Snag Characteristics and Cavity-Nesting Birds in the Unmanaged Post-Fire North- eastern Canadian Boreal Forest

We studied the availability and characteristics of snags and their use by cavity-nesting birds in the northeastern part of the Canadian boreal forest. We built up two long-term (> 200 years) chronosequences following time since the last fire in the unmanaged boreal forest of northeastern Québec, one in the balsam fir-white birch domain (southern region) and one in the spruce-mosses domain (northern region). We then sampled and characterized snags and live trees in 30 stands from each of these two chronosequences. We also looked for nest cavities on all sampled snags, performed bird inventories by point counts, and calculated tree mortality rate from permanent sample plots. Results show that mortality rates follow a U-shaped pattern, with more snags of large diameter (> 20 cm DBH) in young (< 50 years) and in old (> 200 years) forests. In the latter, we also found more nest cavities than in any other age classes. Although abundance of primary cavity nesters (excavating species) did not vary among age classes, secondary cavity nesters (using cavities already available) tend to be more numerous in older forests. Our results highlight the capacity for young and old-growth forests to provide quality habitat for species that are dependent on large snags. Proper forest management should maintain a mosaic of different age forest stands, including snags, to promote biodiversity and provide important resources for resident bird species.


Introduction
In boreal forest ecosystems, snags (dead standing trees or dying trees) contribute to many biological functions and provide valuable habitats for the conservation or enhancement of wildlife (e.g.black-backed woodpecker; Bate et al. 1999, Drapeau et al. 2002, Saint-Germain et al. 2004, Hutto 2006).They are used by a wealth of invertebrate, vertebrate, and plant species.Furthermore, organisms such as fungi and insects that are found in dead trees greatly enhance the value of snags for other wildlife.For example, decomposers invade dead or dying trees, and weaken and soften them, facilitating their use by primary cavity-nesters (i.e.woodpeckers and nuthatches) and secondary cavity-nesters (species that use cavities excavated by primary cavity nesters or natural cavities created by decay; Thomas et al. 1979, Bull 1983, Drapeau et al. 2002).
In the eastern part of the Québec boreal forest, the humid climate lengthens the fire cycle up to 500 years (Bouchard et al. 2008), resulting in a multiaged mosaic with old-growth stands covering 60-65 % of the unmanaged forest territory (Boucher et al. 2003, Côté et al. 2010).In earlier studies, snag abundance has been observed to follow a well-defined U-shaped pattern over stand age.This pattern is characterized by high volume of snags in stands recently disturbed by fire, few snags in actively growing stands, and an increasing snag abundance in ageing stands as canopy trees become progressively senescent (e.g., Harmon et al. 1986, Spies and Franklin 1988, Tyrrell and Crow 1994, Clark et al. 1998).In the course of this long-term stand development, other forest attributes may change, including species composition and tree size distribution.In the northeastern Canadian boreal forest, balsam fir (Abies balsamea (L.) Mill.) generally increases in abundance, while tree diameter distribution becomes increasingly irregular, as black spruce (Picea mariana (Mill.)BSP) dominated forests get older (Pham et al. 2004, Bouchard et al. 2008, Côté et al. 2010).In turn, this could affect snag production since balsam fir is more vulnerable than black spruce to wood decay and spruce budworm (Choristoneura fumiferana (Clem.)),an important defoliation insect causing much mortality in boreal forests of eastern Canada (Whitney 1995, Hennigar et al. 2008).Some studies on bird abundance, species richness, and diversity have compared succession changes originating from fire and logging (Imbeau et al. 1999, Le Blanc et al. 2010), but few have investigated changes in bird communities along a long-term gradient in unmanaged forests.Fire has an immediate positive impact on woodpecker breeding and foraging because of the high number of snags produced (Nappi et al. 2003), confirming the importance of early post-fire disturbance forests for wildlife (Dixon and Saab 2000).In oldgrowth boreal forests, however, results from the western part of the Quebec province indicate relative low snag abundance (Drapeau et al. 2002), while studies of other boreal regions in Canada suggest the high potential for old-growth forests for cavity nesting birds due to high abundance of snags (Schieck et al. 1995).If snag abundance varies from early post-fire years to old-growth, we should expect use of snags to change as well.In addition, little is known about the capacity of old-growth forests to provide adequate habitat for bird species associated with large snags.By exploring these issues, we might be able to provide better recommendations and rationales to manage and harvest forest landscape considering habitat protection for cavity-nesting birds in the boreal forests of eastern Canada.
The general objective of this study was to determine if there is variation in snag availability for cavitynesting birds along two long term chronosequences after fire, and to relate any variation to avian species richness and abundance.First, we quantified snag abundance (basal area) and characteristics (e.g.size classes) as well as tree mortality rate coming from permanent sample plots along chronosequences following fire.Second, we evaluated the effect of tree size and wood degradation levels on the number of bird cavities.Finally, we estimated species richness and bird abundance of cavity nesting birds according to time since fire.

Study Area
Our study was conducted in the North Shore region of Québec, in the northeastern part of the Canadian boreal forest.The topography is characterized by hills and moderate slopes (16-30%), with a maximum recorded altitude of 700 m and a cold maritime climate (Robitaille and Saucier 1998).The climate is boreal humid and is characterized by average annual temperatures between -2.5 and 0.0 °C and annual precipitation ranging from 1000 to 1400 mm (Garet et al. 2009).
Surveys were carried out in two regions of the study area: the southern region (49°42´N, 68°51´W) located a few km north of Baie-Comeau, and the northern region (50°57´N, 68°50´W) located south of the Manic 5 reservoir (Fig. 1).These regions overlap two bioclimatic sub-domains of the boreal zone: the balsam fir -white birch domain (southern region) and the spruce-mosses domain (northern region).Balsam fir characterizes the former, with black spruce dominating stands in poorer sites (MNRF 2003).As latitude increases (shifting to the northern sub-domain), forest stands are increasingly dominated by black spruce.Dominant species found in the region include balsam fir and black spruce, with minor components of white spruce (Picea glauca (Moench) Voss), jack pine (Pinus banksiana Lamb), white birch (Betula papyrifera Marsh.) and trembling aspen (Populus tremuloides Michx.).With time since the last fire, the size structure of live trees becomes more irregular, which is generally accompanied by a gradual increase in balsam fir abundance (De Grandpré et al. 2000, Côté et al. 2010).The region is known for outbreaks of spruce budworm (Choristoneura fumiferana (Clemens)), especially in the southern part of the study area (near Baie-Comeau) between 1975 and 1990 (Bouchard and Pothier 2010).However, in the northern part of the study area, these outbreaks resulted in very little tree mortality (Bouchard and Pothier 2010).Although the southern part was also used for log- ging, numerous fragments of unmanaged forests still remain today (Bouchard et al. 2008) and were used in this study.Fire history has been reconstructed over the area by Bouchard et al. (2008) who mapped burned stands larger than 10 hectares and up to 200 years since last fire.In each of the two regions, 30 stands were identified to form a chronosequence after fire (Fig. 1).For each chronosequence, five independent stands were selected in each of the following age classes: from 0 to 50 years since last fire (class 1), 51 to 100 years (class 2), 101 to 150 years (class 3) and 151 to 200 (class 4).We also selected ten old-growth stands for which the time since fire (from here on TSF) was not determined, but was greater than 200 years (class 5).Overall, we sampled 60 stands (2 chronosequences of 30 stands per chronosequence) that were then inventoried.

Standard Inventory Plots
In each selected stand, we established standard inventory plots during summer 2007 to characterize stand composition and diameter distribution.In each of these circular, 400-m 2 plots, all snags and live trees were identified, along with their diameter at breast height (DBH) for trees larger than 9.1 cm.The state of decomposition of each snag was categorized following Hunter's classification (decay stages, 2 declining to 7 decomposed; Fig. 10.1 in Hunter 1990).

Dead Standing Tree Plots
To quantify snag decomposition, we established one 50 x 20 m plot adjacent to the standard inventory plots in each of the 60 stands selected from the two chronosequences.In each plot, we tallied all snags to increase sample size because standard inventory plots (400 m 2 ) were too small to accurately determine the number of snags available to birds.
For each snag, we determined height (+0.1 m), DBH (+1 cm), and the state of decomposition.For the purpose of this study, only snags with DBH > 9.1 cm and a minimum height of 1.3 m were considered.These characteristics are considered minimal conditions for nesting or foraging birds (Thomas et al. 1979).

Permanent Sample Plots
To estimate the snag production rate at different stand ages in the studied territory, we used a network of permanent sample plots (PSPs) established in 1970 by the Ministère des Ressources Naturelles et de la Faune du Québec (MRNFQ).We overlapped these georeferenced PSPs to the fire map of Bouchard et al. (2008), to determine the time since the last fire for each of the 460 PSPs corresponding to the territory covered by the fire map.Each of these PSPs was inventoried two to four times (depending on the year of establishment) during the snow-free periods between 1970 and 2007 for a total of 1107 plot measurements.The inventory taken in each 400 m 2 circular PSP consisted of measuring the DBH (1.3 m, DBH ± 1 mm) of each tree (live and dead) larger than 9.0 cm.Also, the age and height (± 0.1 m) of at least three live dominant tree species were recorded (Pothier and Mailly 2006).Periodic measurements of each PSP allowed us to quantify the rate of mortality in terms of basal area and to relate this rate of mortality to time since fire as well as to stand characteristics.The average time period between two successive measurements of the PSPs was 10 years, but the mortality rates were calculated on an annual basis.Although the size of these PSPs (400 m 2 ) is rather small to evaluate snag abundance, we believe that the quite long time period covered by these numerous PSPs compensates for this disadvantage.

Bird Nest Cavity
We considered a cavity nest to be a hollowed-out opening in the trunk of a tree, either found naturally in snags or purposely made by birds such as woodpeckers (as long as they were occupied cavities; excluded were the natural holes not used by birds or foraging cavities).Other examples of cavity-nesting birds include chickadees, nuthatches, and creepers (Hunter 1999).We also included brown creepers' nests, built between the trunk and a loose piece of bark, since they are found mostly on dead or dying trees (Hejl et al. 2002).A list of the cavity-nesting species found in the study area is presented in Table 1.From mid-May to July 2007, in addition to dead tree inventories, all 60 stands were surveyed for the presence of cavities in each snag located in the 50 x 20 m plots.

Point Counts
To estimate species richness and bird abundance of cavity nesting birds, we used the point count method (Reynolds et al. 1980, Hutto et al. 1986).One point was located at the center of each standard inventory plot established in the 60 stands from the two chronosequences.All these stands were independent, but in some few cases, one stand was close to another.In these cases, we applied a minimal distance of 200 m between two points while all points were at least 100 m from the edge of a burned site.Therefore, we avoided repetitions in bird detection which could have produced spatial autocorrelation.
Two visits were conducted at each point during the breeding period (late May to late June 2007), to allow detection of a maximal number of species during their peak of singing activity.Observers recorded the number of individuals of each bird species detected by sight or sound for each of the following distance classes from the center of the point: 0-30 m, 30-50 m, 50-75 m, and 75-100 m.In this study, we used the number of species detected in both visits of each point count to a maximum radius of 100 m.
Observers were familiar with bird vocalizations.However, if identification was doubtful, they recorded the bird song using a digital tape recorder with a directional microphone.The two visits to each stand were made by a different person to minimize observer bias.Similarly, we varied the time we visited each point count (e.g.: early vs. later in the morning) to maximize detection probabilities of each species present.At each visit, birds seen or heard were recorded at 10-min intervals for a 20-min period.This is the standard time procedure for boreal forest bird communities when one wants to relate bird richness and abundance with the vegetation characteristics of the area around the point (Drapeau et al. 2000(Drapeau et al. , 2002)).

Statistical Analyses
An analysis of variance (ANOVA) was used to compare mean snag basal area and rate of tree mortality between the two chronosequences and among TSF classes.We weighted the number of snags per plot so that each plot contributed to only one degree of freedom in the model (Desrochers 1992).This technique allows the use of all observations while avoiding pseudo-replication (Machlis et al. 1985).We used TSF as a classification variable rather than as a continuous variable in the statistical analyses, because we were not able to determine the exact value for plots with TSF > 200 years (class 5) from the available fire map (Bouchard et al. 2008).When the ANOVA detected a significant difference for one of the factors, we compared the means of this factor using a Waller-Duncan comparison test.In addition, when the ANOVA detected a significant difference for the interaction between two factors, we compared the means of each level of the interaction using the same comparison test.For all analyses of point count data, we used the maximum number of individuals for each species recorded (resident vs. migrants, primary vs. secondary cavity nesters) at each plot after merging together the observations of the two visits.

Snag Basal Area and Tree Mortality Rate
A total of 1740 snags were measured from 60 stands (30 per chronosequence) covering five TSF classes.Significant differences in snag basal area were observed between the two chronosequences, with a higher value observed in the southern chronosequence (Fig. 2).Snag availability changed significantly among TSF classes (F=3.33,df = 9, P = 0.003).These differences were more structured in the northern than southern chronosequence, showing a defined U-shaped pattern of abundance.Snag basal area was larger in TSF classes 0-50 and > 200 years because of a higher presence of large sized trees.
Tree mortality rate, calculated on a basal area basis using the repeated measurements of the permanent sample plots, differed among TSF classes (F=5.92,df = 4, P = 0.0001) and tended to increase with TSF (Fig. 3).The average annual mortality rate was twice as large in the two oldest plots (> 150 years, rate ≈ 0.40%) compared to the two younger ones (< 100 years, ≈ 0.18%).For  TSF over 200 years, however, the mortality rate tended to decrease.Annual tree mortality rates also differed among TSF classes and tree species groups (Fig. 3).For instance, mortality rates of black spruce increased with TSF, while those of intolerant hardwoods (mostly composed of white birch, and trembling aspen) were highest in TSF classes between 51 and 200 years.Mortality rates of balsam fi r remained relatively constant versus TSF except in the 51-100 class, within which mortality rates were low.No difference in mortality rates were observed in the other species groups dominated by white spruce, and jack pine.

Cavity Abundance
A total of 71 cavities were found in 50 snags, representing 2.87% of all snags tallied in this study (1740).The number of cavities per hectare did not differ signifi cantly among chronosequences and TSF classes, but young stands (0-50 years) and those older than 200 years tended to contain more cavities (Fig. 4a).When snags were in a more advanced stage of decomposition, they seemed to support more cavities (Fig. 4b).Also, a greater proportion of trees with cavities were observed in snags larger than 20 cm DBH compared to smaller snags (8.2% and 1.2% respectively) (Fig. 4c).The number of cavities present in snags larger than 30 cm DBH seemed to be relatively low because such trees were rare on our study sites.The mean number of large snags (> 20 cm) per hectare was signifi cantly different among TSF classes (F=2.83,df = 9, P = 0.0031; Fig. 4d).Snags > 20 cm DBH were generally more abundant in young (0-50 years) and old (> 200 years) stands, compared to those of intermediate ages.

Cavity Nesting Birds
Although cavity nesting bird species were all present in both chronosequences (no differences in cavity nesting bird species composition), we found a signifi cant interaction between TSF and chronosequence that infl uenced the species rich-ness (F = 2.39, df = 9, P = 0.0246).Young forests (age class: 0-50 years) were associated with fewer cavity nesters than other TSF classes in the southern chronosequence but not in the northern one (Fig. 5).Apart from these young stands, no other signifi cant differences were observed between cavity nesters and TSF.Fig. 6.Presence of cavity-nesters according to time since fi re.For each species, line width corresponds to the relative abundance of these birds during point counts (see Methods).Within each TSF class, no line means 0 birds per point count, the thinner black line represents a density of ≤ 0.2 bird/point count whereas the thickest black line corresponds to a density of ≥ 0.8 bird/point count.Scientifi c names for the species listed can be found in Table 1.
Relative abundance (averaged for all point counts) of primary cavity nesters did not vary among age classes.However, secondary nesters seemed to be more abundant in older forests (> 100 years).These included the boreal chickadee and brown creeper (Fig. 6).

Snag Characteristics
Snag availability changed with time since the last fire in the two chronosequences (Fig. 2).In general, our results for the northern chronosequence agree with the typical snag abundance for unmanaged boreal forests, characterized by the U-shaped pattern that represents an early, high level of snags followed by a subsequent transition phase with few snags, and then, a later increase in snag abundance once a forest matures (Harmon et al. 1986, Peet and Christensen 1987, Clark et al. 1998).In the southern chronosequence, this pattern was not observed: rather, during the 101-150 year period, we found an increase in the overall mean basal area for dead trees.Since we inventoried the two chronosequences in 2007, it is possible that the presence of some dead trees in these stands was caused by the late defoliation years of the last spruce budworm outbreak that ended during the early 1990s (Aakala et al. 2007) and that mainly affected the southern area of the studied territory (Bouchard and Pothier 2010).This might also be related to a severe ice storm that occurred during the 2004-2005 winter and that top-killed many dominant trees in stands of the southern chronosequence.These damages on the top of the trees were not observed in the northern chronosequence, likely because the colder temperature resulted in precipitation falling as snow instead of as rain.
According to yield curves for black spruce stands of the study area (Garet et al. 2009), total tree mortality rates (all species) were proportional to the temporal pattern of change of living tree volume as previously observed by Aakala et al. (2008).At the species level, even though balsam fir is thought to be more susceptible to mortality agents than black spruce (Burns andHonkala 1990, Aakala et al. 2007), the increasing snag availability with TSF seems unrelated to the increasing occurrence of balsam fir in old-growth stands (Fig. 3).Indeed, balsam fir and black spruce mortality rates in stands of varying ages seem proportional to their respective abundance that changes with TSF in agreement with the pattern of tree succession observed after fire in the study area (Bouchard et al. 2008, Gauthier et al. 2010).However, species-specific mortality rates are also influenced by the longevity of each species and by secondary disturbances such as spruce budworm defoliation.
It is now widely accepted that early post-fire forests are key habitats for snag recruitment (Fig. 2) because trees killed by fire are a major source of new snags (Wein andMacLean 1983, Gauthier et al. 2001).At the other end of the chronosequence spectrum, though, old-growth forests also provide a large number of snags, mostly with large DBH, at varying stages of decomposition (this study; Bergeron 2000, Bergeron andHarper 2009).These large-diameter snags are more valuable for wildlife than smaller ones because they persist longer (lower falling rate, see Saab and Dudley 1998) and they are suitable for small as well as large cavity-nesting birds (Vaillancourt et al. 2008).Indeed, large diameter snags resulted in an increased number of cavities, and cavity nesters such as woodpeckers are known to prefer these trees (Mannan et al. 1980, Swallow et al. 1986, Drapeau et al. 2002, Nappi et al. 2003, Smith et al. 2008).The number of cavities per unit area also tends to increase in stands composed of numerous highly decayed trees (Fig. 4b).This result is likely related to the residence times of these trees which promote their use by a large variety of cavity-nesting birds, from woodpeckers in lightly decayed snags (at the beginning of the residence time) to weak excavators in highly decayed snags (at the end of the residence time) (Vaillancourt et al. 2008).In our study area, the residence times of dead standing trees is less than five years in the early stage of tree decomposition, but can reach as much as 40 years in the last stage (Aakala et al. 2008).For comparison, the residence times of dead standing Scots pine (Pinus sylvestris L.) trees in drier sites of Finland averaged 81 years whereas some dead trees remained standing as long as 200 years (Rouvinen et al. 2002).

Cavity Nester Abundance
In this study, we used point counts to quantify the occurrence patterns of cavity-nesting birds.We believe that bird abundance should reflect differences observed in their habitat even though this relationship might be better with more detailed data on demography (see Van Horne 2002).Previous studies of the occurrence of primary and secondary cavity nesters have found that some species are more common in recently-burned forests and old-growth stands (Drapeau et al. 2002).In our study, black-backed woodpeckers were not necessarily more numerous in the recently-burned or old-growth forests, but rather were found in all TSF classes.Due to our low number of sample plots per age class for point count purposes, and consequently, low detection of these species, it is difficult to provide abundance estimates by TSF class.Most studies have shown that high productivity in terms of burn use by this woodpecker occurs only 1 to 3 years following a burn (Dixon andSaab 2000, Nappi et al. 2003), but our 1-50 year TSF class included only one such recent fire.This explains why the woodpecker population may not dominate as much as it does in other studies that may have included the period right after fire.Indeed, the present study did not specifically focused on the early years post-fire, it rather focused on the overall long term chronosequences.Meanwhile, in the southern chronosequence, dead tree abundance was not found only in young and old TSF classes, but also in the 101-150 years (Fig. 2).This might explain why black-backed or American three-toed woodpeckers were not necessarily more abundant in any age class (Fig. 6).Nevertheless, similar to the findings of Drapeau et al. (2000Drapeau et al. ( , 2002) ) in the boreal forests of western Québec, cavity nesters such as the red-breasted nuthatch, boreal chickadee and brown creeper were all more abundant in mature and old-growth forests than in younger stands.The high quantity of live and dead trees > 20 cm DBH might explain their abundance in these forests as they provide higher opportunities for nesting and for foraging.

Implications for Management
The boreal forest of eastern Québec is characterized by long fire-return intervals (Bouchard et al. 2008) that lead to the dominance of old-growth forest stands.These stands are characterized by, among other factors, a high level of deadwood, which has a significant ecological function in the ecosystem (Hunter 1999, Imbeau et al. 2001, Bergeron et al. 2006).Our study shows the importance of conserving this forest age structure at the landscape scale and of applying silvicultural practices, such as the irregular shelterwood system (Ruel et al. 2007, Raymond et al. 2009), in exploited old-growth forests (> 150 year).These management methods should help retain structural elements essential to the maintenance of biological diversity, such as snags throughout the regeneration cycle.This would maintain nesting and foraging opportunities for cavity nesters, whereas secondary nesters will also benefit from such management.Our results therefore highlight the strong capacity for old-growth forests to provide quality habitat for species dependent on large snags.Likewise, since non-salvaged postfire stands were also characterized by numerous large-diameter snags, clear-cutting with snag retention should help preserve important features of cavity-nesting bird habitats.

Fig. 1 .
Fig. 1.Location of the study area and fires mapped for the period 1800-2000.Fire dates are shown only for fires > 200 km 2 .

Fig. 2 .
Fig. 2. Snag basal area according to the time since last fire.Significant differences (P < 0.05) for means values (± S.E; n = 30 per chronosequence) among years are indicated by different letters.

Fig. 3 .
Fig. 3. Annual tree mortality rates (%) calculated on a basal area basis for each time since fire (TSF) classes and for four tree species groups.Species groups are balsam fir (ABB), black spruce (PIM), intolerant hardwoods (IH) and the other remaining species (OTH).

Fig. 4 .
Fig. 4. Snag availability and cavity numbers following the two chronosequences.The four diagrams show (A) average number of cavities according to the time since fi re; (B) average number of cavities by state of snag decomposition; (C) proportion of trees holding cavities by DBH class; and (D) average number of snags > 20 cm DBH per ha according to time since fi re.

Fig. 5 .
Fig. 5. Average number of bird species using cavities per plot according to the time since last fi re (TSF).Signifi cant differences (P < 0.05) among mean values (± S.E; n = 30 per chronosequence) of TSF classes are indicated by different letters.

Table 1 .
List of cavity-nesting bird species found in the study area.