| Chapitre 2 Are mixedwood forests perceived by birds as a distinct forest type? | ||
|---|---|---|
|
|
|
| Chapitre 2 Are mixedwood forests perceived by birds as a distinct forest type? | ||
|---|---|---|
|
|
|
|
Table des matières
Ce chapitre a été publié en 2004 dans le 34ieme numéro du «Canadian Journal of Forest Research», aux pages 1895 à 1907. Les auteurs sont : Caroline Girard, Marcel Darveau, Jean-Pierre L. Savard et Jean Huot. Quelques modifications ont toutefois été apportées à la présente version en réponse aux demandes du pré-lecteur de la thèse.
Nous avons évalué à trois échelles spatiales l’influence sur la présence des oiseaux forestiers des peuplements de conifères, de feuillus et mixtes à l’intérieur du domaine de la sapinière à bouleau jaune au Québec. La présence de deux, quatre et quatre espèces était reliée positivement à la forêt mixte respectivement aux échelles de 50 m, 100 m et 1000 m. Parmi elles, les parulines à gorge orangée, à collier, bleue et couronnée répondaient à plus d’une échelle. Certaines espèces étaient également influencées par les forêts de conifères et de feuillus aux trois échelles spatiales. La forêt mixte était associée positivement à la présence de plusieurs espèces et était favorisée au détriment des forêts de conifères ou de feuillus dans certains cas. Ces résultats supportent notre hypothèse que la forêt mixte est un habitat distinct et procurent une nouvelle justification pour préserver ce type forestier.
We evaluated at three spatial scales the influence on forest bird species occurrence of coniferous, deciduous and mixedwood stands within the Quebec balsam fir-yellow birch domain. At the 50 m, 100 m and 1000 m scales, the occurrence of two, four and four species respectively was positively related to mixedwood forests. Among them the blackburnian warbler, northern parula, black-throated blue warbler and ovenbird responded at more than one scale. Some species were also influenced by coniferous and deciduous forests at the three spatial scales. Mixedwood forests were positively associated with the occurrence of many species and moreover were favoured over coniferous or deciduous forests by some species. These results support our hypothesis that mixedwood forests are distinct habitats and provide a new justification to preserve this forest type.
Mixedwood forests are recognized for their diverse flora and fauna (Welsh 1987, Baker et al. 1995, Kirk et al. 1996, Niemi et al. 1998). However, because of their heterogeneity and complexity, they have become a challenging issue for scientists and managers (MacDonald 1995, Wang et al. 1995, Grover and Greenway 1999, Welham et al. 2002). In the past, forestry practices used in mixedwood forests often have resulted in the decline of mixedwood stand abundance or in the modification of their composition and consequently have simplified the general forest mosaic. For instance, in boreal mixedwoods usually naturally governed by a fire-dominated disturbance regime, large clearcuts have facilitated an increase in shade intollerant deciduous trees to the detriment of coniferous trees that regenerate after fire (Harvey and Bergeron 1989, Carleton and MacLellan 1994, Jackson et al. 2000). In temperate mixedwoods naturally governed by a tree-fall regenerating system, both partial harvests and clearcuts have been detrimental to the regeneration of the two dominant species, balsam fir ( Abies balsamea (L.) P. Mill.) and yellow birch ( Betula alleghaniensis Britton) and thus have modified considerably the composition of managed stands (Archambault et al. 1998). Moreover, in boreal and temperate mixedwood forests, conifer plantations have often replaced mixedwood stands (Bergeron and Harvey 1997, Grover and Greenway 1999).
Those problems have recently been acknowledged, and scientists and managers have expressed through the first criterion for sustainable management in Canada (Canadian Council of Forest Ministers 2000) the urgency of developing practices that would maintain forest biodiversity. However, to reach this goal within mixedwood forests, research is still needed. Basic processes such as wildlife habitat selection are not well understood and only a few studies, all on boreal birds, have addressed the influence of mixed coniferous-deciduous forests on wildlife distribution. At the landscape scale, Enoksson et al. (1995) showed that for several birds species the presence of deciduous patches within a coniferous forest landscape was important. At the stand scale, Hobson and Bayne (2000) found a greater avian diversity and a higher abundance of certain species in boreal mixedwood stands than in “pure” deciduous or coniferous stands. Kirk et al. (1996) showed that some bird species occurred more frequently in mixedwood stands than in deciduous or coniferous stands. In a managed forest context, Drapeau et al. (2000) reported that bird communities within mixedwood forests ruled by a regime of natural disturbances differed from those within a pre-industrial landscape and from those within an industrial timber landscape dominated by deciduous trees.
However, these investigations were conducted in a boreal environment ruled by a fire-dominated disturbance regime, where forest composition mostly consists of 3-4 tree species often totalling 95% of the basal area. The present paper reports a study conducted in temperate mixedwood forests regulated by a single-tree fall dynamic which have a more complex and stable forest composition than boreal forests (Gosselin 2002). We assessed whether mixedwoods are a significant feature in the selection process of habitat by birds. Because habitat selection is interpreted as a hierarchical process involving decisions at more than one scale (Cody 1985, Morris 1987, Wiens 1989c, Orians and Wittenberger 1991), we documented bird-habitat relationships at three spatial scales. Our goals were to investigate: 1) whether mixedwood stands are perceived by birds as a distinct forest type or simply as the juxtaposition of deciduous and coniferous forests, 2) whether they attract specifically certain species of birds and 3) whether their surface area in the landscape influences bird species distribution.
The study area lies within a 15 × 30 km area overlapping the Réserve faunique de Portneuf and the ZEC Batiscan-Neilson (46o 50’ N – 71o 50’ W), Quebec (Canada). It is characterized by a varied topography with a mean annual temperature of 2.5°C and a mean annual rainfall of 1000 to 1600 mm (Robitaille and Saucier 1998). This area is typical of the balsam fir-yellow birch bioclimatic domain, a mosaic of three forest types: 1) coniferous dominated stands covering 25% of the area, composed mostly of balsam fir, red spruce ( Picea rubens Sarg.), and of a few white birch ( Betula papyrifera Marsh.), 2) deciduous dominated stands covering also 25% of the study area, dominated by sugar maple ( Acer saccharum Marsh.), yellow birch, red maple ( Acer rubrum L.), and American beech ( Fagus grandifolia Ehrh.), and 3) mixedwood stands covering 50% of the area, consisting mostly of yellow birch and balsam fir with maples and red spruce. Most of the area has been partially cut or clearcut over the last century and consequently virgin stands were not available for the study. Nevertheless, about 27% of the productive area is covered by stands over 90-year-old (Darveau, unpublished data).
Figure 2.1 Sampling point locations within the study area, Québec, Canada. Symbols indicate observation points located in coniferous (circles), deciduous (squares), and mixedwood (triangles) stands. Solid lines denote limits of a reserve and of an exploitation zone territory, and the star indicates a town.
We used the forest type classification reported on provincial forest maps to position 57 observation points within mature coniferous (N = 23), deciduous (N = 15) and mixedwood (N = 19) stands that were unmanaged since at least 50 years (Figure 2.1). Sampling points were at least 200 m apart. Because forest maps are not sufficiently accurate at the local stand scale (Potvin et al. 1999, Dussault et al. 2001) we conducted a field validation using the criteria of ecological type identification (Gosselin 2002). As each sampling point (50 m radius) fitted within a stand (minimum area of 3 ha), only one forest type was associated with each point.
At each point, we used forest maps to characterized vegetation present within a radius of 100 m and 1000 m by calculating the total area of unmanaged coniferous, deciduous and mixedwood stands, and managed stands (i.e., deciduous and mixedwood selectively cut stands). Selectively cut stands were considered in the analysis at these scales because they were part of the environment and potentially influenced bird distribution as their structure was similar to that of unmanaged stands (Norton and Hannon 1997). The 100 m radius was used because of the small size of the stands within the study area, the heterogeneity of their distribution, and their probable influence on birds. The 1000 m radius was chosen to give an idea of the mosaic present within the extended environment of the bird territories and the influence of that mosaic on bird distribution. We used Arcview 3.2a software to delineate, on a forest map, buffer areas of 100 m and 1000 m of radius centered on each sampling point. We used SAS 8.2 (SAS Institute Inc 1999-2001) to summarize the forest data of the buffer areas.
In June 2000 and 2001, at each observation point, we surveyed twice (≥ 5 d apart), with different observers, birds seen or heard within a radius of 50 m (Ralph et al. 1993). We listened for a period of 10 min and then played recorded calls during approximately 5 min to stimulate less vocal birds such as woodpeckers, chickadees, nuthatches, and creepers (Boulet et al. 2000). Counts were conducted in the morning, between 4:00 and 10:00 EDT, only when the weather was appropriate (no rain; wind speed < 25 km/h). Observations were repeated at different moments of the morning.
We modeled relationships between bird species occurrences at observation points and in forest types within three radius (50, 100, and 1000 m) using the Genmod procedure of SAS. We performed logistic regressions with repeated measures (i.e., Generalized Estimating Equations). We used an unstructured correlation structure because we had two repeated observations per point, one in 2000 and another in 2001 (Stoke et al. 2000). We used the maximum number of observations obtained during the two surveys to define the occurrence of a species within each year. All models included the inventory year as covariate. Only species occurring in more than 5% of the observation points within the study area were analysed. To check the fit of models, we approximated an R-square from the logistic regression (Hosmer and Lemeshow 2000) based on the log-likelihoods (L) using the equation:
R2 logistic = ((Lreduce – Lfull)/ Lreduce)
where the reduced model is the logistic model with only the intercept and the full model contains all the predictors in the model. As the 100 and 1000 m scales compared several models (see below), we used a model that included the four variables of interest to calculate this logistic R-square.
Because only one forest type was present within the 50 m radius, we used a nominal variable with three classes (unmanaged coniferous, deciduous and mixedwood stands) to characterize the forest at that scale. We set mixedwood stands as the reference level in regressions and used a probability threshold of 5%. We did not use Bonferroni adjustment because we were testing different hypotheses (one per species) (Shaffer 1995) and because we did not want to increase the chance of making a type II error (i.e. declaring no effect while in fact there was one) taking into account the consequences such an error could entail on management decisions (Perneger 1998).
As opposed to the 50 m scale, the 100 and 1000 m scales included several forest types. Thus, we used the model averaging technique of the information-theory approach of Burhnam and Anderson (2002) to analyse relationships between bird species occurrence and the surface area of four forest types (unmanaged coniferous, deciduous, mixedwood, and selectively cut deciduous-mixedwoods). This approach has the advantage to be biologically more informative than the null hypothesis testing paradigm and than the traditional model selection process. Indeed, it does not compare observations to a null hypothesis biologically probably meaningless or false (Yoccoz 1991, Cherry 1998, Anderson et al. 2000). Instead, its comparison is based on the principle of parsimony (favouring the smallest possible number of parameters that adequately represent the data) and on multiple models carefully chosen for their biological relevance (Burnham and Anderson 2002). It also allows to determine the effect of parameters from inferences based on the entire set of models which often have a greater precision and reduced bias than inferences obtained from a single model (Anderson et al. 2000). No colinearity problem was present in the analysis since the four forest types chose did generally not represent 100% of the area.
For each scale we first selected nine models which partially included the four factors considered (Table 2.1) and that were expected to fit bird species occurrence based on behavioural knowledge of the species (e.g. MacArthur et al. 1966; Holmes and Robinson 1981; Gauthier and Aubry 1996). For each model, we calculated the second-order Akaike Information Criterion (AICc) using the model log-likelihood, the model number of parameters (including the intercept), and the sample size ( n ). AICc is an estimator of the strength of evidence of a model; it identifies the best model within the selected set and ranks the others relatively to the “best” model. We also calculated the Akaike weights of each model, which is the relative likelihood of a model to be the best model. We used this value to calculate the model-averaged parameters of the four factors considered in the analysis and their unconditional standard errors. These two values give for each factor present in the set of models their marginal influence on bird occurrence.
Due to the clustered spatial distribution of our data at the 1000 m scale, we tested the spatial independence of the dependant variable for that scale. We used the software Passage (Rosenberg 2001) to compute the autocorrelation coefficient of Moran’s I from the residuals of the model that included the four variables of interest. We used a binary weigh matrix and a normal distribution assumption to calculate coefficients.
We compared the models tested at the different scales to assess the effect of scale on the occurrence of each species. We measured AICc differences (∆AICc) between models to identify the best model within the selected set, the one with the smallest AICc, and to rank the rest of the models relatively to the “best” model. We considered as good approximation of the best model, models with an AICc difference inferior to 4 (Burnham and Anderson 2002).
A total of 54 species of birds were detected within the 57 observation points in 2000 and 2001. Species richness was higher in mixedwood than in deciduous stands (Z = -2.18, P = 0.029) but similar in mixedwood and coniferous stands ( Z = -1.15 , P = 0.251). Thirty two species were detected in more than 5% of the point counts over the two years. The American redstart (see annexe A for the latin names of all studied species) and chestnut-sided warbler were absent from mixedwood stands, as were the bay-breasted warbler, ruby-crowned kinglet, solitary vireo and yellow-bellied flycatcher from deciduous stands, and the downy and hairy woodpeckers from coniferous stands (Table 2.2). The Nashville warbler was absent from both mixedwood and deciduous stands. Of the 32 species recorded, five could not be analysed because their model did not converge.
Only the blackburnian warbler selected mixedwood stands over both deciduous ( Z = -2.52, P = 0.012) and coniferous ( Z = -3.26, P = 0.001) stands. The northern parula followed the same pattern, but with less significant results (deciduous Z = -2.52, P = 0.012; coniferous Z = -1.92, P = 0.055). Two species preferred coniferous stands: the bay-breasted warbler ( Z = 3.07, P = 0.002) and magnolia warbler ( Z = 2.64, P = 0.008), while six avoided them: the black-throated blue warbler ( Z = -3.16, P = 0.002), black-throated green warbler ( Z = -2.38, P = 0.017), brown creeper ( Z = -2.69, P = 0.007), ovenbird ( Z = -4.54, P < 0.001), red-eyed vireo ( Z = -4.24, P < 0.001), and yellow-bellied sapsucker ( Z = -2.28, P = 0.023). Only the least flycatcher preferred deciduous stands ( Z = 1.93, P = 0.053) whereas only the golden-crowned kinglet ( Z = -2.36, P = 0.018) and red-breasted nuthatch ( Z = -3.72, P < 0.001) avoided them. The 15 remaining species did not show any relationship with forest types ( P < 0.05).
Nineteen species had a convergent model predicting species occurrence within the 100 m radius (Table 2.3). No species was strictly associated with the proportion of unmanaged mixedwood cover. However, the blackburnian warbler occurred more frequently when both unmanaged mixedwood and managed covers were abundant. Similarly, the black-throated blue warbler, ovenbird, and red-eyed vireo were more frequent at points where unmanaged mixedwood and deciduous covers were abundant in their vicinity. Perhaps as a consequence of its strong association with mixedwood or deciduous covers, the ovenbird was in fact negatively associated with conifers. Deciduous cover influenced positively the occurrence of the black-throated green warbler and of the yellow-bellied sapsucker, and negatively influenced the occurrence of the golden crowned kinglet. Coniferous cover influenced positively the occurrence of the magnolia warbler and negatively that of the downy woodpecker. Managed stands influenced positively the occurrence of the magnolia warbler and negatively the occurrence of the golden crowned kinglet.
Thirty species analysed had convergent models predicting bird occurrence within the 1000 m radius (Table 2.4). The northern parula occurrence was exclusively and positively influenced by a high proportion of mixedwoods in the forest. The occurrence of other species was positively influenced by mixedwoods and other forest types as well. The black-throated blue warbler, red-eyed vireo, and ovenbird were influenced positively by both mixedwood and deciduous covers while the black-and-white warbler was influenced positively by mixedwood and negatively by coniferous covers. Mixedwood and deciduous covers had a negative influence on the occurrence of the Nashville warbler and ruby-crowned kinglet. The proportion of deciduous cover in the landscape had a positive effect on the occurrence of the brown creeper, least flycatcher, and yellow-bellied sapsucker, and a negative effect on the occurrence of the bay-breasted warbler, golden-crowned kinglet, magnolia warbler, and yellow-bellied flycatcher. The occurrence of the Canada warbler and
Table 2.3: Model-averaged parameter estimates (± unconditional standard error) of the occurrence of 19 species of birds and the area of four forest types (unmanaged mixedwood, unmanaged coniferous, unmanaged deciduous, and managed deciduous-mixed stands) within a 100 m radius of the observation point.
|
|
Table 2.4: Model-averaged parameter estimates (± unconditional standard error) of the occurrence of 30 species of birds and the area of four forest types (unmanaged mixedwood, unmanaged coniferous, unmanaged deciduous, and managed deciduous-mixed stands) within a 1000 m radius of the observation point.
|
white-throated sparrow were negatively influenced by the proportion of both deciduous and coniferous covers. The black-capped chickadee was attracted by the presence of selectively cut stands of deciduous and mixedwood composition (managed stands) in the landscape. Furthermore, none of the other species studied at this scale was negatively associated to these stands.
Autocorrelation coefficients of Moran’s I showed no spatial correlation among species occurrences in observation points in 2000. However, significant ( P < 0.05) spatial correlations were found in 2001 for the black-throated blue warbler, the ruby-crowned kinglet, the ovenbird and the Swainson’s thrush. Results obtained for those species must be interpreted with caution since we are unsure of the spatial independence of the dependant variable.
Seventeen species had convergent models predicting bird occurrence for the three studied scales (Table 2.5). The downy woodpecker, magnolia warbler, red-breasted nuthatch, red-eyed vireo and yellow-bellied sapsucker were sensitive to forest types only at the 50 m scales, whereas the black-capped chickadee, Canada warbler, and white-throated sparrow were sensitive only at the 1000 m scale. The black-throated blue warbler, black-throated green warbler, golden-crowned kinglet, ovenbird, and Philadelphia vireo were sensitive to forest types at both 50 and 100 m scales. The blackburnian warbler, Swainson’s thrush, and winter wren were sensitive to the three studied scales.
Ten species had convergent models at 50 and 1000 m scales only. Among these species, the bay-breasted warbler, brown creeper, hairy woodpecker, ruby-crowned kinglet, and solitary vireo were sensitive to forest types at the 50 m scale and not at the 1000 m one, whereas the least flycatcher, northern parula, and yellow-bellied flycatcher were sensitive at those two scales. The Nashville warbler was sensitive to forest types at the 1000 m scale, the only scale analysed for that species. Five species had the null model (the one including only the regression intercept) as one of the best model predicting bird occurrence. The occurrence of those species, the American redstart, black-and-white warbler, chestnut-sided warbler,
hermit thrush, and purple finch, was not influenced by forest types at any of the studied scales.
We detected relationships between the occurrence of some bird species and the proportion of coniferous or deciduous forest types in the landscape. Moreover, we observed that several species were sensitive to these two forest types at the finer scale only, the associations found at the larger scales being an artefact created by the clustered nature of the analyses and not of the true selection of species. This was the case of the red-breasted nuthatch and ruby-crowned kinglet whose occurrences were negatively associated with deciduous stands, of the brown creeper, downy woodpecker, red-eyed vireo, and yellow-bellied sapsucker whose occurrences were negatively associated with coniferous stands, and of the bay-breasted warbler and magnolia warbler which showed double associations, one negative with deciduous stands and one positive with coniferous stands (Table 2.6). Two species, the black-capped chickadee and Canada warbler showed a relationship with forest types only at the 1000 m scale, the former with managed stands and the latter negatively with both deciduous and coniferous stands (Table 2.6). Others were sensitive to forest types at several scales. Some showed constancy through scales: the golden-crowned kinglet and yellow-bellied flycatcher for their negative relationship with deciduous stands, the least flycatcher for its positive relation with deciduous stands or the ovenbird for its negative relation with coniferous stands (Table 2.6). Others responded to cover differently across scales: the black-throated blue warbler and black-throated green warbler which both showed a negative relationship with coniferous forest type at the 50 m scales and a positive relationship with deciduous forest type at the 100 m scale (Table 2.6). We also detected relationships between some bird species and mixedwood cover at several scales. Among them, some were also influenced by the “pure” forest types (the black-throated blue warbler
Table 2.6: Synthesis of the relationships between bird occurrence and forest type obtained at three studied spatial scales and comparison with observations mentioned in reviews: The Birds of North America published by the American Ornithologist’s Union and The Breeding Birds of Québec published by the Canadian Wildlife Service.
|
|
and ovenbird) and thus did not seem to distinguish mixedwoods from these other forest types.
The blackburnian warbler and northern parula both selected particularly mixedwood stands at many scales (Table 2.6). The former showed selection at small scales: 50 m and 100 m radius but not at the 1000 m scale. The latter showed selection at the 50 m and 1000 m scales. Unfortunately, the tested models did not converge for the northern parula at the 100 m scale.
The high occurrence of the blackburnian warbler and northern parula in mixedwood forests corresponded to general observations mentioned in reviews (Table 2.6). Kirk et al. (1996) also associated the blackburnian warbler to old boreal mixedwood forests while Hobson and Bayne (2000) associated it with pure coniferous forests. Overall, the other bird-habitat relationships observed in our study were also generally consistent with those mentioned in reviews (Table 2.6). Only the black-throated green warbler showed little divergence and avoided coniferous cover at the 50 m scale in our study area, whereas it was observed in that forest type in other studies (Table 2.5). These divergences could be linked to the evolutionary processes of the species (subspecies emergence) (Rice 1987) since these birds have a broad distribution.
Our results show that temperate mixedwood forests are perceived by certain species as a distinct forest type. The selection of mixedwood stands was found at the 50 m scale, indicating that mixedwood forests are not only a simple mosaic of coniferous and deciduous forest types where coniferous and deciduous bird communities meet, but instead a distinct habitat with its own bird community. Moreover, the influence of mixedwood cover was also found at larger scales, indicating that this forest type is also a significant feature of the habitat selection process at the neighbourhood and landscape scales. Two species were specifically associated with mixedwood forest types through scales and consequently were identified as mixedwood specialists (the blackburnian warbler and northern parula). The preference for mixedwood cover by these species may be related to their ability to use both coniferous and deciduous trees for nesting and foraging (Holmes and Robinson 1981, Morse 1994, Moldehauer and Regelski 1996). Moreover, these species have been observed in a wide range of habitats (Morse 1994, Moldehauer and Regelski 1996), which suggests some plasticity (Inman et al. 1987, Whelan 1989, Parrish 1995b) in their ability to exploit forests with complex structure and composition similar to that found in mixedwood stands.
Our study tested the potential influence of the mixed composition of mixedwood forests on species occurrence. The next step would be to measure survival rates or breeding success in those forests as well as the use of the different tree types by the birds. We could also consider the potential influence on birds of the spatial configuration of the coniferous-deciduous mixture. Penhollow and Stauffer (2000) demonstrated that at large scales, the distribution of different islands of vegetation in heterogeneous habitats may influence the abundance of certain species in the landscape.
The perception by birds of temperate mixedwood forests as a distinct habitat provides a new justification for the maintenance of the coniferous-deciduous mixture within the mixedwood forest. Although preserving this mixture is encouraged by “new forestry” principles and by voluntary environmental certification policies (Canadian Council of Forest Ministers 2000), management practices have not yet been modified accordingly. Efforts should be devoted to evaluate different harvesting strategies, such as group-selection tested in our study area, and to examine their efficiency in facilitating stand regeneration to pre-harvest composition. Preliminary studies already suggest that the group-selection cutting, which allows a harvest of groups of trees in patches that will be large enough to create gaps and favour the regeneration of target species such as yellow birch, could maintain the bird community of the mixedwood stands. In our study, few bird species responded negatively to the presence of managed stands in their vicinity. Moreover, Norton and Hannon (1997) noted that partial-cut logging in boreal mixedwood forests, keeping 60% and more of residual forest, can maintain, at lower densities, species usually intolerant to clear-cuts. Other silvicultural methods that consider the spatial distribution of residual coniferous and deciduous trees also should be evaluated (Schieck et al. 2000). However, caution is required in interpreting the efficiency of mixedwood harvesting methods until survival rate and breeding success of associated species are evaluated.
|
|
|
|
| Chapitre 1 Regional influence of mixedwood forest on bird distribution in Québec and Ontario |
|
Chapitre 3 Does tree species preference of passerine birds vary across forest types? |
