 1.0e6. http://CRAN.R-project.org/ package¼lme4.

Table A.5 Parameter estimates (±1 SE) of linear mixed models used to analyse data points shown in Figs. 3 and 4. Variable names: “Disturbances” corresponds to the three disturbance scenarios: logging only, logging þ fire (without salvage logging), logging þ fire (with salvage logging); “FMU:Disturbances” corresponds to the three forest management units that are nested in each disturbance scenario. See Figs. 3 and 4 for more details on fixed variables and Tables A.3 and A.4 for details on the model selection procedure. Type of Variable

Variables

Type of random effect

Parameter estimates Model for data on Fig. 3

Model for data on Fig. 4

Random

Disturbances FMU:Disturbances (¼ nested) FMU:Disturbances (¼ nested) Residuals Intercept Amount of conservation areas (%) Reduction in annual timber supply (%)

intercept intercept slope e e e e

207.1 62.2 e 8.6 15.8 ± 8.7 0.95 ± 0.07 e

e 187.0 0.07 3.2 13.3 ± 4.8 e 1.11 ± 0.11

Fixed

References Baskent, E.Z., Keles, S., 2005. Spatial forest planning: a review. Ecol. Model. 188, 145e173. Beguin, J., McIntire, E.J.B., Fortin, D., Cumming, S.G., Raulier, F., Racine, P., Dussault, C., 2013. Explaining geographic gradients in winter selection of

landscapes by boreal caribou with implications under global changes in eastern Canada. PLoS One 8, e78510. Bettinger, P., Boston, K., Siry, J.P., Grebner, D.L., 2009. Forest Management and Planning. Academic Press, Burlington, MA. Bouchard, M., Pothier, D., Gauthier, S., 2008. Fire return intervals and tree species succession in the North Shore region of eastern Quebec. Can. J. For. Res. 38, 1621e1633.

J. Beguin et al. / Journal of Environmental Management 163 (2015) 234e245 Bouchard, M., Pothier, D., 2008. Simulations of the effects of changes in mean fire return intervals on balsam fir abundance, and implications for spruce budworm outbreaks. Ecol. Model. 218, 207e218. Boulanger, Y., Gauthier, S., Gray, D.R., Le Goff, H., Lefort, P., Morissette, J., 2013. Fire regime zonation under current and future climate over eastern Canada. Ecol. Appl. 23, 904e923. Boychuk, D., Martell, D.L., 1996. A multistage stochastic programming model for sustainable forest-level timber supply under risk of fire. For. Sci. 42, 10e26. Conrod, M.D., 2010. A Simulation/Optimization System for Modelling Timber and Old Forest under Stochastic Fire Disturbance. M.Sc. thesis. University of Alberta, Edmonton, AB. Courbin, N., Fortin, D., Dussault, C., Courtois, R., 2014. Logging-induced changes in habitat network connectivity shape behavioral interactions in the wolf-cariboumoose system. Ecol. Monogr. 84, 265e285. Courtois, R., Gingras, A., Fortin, D., Sebbane, A., Rochette, B., Breton, L., 2008. Demographic and behavioural response of woodland caribou to forest harvesting. Can. J. For. Res. 38, 2837e2849. Courtois, R., Ouellet, J.P., Breton, L., Gingras, A., Dussault, C., 2007. Effects of forest disturbance on density, space use, and mortality of woodland caribou. Ecoscience 14, 491e498. Courtois, R., Ouellet, J.P., Dussault, C., Gingras, A., 2004. Forest management
bec. For. Chron. 80, 598e607. guidelines for forest-dwelling caribou in Que Didion, M., Fortin, M.-J., Fall, A., 2007. Forest age structure as indicator of boreal forest sustainability under alternative management and fire regimes: a landscape level sensitivity analysis. Ecol. Model. 200, 45e58. Di Marco, M., Boitani, L., Mallon, D., Hoffmann, M., Iacucci, A., Meijaard, E., Visconti, P., Schipper, J., Rondinini, C., 2014. A retrospective evaluation of the global decline of carnivores and ungulates. Conserv. Biol. 28, 1109e1118. Dyer, S.J., O'Neill, J.P., Wasel, S.M., Boutin, S., 2001. Avoidance of industrial development by woodland caribou. J. Wildl. Manag. 65, 531e542. Environment Canada, 2011. Scientific Assessment to Inform the Identification of Critical Habitat for Woodland Caribou (Rangifer tarandus Caribou), Boreal Population, in Canada: 2011 Update. Ottawa, Ontario, Canada, p. 102. plus appendices. Environment Canada, 2012. Recovery Strategy for the Woodland Caribou (Rangifer tarandus Caribou), Boreal Population, in Canada. Species at Risk Act Recovery Strategy Series. Environment Canada, Ottawa xi±138 pp.


tablissement du Caribou Forestier du Que
bec (ERCFQ), Equipe de Re 2008. Plan de
tablissement du caribou forestier (Rangifer tarandus) au Que
bec ‒ 2005-2012. re re des Ressources naturelles et de la Faune, Faune Que
bec, Direction de Ministe l’expertise sur la faune et des habitats, Canada [In French.].


tablissement du Caribou Forestier du Que
bec (ERCFQ), Equipe de Re 2013. Plan de
tablissement du caribou forestier (Rangifer tarandus caribou) au Que
bec ‒ re re du De
veloppement durable, de l’Environnement, de la 2013-2023. Ministe
bec, Faune Que
bec [In French.]. Faune, Faune et des Parcs du Que Fall, A., Fall, J., 2001. A domain-specific language for models of landscape dynamics. Ecol. Model. 141, 1e18. Fall, A., Fortin, M.-J., Kneeshaw, D.D., Yamasaki, S.H., Messier, C., Bouthillier, L., Smyth, C., 2004. Consequences of various landscape-scale ecosystem management strategies and fire cycles on age-class structure and harvest in boreal forests. Can. J. For. Res. 34, 310e322. ^ te
, S.D., Gunn, A., 2011. Conservation of Festa-Bianchet, M., Ray, J.C., Boutin, S., Co caribou (Rangifer tarandus) in Canada: an uncertain future. Can. J. Zool. 89, 419e434.
, L., Kneeshaw, D., Morin, H., Gauthier, S., Vaillancourt, M.-A., Leduc, A., De Grandpre Drapeau, P., Bergeron, Y., 2009. Ecosystem Management in the Boreal Forest.
du Que
bec, Que
bec, QC, Canada. Presses de l'Universite Hatter, I.W., Bergerud, W.A., 1991. Moose recruitment, adult mortality and rate of change. Alces 27, 65e73. Hauer, G., Cumming, S.G., Schmiegelow, F., Adamowicz, W., Weber, M., Jagodzinski, R., 2010. Tradeoffs between forestry resource and conservation values under alternate policy regimes: a spatial analysis of the western Canadian boreal plains. Ecol. Model. 221, 2590e2603. €nkko €nen, M., Desrochers, A., 2001. Long-term effects of forestry on Imbeau, L., Mo birds of the eastern canadian boreal forests: a comparison with Fennoscandia. Conserv. Biol. 15, 1151e1162. Johnson, E.A., 1996. Fire and Vegetation Dynamics: Studies from the North American Boreal Forest. Cambridge University Press, Cambridge, UK. Le Goff, H., Leduc, A., Bergeron, Y., Flanningan, M., 2005. The adaptative capacity of forest management to changing fire regimes in the boreal forest of Quebec. For. Chron. 81, 582e592. Lindenmayer, D.B., Noss, R.F., 2006. Salvage logging, ecosystem processes, and biodiversity conservation. Conserv. Biol. 20, 949e958. McMullin, R.T., Thompson, I.D., Newmaster, S.G., 2013. Lichen conservation in

245

heavily managed boreal forests. Conserv. Biol. 27, 1020e1030. Millennium Ecosystem Assessment, 2005. Ecosystems and Human Well-being. Island Press, Washington, DC. Morgan, D.G., Abdallah, S.B., Lasserre, P., 2007. A real options approach to forestmanagement decision making to protect caribou under the threat of extinction. Ecol. Soc. 13, 27 [online]. Naidoo, R., Balmford, A., Ferraro, P.J., Polasky, S., Ricketts, T.H., Rouget, M., 2006. Integrating economic costs into conservation planning. Trends Ecol. Evol. 21, 681e687. Nappi, A., Drapeau, P., Savard, J.-P., 2004. Salvage logging after wildfire in the boreal forest: is it becoming a hot issue for wildlife? For. Chron. 80, 67e74.
ry, S., Bujold, F., Chabot, M., Dumont, M.-C., Duval, J., Drapeau, P., Nappi, A., De
colte dans les fore ^ts Gauthier, S., Brais, S., Peltier, J., Bergeron, I., 2011. La re
eseEnjeux et orientations pour un ame
nagement e
cosyste
mique. Direcbrûle ^ts, Ministe re des Ressources tion de l’environnement et de la protection des fore
bec, p. 51. naturelles et de la Faune, Que Nelson, J., 2003. Forest-level models and challenges for their successful application. Can. J. For. Res. 33, 422e429. Paillet, Y., et al., 2010. Biodiversity differences between managed and unmanaged forests: meta-analysis of species richness in Europe. Conserv. Biol. 24, 101e112. Peter, B., Nelson, J., 2005. Estimating harvest schedules and profitability under the risk of fire disturbance. Can. J. For. Res. 35, 1378e1388. Polasky, S., Nelson, E., Lonsdorf, E., Fackler, P., Starfield, A., 2005. Conserving species in a working landscape: land use with biological and economic objectives. Ecol. Appl. 15, 1387e1401. Pothier, D., Savard, F., 1998. Actualisation des tables de production pour les princes forestie res du Que
bec. Ministe re des Ressources naturelles, cipales espe Direction des inventaires forestiers. RN98e3054. [in French]. Raulier, F., Le Goff, H., Gauthier, S., Rapanoela, R., Bergeron, Y., 2013. Introducing two indicators for fire risk consideration in the management of boreal forests. Ecol. Indic. 24, 451e461. Rempel, R.S., Baker, J., Elkie, P.C., Gluck, M.J., Jackson, J., Kushneriuk, R.S., Moore, T., Perera, A.H., 2007. Forest policy scenario analysis: sensitivity of songbird community to changes in forest cover amount and configuration. Avian Conservation Ecol. 2, 5.
glement sur les normes d'intervention dans les fore ^ts du domaine de l'Etat, RNI - Re
bec. Available online: http://www2. 2002. Gouvernement du Que publicationsduquebec.gouv.qc.ca/dynamicSearch/telecharge.php? type¼3&file¼/A_18_1/A18_1R7.HTM. Savage, D.W., Martell, D.L., Wotton, B.M., 2010. Evaluation of two risk mitigation strategies for dealing with fire-related uncertainty in timber supply modelling. Can. J. For. Res. 40, 1136e1154. Schaefer, J.A., 2003. Long-term range recession and the persistence of caribou in the taiga. Conserv. Biol. 17, 1435e1439. Schmiegelow, F.K., Stepnisky, D.P., Stambaugh, C.A., Koivula, M., 2006. Reconciling salvage logging of boreal forests with a natural-disturbance management model. Conserv. Biol. 20, 971e983. Schneider, R.R., Stelfox, J.B., Boutin, S., Wasel, S., 2003. Managing the cumulative impacts of land uses in the western Canadian sedimentary basin: a modeling approach. Conserv. Ecol. 7, 8 [online] URL: http://www.consecol.org/vol7/iss1/ art8. Schneider, R.R., Hauer, G., Dawe, K., Adamowicz, W.L., Boutin, S., 2012. Selection of reserves for woodland caribou using an optimization approach. PLoS One 7, e31672. Swanson, M.E., Franklin, J.F., Beschta, R.L., Crisafulli, C.M., DellaSala, D.A., Hutto, R.L., Lindenmayer, D.B., Swanson, F.J., 2010. The forgotten stage of forest succession: early-successional ecosystems on forest sites. Front. Ecol. Environ. 9, 117e125. Tittler, R., Messier, C., Fall, A., 2012. Concentrating anthropogenic disturbance to balance ecological and economic values: applications to forest management. Ecol. Appl. 22, 1268e1277. Venier, L.A., Thompson, I.D., Fleming, R., Malcolm, J., Aubin, I., Trofymow, J.A., Langor, D., Sturrock, R., Patry, C., Outerbridge, R.O., Holmes, S.B., Haeussler, S.,
, L., Chen, H.Y.H., Bayne, E., Arsenault, A., Brandt, J.P., 2014. Effects of De Grandpre natural resource development on the terrestrial biodiversity of Canadian boreal forests 1. Environ. Rev. 22, 457e490. Verkerk, P., Zanchi, G., Lindner, M., 2014. Trade-offs between forest protection and wood supply in Europe. Environ. Manag. 53, 1085e1094. Vors, L.S., Schaefer, J.A., Pond, B.A., Rodgers, A.R., Patterson, B.R., 2007. Woodland caribou extirpation and anthropogenic landscape disturbance in Ontario. J. Wildl. Manag. 71, 1249e1256. Wasser, S.K., Keim, J.L., Taper, M.L., Lele, S.R., 2011. The influences of wolf predation, habitat loss, and human activity on caribou and moose in the Alberta oil sands. Front. Ecol. Environ. 9, 546e551.