Incorporating Insect and Wind Disturbances in a Natural Disturbance-Based Management Framework for the Boreal Forest

Natural disturbances are fundamental to forest ecosystem dynamics and have been used for two decades to improve forest management, notably in the boreal forest. Initially based on fire regimes, there is now a need to extend the concept to include other types of disturbances as they can greatly contribute to forest dynamics in some regions of the boreal zone. Here we review the main descriptors—that is, the severity, specificity, spatial and temporal descriptors and legacies, of windthrow and spruce bud worm outbreak disturbance regimes in boreal forests—in order to facilitate incorporating them into a natural disturbance-based forest management framework. We also describe the biological legacies that are generated by these disturbances. Temporal and spatial descriptors characterising both disturbance types are generally variable in time and space. This makes them difficult to reproduce in an ecosystem management framework. However, severity and specificity descriptors may provide a template upon which policies for maintaining post harvesting and salvage logging biological legacies can be based. In a context in which management mainly targets mature and old-growth stages, integrating insect and wind disturbances in a management framework is an important goal, as these disturbances contribute to creating heterogeneity in mature and old-growth forest characteristics.


Introduction
Natural disturbances are fundamental components of forest ecosystems [1].They range from frequent, low-severity, small-scale (e.g., gap forming) disturbances to infrequent, large-scale, high-severity events that can markedly alter forest structure and function [2][3][4].An increasing body of evidence indicates that plant and animal species have evolved with and are adapted to ecosystem-specific natural disturbance regimes [5,6].Forest ecosystems are therefore expected to be resilient to such disturbances, that is, to have the capacity to recover functionally, even if local characteristics such as structure and species composition can be somewhat different from pre-disturbance conditions [7].Thus, it has been suggested that forest management that creates conditions similar to those following natural disturbances will have a smaller negative impact on biodiversity and ecological processes.
Over the last two decades, a growing understanding of the integral role of natural disturbances in forest ecosystem dynamics has motivated the development of forest management approaches based on natural disturbances [8][9][10].The emulation of natural disturbances has now become a dominant paradigm in forest management in parts of the world [9,11,12].Although it is recognized that forest management does not recreate exactly the same conditions and ecosystem processes as natural disturbances [13], it is generally assumed that human-made disturbances that retain key legacy attributes will have a smaller negative effect on biodiversity and forest ecosystem processes [8,11].
In boreal forests, the natural disturbance-based management approach has been largely inspired by the impact of fire on forested landscapes [10,14,15].This is understandable given the dominant role of wildfire in driving forest dynamics in a wide array of ecosystems all around the globe and more specifically in the boreal zone [16].However, non-fire disturbances such as insect outbreaks and windthrow are also important components of boreal forest ecosystem dynamics [17][18][19][20][21][22].The role of such disturbances on forest dynamics is especially important in regions where the fire recurrence is low [23].
In order to integrate fire, windthrow and insect outbreaks into a natural disturbance-based forest management frameworks, a sound understanding of the characteristics of these disturbance regimes is required [24].Spatial and temporal components (e.g., patch size and disturbance frequency) of disturbance regimes are major determinants of forest structure and composition at stand-and landscape-scales [25].The spatial and temporal aspects of disturbance regimes are also fundamental characteristics of forest management planning and silvicultural treatment design [26].When combined with an assessment of disturbance severity, such disturbance descriptors can be used to assess the range of variability associated with natural or human-made disturbances [14,27].Although these three types of descriptors (extent, frequency and severity) were found to be useful to describe disturbance regimes, especially in the case of wildfires [14], when other types of disturbances (e.g., windthrow and insect outbreaks) are of interest, a different set of descriptors may need to be considered as well.For instance, in the case of biotic disturbances such as insect outbreaks, variation in stand vulnerability due to the host specificity of the insect and associational interactions among tree species can influence the impact of an outbreak at stand-and landscape-scales [17,[28][29][30].
An adequate knowledge of biological legacies produced by natural disturbances is also essential in order to guide forest management planning and silvicultural interventions, particularly in terms of retention policies [31].Structural legacies include standing dead trees (snags), downed logs and other woody debris, tip up mounds associated with uprooted trees, remnant live trees and patches of undisturbed understory [13].Their abundance, characteristics (e.g., decay stage for snags or woody debris) and distribution (e.g., spatial pattern) play an important role in the recovery of forest ecosystems after a disturbance [7,32,33] and can represent important habitats for a range of plant and animal species.
Here we review the main characteristics of windthrow and insect outbreak disturbance regimes in boreal forests as an important first step to incorporating these disturbances into a natural disturbance-based forest management framework.Because fire regimes were extensively described elsewhere [10,11,14,15,34,35], they will not be examined specifically in this review except as a reference to emphasise the distinctiveness of non-fire disturbances.More specifically, we review the temporal and spatial components of non-fire disturbance regimes, as well as information regarding the effect of variation in severity and vulnerability.We then identify the main structural legacies found after these disturbances.We conclude by highlighting the key issues that need to be considered if windthrow and insect outbreak disturbances are to be incorporated in a natural disturbance-based forest management framework.Although our intent is for this review to be relevant to a wide range of boreal forest ecosystems, we will focus on northeastern North American boreal forests and we will use spruce budworm (SBW; Choristoneura fumiferana (Clemens)) outbreaks as an example of insect disturbance.

Disturbance Regime Descriptors
The temporal and spatial characteristics of a disturbance regime are probably the most frequently studied descriptors.Together, they are major determinants of forest structure and composition at stand-and landscape-scales [14,25,26].The temporal characteristics of disturbance events and regimes include duration of the event, rotation period or cycle (mean time to disturb an area equivalent to the study area), return interval (time between two disturbance episodes at the same place), frequency (mean proportion of the area affected annually) and periodicity (recurrence properties) [8,24,25,36].Spatial characteristics of disturbances include total area affected by a disturbance, mean or median size of affected patches, size frequency distribution, patch shape characteristics and connectivity among patches [8,25,37].
A third important descriptor is disturbance severity and how it varies spatially, temporally and as a function of multiple biotic and abiotic factors [38].The severity of a disturbance refers to the magnitude of the impact of a disturbance event on the organisms (e.g., percentage of tree mortality) and other abiotic components (e.g., soil disturbance) of an ecosystem [8,24,36].The severity is generally a function of disturbance intensity, which is the physical energy associated with a disturbance (e.g., wind speed during a storm or defoliation during an outbreak) [25].While disturbance intensity is an intrinsic characteristic of a disturbance event, disturbance severity is a measure of its ecological effect [25].Disturbance severity is a major determinant of post-disturbance recovery patterns and forest species composition [35].
While assessment of disturbance severity is often focused on tree mortality, Roberts [39] developed a three-axis model of disturbance severity in which percentage of tree mortality is considered in conjunction with the percentage of understory vegetation affected by a disturbance and the percentage of forest floor/soil removed or disrupted.These three components of disturbance severity will be affected differently depending on disturbance type, which will influence the post-disturbance response of a forest ecosystem [39,40].
Although these three descriptors can capture a large amount of information regarding disturbance regime impacts on forest ecosystems, other characteristics may prove useful for describing specific types of disturbances.For example, specificity of the disturbance is especially important to consider in the case of SBW outbreaks because of important differences in host susceptibility and vulnerability to the insect [41].Specificity relates to the selective nature of a disturbance towards particular species, forest type, seral stage or stand characteristics [8].Specificity to windthrow also varies amongst species and size classes of trees [42][43][44].Variation in severity (tree mortality) related to disturbance specificity is important to understand because it can play a major role in modulating the impact of the disturbance on harvest volume and on the implementation of salvage logging strategies (e.g., [45]).See Table 1 for a list of these descriptors for both windthrow and SBW.

Temporal Descriptors
Tree mortality caused by severe windthrow tends to occur over a relatively short time, although trees at the edge of windthrow openings can continue to fall if the storm prolongs [46,47] and during the years following the windthrow episode (Table 1).This phenomenon, sometimes referred to as the "domino effect" [64,65], can be explained by the fact that wind can penetrate into an open stand more easily and also because falling trees can break or uproot neighboring trees [22,64].Moreover, there may be a lag of several years between the occurrence of windthrow and death of the damaged trees [66].
Disturbance by wind is considered particularly difficult to characterize [1].The two phenomena mentioned earlier (i.e., spreading of a windthrow patch and lag in tree mortality) contribute to the difficulty in establishing the time at which a windthrow event occurred.Another reason why windthrow recurrence is difficult to establish is that many windthrow gaps are small and difficult to assess [67].Failure to consider the smaller windthrow patches in a windthrow survey could introduce bias in windthrow recurrence estimation [64].
Despite these limitations, estimates of windthrow cycle have been reported for both stand-replacing and partial windthrow in boreal forests.For stand-replacing windthrow, return intervals of 3570 to over 8000 years have been reported in northeastern boreal forest [22,[48][49][50] (Table 1).Severe windthrows are therefore relatively infrequent and their return intervals are much longer than the tree longevity of boreal species [46,65].For this reason, they are considered to have a negligible influence on the age structure of forests in Quebec [22].However, non-stand replacing windthrows that generate small-scale mortality (<4 ha) occur much more frequently in boreal forests (return interval of 71 to 450 years [22,51]) (Table 1).Such return intervals compare to what is observed with fire return intervals in the boreal forest [35,68,69].

Spatial Descriptors
Stand-replacing windthrow in northeastern boreal forest were shown to range in size from 80 ha to approximately 60,000 ha and although many had an indistinct shape some were linear and characteristic of a major windstorm [48] (Table 1).In the case of partial windthrow, patch size is difficult to evaluate due to relatively low-contrast (or "soft edges") between the disturbed patches and undisturbed forest matrix [52,70].Furthermore, a windthrow patch can increase in size with time because of the domino effect [64,65].Studies from eastern boreal forest revealed that windthrows varied from 0.5 ha to 100 ha and most of the patches (i.e., 92%) were ≤4 ha or less in size [53,54].However, even if small windthrow patches predominate in number, larger ones cover the highest percentage of the affected stands [53,55,71].In order to link spatial characteristics of windthrow and their corresponding effects on forest dynamics, Ulanova [56] used three scales of study.Catastrophic large-scale windthrows were considered at the landscape and associated to secondary succession.Single or multiple smaller events affecting the forest community were associated to gap phase while individual tree fall to micro-succession dynamics [56].

Severity
Windthrow severity can be described according to the percentage of dead trees and amount of the forest floor and understory vegetation disrupted by uprooting of trees [39,40].Aerial surveys usually quantify windthrow as percentage of fallen or broken trees at the stand scale.Catastrophic windrows, characterised by high wind speeds will generate the highest severity along these axes by killing most canopy trees, some of regeneration and creating major soil disturbances [56,57] (Table 1).On the other hand, severity of smaller windthrow events can be highly variable but usually results in partial canopy mortality that may extend over several years after the main event with some soil disturbance resulting from tree uprooting [58,59].

Specificity
Even though the specificity of windthrow cannot be compared to a biotic disturbance in that there is no host-pest relationship, tree-, stand-and site-level factors influence windthrow specificity [22,60,61,72].Tree species is an important determinant of windthrow risk [43,60,62].Among conifers, balsam fir (Abies balsamea) is often reported to be more vulnerable to wind damage than Picea sp., mainly because it is more prone to root rot [61,63,73].Indeed, root or bark pathogens could influence windthrow risk, by affecting tree stability.Another example is the beech bark disease that increases the probability of American beech (Fagus grandifolia Ehrh.) stem breaks [62].Other factors such as stand age, tree height and slenderness ratio (the ratio of the height to diameter at 1.3 m above ground) play an important role in determining the vulnerability of trees and stands to wind damage [43,[74][75][76] (Table 1).The type of damage (uprooting vs. stem breakage) can itself vary among species and as a function of tree size [62].Stand structure and edges (proximity to a forest edge, as well as edge orientation) interact with wind dynamics to influence windthrow risk [4,57,[77][78][79].Site variables such as slope, topographic exposure, soil characteristics and surficial deposit can also influence windthrow risk [22,48,72,[80][81][82].For example, stands on poorly drained sites or on thin deposits tend to be more prone to windthrow [4,22,48,72,74].However, the significance and relative importance of predictors are often found to vary with the type of windthrow (partial vs. stand-replacing) and the scale considered [48].Various aspects of wind disturbance regime (e.g., rotation period, severity, etc.) are affected by specificity.For instance, stands that are more prone to windthrow tend to be associated with shorter windthrow return intervals [50].

Temporal Descriptors
Outbreaks are known to occur periodically, at about 30 year intervals on a supra-regional scale, at least during the 20th century [18,[83][84][85][86] (Table 1).In eastern North America, three major outbreaks occurred during the 20th century: the first in 1909-1925, the second from 1947 to 1957 and the last one between 1966 and 1992 [87,88].However, the temporal cycles of outbreak and their amplitude (e.g., severity) varied among landscapes being influenced by stand composition and their spatial arrangement [86].In the beginning of the 21th century, budworm populations were on the rise again and signs of defoliation were observed in the Quebec North-Shore region in 2006.In 2017, the outbreak affected over 7 Mha of forest mainly in northeastern Quebec [89] and it has recently reached New Brunswick and Maine (USA).
An important characteristic of SBW outbreaks is that tree mortality occurs gradually over the years, therefore leading to a gradual opening of the canopy [90,91] (Table 1).Balsam fir usually start to die after four or five years of severe defoliation and mortality is usually complete within 10 years of the onset of the outbreak [41].MacLean and Piene [92] reported mortality starting during the fourth to the sixth year of defoliation, with mortality occurring earlier in severely defoliated plots.

Spatial Descriptors
As SBW is host-species dependent, the range and distribution of the host species can either limit or allow spatial extent of outbreaks [17,86].The spatial extent of an outbreak will also be influenced by the dynamics of insect populations (e.g., whether the outbreak starts at an epicenter and spreads, or starts synchronously at several locations in a region as a result of climatic and/or biotic influences) [17,83,93].However, recent findings also suggest a feedback between forest structure and outbreak spatial dynamics [86].Such controls of outbreak dynamics will result in large-scale spatial variability in outbreak severity from one outbreak to the next.
When stand-replacing spruce budworm outbreaks are considered, disturbed patches ranging in size from 2 to >50 ha are reported [17,54,94] (Table 1).However, a large proportion of trees remained alive in the openings.Similar to partial windthrow, SBW-caused gaps can be difficult to evaluate, because the contrast between the disturbed and undisturbed portions of the forest matrix can be blurred by (i) the presence of remnant living trees within affected patches [94] and (ii) a partially opened canopy in the undisturbed area [90].Considering these limitations, many studies measured gap fraction at the stand scale [95][96][97] and concluded that SBW outbreaks create complex patterns of defoliation, a high level of heterogeneity in canopy openness and high ratios of edge to interior conditions [94,96,98,99].

Severity
Spruce budworm outbreak severity can be described according to the percentage of dead host trees in the canopy and the amount of understory regeneration killed.In the initial stages of an outbreak, defoliation occurs primarily in the overstory.As population increases and defoliation reaches high levels in the tree canopy, larvae spin silk threads from defoliated upper canopy tree branches onto advance regeneration [100].Following very severe defoliation, only smaller seedlings (<50 cm in height) survive [101].Forest stands with little regeneration or in which a tall seedling or sapling bank was killed tend to experience the greatest transition to non-host species [102].Bergeron et al. [103] showed that canopy mortality is highly variable, ranging from 0% to 100% and depends on tree and stand characteristics as well as the scale of analysis.

Specificity
Specificity to SBW varies as a function of tree, stand and site level characteristics and strongly modulate outbreak severity.It can also be influenced by forest mosaic conditions that is the density of host species in a landscape [103,104].At the tree-level, vulnerability increases with tree size and species, with balsam fir being the most vulnerable host [103,105].Characteristics such as stand age, species composition and stand location in the surrounding forest landscape are factors that determine spatial variability in tree mortality [103,106].Host tree mortality may also vary with site characteristics such as slope [107], surficial deposit and moisture regime [108] but not always [103,106] (Table 1).
One important difference between SBW and windthrow is that mortality resulting from SBW is more strongly species-specific, due to the tight host-specificity of the insect.SBW affects balsam fir severely and, to a lesser extent, white red and black spruce, in that order [105,106].Stand composition thus influences disturbance severity (and hence post-outbreak stand structure), with higher defoliation and mortality observed in mature fir stands [18,30,41,103,106], while young stands or those with a high proportions of hardwood species being much less vulnerable [109].For example, MacLean [17] reports mortality levels averaging 85% in mature balsam fir stands, 42% in immature balsam fir stands, 36% in mature spruce stands and 13% in immature spruce stands.The composition of the surrounding forest can also influence SBW outbreak severity at landscape-level [103].An increase in hardwood content in the surrounding forest reduces defoliation and subsequent growth losses in balsam fir and spruce during the initial years of a SBW outbreak [18,103].

Biological Legacies
Biological legacies are recognized as important outcomes of disturbances [7,13,110].They can be organisms, organic materials and biologically created patterns that are modified by or persisting after disturbance [110].Biological legacies help enhance, maintain or restore the compositional, structural and functional diversity in the post-disturbance ecosystem [110].The abundance, characteristics (e.g., decay state in the case of coarse woody debris) and spatial distribution of biological legacies influence the rate and pattern of post-disturbance ecosystem recovery [3,7].Structural legacies include standing dead trees (snags), downed logs and other woody debris, tip up mounds associated with uprooted trees, remnant live trees and patches of undisturbed understory, seed and seedling banks and vegetative reproductive parts (e.g., root systems) [13].Knowledge on dead wood dynamics is critical for better predicting and monitoring carbon storage and release.

Snags and Coarse Woody Debris
Snags and coarse woody debris (CWD) are particularly important legacies in forested ecosystems [111].Snags can be relatively long-lasting structures (e.g., with a half-life of 30 to 40 years in Abies balsamea-Picea mariana forests).They represent important pools of dead wood in forest ecosystems, their rates of decay and fall being key determinants of the abundance and quality of CWD [112][113][114].Snags and CWD are significant for biodiversity as numerous organisms are dependent on dead wood [111,115,116].They can serve as habitat for a number of vertebrates, invertebrates, fungi and lichens [33,117,118] and contribute to community assembly [33].When sufficiently decomposed and/or covered by moss, they constitute a preferred establishment microsite for some tree species [119].
Because, tree mortality in windthrow occurs through stem breakage and tree uprooting, generally few snags are left (Table 2) [20].However, some storms (heavy wet snow and ice storms) may cause stem breakage higher in the tree and generate many snags.In situations where tree uprooting is more common than stem breakage (e.g., due to species and/or site characteristics), snag density might not necessarily increase with windthrow severity [59].Woody debris of all size can be particularly abundant after windthrow (Table 2) and downed dead wood volume is positively correlated to windthrow severity [50,59].The distribution of coarse woody debris (CWD) among decay classes can change rapidly during the years after windthrow because of wood degradation, which depends, in part, on species and climate but also on the position of the fallen trees (directly touching the forest ground, on a top of a fallen stems pile, etc.) [120].Because of windthrow patch expansion, dead wood recruitment may continue over the years after an initial windthrow episode, leading to the presence of dead wood in different decay stages [59,114,121].
In the case of SBW outbreaks, trees are killed standing, which has the potential to generate a large number of snags (Table 2) [20,90].Due to the host specificity of the SBW, snags will be much more abundant among host species [20].Snags may break or be blown down from a few up to 25 years after death and thus, ensure down dead wood recruitment for a long period [104,113,122].Therefore, in the case of SBW outbreaks (in comparison to windthrow), CWD inputs tend to lag behind the actual time of the outbreak (Table 2) [123,124].Furthermore, as SBW tree mortality can spread over several years [103,125], this contributes to a prolonged accumulation of snags and CWD.Such a characteristic in dead wood production distinguishes SBW outbreaks from both windthrow and fire.Abundant CWD and moss cover after SBW outbreaks provides good substrates for germination and survival of tree regeneration [126,127].Preserve some vertical structure in the post-disturbance stand [13,91] Coarse woody debris Abundant (short term) Increase with snag fall rate When sufficiently decomposed and/or moss covered, substrate for germination [126] Tip up mounds Many Unusual Source of heterogeneity in establishment microsites [129] Germination microsites for some species (e.g., small seeded Picea sp. and Betula sp.) [130] Influence the spatial distribution of regeneration (and eventually of trees) in the post-disturbance stand [56] Contribute to maintain (or increase) species diversity [56] Live remnant trees Few to many Depending on stand species composition Possible seed source for recolonization of area [7] Limit soil erosion [7] Preserve some vertical structure in the post-disturbance stand [13] Intact understory patches Possible (patchy) Common Maintain pre-established regeneration (seedling banks) [131,132] Patches of intact forest floor associated with maintenance of some shrub species [131,132]

Tip-Up Mounds
Tree uprooting induces a pit-and-mound microtopography very characteristic of stands after windthrow (Table 1) [56,129,133,134].Ulanova (2000) [56] reported that 15% to 25% of the forest floor area was in pits and mounds in Russian old-growth forests but this proportion can vary markedly depending on forest type and severity of windthrow.Kuuluvainen and Juntunen [130] found that tip-up mounds covered only 3.4% of the ground area in a mature Pinus sylvestris forest in Finland and Waldron et al. [59] observed even lower proportions (approximately 2%) in eastern Canada black spruce boreal forests.Such structural legacies can persist over 100 years in forest ecosystems [56].The microtopography associated with pits and mounds creates heterogeneity in soil characteristics (e.g., temperature and humidity), litter accumulation and light availability [129,135].Soil turnover associated with tree uprooting contributes to expose mineral soil and provide seedbed heterogeneity following windthrow [129].Such structural changes play important roles in tree regeneration processes and contribute to the maintenance of plant diversity.For instance, the germination of small-seeded species (e.g., Pinus sp. and Betula sp.) is favored on windthrow generated tip-up mounds [130].Waldron et al. [131] found a positive association between birch (Betula papyrifera) regeneration and windthrow mounds.Jonsson and Esseen [136] showed that in Scandinavian Picea abies forests, bryophyte diversity was favored by the microsite diversity created by tree uprooting.

Remnant Live Trees
Remnant live trees help maintain some vertical structure in the post-disturbance stand, limit soil erosion and facilitate the recolonization of late-successional species and the recovery of total ecosystem carbon stocks after disturbance [7].Remnant live trees and undisturbed patches provide habitat, modify micro-environmental conditions and maintain connectivity [13].In stand-replacing windthrows, it would be very rare to observe uprooting or stem breakage of all trees.The post-disturbance stands will more likely consist of a mosaic of completely opened area interspersed with patches of remnant trees [55].As windthrow severity decreases, remnant trees and undisturbed forest mosaic will dominate the post-disturbance.After SBW outbreaks, due to the host specificity of the insect, remnant live trees will mostly be of non-host species (Table 2) [94] but also comprise some surviving host trees [20].

Undisturbed Understory Patches
While newly created microsites such as those associated with pits and mounds are important for the regeneration of some species, intact understory patches also play a role in maintaining other species.These patches have intact forest floor and thus they maintain intact seed and seedling banks [13].Undisturbed understory patches were found to be associated with pre-established Picea mariana and Abies balsamea regeneration, Sphagnum mosses and ericaceous shrubs in a study of windthrow in the Quebec North Shore region [131].The pre-established regeneration that persists in undisturbed patches often plays an important role in post-windthrow stand recovery [132].Due to the nature of the disturbance, SBW outbreaks tend to leave an intact forest floor and understory layer, at least for the shrubs and herbaceous species [137].

Incorporating Knowledge of Windthrow and Insect Disturbance into Boreal Forest Management
Knowledge of spatial and temporal characteristics of fire regimes, combined with notions of severity, have been used to develop a natural disturbance-based forest management framework for boreal forests [8,34].This review provides a complementary framework that considers other disturbance types prevailing in regions with long fire return interval and where old-growth forests proportion is high [19].Considering that mature and old-growth forests are targeted preferentially by both harvesting and partial disturbances (i.e., non stand replacing), understanding the patterns and biological legacies that windthrows and SBW outbreaks generate is imperative, in order to limit the impacts of harvesting on forest dynamics and biodiversity.Moreover, this understanding can also contribute to increase wood production and value by guiding future interventions.This could be achieved either by lowering forests' vulnerability to these disturbances or retaining biological legacies to facilitate forest recovery and minimize the costs of silviculture.It is therefore relevant to think about how such disturbances affect boreal forests and consider how current knowledge may guide forest management planning, silviculture development and salvage logging policies.
We acknowledge that a complete understanding of the factors that influence forest dynamics cannot be reached unless we consider the whole set of disturbance agents in a given ecosystem [24].Thus, the focus on these disturbance types brings knowledge that was lacking in order to propose a comprehensive ecosystem based-management framework.Both windthrow and insect outbreaks are drivers of forest complexity, ecosystem change and self-organization and in the face of global change, these inherent ecosystem properties are paramount to maintain resilient forests [138].The integration of knowledge about the response of forests to these disturbances for use in forest management may, however, be more challenging than it is for fire.With the exception of stand-replacing windthrow, the spatial imprints of these disturbances are much more diffuse, and their temporal characteristics, which are highly variable, may be difficult to integrate into a management framework.For example, patch size could be defined as total area covered by a windthrow or a SBW outbreak event.However, within a disturbance event, some areas are intact while others are severely affected.Thus, patch definition could also refer to the within-disturbance severity pattern [53,139].Temporal and spatial descriptors could also be used to guide the timing and the extent of interventions.Temporal patterns of tree mortality could provide insights into the number and temporal distribution of entries in a stand with spatial characteristics providing insights on t the size and distribution of harvest openings.
Disturbance severity and specificity descriptors, on the other hand, may provide a template upon which policies for maintaining post harvesting and salvage logging biological legacies can be based.These descriptors will inform as to the percentage of canopy removal and the proportion of harvest by species.Additionally, a better knowledge of forest conditions (composition, site and stand characteristics) can help identify conditions that need to be restored [39,40] as well as biological legacies to maintain following harvesting in order to hasten forest recovery and maintain resilience.
Although managers can not consider all disturbance agents involved in a forested region, in addition to windthrow and SBW, fire should obviously be considered [14].In addition to SBW, other insects could be considered, if relevant, depending on the region and the ecological context (e.g., Hemlock looper or Mountain Pine Beetle).The relative importance of the role played by each type of disturbance could also be evaluated.Fire disturbance regime characteristics in a region influence the age structure of the forest and where fires are stand reinitiating their frequency can guide the portion of the territory to be managed using an even aged approach [54].The remaining portion of the territory could correspond to the area where partial disturbances will drive forest dynamics (windthrow, insect outbreak and senescence).In the North American boreal forest, we now have a good idea of the current fire recurrence as well as how it is predicted to change in the future with climate change [140].Once proportions of a territory corresponding to severe (e.g., fire), intermediate (e.g., partial windthrow and insect outbreaks) and small-scale disturbances are defined, silvicultural options can be selected and/or developed while bearing in mind the stand-level effects of the various disturbances [141].Biological legacies should also be considered to guide the implementation of variable retention approach.
Although the examples presented in this paper come primarily from North Eastern North America, forest insect pests and wind disturbance are ubiquitous in forests throughout the world.The goal of this work is thus to provide a template for reflection for readers and users throughout other parts of the globe.The approach proposed in this review fits within a more general conceptual framework of understanding, emulating and modelling natural disturbances for guiding forest management or restoration [9,[142][143][144][145].

Knowledge Gaps and Avenues for Research
Both windthrow and SBW outbreak generate large pulse of dead wood and compared to fire little knowledge exists on the ecological importance dead wood for biodiversity and ecological processes (e.g., regeneration) following these disturbances.How much dead wood is enough to maintain dependent species and associated processes and at what spatial scale?SBW outbreaks are regional events, creating a large periodic influx of dead wood with long intervals without dead wood contribution.Bouchard et al. [146] suggested that such pulses in mortality and thus dead wood may be an important driver in the movement of faunal species in forests subject to SBW disturbance.Attempting to produce such pulses would thus be a challenge for forest managers who aim to produce a continual wood flow.This may be an example of where the goal should be not to emulate disturbances but rather to ensure that salvage logging does not remove all of the dead wood [147].However, effect of the amount and distribution of dead wood retained after salvage logging on biodiversity is not fully understood [148].
A long-term perspective is needed to understand the effects of windthrow and SBW on forest attributes and processes as most studies consider only the first few years after disturbance [20,149].Mid-and long-term post-disturbance studies may add information on legacies and stand recovery pattern that could inform forest management strategies.
Interactions among disturbance types have recently become a growing research interest [25,[150][151][152].Taking into account interactions among disturbances is important both when two or more natural disturbances interact and when natural and human-made disturbances interact.For example, although in this review windthrow and SBW were treated separately for the sake of clarity and simplicity, in reality, there are many occasions where these disturbances interact.For instance, windthrow occurring in stands previously affected by the SBW seems to be relatively common and has both ecological and silvicultural impacts [153].Increasing windthrow risk is also observed following partial harvesting [61,154].While organisms have evolved with and become adapted to multiple natural disturbance regimes, they might not be adapted to novel combinations of natural and human-caused disturbances [155].Multiple interacting disturbances could lead to unexpected results, including shifts in post-disturbance recovery patterns and reduced ecosystem resilience [150,156].Compounded effects of human-caused and natural disturbances should especially be of concern when human interventions are severe and closely follow the occurrence of natural disturbances, as is generally the case with salvage logging performed after fire, windthrow, or insect outbreaks [155,157].
Understanding how climate change will influence these disturbance regimes and how this will interact with forest management will also be crucial for forest managers in the future [158].Insects are projected to expand or shift their ranges into new areas and perhaps also into new or previously considered less-vulnerable tree species [159].Furthermore, interaction between climatic stress and SBW defoliation may contribute to increase the rate of tree mortality as climate changes [125].Windthrow may also increase in some areas [160], although, due to considerable uncertainty a risk management approach may be a first step to dealing with this changes in its occurrence and severity.In all cases, species will be under increasing pressure from global changes in habitat due to forest management and climate change.Providing some conditions within historical ranges will help the most vulnerable species by giving them opportunities to maintain regionally and eventually migrate to suitable sites.More importantly, it will be crucial to integrate climate change into an ecosystem based forest management framework to be able to understand vulnerabilities and propose mitigation measures [161][162][163][164].

Conclusions
The objective of this review was to identify the main characteristics of the disturbance regimes associated with windthrows and SBW outbreaks in northeastern North American boreal forests in order that they could be used to inform forest management and potentially reduce deviations from natural conditions.To do so we reviewed several disturbance regime components, that is, not only temporal and spatial aspects but also how variation in disturbance severity and specificity modulate the effect of such disturbances at stand-and landscape-scales.We also reviewed biological legacies left by these disturbances, which play a major role in ecosystem recovery.These disturbances help maintain a dynamic old growth stage.While some more severe events can bring the system to a re-initiation stage, most events are diffuse in space and time and help maintain and enhance old-growth characteristics.Thus, incorporating knowledge of windthrow and SBW disturbances and including information about biological legacies in a management framework could help managers address ecosystem management objectives, particularly in forest areas where post-disturbance salvage logging could temporarily represent a major proportion of managed stands.

Table 1 .
Disturbance regime main descriptors of windthrow and spruce budworm outbreaks in the boreal forest of eastern Canada.