THE EFFECTS OF SITE PRODUCTIVITY AND RICHNESS

THE EFFECTS OF SITE PRODUCTIVITY AND RICHNESS
THE EFFECTS OF SITE PRODUCTIVITY AND
HETEROGENEITY ON BIRD HABITAT QUALITY AND SPECIES
RICHNESS
By
Neal Phillip Perry Simon
B.Sc., Memorial University of Newfoundland, 1996
M.Sc.F. University of New Brunswick, 1998
A Dissertation Submitted in Partial Fulfillment of
the Requirements for the Degree of
Doctor of Philosophy
in the Graduate Academic Unit of
Forestry and Environmental Management
Supervisor:
A.W. Diamond, Ph.D., Biology/Forestry
Examining Board:
Gwendolyn Davies, Ph.D., Graduate Studies, Chair
Thom Erdle Ph.D., Foresty
Jeff Houlahan, Ph.D., Biology
Van Lantz, Ph.D., Forestry
External Examiner:
Pierre Drapeau, Ph.D., Biology, Université du Québec
à Montréal
This dissertation is accepted by the
Dean of Graduate Studies
THE UNIVERSITY OF NEW BRUNSWICK
November 2005
© Neal Philip Perry Simon, 2006
Abstract
I assessed factors affecting timber productivity, plant abundances and
their effects on songbird habitat quality and species richness at 220 sites in
central Labrador. Arthropod abundance, landscape content, heterogeneity and
microclimate (e.g., temperature, radiation) were also examined as potential
influences on bird habitat quality and richness. Elevation, slope and drainage
were dominant features influencing timber volume and plant abundances.
Timber volume and richness of understory vegetation increased with elevation
and slope and were negatively associated with rapidly drained soil. Hermit
thrush (Catharus ustulatus) abundance was inversely related to timber volume.
The remaining 14 bird species showed few relationships with vegetation but,
where relationships existed, landscape variables were usually more important
than local variables.
Site occupation was inconsistent between years suggesting that birds did
not deterministically choose ‗optimum‘ breeding territories, but instead,
stochastically occupied them when confronted with sites that are overall
satisfactory in quality. I attribute this, in part, to birds having difficulty assessing
resources within a site since arthropod biomass varied between years and within
the breeding season. The high yearly variance in total bird abundance relative to
the sum of the variances of all species signifies parallel fluctuations in species
densities indicating low site-specific competition. Low site-specific competition
can broaden species‘ niches, potentially contributing to poor bird-habitat
relationships.
ii
My only observed bird species richness-productivity patterns occurred at
the stand grain and showed 3 different patterns: linear decline, a linear increase
and curvilinear decline with biomass, temperature and radiation respectively. A
covariation between heterogeneity and productivity may explain the richnessvolume and richness-temperature curves but not the richness-radiation curve. I
suggest that conflicting richness-productivity patterns in the literature could result
from different patterns of covariation between heterogeneity and common
productivity surrogates where richness is driven by heterogeneity.
iii
Acknowledgements
I thank Directors, Ken Colbert, Keith Deering and Len Moores from the
Newfoundland and Labrador Department of Natural Resources for allowing me to
attend graduate school while employed as the Regional Ecologist. The
Department also provided support, both cash and in-kind to implement the study
and attend committee meetings. Human Resources and Skills Development
Canada provided funding for summer student assistants and the Innu Nation
provided in-kind support. Thanks are extended to all those who assisted in data
collection, notably, B. Denty, D. Blake, B. Campbell, J. Colbert, L. Elson, R.
Flynn, B. Hope, M. Michelin, K. Mitchell, R. Neville, F. Phillips, B. Rodrigues, F.
Taylor. D. Goulding and D. Jennings assisted in GIS while administrative staff,
P. Churchill, R. Hope, M. Johnson and G. Tee, assisted with purchases and
student hiring.
Sincere thanks are extended to my supervisor, Dr. Tony Diamond for his
encouragement and guidance throughout the entire project. I also thank Drs.
Graham Forbes and Dan Keppie for being on my committee and for giving
suggestions and criticisms throughout the project. Fellow graduate students,
Matt Betts, Keith Chaulk and Joe Nocera, provided excellent suggestions,
criticisms and ideas throughout all stages of my degree. Dr. Charles Bourque
provided advice and a program to calculate solar radiation and temperature.
Bruce Roberts critiqued drafts and provided advice on chapter 2.
iv
In addition to her role in data collection, entry and document formatting, I
also thank my fiancée Leanne Elson for being there and providing
encouragement throughout my degree.
v
Table of Contents
Abstract .............................................................................................................. ii
Acknowledgements ........................................................................................... iv
Table of Contents .............................................................................................. vi
List of Tables ................................................................................................... viii
List of Figures ...................................................................................................... xi
Chapter 1 – General Introduction ........................................................................ 1
Introduction.................................................................................................... 1
References ..................................................................................................... 8
Chapter 2 - Evaluation of environmental factors influencing vegetation structure
in mature Picea mariana forests using constrained ordination and constrained
classification. ..................................................................................................... 19
Abstract ........................................................................................................ 20
Introduction.................................................................................................. 21
Study Area.................................................................................................... 23
Methods ........................................................................................................ 24
Results ......................................................................................................... 27
Ordination .................................................................................................. 27
Multivariate regression trees ...................................................................... 28
Discussion ................................................................................................... 30
Acknowledgements ..................................................................................... 36
References ................................................................................................... 36
Chapter 3 - Songbird habitat quality across a timber productivity gradient within
an unfragmented northern boreal forest: local and landscape effects. ............... 49
Abstract ........................................................................................................ 50
Introduction.................................................................................................. 51
Study Area.................................................................................................... 55
Methods........................................................................................................ 56
Bird census ................................................................................................ 56
Local Vegetation ........................................................................................ 58
Landscape variables .................................................................................. 59
Detectability ............................................................................................... 59
Bird vegetation models ............................................................................... 61
Post-hoc exploratory data analysis ............................................................. 66
Results ......................................................................................................... 66
Relationships between local vegetation and forest types ............................ 66
Bird Detectability ........................................................................................ 67
Effect of forest type and vegetation variables on bird abundance ............... 67
Effect of forest type and vegetation variables on bird site reoccupancy and
reproductive activity.................................................................................... 69
Post-hoc exploratory analysis ..................................................................... 69
Discussion ................................................................................................... 69
Acknowledgements ..................................................................................... 74
References ................................................................................................... 75
vi
Chapter 4 – The contribution of climate, arthropods and species interactions in
patterns of bird occupancy and habitat quality. ................................................ 101
Abstract ...................................................................................................... 102
Introduction................................................................................................ 103
Study Area.................................................................................................. 106
Methods ...................................................................................................... 107
Microclimate variables .............................................................................. 108
Arthropod sampling .................................................................................. 109
Bird relationships with microclimate variables .......................................... 109
Arthropod-forest type relationships ........................................................... 112
Community stability and competition ........................................................ 112
Results ....................................................................................................... 113
Bird-microclimate relationships ................................................................. 113
Arthropod Mass ........................................................................................ 114
Community stability and competition ........................................................ 114
Discussion ................................................................................................. 115
Acknowledgements ................................................................................... 121
References ................................................................................................. 121
Chapter 5 – An evaluation of the environmental heterogeneity hypothesis and its
relation to site productivity. .............................................................................. 137
Abstract ...................................................................................................... 138
Introduction................................................................................................ 139
Study area .................................................................................................. 143
Methods ...................................................................................................... 144
Bird survey ............................................................................................... 144
Vegetation survey .................................................................................... 145
Temperature and Solar Radiation............................................................. 145
Data Analysis ........................................................................................... 146
Results ....................................................................................................... 151
Discussion ................................................................................................. 152
Acknowledgements ................................................................................... 156
References ................................................................................................. 156
Chapter 6 – General discussion and conclusions. .......................................... 173
References ................................................................................................. 179
vii
List of Tables
Table 2.1. Vegetation species and codes .................................………………...42
Table 2.2. Descriptions and codes of environmental variables used
in redundancy analysis and multivariate regression tree models............……….43
Table 2.3. Test statistics of environmental variables in order of their inclusion
into stepwise redundancy analysis…………………………………………………..44
Table 3.1. Descriptions and codes of variables used in bird-habitat
models……………………………………………………………………………….....86
Table 3.2. The effect of forest type on songbird detectability as determined by
SURVIV...........................................................………………..............................87
Table 3.4. Best models (i.e., lowest AIC) and partial R2e due to local and
landscape variables as determined by logistic and autologistic regression for bird
species-years where R2e ≥ 0.15 for the global model..........……………………...88
Table 4.1. Descriptions and codes of variables used in bird-habitat models
from Chapter 3 ................................................................................................130
viii
Table 4.2. Variance explained by environment for yearly models of bird
abundance, reproductive activity and site reoccupancy in relation to predicted
vegetation features...........................................................................................131
Table 4.3. Nested models of arthropod mass with associated Akaike‘s
Information Criteria (AIC), model weights calculated by linear mixed effects
models using maximum likelihood....................................................................132
Table 4.4. The importance of microclimate for bird-habitat quality
determined by logistic and autologistic regression models..............................133
Table 5.1. General description of dominant local vegetation (crown cover) and
landscape composition (% of circular buffer of varying radii) at 3 grain
sizes.................................................................................................................164
Table 5.2. Scale-specific correlations (Pearsons) between the 3 productivity
measures, volume (V), solar radiation (SR), Temperature (T).........................165
Table 5.3. The effects of volume on heterogeneity [foliage height diversity (FHD)
and horizontal diversity (HD) and their interaction (FHD:HD)] and species
richness as determined by least-squares and Poisson (species richness)
regression.........................................................................................................166
ix
Table 5.4. The effects of mean daily temperature (T) on heterogeneity [foliage
height diversity (FHD), horizontal diversity (HD) and their interaction (FHD:HD)]
and species richness as determined by least-squares and Poisson (species
richness) regression..........................................................……………..............167
Table 5.5. The effects of mean daily solar radiation (SR) on species richness
and heterogeneity [foliage height diversity (FHD), horizontal diversity (HD) and
their interaction (FHD:HD)], as determined by Poisson (species richness) and
least-squares regression..........................................……………………………..168
Table 5.6. Poisson regressions of species richness predicted by structural (FHD,
HD) and landscape diversity..............................................................................169
x
List of Figures
Figure 1.1. Conceptual framework of the abiotic and biotic relationships of plants
and songbirds considered in this dissertation…..............…………………………17
Figure 1.2. Map of Atlantic Canada showing study area.....................................18
Figure 2.1. Dominant environmental vegetation gradients as determined by
principal component analysis…………….............................................................45
Figure 2.2. Plant-environmental relationships as assessed by stepwise
redundancy analysis...........................………………………………………………46
Figure 2.3. Vegetation groups in relation to environmental variables as
determined by a multivariate regression tree.................…………………………..47
Figure 3.1. Mean and standard error of vegetation features in relation to forest
type.......................................…………………………….……………......………..89
Figure 3.2. Species specific detection probabilities and standard errors over a
10 minute point count.…..............................………………………..…………......90
Figure 3.3. Bird abundance and SE in relation to forest type...........................91
xi
Figure 3.4. Reproductive activity and site occupancy (probability, SE) in
relation to forest type..........................................................................................92
Figure 4.1. Arthropod biomass (mean, SE) in relation to branch height...........134
Figure 4.2. Arthropod biomass (mean, SE) from water traps in relation to
collection week and forest type..........................................................................135
Figure 5.1. Relationships between timber volume, vegetation heterogeneity and
songbird species richness..........……………………………………………………170
Figure 5.2. Relationships between temperature, heterogeneity
and songbird species richness...……………………………………………….......171
Figure 5.3. Relationships between mean dailt solar radiation, vegetation
heterogeneity and songbird species richness...................................................172
xii
Chapter 1 – General Introduction
Introduction
The term ecology originated from the Greek oikos meaning ‗household‘,
‗home‘, or ‗place to live‘. Since Ernst Haeckel first defined ecology as ―the body
of knowledge concerning the economy of nature—the investigation of the total
relationships of the animal to its inorganic and organic environment…‖ several
variations of this original definition have been proposed (Smith 1996, p 4). Smith
(1996) criticized many of these definitions, including the original, as either too
restrictive or too vague. He proposed a wider working definition in which
―Ecology is the study of structure and function of nature. Structure includes the
distribution and abundance of organisms as influenced by the biotic and abiotic
elements of its environment; and function includes all aspects of growth and
interaction‖ (Smith 1996, p 4) As ecology covers a wide range of disciplines,
e.g., from marine to terrestrial, bacteria to mammals, individuals to ecosystems, it
involves, somewhat by necessity, isolated groups of specialists. In recognizing
this need, the British Ecological Society attempted to centralize animal ecology
(Anonymous 1932) by creating the Journal of Animal Ecology. The Journal of
Ecology, from which the Journal of Animal Ecology was derived, became focused
on plants. Smith (1996) criticized this separation, arguing that it reduced
communication within ecology. Insufficient communication between ecological
disciplines has been a frequent complaint among ecologists (Macfadyen 1975,
Austin 1985, Weiner 1995, Ford and Ishii 2001) (see however Nobis and
Wohlgemuth 2004), but this complaint is not new. Moore (1920) emphasized the
1
synthesis of various ecological disciplines and challenged ecologists to think
outside one‘s specific field when he described the scope of the Ecological
Society of America‘s ‗new‘ journal Ecology. As Moore (1920) suggested, human
limitations require ecologists to focus on a particular ecological discipline, but one
should take a broad ecological view of its issues.
An enormous amount of ecological understanding has resulted from
research on birds (Newton 1995). This is because birds are relatively easy to
identify, observe, census, capture and mark (Newton 1995, Hutto 1998, Venier
and Pearce 2004). In terrestrial systems, birds, particularly songbirds, have been
recommended as indicators of environmental change (Hutto 1998, Venier and
Pearce 2004) and form the basis of many extensive and long-term monitoring
programs e.g., the breeding bird survey (Sauer et al. 2004). As a result,
songbird-habitat relationships have been studied extensively. Traditionally,
vegetation variables, e.g., plant species occurrence, stem densities, stratification
and needle architecture have been used to predict songbird abundance, foraging
and nest sites (Lack 1933, MacArthur and MacArthur 1961, Roth1976, Franzreb
1978, Morse 1989, Parrish 1995). Forest fragmentation concerns (Hagan et al.
1996, Saunders et al. 1991, Wade et al. 2003) and the recognition that habitat
selection occurs over several spatial scales (Wiens 1989, Wiens et al. 1993,
Lichstein et al. 2002a) have led research to the landscape scale (i.e., extents up
to several kilometres for songbirds). Still, variables used to predict bird
abundances are almost exclusively vegetation that result from different intensities
and ages of large scale disturbances, usually fire or logging (Schwab and Sinclair
2
1994, Hagan et al. 1996, Drapeau et al. 2000, Simon et al. 2000, Simon et al.
2002). Though understanding the effects of large-scale disturbances, particularly
anthropogenic disturbances, are critical for conservation, the focus on
disturbance-induced vegetation change may be deflecting attention from other
habitat variables that could be important for birds.
Fire, timber harvesting, and to lesser extents insects and wind, are the
most commonly studied agents of vegetation change for boreal and temperate
bird-habitat studies. However, abiotic factors, (e.g., edaphic and topographic), as
determinants of plant community structures are common in the plant literature
(Foster 1984, Foster and King 1986, Meades and Roberts 1992, Wang and
Klinka 1996, Wang 2000) but rarely considered by bird ecologists (however, see
Folkard and Smith 1995, Seagle and Sturtevant 2005). Disturbance age and
intensity are sometimes the only descriptors reported by bird ecologists in
disturbance and succession studies (e.g., Derleth et al. 1989, Patterson et al.
1995). The failure to include detailed vegetation descriptions in such studies is
problematic because age is taken to be a surrogate for vegetation structure, but
that structure varies with edaphic and climatic conditions (Foster 1984, Foster
and King 1986, Meades and Roberts 1992, Oliver and Larsen 1996, Ohmann
and Spies 1998). This information is critical to interpretation and failing to include
detailed vegetation descriptions of sites makes it difficult to explain discrepancies
in results of bird-disturbance studies conducted under different conditions.
Further, detailed vegetation descriptions are critical in studies attempting to
3
landscape level habitat selection because it is necessary to partial out local
vegetation effects to isolate critical landscape features.
Apart from their direct effects on vegetation structure, there are good
reasons to consider abiotic factors in bird habitat selection studies. Ecologists
have known for at least 30 years that species diversity is related to ecosystem
productivity, i.e., the amount of energy moving through a system. Since Connell
and Orias (1964) proposed productivity as an explanation for the
tropic/temperate diversity gradient, several researchers have found correlations
between bird diversity and different surrogates of productivity at large spatial
extents (Nilsson and Nilsson 1978, Nilsson 1979, Currie 1991, Wright et al.
1993). In fact, considering only the abiotic environment, Currie (1991) was able
to account for nearly 80 % of the regional variation in bird species richness.
Similarly, O‘Connor et al. (1996) determined that climate was more important
than land cover metrics in predicting bird species richness at a continental scale.
Northern boundaries of bird geographic ranges are determined almost
exclusively by climate (Root 1988, Venier et al. 1999, Venier et al. 2004,
Böhning-Gaese and Lemoine 2004) and climatic influences on habitat selection
tend to be greater near geographic range boundaries (Williams et al. 2003). At
smaller spatial extents, microclimate determines nest placement (Walsberg 1981,
With and Webb 1993). At intermediate extents, weather influences feeding
behaviour in winter and spring (Wachob 1996, Smith et al. 1998) and forest
productivity gradients influence invertebrate biomass which in turn affects bird
reproduction (Seagle and Sturtevant 2005). Further, Smith et al. (1998)
4
suggested that early arriving migrants experience food limitation that may be
offset by foraging in warmer microclimates producing abundant food. Similarly,
elevation has been linked to bird population trends and habitat selection (James
et al. 1996, Lichstein et al. 2002a). This has lead to recommendations towards
the inclusion of abiotic factors such as climatic parameters and elevation in birdhabitat quality models (James et al. 1996, Irwin 1998, Martin 2001).
Habitat quality is a measure of the importance of a habitat type for an
individual‘s fitness (Van Horne 1983). The assumption that density reflects
habitat quality may be valid where an ―ideal free‖ distribution occurs (Fretwell and
Lucas 1970). However, factors such as habitat patchiness, territoriality, site
tenacity and migratory behaviour can lead bird distributions away from an ideal
free distribution (Van Horne 1983; Pulliam 1988; Pulliam and Danielson 1991;
Vickery et al. 1992). Similarly, ideal free models imply strong deterministic
habitat selection, an assumption criticized for northern birds (Haila et al. 1996).
Despite knowledge of these factors, density remains the dominant measure of
habitat quality because comparing fitness between sites requires long-term
population monitoring unfeasible for most studies, particularly those of a large
scale. If several measures of habitat quality (including density) show similar
patterns, researchers can be more confident in their assessment of habitat
quality. Discrepancies between measures could provide insight to different
population processes (e.g., source and sink dynamics). Many theories of habitat
selection imply animals occupy the highest quality habitats more consistently
(Sutherland 1997, Rodenhouse et al. 1997). Similarly, breeding site return rates
5
are often (though not always) higher for successful than unsuccessful breeders
(Haas 1997, Howlett and Stutchbury 2003, Porneluzi 2003). This indicates that
the frequency that a site is occupied would be a good indicator of habitat quality
(e.g., Hames et al. (2001)). In addition, some researchers recommend including
indices of breeding behaviour into studies of bird habitat quality (e.g., Vickery et
al. 1992, Gunn et al. 2000).
Abiotic factors such as climate can also affect habitat selection indirectly.
A harsh climate can reduce species densities, reducing competition which can
weaken bird-vegetation relationships (Rosenzweig 1991, Hailia et al. 1996).
Haila et al. (1996) found stochastic rather than deterministic territory selection in
a landscape where differences in forest types were subtle gradients. Haila et al.
(1996) further suggested that stochastic territory selection could result from
violations in assumptions of optimality theories of habitat selection (summarized
in Sutherland 1997): resource levels adjust immediately to consumer population
levels and the consumers recognize these levels. Because of the short breeding
season, northern songbirds acquire territories in the partially snow-covered
spring and may be unable to assess resource levels accurately. In such timelimited breeding environments, habitat selection may be driven by heterospecific
attraction: later establishing species (migrants) use the presence of earlier
established species (residents) as cues to profitable breeding sites (Mönkkönen
et al. 1997, 1999, Forsman et al. 1998, Thomson et. al. 2003). This process may
also operate within species, i.e., conspecific attraction, where later arriving
6
migrants use the presence of earlier arrivals within the same species to indicate
profitable breeding sites (Lichstein et al. 2002b, Ward and Schlossberg 2004).
In this dissertation, I take a broad ecological approach to assessing bird
habitat quality in mature forests (Fig. 1.1) of central Labrador (Fig. 1.2). The
dissertation is written in articles format, with 4 main chapters linked with General
Introduction (Chapter 1) and General Discussion and Conclusions chapters
(Chapter 6). The principle focus of the thesis is in Chapter 3 and is an
investigation of the effects of site productivity (i.e., the site‘s ability to produce
timber) on bird habitat quality. Within a successional stage, stem densities,
canopy cover and arthropod biomass, per unit foliage should increase with site
productivity, leading to the prediction that habitat quality for canopy and stem
feeding species will increase with productivity. This prediction is contingent on
the assumption that birds have sufficient knowledge about the resources to
select the ‗optimum‘ habitat (Rosenzweig 1991, Haila et al. 1996). This
assumption may not be valid where the breeding season is short (Haila et al.
1996), resources vary erratically (Wiens 1985) or subtly along gradients (Haila et
al. 1996). Under these conditions, I predict selection patterns that correlate
poorly with resources and sites occupied are unpredictable among years; this is
assessed in Chapter 4. An additional focus of the thesis is in Chapter 5 which
investigates the covariation between heterogeneity and productivity and how the
productivity-heterogeneity relationship may be a driver for the richnessproductivity relationship, dubbed the ‗environmental heterogeneity hypothesis‘.
Under this hypothesis, I predict: (1) the shape of the productivity-richness curve
7
is determined by the shape of the productivity-heterogeneity curve, (2) bird
species richness increases with heterogeneity (3) heterogeneity explains more of
the variation in species richness than productivity. A common feature of these
three main chapters is the effect of vegetation structure on birds. Chapter 2
assesses the importance of different soil and topographic features in creating
different vegetation structures that likely affects bird habitat quality and species
richness.
References
Anonymous. 1932. Forward. – J. Anim. Ecol. 1: 1-2.
Austin, M.P. 1985. Continuum concept, ordination methods, and niche theory. –
Ann. Rev. Ecol. Syst. 16:39-61.
Böhning-Gaese, K. and Lemoine, N. 2004. Importance of climate change for the
ranges, communities and conservation of birds. – Adv. Ecol. Res. 35:211236.
Connell, J.H. and Orias, E. 1964. The ecological regulation of species diversity. –
Am. Nat. 98:399-414.
Currie, D.J. 1991. Energy and large-scale patterns of animal and plant-species
richness. – Am. Nat. 137:27-49.
Derleth, E.L., McAuley, D.G. and Dwyer, T.J. 1989. Avian community response
to small-scale habitat disturbance in Maine. – Can. J. Zool. 67:385-390.
Drapeau, P., Leduc, A., Giroux, J-. F., Savard, J-. P.L., Bergeron, Y. and Vickery,
W. L. 2000. Landscape-scale disturbances and changes in bird
communities of boreal mixed-wood forests. – Ecol. Monogr. 70: 423-444.
8
Folkard, N.F.G. and Smith, J.N.M. Smith. 1995. Evidence for bottom-up effects
in the boreal forest: Do passerine birds respond to large-scale
experimental fertilization? – Can. J. Zool. 73: 2231-2237.
Ford and Ishii 2001. The method of synthesis in ecology. – Oikos 93: 153-160.
Forsman, J.T., Mönkkönen, M., Helle, P. and Inkeröinen, J. 1998. Heterospecific
attraction and food resources in migrants‘ breeding patch selection in
northern boreal forest. – Oecologia 115:278-286.
Foster, D.R. 1984. Phytosociological description of the forest vegetation of
southeastern Labrador. – Can. J. Bot., 62: 899-906.
Foster, D.R. and King, G.A. 1986. Vegetation pattern and diversity in S.E.
Labrador, Canada: Betula papyrifera (birch) forest development in relation
to fire history and physiography. – J. Ecol. 74: 465-483.
Franzreb, K.E. 1978. Tree species used by birds in logged and unlogged mixedconiferous forests. – Wilson Bull. 90:221-238.
Fretwell, S.D. and Lucas, H.L. 1970. On territorial behaviour and other factors
influencing habitat distribution in birds. – Acta Biotheor. 19:16-36.
Gunn, J.S., DesRochers, A., Villard, M.-A. and Ibarzabal, J. 2000. Playbacks of
mobbing calls of black-capped chickadees as a method to estimate
reproductive activity of forest birds. – J. Field Ornithol. 71: 472-483.
Haas, C.A. 1997. What characteristics of shelter belts are important to breeding
success and return rates of birds? – Am. Midl. Nat. 137:225-238.
Hagan, J.M., Vander Haegen, M.W., and Mckinley, P.S. 1996. The early effects
of forest fragmentation on birds. – Conserv. Biol. 10:188-202.
9
Haila, Y., Nicholls, A.O., Hanski, I. and Raivio, S. 1996. Stochasticity in bird
habitat selection: year-to-year changes in territory locations in a boreal
forest bird assemblage. – Oikos 76:536-552.
Hames, R.S., Rosenberg, K.V., Lowe, J.D. and Dhondt, A.A. 2001. Site
reoccupation in fragmented landscapes: testing predictions of
metapopulation theory. – J. Anim. Ecol. 70:182-190.
Howlett, J.S. and Stutchbury, B.J.M. 2003. Determinants of between-season
site, territory, and mate fidelity in hooded warblers (Wilsonia citrina). – Auk
120:457-465.
Hutto, R.L. 1998. Using landbirds as an indicator species group. – In: Marzluff,
J.M. and Sallabanks, R. (eds.), Avian Conservation. Research and
Management. Island Press, pp. 75-92.
Irwin, L.L. 1998. Abiotic influences on bird-habitat relationships. – In: J. Marzluff
and Sallabanks, R. (eds.), Avian conservation and management, Island
Press, Washington, pp. 209-217.
James, F.C., McCulloch, C.E. and Wiedenfeld, D.A., 1996. New approaches to
the analysis of population trends in landbirds. - Ecology 77:13-27.
Lack, D. 1933. Habitat selection in birds. – J. Anim. Ecol 2:239-262.
Lichstein, J. W., Simons, T.R., and Franzreb, K.E. 2002a. Landscape effects on
breeding songbird abundance in managed forests. – Ecol. Appl. 12:836857.
10
Lichstein, J. W., Simons, T.R., Shriner, S.A. and Franzreb, K.E. 2002b. Spatial
autocorrelation and autoregressive models in ecology. – Ecol. Monogr.
72:445-463.
MacArthur, R.H. and MacArthur, J.W. 1961. On bird species diversity. – Ecology
42:594-598.
MacFayden, A. 1975. Some thoughts on the behaviour of ecologists. – J. Ecol.
63: 379-391.
Martin, T.E. 2001. Abiotic vs. biotic influences on habitat selection of coexisting
species: climate change impacts? – Ecology 82:175-188.
Meades, W.J. and Roberts, B.A. 1992. A review of forest site classification
activities in Newfoundland and Labrador. – For. Chron. 68: 25-33.
Mönkkönen, M., Härdling, R., Forsman, J.T. and Tuomi, J. 1999. Evolution of
heterospecific attraction: using other species as cues in habitat selection.
– Evol. Ecol. 13: 91-104.
Mönkkönen, M., Helle, P., Niemi, G.J. and Montgomery, K. 1997. Heterospecific
attraction affects community structure and migrant abundances in northern
breeding bird communities. – Can. J. Zool. 75: 2077-2083.
Moore, B. 1920. The scope of ecology. – Ecology 1:3-5.
Morse, D.H. 1989. American warblers: An ecological and behavioral perspective.
– Harvard University Press, Cambridge.
Newton, I. 1995. The contribution of some recent research on birds to ecological
understanding. – J. Anim. Ecol. 64:675-696.
Nilsson, S.G. 1979. Density and species richness of some forest bird
11
communities in south Sweden. – Oikos 33:392-401.
Nilsson, S.G. and Nilsson, I.N. 1978. Breeding bird community densities and
species richness in lakes. – Oikos 31:214-221.
Nobis, M. and Wohlgemuth, T. 2004. Trend words in ecological core journals
over the last 25 years (1978-2002). – Oikos 106:411-421.
O‘Connor, R.J., Jones, M.J., White, D., Hunsaker, C., Loveland, T., Jones, B.
and Preston, E. 1996. Spatial partitioning of environmental correlates of
avian biodiversity in the conterminous United States. – Biodiversity Letters
3:97-110.
Ohmann, J.L., Spies, T.A. 1998. Regional gradient analysis and spatial pattern
of woody plant communities of Oregon forests. – Ecol. Monogr. 68:1151182.
Oliver, C.D., and B.C. Larson. 1996. Forest stand dynamics. – John Wiley &
Sons.
Parrish, J.D. 1995. Effects of needle architecture on warbler habitat selection in
a coastal spruce forest. – Ecology 76:1813-1820.
Patterson, I.J., Ollason, J.G. and Doyle, P. 1995. Bird populations in upland
spruce plantations in northern Britain. – For. Ecol. Manage. 79:107-131.
Porneluzi, P.A. 2003. Prior breeding success affects return rates of territorial
male ovenbirds. – Condor 105:73-79.
Pulliam, H.R. 1988. Sources, sinks and population regulation. – Am. Nat.
132:652-661.
12
Pulliam, H.R. and Danielson, B.J. 1991. Sources, sinks, and habitat selection: a
landscape perspective on population dynamics. – Am. Nat. 137: S50-S66.
Rodenhouse, N.L. Sherry, T.W. and Holmes, R.T. 1997. Site-dependent
regulation of population size: A new synthesis. – Ecology 78:2025-2042.
Root, T. 1988. Energy constraints on avian distributions and abundances. –
Ecology 69:330-339.
Rosenzweig, M.L. 1991. Habitat selection and population interactions: The
search for mechanism. – Am. Nat. 137:S5-S28.
Roth, R.R. 1976. Spatial heterogeneity and bird species diversity. – Ecology
57:773-782.
Sauer, J. R., Hines, J.E. and Fallon, J. 2004. The North American Breeding Bird
Survey, Results and Analysis 1966 - 2003. Version 2004.1. – US USGS
Patuxent Wildlife Research Center, Laurel, MD. [URL: http://www.mbrpwrc.usgs.gov/bbs /bbs.html].
Saunders, D.A., Hobbs, R.J. and Margules, C.R. 1991. Biological consequences
of ecosystem fragmentation: A review. – Conserv. Biol. 5:18-32.
Schwab, F.E. and Sinclair, A.R.E. 1994. Biodiversity of diurnal breeding bird
communities related to succession in the dry Douglas-fir forests of
southeastern British Columbia. – Can. J. For. Res. 24:2043-2040.
Seagle, S.W. and Sturtevant, B.R. 2005. Forest productivity predicts invertebrate
biomass and ovenbird (Seiurus aurocapillus) reproduction in Appalachian
landscapes. – Ecology 86: 1531-1539.
13
Simon, N.P.P., Schwab, F.E. and Diamond A.W. 2000. Patterns of bird
abundance in relation to logging in western Labrador. – Can. J. For. Res.
30: 257-263.
Simon, N.P.P., Schwab, F.E. and Otto, R.D. 2002. Songbird abundance in clearcut and burned stands: a comparison of natural disturbance and forest
management. – Can. J. For. Res. 32:1343-1350.
Smith, R.L. 1996. Ecology and Field Biology, fifth edition. – HarperCollins.
Smith, R., Hamas, M., Dallman, M. and Ewert, D. 1998. Spatial variation in
foraging of the black-throated green warbler along the shoreline of
northern Lake Huron. – Condor 100:474-484.
Sutherland, W.J. 1997. From individual behaviour to population ecology. Oxford
Series in Ecology and Evolution. – Oxford University Press.
Thomson, R.L., Forsman, J.T. and Mönkkönen, M. 2003. Positive interactions
between resident birds: testing the heterospecific attraction hypothesis. –
Oecologia 134:431-438.
Van Horne, B. 1983. Density as a misleading indicator of habitat quality. – J.
Wildl. Manag. 47:893-901.
Venier, L.A. and Pearce, J.L. 2004. Birds and indicators of sustainable forest
management. – For. Chron. 80:61-66.
Venier, L.A., McKenney, D.W., Wang, Y. and McKee, J. 1999. Models of largescale breeding-bird distribution as a function of macro-climate in Ontario,
Canada. – J. Biogeogr. 26: 315-328.
14
Venier, L.A., Pearce, J.L., Mckee, J.E., McKenney, D.W. and Niemi, G.J. 2004.
Climate and satellite-derived land cover for predicting breeding bird
distribution in the Great Lakes Basin. – J. Biogeogr. 31:315-331.
Vickery, P.D., Hunter, Jr., M.L. and Wells, J.V. 1992. Is density an indicator of
breeding success? – Auk 109:706-710.
Wachob, D.G. 1996. The effect of thermal microclimate on foraging site selection
by wintering mountain chickadees. – Condor 98:114-122.
Wade, T.G., Riitters, K.H., Wickham, J.D. and Jones, B.K. 2003. Distribution and
causes of global forest fragmentation. – Conserv. Ecol. 7:7. [online]
URL:http//www.consecol.org/vol7/iss2/art7.
Walsberg, G.E. 1981. Nest-site selection and the radiative environment of the
warbling vireo. – Condor 83:86-88.
Wang, G. G. 2000. Use of understory vegetation in classifying soil moisture and
nutrient regimes. – For. Ecol. Manage., 129: 93-100.
Wang, G. G. and Klinka, K. 1996. Use of synoptic variables in predicting white
spruce index. – For. Ecol. Manage., 80: 95-105.
Ward, M.P. and Schlossberg, S. 2004. Conspecific attraction and the
conservation of territorial songbirds. – Conserv. Biol. 18: 519-525.
Weiner, J. 1995. On the practice of ecology. – J. Ecol. 83:153-158.
Wiens, J.A. 1985. Habitat selection in variable environments: Shrub-steppe birds.
– In: Cody, M.L. (ed.), Habitat Selection in Birds. Academic Press, Inc.,
pp. 227-251.
Wiens, J.A. 1989. Spatial scaling in ecology. – Funct. Ecol. 3:385-397.
15
Wiens, J.A. Stenseth, N.C., Van Horne, B., and Ims, R.A. 1993. Ecological
mechanisms in landscape ecology. – Oikos 66:369-380.
Williams, C.K., Ives, A.R. and Applegate, R.D. 2003. Population dynamics
across geographical ranges: time-series analyses of three small game
species. – Ecology 84:2654-2667.
With, K.A. and Webb, D.R. 1993. Microclimate of ground nests: the relative
importance of radiative cover and wind breaks for three grassland species.
– Condor 95: 401-413.
Wright, D. H., Currie, D.J, and Maurer, B.A. 1993. Energy supply and patterns of
species richness on local and regional scales. –In: Ricklefs R.E., Schluter
D. (eds) Species diversity in ecological communities, pp: 66- 74.
16
Figure 1.1. Conceptual framework of the abiotic and biotic relationships of plants
and songbirds considered in this dissertation.
17
Figure 1.2. Map of Atlantic Canada showing study area.
18
Chapter 2 - Evaluation of environmental factors influencing vegetation structure
in mature Picea mariana forests using constrained ordination and constrained
classification1.
1
I intend to submit this chapter to Canadian Journal of Forest Research as: Simon, N.P.P., and
Diamond, A.W. Evaluation of environmental factors influencing vegetation structure in mature
Picea mariana forests using constrained ordination and constrained classification.
19
Abstract
We assessed the influence of soil and topography on mature Picea
mariana forests across 223 sites in central Labrador, NL, Canada using stepwise
redundancy analysis (RDA) and multivariate regression trees (MRT). Elevation
and water movement, i.e., slope and drainage were dominant features of the
RDA. The MRT explained a similar proportion of variation as the RDA, but used
only 3 variables: drainage, slope/aspect and elevation, where the RDA used 8.
Plant associations generally followed known edaphic associations. Kalmia
angustifolium and Cladina spp. were abundant on flat, rapidly drained sites at low
elevations. Picea mariana volumes were highest on moderately sloped and
elevated sites while steeper slopes with thick humic layers were dominated by
Abies balsamea, Alnus crispa and Dryopteris spp. Alnus rugosa and Sphagnum
spp. were most associated with very poorly drained sites. Despite the
domination of our site by mature Picea mariana, vegetation structure was quite
different due to soil and topography. These different forest structures are subject
to different economic pressures and represent different conservation values.
20
Introduction
Canada currently protects only 7.4 % of its closed forests (11 th of the top
15 countries with closed forest) (UNEP 2001). Most of Canada‘s closed forests
are boreal, which are largely intact worldwide (Zasada et al. 1997, Korovin et al.
1998, Burton et al. 2003). However, commercial forestry has expanded rapidly
within the boreal forest during the past 2 decades and fibre shortages are
causing harvesting to expand further into more remote boreal forests (Burton et
al. 2003, Department of Forest Resources and Agrifoods 2003). Central
Labrador an example of this phenomenon; commercial logging is recent (1970s)
and small-scale but is expected to increase by 400% in the next few years
(Department of Forest Resources and Agrifoods 2003).
Unfortunately, Labrador lacks a comprehensive forest classification
system necessary for ensuring all mature forest types are protected. Coarse
vegetation classifications (e.g., Hustich (1949), Lopoukhine et al. (1978), Canada
Committee on Ecological Land Classification 1989) and vegetation –
soil/topography relationships exist (Wilton 1959, 1964), but they are qualitative
and focus on overstory trees and timber production. The phytosociological
classification of Foster (1984) for south-eastern Labrador is the only quantitative
description of plant-soil associations for Labrador forests. Although useful, the
arbitrary boundaries between groups inherent in TWINSPAN may not accurately
reflect associations between species and environmental variables (Legendre and
Legendre 1998). Further, differences in soil conditions, climate and disturbance
frequencies alter plant responses to local environmental gradients (Meades and
21
Roberts 1992, Oliver and Larsen 1996, Ohmann and Spies 1998). Meades and
Roberts (1992) state these differences prohibit the extrapolation of existing
classifications in adjacent areas (e.g., Damman 1967, Foster 1984, Meades and
Moores 1989) to central Labrador. Notwithstanding their limitations, studies
within Labrador and adjacent areas generally conclude that plant abundances
are driven by moisture and nutrients. These factors are determined by
slope/aspect, soil drainage and texture and thus, is seems reasonable they
would be dominant features in a quantitave community level vegetation model.
Many studies of plant-environment relationships use either unconstrained
ordination (e.g., PCA), or unconstrained clustering techniques with arbitrary or
absolute memberships, (e.g., TWINSPAN) to infer plant-environment
relationships. While those techniques have successfully elucidated patterns in
ecological data for decades, more recently developed methods reveal speciesenvironment relationships more efficiently. Unlike their unconstrained
counterparts, constrained ordination or clustering relates species to
environmental variables without inferring environmental gradients from species
composition (ter Braak 1987, Palmer 1993, Legendre and Legendre 1998, De‘ath
2002). Also, some clustering methods have been criticized for imposing rigid or
arbitrary boundaries between ecological groups (Legendre and Legendre 1998).
Discrete homogenous groupings are rare in ecology and occur in plant
communities only along extremely steep environmental gradients (Maycock and
Curtis 1960, Stohlgren et al. 1998, Schaefer and Wilson 2002). Thus, blurred
discontinuities among groups better reflect most community structures (Legendre
22
and Legendre 1998, Stohlgren et al. 1998, Schaefer and Wilson 2002).
Our objectives are (1) to evaluate plant-soil relationships in mature Picea
mariana forests in central Labrador and, (2) compare two statistical techniques,
constrained ordination (redundancy analysis, RDA), and constrained clustering
(multivariate regression trees, MRT) , in their efficacy in elucidating these
relationships and usefulness in forest management planning. These techniques
are complementary because each emphasizes different data structures and
assumes different relationships between species and environmental variables
(De‘ath 2002). Redundancy analysis assumes a linear or unimodal (with
appropriate data transformations) relationship between environmental variables
and emphasizes global structure (overall associations) while MRT makes no
assumptions about species-environmental relationships and emphases local
interactions local (within group) structure.
Study Area
The study area is within the perhumid high boreal ecoclimatic region
(Canada Committee on Ecological Land Classification 1989). Humo-ferric and
ferro-humic podzols occur on morainic and outwash deposits, frequently with
hardpan. Lowlands are typified by organic soils and gleysols are common in bog
edges and depressions (Lopoukhine et al. 1978). Bogs dominated by Sphagnum
mosses are common on marine clays and Cladina lichens with sparse Picea
Mariana and are common on elevated sand terraces. P. mariana and to a lesser
extent Abies balsamea, forests occur on shallow upland soils while A.
balsamea/Betula papyriferia forests occur on well drained alluvium (Lopoukhine
23
et al. 1978). This area experiences a mean annual temperature of -0.5ºC
(monthly mean range: -18.1ºC-15.4ºC) and precipitation amounts of 949 mm, half
of which falls as snow (Environment Canada Climate Normals: http://www.mscsmc.ec.gc.ca; viewed 27 September 2004). Snow generally remains on the
ground from October through June.
Methods
A total of 223 sites, ≥ 250 m apart, were distributed in mature (> 130
years) Picea mariana forests within 50 km of Goose Bay, NL, Canada (53○ 20‘ N,
60○ 25‘ W). To ensure that we measured a range of forest conditions, we used
forest cover maps (1:12 500) to classify stands based on height and crown cover
classes that reflected a range of timber volumes. We concentrated our efforts in
five classes that accounted for 85% of the mature forests. At each site, five, 8 x
10 m subplots were established, one at the centre and one at each of the 4
cardinal directions approximately 60 m from the site centre. These data were
also used to assess bird-vegetation relationships, so subplots were distributed to
characterize vegetation within a 100 m radius circle to coincide with bird pointcount stations. Within each subplot, we recorded the species and diameter at
breast height (dbh) of all trees > 7 cm dbh and the height and dbh of all snags >
1.5 m tall and > 7 cm dbh. The crown cover of trees and tall shrubs was
estimated using Emlen (1967). Within each subplot, tree and tall shrub height
was recorded above 20 systematically placed points (n = 100 per site). Points
were established 1 m apart along two transects that were 4 m apart. We
expressed conifers ≥ 5 m as timber volume while those < 5m were expressed as
24
percent cover by height class. Timber volume was calculated using tree
diameter data from our vegetation plots and local volume tables (Newfoundland
and Labrador Department of Natural Resources, unpublished data). The range
of heights expressed by other tree species was small so all height classes were
combined. We estimated crown cover of all herbaceous plants and low shrubs
using Daubenmire (1968). A 0.1 m2 quadrat was placed at each corner of each
subplot (n = 20 per site) and vegetation cover was recorded into 1 of 7 classes:
0, 0 to <5, 5 to < 25, 25 to < 50, 50 to < 75, 75 to < 95, and 95 to 100%. The
mid-points of these classes were used in analyses (Table 2.1). The standard
errors on our plant abundance estimates, by site, were generally < 10% for our
most variable species, suggesting our sampling was sufficient. Plants occurring
on < 5% of the study plots were excluded from analyses. As the vegetation data
contained two different measures of abundance, crown cover and volume, we
standardized vegetation data by dividing the abundance of each plant species by
the maximum value recorded for that species (Legendre and Legendre 1998). At
the centre subplot, we recorded the following environmental variables: soil
drainage (6 ordinal classes), soil type classified to great group (Agriculture
Canada Expert Committee on Soil Survey 1987), slope, aspect, % rockiness and
the presence of features that would restrict root penetration (e.g., water, fragipan)
within 50 cm of the soil surface. Aspect was combined with slope and
transformed such that north-facing aspects produced high positive values which
increased with slope [northness = COS(aspect) X TAN(slope)] (adapted from
Beers et al. 1966). A similar index was used that produced high values for
25
northeast slopes [northeastness = COS(aspect - 45○) X TAN(slope)] since
northeast aspects in the northern hemisphere can be the most mesic and
correlate well with site index (Stage 1976) (see Table 2.2).
Redundancy analysis (RDA) is inappropriate for unimodal data distributed
across an environmental gradient and containing many zeros, so the usual
technique for such data is canonical correspondence analysis (CCA) (Legendre
and Legendre 1998, Legendre and Gallagher 2001). However, the Chi-square
distance preserved in CCA is not unanimously favoured among ecologists
(Legendre and Gallagher 2001). We used RDA on Hellinger transformed data
because it circumvents problems associated with abundance distributions and
Chi-square distances (Legendre and Gallagher 2001).
To examine the possibility of a broad-scale spatial trend in our vegetation
data, we included the geographic coordinates of our sites and their polynomials
(b1x, b2y, b3x2, b4xy, b5y2, b6x3, b7x2y, b8xy2, b9y3) into our ordinations. We
partitioned the variation in the plant community into 4 components: pure
environment, pure space, spatially structured environment and unexplained
through partial RDA (Borcard et al. 1992, Legendre and Legendre 1998). The
RDA was performed in CANOCO Version 4 (ter Braak and Šmilauer 1999), and
all categorical and ordinal explanatory variables (soil type and drainage) were
transformed into dummy variables (Legendre and Legendre 1998). Both spatial
and environmental variables were subjected to stepwise selection (α = 0.05) with
statistically significant variables retained for ordination. Statistical significance
was determined using 999 Monte Carlo simulations (ter Braak and Šmilauer
26
1999). We compared the position of plant species in the first two canonical axes
of the RDA with the first two axes of a Principal Component Analysis (PCA), also
performed in CANOCO. If the species positions are similar in both analyses it
suggests the environmental variables captured the main gradients in the
vegetation (Brown et al. 1993, Legendre and Legendre 1998).
We also related vegetation to soil and topographic features using
multivariate regression trees (MRT) in S-PLUS using the treeplus library of
De‘ath (2002). The analysis constructs trees by repeatedly splitting the data
based on a single explanatory variable. The split results in two mutually
exclusive groups, each of which is as homogenous as possible, i.e., with the
lowest sums of squares about the multivariate mean. The splitting procedure is
then re-applied to each group separately (De‘ath 2002). Splitting continues until
an overly large tree is grown which is later ―pruned‖ to an appropriate size. Tree
size was chosen using the 1-standard error rule (1-SE), i.e., the largest tree
whose cross-validated error is within 1-SE of the minimum.
Results
Ordination
The 12 environmental variables retained by stepwise selection actually
represented 8 of the original environmental variables (Table 2.3) because ordinal
and categorical variables were converted to dummy variables (i.e., each class of
drainage and soil was considered a different variable in the analysis). The sum
of all canonical axes was statistically significant (F = 10.93, p < 0.005) and
accounted for 37.5% of the variation in species abundances. Of this 37.5%
27
environmental variation, 58% of the explained variance (21.5% of the total
variation) was spatially structured, i.e. was also explained by the spatial
coordinates. The spatial coordinates exclusively explained 9% of the total
variation.
The first and second canonical axes accounted for 23.9% and 6.3% of the
variation in plant abundance respectively. This is only slightly lower than to the
variation captured in the PCA, (34.6% and 13.8% respectively) and the positions
of species in the ordinations were similar in both analyses (Fig. 2.1, 2.2).
Elevation and water movement are dominant features of the RDA ordination with
rapidly drained sites negatively associated with elevation and slope (Fig. 2.2).
Kalmia angustifolium, Cladina spp. and Vaccinium vitis-idaea, were most
abundant on flat, rapidly drained sites of low elevation. Picea mariana < 5 m,
Ledum groenlandicum and Vaccinium angustifolium were most associated with
flat, well drained ferro-humic podzols. Pleurozium schreberi was most abundant
on moderately well drained sites with the presence of root restricting layer,
regardless of slope. Picea mariana volume and Empetrum nigrum were on
moderately sloped elevated sites with root restricting layer. Abies balsamea of
all sizes, Alnus crispa, Betula papyrifera, Dryopteris spp, and Ptilium cristacastrensis were on steeper slopes with thick humic layers. Alnus rugosa, Rubus
chamaemorus and Sphagnum spp. were most abundant on very poorly drained
north facing slopes.
Multivariate regression trees
Using three variables, the 1-SE rule yielded a 6 split, 7 group tree that
28
explained 35% of the variation in plant abundances (Fig. 2.3). Sites of elevations
<153 m were subdivided by drainage; the poorly drained group had more Abies
balsmaea, Gaultheria hispidula and Sphagnum spp while the other drainage
groups had abundant Cladina spp. The other drainage classes were further
subdivided by elevation again. The <69 m elevation group was similar to 69 -153
m except that many of the uncommon species from 69 -153m were completely
absent and the shrub layer was dominated by Kalmia angustifolia.
The 4 groups with elevations > 153 m had more species and conifer
volumes than those at lower elevations. The group with a high northness index
produced the greatest abundance of Abies balsamea and relatively even
abundances of broad-leaved trees and herbs. The shrub and moss layers were
well represented with abundant Cornus canadensis, Gaultheria hispidula,
Hylocomium splendens, and Ptilium crista-castrensis. Sites with a low northness
index (other aspects and shallow or flat slopes) were further subdivided by soil
drainage: those with slower water movement (imperfect, poor and very poor) and
those with faster water movement (moderate, well and rapid). The group with
slower water movement was similar to the previously described group but with
less Abies balsamea and herbs, more shrubs and the broad-leaved trees were
mostly Alnus rugosa. Sites with faster water movement were further subdivided
into 2 groups by elevation. The group defined by > 295 m in elevation was
similar to the group with the high northness index but with more broad-leaved
trees and herbs. The group defined by 153 - 295 m elevation with moderate to
well drained soils was quite different from other groups within the > 153 split and
29
more similar to those with < 153 m elevation – fewer Abies balsmaea, broadleaved trees and herbs while Bazzania trilobata and Cladina spp. were common
in the moss layer.
Discussion
Although our sites were dominated by mature Picea mariana, their
abundance and size, and those of other species, differed with soil/topographic
features—mostly elevation. Rowe (1972) stated that productive forests in this
region were limited to sheltered lowlands. It was, therefore, reasonable to expect
a negative relationship with elevation and understory richness and timber
volume; however, we found the opposite. The range in elevation in our study
was 335 M; using Bourque and Gullison‘s (1998) model this corresponds to a 2.2
○
C range in the average daily temperature during the growing season. This
temperature difference corroborates personal observations of slower snowmelt at
higher elevations. Siccama et al. (1970) found overstory productivity decreased
while understory abundance and diversity increased with elevation across a
similar elevation range as ours. They interpreted this change to be edaphic
rather than climatic. Our RDA supports this suggestion because slope and rapid
soil drainage (surrogates for nutrients since moisture carries nutrients down
slope) are positively and negatively related to elevation, respectively. After
elevation, soil drainage and slope/aspect are other dominant predictors in the
RDA and the MRT. Others (Wilton 1959, 1964, Damman 1967, Foster 1984)
also found moisture and drainage to be the major determinants of plant
communities. In northern boreal areas, productive forests are often on moist
30
slopes at intermediate to high elevations (Wilton 1959, 1964, Foster 1984).
Similarly, we found high Picea mariana volumes tended to occur on moderately
elevated slopes, while Abies balsamea volumes were more restricted to higher
elevations and steeper slopes. Picea mariana tended to be the dominant
understory conifer except on sites with high timber volumes, where Abies
balsamea occupied a similar or greater proportion of the conifer saplings. This
likely reflects the relative shade tolerance of Abies balsamea and its susceptibility
to moisture stress when growing in open or dry areas (Frank 1990, McLaren and
Janke 1996).
The plant-environmental relationships are clearly illustrated in the RDA,
and generally correspond to known plant-edaphic relationships. The prevalence
of Kalmia angustifolium, Cladina spp, Vaccinium vitis-idaea, Vaccinium
angustifolium, and Ledum groenlandicum on flat, rapidly drained sites
corresponds to Damman (1967), De Grandpré et al. (2003), and Mallik (2003).
Our moderately sloped, elevated sites with moderate drainage likely produced
the intermediate moisture condition required to produce higher Picea mariana
volumes and Gaultheria hispidula abundances (Meades and Moores 1989, De
Grandpré et al. 2003). However, Maycock and Curtis (1960) found these species
and Vaccinium angustifolium, Ledum groenlandicum to follow a bimodal
abundance distribution with moisture. The presence of root restriction within 50
cm, did not negatively influence tree growth. Fragipan was the dominant root
restricting agent but this can be benifical if it occurs on slopes because it
promotes water seepage whereas it causes stagnation on flat areas (Damman
31
1967).
Our finding that Abies balsamea, Ptilium crista-castrensis, Betula
papyrifera, Dryopteris spp and Alnus crispa were more associated with steeper
slopes with north and northeastern aspects and thick humic layers is consistent
with Damman (1967) Foster (1984), Foster and King (1986) and Simon and
Schwab (2005). Alnus rugosa, Vaccinium oxycoccus, Rubus chamaemorus and
Sphagnum spp., known associates of hydric soil conditions (Damman 1967,
Foster 1984, De Grandpré 2003), dominated our very poorly drained sites. Sites
and species associated with very poor drainage also appear positively related to
slope. Most of our very poorly drained sites were characterized by slight slopes
but because they were generally near the bottom of hills, they likely experienced
extensive water pooling which produced the very poor drainage conditions.
The MRT split non-poorly drained sites < 153 m in elevation into two
groups according to elevation, both of which are similar to groups described by
Hustich (1949) and Linteau (1955). The higher elevation group contains more
moss, in particular Sphagnum, and is similar to the transition forests of sprucelichen to the spruce-Sphagnum and feather moss forests of Wilton (1959).
Wilton (1959, 1964) describes a black spruce-balsam fir-Sphagnum type which
occurs on clay soils at approximately 90 m elevation. This corresponds to our
poorly drained group < 153 m elevation. Our higher elevation (> 153 m) sites
contain most of the species and timber volume in the study area, corresponding
to the fir-spruce/rich herb type of Wilton (1959) and Bajzak (1973) and the firspruce forest of Foster (1984). Although others have found Betula papyrifera to
32
be widespread (Maycock and Curtis 1960), they suggest the distribution pattern
results from disturbance. In our study, Betula papyrifera was restricted to a few
openings within a predominately Abies balsamea/Picea mariana forest.
The variation explained by both the RDA and the MRT is similar to that of
other studies using constrained ordination (Schaefer and Messier 1995,
Cushman and Wallin 2002). Although the RDA explained somewhat less
variation than was captured in the first two PCA axes, the similarity of species‘
positions between both ordination diagrams indicates the environmental factors
explained vegetation abundance well. The discrepancy of variance explained in
the PCA and RDA may be due to the coarseness of our environmental variables.
Drainage (or moisture) is one of the most important factors influencing plant
distribution in boreal forests (Damman 1967, Maycock and Curtis 1960, Foster
1984) yet we analyzed it as a categorical variable. We did not measure soil
chemistry, e.g., nutrient availability and pH. These variables could have
explained additional variation in plant distribution, but they tend to be related to
drainage and topography (Fralish 1994, Roberts et al. 1998).
The spatially structured environmental variation (21.5%) was likely do to
spatial structuring of the environmental variables, rather than a spurious
relationship with an unmeasured spatially structured variable, as elevation
increased along a northwest trend. The 9% of the variation attributed exclusively
to space could be due to an unmeasured spatially structured variable or inherent
biological processes, e.g., seed dispersal. However, because it was so small it is
unlikely that we missed any fundamental spatially structuring processes.
33
Although the variance explained by both statistical methods was similar,
the MRT used fewer variables. We suspect this is because MRT is more efficient
at modeling of interactions and does not assume a particular relationship
between the species and environmental data. By splitting the data into mutually
exclusive groups which are further split, the MRT inherently models interactions
between variables but for RDA these interactions must be identified beforehand
and modeled explicitly (De‘ath 2002). We did not include interactions in our RDA
because none were immediately obvious prior to analysis and including all
possible interactions would lead to overfitting. Redundancy analysis assumes a
linear relationship between variables (Legendre and Legendre 1998) but can
accommodate unimodal relationships when the data are Hellinger transformed
(Legendre and Gallagher 2001). Thus, RDA would not adequately describe
species-evironment relationships for species such as those distributed bimodally
across a resource gradient which are not uncommon in plants (e.g, Maycock and
Curtis 1960). Where species abundances are strongly influenced by interactions
among environmental variables and, show complex distributions across
resources, MRT should better describe the plant-environment relationships. The
dominant relationships in both of our MRT and RDA agree, but there are
discrepancies which we attribute to bimodal distributions. For example, the RDA
shows a strong negative relationship between elevation and Cladina spp. The
MRT shows this species to be most abundant on lower elevations as well, but it
also is relatively common at higher elevations—this latter relationship was not
apparent in the RDA.
34
Species were not exclusive to particular groups, supporting those who
conclude that discrete homogenous groupings are rare in ecology (Maycock and
Curtis 1960, Stohlgren et al. 1998, Schaefer and Wilson 2002). This
demonstrates the importance of using analytic techniques that can accommodate
indistinct discontinuities among groups (Legendre and Legendre 1998, Stohlgren
et al. 1998, Schaefer and Wilson 2002). The lack of discrete groups suggests
that eliminating a particular group on the landscape would not likely cause major
species extirpation. However, plant species often showed very different
abundances among groups, so elimination of a particular group could drastically
reduce the abundance of some species on the landscape.
The RDA and MRT showed soil and topographic features create different
vegetation assemblages. When protecting mature Picea marina forests, it is
insufficient to protect only by age class because these forests exhibit very
different overstory and understory structures and composition. These different
mature forest structures likely represent different conservation values, and are
subject to different economic pressures. One obvious forest managementconservation conflict is that sites with highest timber volumes are usually a high
priority for logging. These forests also posess abundant broad leaved trees,
ample cover of shorter conifers (suggesting high structural diversity), high
understory plant abundance and diversity. For conservation planning, managers
can use the predictive model produced by the MRT to predict understory
conditions, allowing them to ensure a variety of plant species and structures are
maintained. Because the model relies heavily on variables derived from digital
35
elevation models, the model would be useful to predict conditions when field is
logistically difficult to collect.
Acknowledgements
Funding was provided by the Newfoundland and Labrador Department on
Natural Resources, and Human Resources and Skills Development Canada.
Thanks are extended to those assisting in data collection, notably, B. Campbell,
J. Colbert, L. Elson, R. Flynn, M. Michelin, K. Mitchell, R. Neville, F. Phillips, B.
Rodrigues, F. Taylor. D. Goulding and D. Jennings assisted in GIS work. M.
Betts, K Chaulk, G. Forbes, D. Keppie and B. Roberts reviewed earlier drafts of
the manuscript.
References
Agriculture Canada Expert Committee on Soil Survey. 1987. The Canadian
system of soil classification, 2nd ed. – Agric. Can. Publ. 1646, Ottawa.
Bajzak, D. 1973. Bio-physical land classification of the lake Melville area,
Labrador. – Inf. Rep. N-X-88. Nfld. For. Res. Cent., St. John‘s.
Beers, R.W., Dress, P.E. and Wensel, L.C. 1966. Aspect transformation in site
productivity research. – J. For. 64:691-692.
Borcard, D., Legendre, P. and Drapeau, P. 1992. Partialling out the spatial
component of ecological variation. – Ecology 73:1045-1055.
Bourque, C. P-.A. and Gullison, J.J. 1998. A technique to predict hourly potential
solar radiation and temperature for a mostly unmonitored area in the Cape
Breton Highlands. – Can. J. Soil Sci. 78:409-420.
36
Brown, A., Birks, H.J.B. and Thompson, D.B.A. 1993. A new biogeographical
classification of the Scottish uplands. II. Vegetation-environment
relationships. – J. Ecol. 81:231-251.
Burton, P.J., Messier, C., Weetman, G.F. Prepas, E.E. Adamowicz, W.L. and
Tittler, R. 2003. The current state of the boreal forest and the drive for
change. In: Burton, P.J. Messier, Smith, D.W. and Adamowicz, W.L.
(eds.), Towards sustainable management of the boreal forest. NRC
Research Press, Ottawa, pp. 1-40.
Canada Committee on Ecological Land Classification. 1989. Ecoclimatic Regions
of Canada: First approximation. – Environment Canada, Ecological Land
Classification Series No. 23. Canadian Wildlife Service.
Cushman, S.A. and Wallin, D.O. 2002. Separating the effects of environmental,
spatial and disturbance factors on forest community structure in the
Russian Far East. – For. Ecol. Manage. 168:201-215.
Damman, A.W.H. 1967. The forest vegetation of western Newfoundland and site
degradation associated with vegetation change. – Ph.D. thesis, University of
Michigan, MI.
Daubenmire, R. 1968. Plant communities. A textbook of plant synecology. –
Harper and Row Publishers, London.
De‘ath, G. 2002. Multivariate regression trees: A new technique for modeling
species-environment relationships. – Ecology 83:1105-1117.
De Grandpré, L., Bergeron, Y., Nguyen, T., Boudreault, C. and Grondin. P. 2003.
Composition and dynamics of the understory vegetation in the boreal
37
forests of Quebec. – In: Gilliam, F.S. and Roberts, M.R. (eds.), The
herbaceous layer in forests of eastern North America. Oxford University
Press, New York, pp. 238-261.
Department of Forest Resources and Agrifoods. 2003. Five year operating plan
for forest management district 19A (Goose Bay). – Newfoundland and
Labrador Department of Forest Resources and Agrifoods, Goose Bay.
Emlen, J.T. 1967. A rapid method for measuring arboreal canopy cover. –
Ecology 48:158-159.
Foster, D.R. 1984. Phytosociological description of the forest vegetation of
southeastern Labrador. – Can. J. Bot. 62:899-906.
Foster, D.R. and King, G.A. 1986. Vegetation pattern and diversity in S.E.
Labrador, Canada: Betula papyrifera (birch) forest development in relation
to fire history and physiography. – J. Ecol. 74:465-483.
Fralish, J.S. 1994. The effect of site environment on forest productivity in the
Illinois Shawnee Hills. – Ecol. Appl. 4:134-143.
Frank, R.M. 1990. Balsam Fir. – In: Burns, R.M. and B.H. Honkala (tech. cords.),
Silvics of North America: Volume 1. Conifers, Agric. Handb. 654. USDA
For. Serv., Washington, DC, pp. 26-35.
Hustich, I. 1949. Phytogeographical regions of Labrador. – Arctic 2:34-42.
Korovin, G., Karpov, E., Islaev, A., Nefedjev, V., Efremov, D., Sedych, V.,
Sokolov, V., Schmidt, T., Blauberg, K., Ljusk Eriksson, O., Nilsson, S.,
Raile, G., Sallnäs, O. and Shvidenko, A. 1998. Siberia and Far East
38
Russia‘s future wood supply: an analysis. – International Institute for
Applied Systems Analysis, Interim Report IR-97-024, Laxenburg, AT.
Legendre, P. and E.D. Gallagher. 2001. Ecologically meaningful transformations
for ordination of species data. – Oecologia 129:271-280.
Legendre, P. and Legendre, L. 1998. Numerical Ecology, second English edition.
– Elsevier, New York.
Linteau, A. 1955. Forest site classification of the northeastern coniferous section
boreal forest region Quebec. – Dept. North. Aff. Nat. Resour. For. Br. Bull.
118, Ottawa.
Lopoukhine, N., Prout, N.A., and Hirvonen, H.E. 1978. Ecological land
classification of Labrador. Ecological Classification Series, No. 4. – Lands
Dir. (Atl. Reg.), Fish. Environ. Can., Halifax.
Mallik, A.U. 2003. Conifer regeneration problems in boreal problems in boreal
and temperate forests with Ericaceous understory: role of disturbance,
seedbed limitation, and keystone species change. – Crit. Rev. Plant Sci.
22:341-366.
Maycock, P.F. and Curtis, J.T. 1960. The phytosociology of boreal coniferhardwood forests of the Great Lakes region. – Ecol. Monogr. 30:1-35.
Mclaren, B.E. and Janke, R.A. 1996. Seedbed and canopy cover effects on
balsam fir seedling establishment in Isle Royale National Park. – Can. J.
For. Res. 26:782-793.
Meades, W.J. and Moores, L. 1989. Forest site classification manual. A field
guide to the Damman forest types of Newfoundland. – For. Resour. Dev.
39
Agree. FRDA Report 003.
Meades, W.J. and Roberts, B.A. 1992. A review of forest site classification
activities in Newfoundland and Labrador. – For. Chron. 68:25-33.
Ohmann, J.L. and Spies, T.A. 1998. Regional gradient analysis and spatial
pattern of woody plant communities of Oregon forests. – Ecol. Monogr.
68:1151-182.
Oliver, C.D. and Larson, B.C. 1996. Forest stand dynamics. – John Wiley and
Sons, Inc., New York.
Palmer, M.W. 1993. Putting things in even better order: the advantages of
canonical correspondence analysis. – Ecology 74:2215-2230.
Roberts, B.A., Deering, K.W. and Titus, B.D. 1998. Effects of intensive harvesting
on forest floor properties in Betula papyrifera stands in Newfoundland. – J.
Veg. Sci. 9:521-528.
Rowe, J.S. 1972. Forest Regions of Canada. – Env. Can. For. Serv., Publ. No.
1300, Ottawa.
Schaefer, J.A. and Messier, F. 1995. Scale-dependent correlations of arctic
vegetation and snow cover. – Arct. Alp. Res. 27:38-43.
Schaefer, J.A. and Wilson, C.C. 2002. The fuzzy structure of populations. – Can.
J. Zool. 80:2235-2241.
Siccama, T.G., Bormann, F.H. and Likens, G.E. 1970. The Hubbard Brook
ecosystem study: productivity, nutrients, and phytosociology of the
herbaceous layer. – Ecol. Monogr. 40:389-402.
40
Simon, N.P.P. and Schwab, F.E. 2005. Plant community structure after wildfire in
the subarctic forests of Western Labrador. – North. J. Appl. For. In press
Stage, A.R. 1976. An expression for the effect of aspect, slope, and habitat type
on tree growth. – For. Sci. 22:457-460.
Stohlgren, T.J., Bachand, R.R., Onqami, Y. and Binkley, D. 1998. Speciesenvironment relationships and vegetation patterns: effects of spatial scale
and tree life-stage. – Plant Ecol. 135:215-228.
ter Braak, C.J.F. 1987. The analysis of vegetation-environment relationships by
canonical correspondence analysis. – Vegetatio 69:69-77.
ter Braak, C.J.F. and Šmilauer, P. 1999. Canoco reference manual and users
guide to Canoco for Windows: software for canonical community
ordination version 4. .
UNEP. 2001. An assessment of the status of the World‘s remaining closed
forests. – UNEP/DEWA/TR 01-2.
Wilton, W.C. 1959. Forest types of the Grand Lake and northwestern Lake
Melville areas of Labrador. – Dept. For. Publ., For. Resour. Div. Tech.
Note No. 83, Ottawa.
Wilton, W.C. 1964. The forests of Labrador. – Can. Dept. For., Dept. For. Publ.
No. 1066, Ottawa.
Zasada, J.C., Gordon, A.G., Slaughter, C.W. and Duchesne, L.C. 1997.
Ecological considerations for the sustainable management of the North
American boreal forests. – International Institute for Applied Systems
Analysis, Interim Report IR-97-024, Laxenburg, AT.
41
Table 2.1. Vegetation species and codes.
Species
Coniferous trees
Abies balsamea volume
Abies balsamea 2-5 m tall
Abies balsamea 0-2 m tall
Picea mariana volume
Picea mariana 2-5 m tall
Picea mariana 0-2 m tall
Broad-leaved trees
Alnus crispa
Alnus rugosa
Betula papyrifera
Woody shrubs
Cornus canadensis
Empetrum nigrum
Gaultheria hispidula
Kalmia angustifolium
Ledum groenlandicum
Linneae borealis
Lycopodium annotinum
Rubus chamaemorous
Vaccinium angustifolium
Vaccinium cespitosum
Vaccinium oxycoccus
Vaccinium vitis-idaea
Herbaceous plants
Coptis groenlandica
Dryopteris spp.
Equisetum sylvaticum
Glyceria canadensis
Maianthemum canadense
Smilacina trifolia
Mosses\liverwarts\lichens
Bazzania trilobata
Cladina spp.
Dicranum majus
Hylocomium splendens
Peltigera spp.
Pleurozium schreberi
Polytrichum commune
Ptilium crista-castrensis
Sphagnum spp.
Code
Abi.bal.v
Abi.bal.5
Abi.bal.2
Pic.mar.v
Pic.mar.5
Pic.mar.2
Aln.cri
Aln.rug
Bet.pap
Cor.can
Emp.nig
Gau.his
Kal.ang
Led.gro
Lin.bor
Lyc.ann
Rub.cha
Vac.ang
Vac.ces
Vac.oxy
Vac.vit
Cop.gro
Dry.spp
Eqi.syl
Gly.can
Mai.can
Smi.tri
Baz.tri
Cla.spp
Dic.maj
Hyl.spl
Pel.spp
Ple.sch
Pol.com
Pti.cri
Sph.spp
42
Table 2.2. Descriptions and codes of environmental variables used in
redundancy analysis and multivariate regression tree models.
Variable
Type
Description
Elevation
Humic
Northness
Continuous
Continuous
Continuous
Northeastness
Continuous
Elevation, range = 5-340m (Elev)
Thickness of humic layer, range = 0.-42 cm (Humic)
Tan(slope) * Cos(aspect), gives highest positive
values for north slopes, range = -0.186 – 0.397
(North)
Tan(slope) * Cos(aspect - 45○) gives highest positive
values for northeast slopes, range = -0.209 – 0.483
(NEast)
Degrees, range = 0 - 25○
Very poor (VP), Poor (P), Imperfect (I), Moderate (M),
Well (W), Rapid (R)
Brunisol (B), Gleysol (G), Ferro-humic podzol (FHP),
Regosol (R), Humic gleysol (HG), Organic (O), Humic
Podzol (HP), Humo-Ferric podzol (HFP)
Presence of a root restriction within 50 cm of soil
surface (Restrict)
% rockiness, range = 0 – 85 (Rock)
Slope
Drainage
Categoricala
Soil type
Categorical
Restriction binomial
Rockiness
Continuous
a
Measured as an ordinal variable but analyzed as categorical.
43
Table 2.3. Test statistics of environmental variables in order of their inclusion
into stepwise redundancy analysis.
Variable
Elevation
Northness
Humo-Ferric podzol
Ferro-humic podzol
Northeast
Restriction
Rapid drainage
Well drainage
Moderate drainage
Very poor drainage
Humic
Slope
Regisol
Humic gleysol
Humic podzol
Brunisol
Imperfect drainage
Poor drainage
Rockiness
Gleysol
Organic
F
49.29
12.63
9.76
7.82
6.82
4.54
3.87
3.25
3.34
2.31
2.05
1.99
1.67
1.20
1.01
0.86
0.99
0.99
1.02
0.68
0.68
P
<0.005
<0.005
<0.005
<0.005
<0.005
<0.005
<0.005
<0.005
<0.005
0.015
0.025
0.045
0.060
0.295
0.460
0.570
0.380
0.380
0.415
0.77
0.77
44
Figure 2.1. Dominant environmental vegetation gradients as determined by
principal component analysis. Species codes in Table 2.1.
45
Figure 2.2. Plant-environmental relationships as assessed by stepwise
redundancy analysis. Environmental variable codes in Table 2.2, species codes
in Table 2.1.
46
Figure 2.3. Vegetation groups in relation to environmental variables as
determined by a multivariate regression tree. Species codes in Table 2.1,
environmental variable codes in Table 2.2. Tree branch lenghts are proportional
to the variance explained. Bar graphs reflect species abundances, numbers in
parentheses refer to the number of sites within the respective group
47
48
Chapter 3 - Songbird habitat quality across a timber productivity gradient within
an unfragmented northern boreal forest: local and landscape effects 2.
2
I intend to submit this chapter to Landscape Ecology as: Simon, N.P.P., and Diamond, A.W.
Songbird habitat quality across a timber productivity gradient within an unfragmented northern
boreal forest: local and landscape effects.
49
Abstract
We assessed bird habitat quality relative to the site‘s ability to produce
timber (site productivity), in mature black spruce (Picea mariana) forests of central
Labrador, NL, Canada. Only hermit thrushes (Catharus guttatus) were strongly
related to forest type, with abundance inversely related to productivity class. For
the remaining 14 species, our habitat quality measures—abundance, site
occupancy and reproductive activity showed few relationships with vegetation at
local (100 m) and landscape extents (500-2000 m). The lack of a positive
relationship between bird habitat quality and productivity may be because most
species were either typical of early successional forests (e.g., white-throated
sparrow, Zonotrichia albicollis) or those known to be tolerant of canopy openings
(e.g., Yellow-rumped warbler, Dendroica coronata). Where relationships existed,
landscape variables were often of similar or greater importance than local
variables. This was particularly true for species typical of early successional
forests which were uncommon on the landscape. Since all 3 habitat quality
measures did not, in general, show a positive association with site productivity, a
reduction of highly productive forests would likely not have a long-term detrimental
effect on forest birds in this system.
50
Introduction
The boreal forest is one of the world‘s largest reserves of underexploited
wood fibre (Zasada et al. 1997, Korovin et al. 1998, Burton et al. 2003). Despite
a history of intense harvesting in certain boreal regions, e.g., the Nordic countries
(Duinker et al. 1998, Larsson and Danell 2001), the boreal forest has received
less forestry pressure than other forest biomes (Duinker et al. 1998, Wade et al.
2003). However, commercial forestry has expanded rapidly within the boreal
forest during the past 2 decades (Burton et al. 2003) and many areas are
currently experiencing or projecting fibre shortages (Nilsson et al. 1999, Apsey et
al. 2000, Burnett et al. 2003, Department of Forest Resources and Agrifoods
2003, Schneider et al. 2003, Fall et al. 2004). Within the boreal forest, most
commercial forestry, and forest ecology research, is concentrated in the more
productive southern region (Korovin et al. 1998, Burton et al. 2003). Fibre
shortages are increasing pressure to expand harvesting further into the sparser
forests of the northern boreal and taiga regions (Burton et al. 2003, Department
of Forest Resources and Agrifoods 2003). These regions tend to have a high
proportion of low-volume forests with commercially viable stands confined to
slopes and river valleys (Wilton 1965, Rowe 1972, Gray 1995, Chapter 2). This
situation implies that most of these landscapes would remain forested after
logging, but would consist mostly of non-commercial, low-volume stands.
Nevertheless, the potential for change, and limited knowledge of harvesting
effects in these systems, warrants additional research (Burton et al. 2003).
Further, conservation efforts directed towards species-rich areas may be
51
deflecting attention from species-poor regions like northern boreal forests
(Kareiva and Marvier 2003). To assess the environmental effects of forest
harvesting, researchers have recommended using indicator groups—several
species with different niche requirements that can be surveyed with a single costeffective protocol (Hutto 1998, Nilsson et al. 2001), such as songbirds (Hutto
1998, Venier and Pearce 2004).
Songbird-habitat relationships can be grouped into 2 study extents, local
and landscape (Diamond 1999, Lichstein et al. 2002a). Most local-extent
research directs measurements at extents smaller than, or of similar magnitude
to, the territory size; vegetation variables (e.g., stem densities, stratification) have
been used to predict songbird abundance, foraging and nest sites (Lack 1933,
MacArthur and MacArthur 1961, Roth 1976, Franzreb 1978, Morse 1989).
Landscape variables tend to occur at larger extents (i.e., several kilometers for
songbirds) and include vegetation types surrounding the territory and measures
of spatial associations, e.g., isolation and patch-size (Saunders et al. 1991,
Mazerolle and Villard 1999). Local variables tend to be most important but
landscape features can significantly influence abundances on oceanic islands or
forest in urban/agricultural settings (Andrén 1994, Mazerolle and Villard 1999).
However, the contrast between patches and the matrix is less in forests than with
the oceanic and agricultural settings (Drolet et al. 1999), suggesting landscape
effects should be less important. Research in forests supports this suggestion
but weak landscape effects still exist, even in relatively unfragmented forests
(McGarigal and McComb 1995, Trzcinski et al. 1999, Drapeau et al. 2000,
52
Lichstein et al. 2002a). As logging usually targets higher productivity (productive
from a commercial timber perspective) stands first, logging northern forests will
result in low-productivity stands being the dominant forest cover on the
landscape. These low-productivity stands may become the major patches of
mature forest for many species and a major component of the matrix for others.
Therefore, to study the effects of logging on songbirds in northern boreal forest
landscapes requires a multi-scale approach to assess the differences in habitat
quality in relation to site productivity.
Given the prevalence of low-productivity stands on some landscapes, the
tendency for protected areas to be skewed towards such areas (Niemelä et al.
2001, Raivio et al. 2001), and the paucity of research relating bird habitat quality
with productivity is a concern (Niemi et al. 1998). Within a successional stage,
timber volume increases with tree diameter and height, providing greater
potential foraging area for foliage gleaners and stem feeders. Further, higher
nutrient availability should increase canopy arthropod densities (Herms and
Mattson 1992). Consequently, habitat quality for canopy and stem feeding birds
should increase with timber volume. Ground and shrub feeders are often
associated with low growing trees and shrubs that might be limited on less
productive sites (Chapter 2), unless light becomes a limiting factor. Further,
forest productivity increased ground invertebrate biomass producing higher
reproduction for a ground feeding bird (Seiurus aurocapillus; Seagle and
Sturtevant 2005). Following this reasoning, habitat quality of ground-nesters and
-feeders should also increase with productivity. Alternatively, highly productive
53
forests are relatively rare on the landscape in most northern boreal forests
(Wilton 1965, Rowe 1972, Gray 1995). Hence, it is also possible that insufficient
amounts productive forest exists to maintain viable populations of species that
would otherwise select these areas.
Habitat quality is a measure of the importance of a habitat type for an
individual‘s fitness (Van Horne 1983). The assumption that density reflects
habitat quality may not be valid in situations of habitat patchiness, territoriality,
site tenacity and migratory behaviour (Van Horne 1983; Pulliam 1988; Pulliam
and Danielson 1991; Vickery et al. 1992). However, if other measures of habitat
quality (including density) show similar patterns, researchers can be more
confident in their assessment of habitat quality. Many theories of habitat
selection imply that the highest quality habitats will be occupied more
consistently (Rodenhouse et al. 1997, Sutherland 1997). Similarly, breeding site
return rates are often higher for successful than unsuccessful breeders (Haas
1997, Howlett and Stutchbury 2003, Porneluzi 2003), indicating site reoccupancy (persistence) should be a good indicator of habitat quality (e.g.,
Hames et al. 2001). Indices of breeding activity would also provide information
on the value of a particular habitat to an individual‘s fitness (e.g., Vickery et al.
1992, Gunn et al. 2000). Hence, these 3 measures (density, re-occupancy and
reproductive activity) provide different but complemantary information that allows
a more complete assement of habitat quality.
We assessed the differences in site-specific habitat quality at multiple
extents in relation to stand productivity (i.e, the sites‘ ability to produce
54
commercial timber) within a relatively unlogged region near the boreal/subarctic
forest boundary. We assess this relationship at multiple extents because
landscape variables, in addition to local variables, can influence habitat quality,
even in unfragmented forests. We use 3 measures of songbird habitat quality:
bird abundance, site re-occupancy and breeding activity, as these provide
different but complimentary information about the quality of a particular habitat to
increase individual‘s fitness. A strong positive association between habitat
quality and productivity suggests a significant conservation issue since highly
productive stands are uncommon in these landscapes and are usually a high
priority for logging.
Study Area
The study area was within the perhumid high boreal ecoclimatic region
(Canada Committee on Ecological Land Classification 1989). The study was
conducted among mature (> 130 years) natural black spruce (Picea mariana)
stands within 50 km of Goose Bay, NL, Canada (53○ 20‘ N, 60○ 25‘ W). Sites
were dominated by black spruce and to a lesser extent balsam fir (Abies
balsamea). Feather moss (Pleurozium schreberi) occurs commonly on well
drained sites and Sphagnum spp. on poorly drained sites. Approximately 34% of
the study area is heavy timver (≥ 100 m3/ha), 45% is sparse timber (<100 m3/ha),
13% is old burns (≥15 years), 5% is old cuts (≥15 years) and 3% is young cuts
(<15 years). Most commercial harvesting occurred during the mid 1970s, and
was virtually absent prior to this and during the 1980s. Approximately 50 000 m3
of timber has been extracted annually from the mid 1990s to present. Expected
55
highway construction will increase access with a projected quadrupling of the
timber harvest (Department of Forest Resources and Agrifoods 2003).
Anticipated hydroelectric developments will harvest 1.2 million m3 of timber from
the flood zone.
Incomplete sampling of an environmental gradient can lead to erroneous
species-environment relationships (Vaughan and Omerod 2003). To ensure
complete sampling of the mature forest timber volume gradient, we stratified
forest stands according to volume. Stratification was based on height and crown
cover, using forest inventory maps at 1:12 500 scale (Department of Natural
Resources, unpublished data). Approximately 85% of mature forests were
captured in 5 forest types: lichen woodland (LW), open spruce (OS), low volume
spruce (LS), medium volume spruce (MS) and high volume spruce (HS); we
therefore concentrated our efforts in those forest types. Of the 15 % that were
not represented by these 5 forest types, 12% produced volumes within the
bounds of our sampled timber volume gradient (i.e., higher volumes than LW but
lower volume than HS). The remaining 3% produced volumes slightly higher
than HS but were either in stands too small to meet our sampling criteria or too
remote to sample.
Methods
Bird census
Birds were enumerated using 220 point counts of 10-minute duration,
divided into 3-, 2-, and 5-minute intervals in which all birds seen or heard were
recorded (Bibby et al. 1993, Smith et al. 1997). We attempted to have 50 point
56
counts per treatment, based on Smith et al. (1997) who suggested this level of
sampling was sufficient to discern any significant patterns. Availability,
accessibility and the requirement for sites to remain unlogged for the duration of
the study limited our sampling within each forest type as follows: LW = 50, OS =
47, LS = 44, MS = 49, HS = 30. Nevertheless, the standard errors were usually
within 10% of the abundance estimate for our most abundant species and within
15% of our less abundant species suggesting this sampling intensity was
sufficient. We used the maximum number of males recorded within 100 m radius
on a single visit as our index of abundance because the number of birds
recorded is the minimum number present (Bibby et al. 1993). Individuals within
100 m that were flying over were excluded from analyses. Observations were
conducted between 04:00-11:00 AST from June 01 - July 31 in 2000-2002. Point
count stations were distributed among the 5 stand types described above, were ≥
250 m apart and ≥ 100 m from the nearest edge. Two visits were made to each
point count station and observers were alternated between visits to reduce
observer bias. To establish a consistent protocol, all observers conducted
simultaneous point counts with an experienced observer before collecting data
independently. An index of breeding behaviour was used following Gunn et al.
(2000), which provides a reasonable indicator of reproductive success at coarse
scales (Doran et al. 2005). In all 3 years of the study, 5 minute play-backs of
black-capped chickadee (Poecile atricapillus) mobbing calls, following the point
counts, were used to attract songbirds to facilitate observations of songbird
breeding activity. We ranked breeding activity 0 = absent, 1= male only, 2 =
57
pairs, 3 = adults carrying nesting material, 4 = adults carrying food and 5 = the
presence of juveniles (adapted from Gunn et al. 2000). For additional data on
reproductive activity, after the completion of both point count rounds, we
conducted 1 extra visit to each station in 2001 and 2 extra visits in 2002 using
only the mobbing play-back.
Local Vegetation
At each point count station, five, 8 x 10 m plots were established, one at
the centre and one at each of the 4 cardinal directions approximately 60 m from
the station centre. Within each subplot, we recorded the species and diameter at
breast height (dbh) of all live trees > 7 cm dbh and the height and dbh of all
snags > 1.5 m tall and > 7 cm dbh. Timber volume was calculated using tree
diameter data from our vegetation plots and local volume tables (Newfoundland
and Labrador Department of Natural Resources, unpublished data). We
estimated the crown cover of trees and tall shrubs using Emlen (1967). Within
each plot, we recorded the height of the top and bottom of live crown above 20
systematically placed points (n = 100 per point count station) that were
established 1 m apart along two transects 4 m apart. All herbaceous plants and
low shrubs were estimated using Daubenmire (1968). A 0.1 m2 quadrat was
placed at each corner of each subplot (n = 20 per point count station).
Vegetation cover was recorded into 1 of 7 classes: 0, > 0 - 5, > 5 - 25, >25 - 50, >
50 - 75, > 75 - 95, and > 95 - 100. The mid-points of these classes were used in
analyses. Within a point-count station, the standard errors on our abundance
58
estimates were usually < 10%, indicating a relatively precise description of the
vegetation.
Landscape variables
Landscape composition was quantified at 3 spatial extents using circular
buffers around each point count station of 500 m, 1000 m, and 2000 m radii.
Variables (listed in Table 3.1) were calculated with ARCVIEW 3.2 (Environmental
Systems Research Institute 1999) using 1992 forest inventory data (Department
of Natural Resources, unpublished data) digitized from 1:12 500 scale aerial
photographs. The inventory was updated to include disturbances (fire,
harvesting and new roads) since 1992. Our landscape was predominantly
mature black spruce forests that varied by height and density. Therefore, the
usual landscape components, (i.e., patch, corridor and matrix) are less applicable
because forest types change subtly along gradients rather than with hard edges
typically created by stand-replacing disturbances. It is therefore unlikely that
landscape configuration (e.g., patch size, isolation) was important in our study.
Further, landscape configuration is usually confounded with landscape
composition (Fahrig 2003). We therefore did not consider landscape
configuration separately.
Detectability
As songbird detectability can vary with forest cover, Schieck (1997)
recommended researchers demonstrate that results are not affected by biased
detections. We used the approach of Farnsworth et al. (2002) to evaluate
detection probabilities among the different forest types. For each visit, we
59
recorded in which interval a particular individual was first detected (3, 2 or 5
minutes). For a highly detectable species, most first detections should occur in
the first interval. For a less detectable species, a substantial number of first
detections should occur in the later intervals. Farnsworth et al. (2002) used
SURVIV (White 1983) to estimate parameters describing the conditional
probability that a species was detected in each of the 3 intervals given that it was
detected in the entire 10 minutes.
This method can be used to calculate the probability that a randomly
selected individual is detected during the entire 10 minute sampling period. By
summing the number of first detections in each interval by forest type, differences
in detection probability by forest type can be estimated (G. Farnsworth, personal
communication). We adapted code used in Farnsworth et al. (2002), available
online (URL:www.mbr-pwrc.usgs.gov/software/ CountRemoval.html) and fit 4
models using SURVIV. The most general model MFC contained parameters for
each forest type, and incorporated heterogeneity in detectability among the count
intervals due to their unequal length. Model MF contained parameters for forest
type only, model MC incorporated heterogeneity only, while model M did not
contain parameters for forest type or incorporate heterogeneity. We used AIC to
evaluate the strength of evidence for each model. For models without a large
sample size: parameter ratio, subtle effects are difficult to detect and there is
increased potential of overfitting (Burnham and Anderson 2002, Vaughn and
Ormerod 2003). Therefore, we combined the five forest types into 2 categories,
non-commercial (LW and SS) and commercial (the remaining 3 forest types) for
60
all species. For common species we also analyzed detectability using all 5 forest
types. Common and scientific names for species and their codes are in Table
3.2.
Dividing the number of individuals detected within fixed radius point counts
by the detectability of the species will estimate density (Farnsworth et al. 2002).
This is more meaningful where the desired spatial extent of the density estimate
contains many point count stations. If this conversion was conducted on
individual point counts, it would underestimate density where no species were
detected because zeros would be divided by detectability. Because we analyzed
our data by individual point counts, we did not convert to density and computed
detectability only to confirm there was no auditory bias due to forest type.
Bird vegetation models
Spatial dependence in observations violates the assumption of
independence, and can cause false correlations unless spatial pattern is
accounted for in statistical models (Legendre and Legendre 1998, Keitt et al.
2002, Lichstein et al. 2002b). One obvious source of spatial dependence in our
study is the spatial overlap of landscape variables in adjacent point counts
stations. Hence, birds recorded at different point count could be responding to
identical landscape features. Thus, we modelled species-environment
relationships using autoregressive models to account for spatial pattern (Keitt et
al. 2002, Klute et al. 2002, Lichstein et al. 2002b). For Gaussian data,
conditional autoregressive models (CAR) are appropriate when residual
abundance patterns depend upon unmeasured environmental variables or
61
endogenous processes (e.g., conspecific attraction) (Keitt et al. 2002, Lichstein
et al. 2002b). For binomial variables, autologistic regression includes within a
logistic regression model a term, the autocovariate, which conditions the
response for a given location on the values of the response at neighbouring
locations (Gumpertz et al. 1997, Klute et al. 2002).
Count data like our songbird abundances often violate assumptions
required by Gaussian models and are typically analyzed using generalized linear
models with a Poisson distribution (McCullagh and Nelder 1989). Unfortunately,
autoregressive models with a Poisson response distribution are impractical
because they can incorporate only negatively correlated errors (Cressie 1993).
As a compromise between adhering to distribution assumptions and accounting
for spatial pattern, we used either Gaussian or logistic regression depending
upon the species‘ abundance. Yellow-rumped warblers, dark-eyed juncos and
ruby-crowned kinglets occurred on 55-87% of the study points with observations
> 1 occurring on 10-55 % of the point counts. These species‘ abundances were
log + 1 transformed and analyzed using Gaussian models. Similar to Lichstein et
al. (2002b), we compared non-spatial Gaussian models with Poisson models and
the results were qualitatively similar, suggesting the Gaussian models described
bird-vegetation associations adequately. The remaining less common species
rarely (<10%) had observations > 1; reducing these species to presence/absence
would result in little information loss. These species were analyzed with logistic
models that would better satisfy distribution assumptions (McCullagh and Nelder
1989, Klute et al. 2002). We followed the approach of Keitt et al. (2002) and
62
Lichstein et al. (2002b) to incorporate spatial pattern in our songbird analyses: 1)
we tested for spatial pattern in the residuals of non-spatial statistical models
(ordinary least-squared regression or logistic regression) using code in
Ecological Archives M072-007-S1 from Lichstein et al. (2002b) to compute
significance tests for correlograms. The code uses 999 permutations to generate
the null distribution and the global significance was assessed at α=0.05,
Bonferroni corrected for 20 lags at 250 m intervals. 2) If pattern was detected in
the residuals, we reanalyzed the data using autoregressive models (autoGaussian or autologistic). The distance to which space was accounted for in
models was the maximum distance that autocorrelation appeared in the
residuals. We then re-examined the residuals of the spatial models for pattern.
Prior to fitting autoregressive models, we examined directional correlograms of
residuals and determined that autocorrelation was isotropic (Legendre and
Legendre 1998). Relative contributions of environment and space to particular
models were assessed through partial R2 (Nagelkerke 1991) for the proportion of
variance explained by i) non-spatial model, ii) space only, iii) environment and
space (autoregressive or autologistic models) (Borcard et al. 1992, Legendre and
Legendre 1998, Lichstein et al. 2002b).
Within the framework described above, we constructed models to evaluate
the effects of forest type on bird abundance and the effects of local, landscape
and abiotic variables on our 3 measures of habitat quality (abundance, frequency
of site occupation, and probability of breeding behaviour). All predictor variables
were subjected to a Spearman‘s correlation. When │r│ was > 0.7, we retained
63
the characteristic that we perceived to have the greater biological relationship
with our species (based on life history and published accounts in the general
area, e.g., Simon et al. 2002). The exceptions were variables we expected
would be difficult for managers to obtain. For example, we retained crown cover
of trees > 10 m tall instead of canopy depth since crown cover can be obtained
from forest inventory.
We were interested the effect of timber productivity on the entire bird
community that occurred within our mature black spruce forests. Hence, rather
than restricting analyses to species typically associated with closed forests we
analyzed all species that occurred on ≥ 5% of the point counts in a single year.
We constructed two types of models; first we used only the forest types as
predictors because forest planning is based on these categories. Second, we
used detailed vegetation descriptions because birds may be responding to more
specific features that are concentrated in more than one forest type. For the
latter, we constructed global models using only variables believed to be important
for our species; these were based on feeding and nesting guilds and bird-habitat
studies conducted in the same study area (Simon et al. 2002) or nearby (Simon
et al. 2000, Schwab et al. 2001). Variables for 1) canopy feeders included,
con>10, con2-10, HT500, 2) snag nesters and/or feeders included, snagT, snag715, snag>15, OB500, EB2000 , 3) stem feeders included, vol, 4) ground feeders
included, blt0-2, ES500, ES2000, 5) edge associated species RD500, RD1000, and 6)
those feeding near water edges, W 500, W 2000. Variance explained was usually
evaluated only on the global model, but partial R2 of other models were computed
64
for comparisons with other studies. If residuals of the global model were
autocorrelated we proceeded with spatial models (described above). We
constructed nested models by eliminating variables from the global model,
beginning with the large extent landscape variables, as we perceived selection
for local vegetation as the most parsimonious explanation of bird habitat
selection (Lichstein 2002a, Fahrig 2003). If the elimination of a variable reduced
the AIC, it was excluded from further models, otherwise it was retained. The
strength of evidence for each model was evaluated by the AIC weight (Burnham
and Anderson 2002). If the partial R 2 due to environmental factors was < 0.15,
we felt the explained variance was too low to be biologically informative, and did
not investigate further models.
Our site reoccupation data were counts (number of years a site was
occupied), but with a finite upper bound (i.e., 3) which violates the assumption of
a Poisson distribution (McCullagh and Nelder 1989). No transformation could
normalize it so we converted site occupation to a binomial variable. For common
species (yellow-rumped warbler, ruby-crowned kinglet and dark-eyed junco) most
sites were occupied in 2 of the 3 years so we considered sites occupied 3 years
as ―reoccupied‖. For less common species (Swainson‘s thrush, gray jay, fox
sparrow) 3 years of occupation was rare so we considered a site ―reoccupied‖ if it
was occupied > 1 year. Most of our reproductive activity observations were of
pairs, so we reduced the ordinal response to binomial categories (0 and 1 = no
evidence, 2-4 = evidence of reproductive activity). These reductions to binomial
variables also facilitated our spatial modelling framework.
65
Post-hoc exploratory data analysis
When vegetation variables that we predicted would be most important for
bird habitat quality (our autoregressive models) explained habitat quality poorly,
we did post-hoc exploratory data analyses. One should interpret these analyses
with caution as they represent hypotheses to be confirmed with independent data
sets (Burnham and Anderson 2002). We used Classification and Regression
Trees (CART) to model bird-habitat quality using 40 uncorrelated local and
landscape variables (Table 3.1). Classification and Regression Trees were
performed on S-plus 2000 (Mathsoft 1999) using the RPART2 library of Atkinson
and Therneau (2000) distributed by Statlib (URL:http://lib.stat.cmu.edu/S/). We
chose CART since it can model higher order interactions and nonlinear
relationships, and makes no assumptions about distribution of variables (De‘ath
and Fabricus 2000). Thus, all variables were the original data, i.e., with no
transformations or conversions to binomial variables. Ordinal variables were
analyzed as classes. Tree size was chosen using the 1-standard error rule, i.e.,
the smallest tree producing a cross validation error within 1 standard error of the
minimum was retained (De‘ath and Fabricus 2000).
Results
Relationships between local vegetation and forest types
The crown cover of conifers > 10 m, densities of stems > 25 cm dbh and
snags > 15 cm dbh increased with productivity class (Fig. 3.1). Crown cover of
conifers > 2-10 m and > 0.5-2 m tall both peaked at intermediate productivities,
with the latter being more pronounced. Crown cover of conifers ≤ 0.5 m and
66
bottom live crown ≤ 0.5 m showed little pattern with productivity class. Broadleaved trees were uncommon and almost exclusive to the high volume spruce
forests. Densities of stems ≥ 7-15 cm dbh were similar among the noncommercial forests (lichen woodland, open spruce) but declined through the
commercial forests. Conversely, densities of stems > 15-25 dbh increased
among the non-commercial forests but remained similar among the commercial
stands. Densities of snags ≥ 7-15 cm dbh were lower on the LW sites than the
remaining forest types.
Bird Detectability
Overall detectability was high, ranging from 0.71 - 0.99% (Fig. 3.2). The
lowest AICs for all species except pine grosbeak were for models that did not
distinguish among forest types (i.e., Mc or M; Table 3.2). For pine grosbeak, the
model that distinguished among forest types (MF) received weak support (Δ AIC
= 0.3 over the next best model), but the estimated detectability in the noncommercial and commercial forests was quite different: 0.89, SE = 0.03; 0.54, SE
= 0.12, respectively.
Effect of forest type and vegetation variables on bird abundance
The abundances of pine siskin in 2001 and Tennessee warbler in 2002
tended to increase with productivity class (R2e = 0.22 and 0.20, respectively) (Fig
3.3), but showed little pattern in other years. All other foliage gleaners and stem
feeders showed little pattern with productivity class (R2e ≤ 0.14; Appendix 3.1).
Of the ground-gleaning species, only hermit thrush demonstrated strong forest
type selection; they were most abundant in lichen woodlands in 2000 and 2001
67
(R2e = 0.46 and 0.26, respectively). White-throated sparrow showed a weak
pattern in 2001; it was absent from the low and medium volume spruce stands
but occurred in the other forest types (R2e = 0.15).
Models of predicted bird abundance-vegetation relationships generally
provided poor explanations of the data and varied within the same species
among years (Appendix 3.2). Where birds showed some relationship with
vegetation, landscape composition tended to be relatively important (Table 3.3).
The best model for white-throated sparrow abundance in 2000 consisted only of
a landscape composition variable and pine siskin in 2001 had a landscape partial
R2e of 0.21. For species where R2e ≥ 0.15, northern waterthrush in 2000 and
2001 was the only species without a landscape variable in the best model. But,
other models for that species which received weaker support did contain
landscape composition variables. Despite being selected in the best model, the
partial R2 for landscape composition only models were low for hermit thrush (≤
0.01). For the remaining species, 55-90% of the variance explained by the global
models can be attributed to landscape variables. This indicates that landscape
composition is of equal or of greater importance than local vegetation.
The residuals of our global bird-vegetation models were significantly
autocorrelated for 14 species-year combinations at distances between 250 m
and 4375 m (Appendix 3.3). The residuals of the autoregressive and autologistic
models showed no autocorrelation indicating the assumptions of independence
were valid. Partial R2 for the spatial parameters of these models ranged from
0.02 – 0.48, often higher than the vegetation variables within those models.
68
Effect of forest type and vegetation variables on bird site reoccupancy and
reproductive activity
Reproductive activity did not show any strong consistent trends across
forest types for any species (Fig 3.4) and vegetation models explained little
variation in reproductive activity, (R2 < 0.15) (Appendix 3.4). Yellow-rumped
warbler, in 2002 was the only species where reproductive activity was highest in
high volume spruce. However, this pattern was weak and, in the previous year,
reproductive activity was lowest in that stand type. Despite poor model fit,
yellow-rumped warbler site reoccupancy was highest in high volume spruce while
ruby-crowned kinglet occupancy declined from lichen woodland to high volume
spruce (Fig. 3.4, Appendix 3.4). Vegetation models explained little variation in
site occupancy (Appendix 3.5).
Post-hoc exploratory analysis
Our post-hoc exploratory analysis failed to provide adequate explanations
of the data for most species. Hermit thrush abundances were positively
associated with lake edges within 2000 m while fox sparrow re-occupancy was
negatively related to heavy timber within 500 m and positively related to elevation
(R2 = 0.38 and 0.35 respectively).
Discussion
The most striking result of this study was the poor relationships between
birds and vegetation; relationships were either non-existent or weak and
inconsistent among years. While forest types were somewhat similar in that they
were all dominated by black spruce, there were structural differences. It seemed
69
reasonable to expect canopy and stem feeders to be present in all forest types
but with their densities increasing in proportion to tree height, crown cover and
stem densities and timber volume. These features vary on average 300 – 500 %
across forest types, suggesting a similar variation in canopy and stem feeding
birds would be observed. Our weak bird-habitat relationships could be related to
the life history characteristics of the bird species analyzed. Many of our canopy
and stem feeding species (e.g., yellow-rumped warblers) can tolerate canopy
openings (Hunt and Flashpohler 1998), suggesting they may not require dense
crowns of tall trees. Many of the species known to have strong associations with
tall conifers (e.g., brown creepers, Certhia americana) were rare or absent on our
landscape. This may be because the amount of proportion of high volume
spruce on our landscape was too low to support viable population of these
species. Others (e.g. boreal chickadee and three-toed woodpeckers) are also
considered specialized and sensitive to mature forest loss (Imbeau et al. 2001,
Schmeigelow and Mönkkönen 2002) hence, we expected to see stronger
associations with these species. Many species analyzed were ground and shrub
foragers that tend to be associated with open spaces having low growing broadleaved shrubs (e.g., white-throated sparrow). These were included because we
were interested in productivity effects on the entire bird community that occurred
with Labrador‘s black spruce forests. There was very little variation in low
growing broad-leaved shrubs across our productivity gradient, thus weak
vegetation associations for these species is not surprising.
70
It is difficult to conclude firmly that birds were not associated with any
environmental variables because it is always possible that our field and analytical
methods biased our results. We feel that methodological bias is low in our study
and outline our reasons below. Point counts have been criticized because
detections can vary with vegetation cover (Schieck 1997). Except for pine
grosbeak, we found no evidence of bias, indicating that abundances we derived
from point counts reflect the densities of the remaining species.
Failing to include key habitat variables and autocorrelation or modeling at
the wrong scale can lead to poor model performance (Vaughn and Ormerod
2003), but the vegetation variables used in our regression models were based on
the species‘ life history. These and similar variables have successfully predicted
bird abundance in studies of close geographic proximity to ours (Schwab et al.
2001) and in more distant locations (Schwab and Sinclair 1994, Drapeau et al.
2000, Lichstein et al. 2002a). Further, our post-hoc exploratory data analysis
(CART models) used these and many other uncorrelated variables that we
thought could influence bird distribution. Both procedures explained little
variation in bird habitat quality so it is improbable that we missed key habitat
features for all species within our range of forest types. We used 3 statistical
techniques to evaluate bird-vegetation relationships: Poisson regression (not
reported), Gaussian/logistic regression (incorporating spatial autocorrelation
where detected) and Classification and Regression Trees. These are common
and successful analytical techniques in ecological literature, suggesting our
results were not an artifact of the modeling techniques or the failure to
71
incorporate autocorrelation. It is doubtful that our data were recorded at the
wrong scale since there is abundant literature describing local and landscape
bird-vegetation relationships at 100 m and 500- 2000 m extents respectively
(Hagan et al. 1996, Drapeau et al. 2000, Lichstein et al. 2002a). There is no
reason to expect that birds in our study respond to different scales than birds
elsewhere.
Vaughn and Ormerod‘s (2003) concern of incomplete coverage of
environmental space is related to the weak bird-vegetation relationships. We are
confident that we adequately sampled the range of timber volume among natural
mature forests in Labrador since only 3% of Labrador‘s inventoried land base is
outside of our sampled timber volume gradient. Yet our sites were dominated by
a single tree species and vary mostly in height and densities. This relatively low
contrast among these forests may be partly responsible for the weak birdvegetation relationships. If a broader range of forest structures were included
(e.g., areas with no trees) our models would likely have higher explained
variance, (e.g., Schwab et al. 2001).
Despite the general weak bird-vegetation relationships, where birdvegetation relationships existed there was a relatively strong landscape effect.
The influence of landscape content is expected to be inversely proportional to the
amount of habitat on the landscape (Andrén 1994, Drolet et al. 1999, Fahrig
2003). Early successional forests and watercourses were uncommon in our
study sites, suggesting a landscape effect would be more likely for species
associated with these areas. Consistent with this reasoning, 3 of the 5 of our
72
species whose abundances were related to landscape composition variables,
black-backed woodpecker, Tennessee warbler, and white-throated sparrow, are
characteristic of early successional forests while another is characteristic of water
edges (northern waterthrush). Where vegetation was able to explain some
variation in abundance, partial R2 values indicate landscape variables were more
important than local variables in determining bird abundances. This result is
contrary to other studies in forested areas (Hagan and Meehan 2002, Lichstein et
al. 2002a) but similar to Drapeau et al. (2000) who concluded that local and
landscape variables are similar in importance. However, the potential remains
for this result to be somewhat confounded by the different variables
characterizing local and landscape vegetation. It would be logistically impossible
to sample vegetation extents of 2000 m at the same intensity as our local
vegetation plots. Thus, we had to rely on aerial photography as most other
researchers (e.g., Hagan and Meehan 2002, Drapeau et al. 2000, Lichstein et al.
2002a). Our landscape variables were coarser than our local variables,
suggesting our reported landscape effects were conservative.
It is difficult to identify the source of autocorrelation in the residuals of our
non-spatial models. It is possible that unmeasured, spatially structured
environmental variables contributed to the observed autocorrelation, despite our
efforts to include all important vegetation variables. Alternatively, conspecific
attraction has been documented in songbirds (Ward and Schlossberg 2004) and
could have caused the autocorrelation in our bird-vegetation models as
suggested by Lichstein et al. (2002a).
73
Our 3 measures of habitat quality did not show a strong positive increase
with productivity for any species suggesting that low productivity forests will
support most forest birds in this region. We are confident in this result because
these different, but complimentary measures show similar patterns. Our results
imply that reducing the amount of high volume mature stands would not likely
have long-term negative impacts on birds in this landscape. This is partly
because many species in this system are characteristic of open areas or can
tolerate canopy openings. Other species typically associated with closed forests
(e.g., boreal chickadees) showed weak responses to vegetation. However, the
relatively strong relationship between pine siskins and the proportion of heavy
timber within 500 m, suggesting that the effects of cutting heavy timber could
reduce pine siskin numbers beyond the actual harvesting area. Where
vegetation models explained some variation in bird habitat quality, there was
often a strong influence of landscape relative to local extents. However, this
effect was largely confined to species typical of early successional forests.
Acknowledgements
Funding was provided by the Newfoundland and Labrador Department on
Natural Resources and Human Resources and Skills Development Canada.
Thanks are extended to those assisting in data collection, notably, B. Campbell,
J. Colbert, L. Elson, R. Flynn, M. Michelin, K. Mitchell, R. Neville, F. Phillips, B.
Rodrigues, F. Taylor. D. Goulding and D. Jennings assisted in GIS work and C.
Bourque assisted with, and provided a program to compute abiotic variables.
Thanks are also extended to the Innu Nation. J. Lichstein provided statistical
74
advice and G. Farnsworth provided code to assess forest type-detectability
differences. M. Betts, G. Forbes, D. Keppie, and F. Phillips reviewed earlier
drafts of the manuscript.
References
Andrén, H. 1994. Effects of habitat fragmentation on birds and mammals in
landscape with different proportions of suitable habitat: a review. – Oikos
71:355-366.
Apsey, A., Laishley, D., Nordin, V. and Paillé, G. 2000. The perpetual forest:
Using lessons from the past to sustain Canada‘s forests in the future. –
For. Chron. 76:29-53.
Bibby, C.J., Burgess, N.D. and Hill, D.A. 1993. Bird census techniques. Academic Press, Toronto.
Borcard, D., Legendre, P. and Drapeau, P. 1992. Partialling out the spatial
component of ecological variation. – Ecology 72: 1045-1055.
Burnett, C., Fall, A., Tomppo, E. and Kalliola, R. 2003. Monitoring current status
of and trends in boreal forest land use in Russian Karelia. – Conserv.
Ecol. 7:8. [online] URL: http://www.consecol.org/vol7/iss2/art8.
Burnham, K.P. and Anderson, D.R. 2002. Model selection and multimodel
inference. A practical information-theoretic approach, Second Edition. –
Springern, New York.
Burton, P.J., Messier, C., Weetman, G.F. Prepas, E.E. Adamowicz, W.L. and
Tittler, R. 2003. The current state of the boreal forest and the drive for
change. – In: Burton, P.J. Messier, Smith, D.W. and Adamowicz, W.L.
75
(eds.), Towards sustainable management of the boreal forest. NRC
Research Press, Ottawa, pp. 1-40.
Canada Committee on Ecological Land Classification. 1989. Ecoclimatic Regions
of Canada: First approximation. – Environment Canada, Ecological Land
Classification Series No. 23. Canadian Wildlife Service.
Cressie, N.A. 1993. Statistics for spatial data. – John Wiley and Sons, New York.
Daubenmire, R. 1968. Plant communities. A textbook of plant synecology. –
Harper & Row Publishers, London
De‘ath, G. and Fabricius, K.A. 2000. Classification and regression trees: A
powerful yet simple technique for ecological data analysis. – Ecology
81:3178-3192.
DeGraaf, R.M. and Chadwick, N.L. 1987. Forest type, timber size class, and
New England breeding birds. – J. Wildl. Manag. 51:212-217.
Department of Forest Resources and Agrifoods. 2003. Newfoundland and
Labrador sustainable forest management strategy. – Newfoundland and
Labrador Department of Forest Resources and Agrifoods.
Diamond, A.W. 1999. Concluding remarks: content versus context in forest bird
research. – In: Diamond, A.W. and Nettleship, D.N. (eds.), Biology and
conservation of forest birds. Society of Canadian Ornithologists Special
Publication No. 1, pp. 139-143.
Drapeau, P., Leduc, A., Giroux, J-. F., Savard, J-. P.L., Bergeron, Y. and Vickery,
W. L. 2000. Landscape-scale disturbances and changes in bird
communities of boreal mixed-wood forests. – Ecol. Monogr. 70: 423-444.
76
Doran, P.J., Gulezian, P.Z. and Betts, M.G. 2005. A test of the mobbing
plackback method for estimating bird reproductive success. – J. Field.
Ornithol. 76:227-233.
Drolet, B., Desrochers, A. and Fortin, M.-J. 1999. Effects of landscape structure
on nesting songbird distribution in a harvested boreal forest. – Condor
101:699-704.
Duinker, P.N., Nilsson, S. and Chipeta, M.E. 1998. Foresty for sustainable
development and global fibre supply. – Unasylva 49:3-10.
Ehrlich, P.R., Dobkin, D.S. and Wheye, D. 1988. The birder‘s handbook. – Simon
and Schuster Inc., New York.
Emlen, J.T. 1967. A rapid method for measuring arboreal canopy cover. –
Ecology 48:158-159.
Environmental Systems Research Institute (ESRI). 1999. ArcView GIS 3.2. –
ESRI.
Fahrig, L. 2003. Effects of habitat fragmentation on biodiversity. – Annu. Rev.
Ecol. Evol. Syst. 34:487-515.
Fall, A., Fortin, M-.J., Kneeshaw, D.D., Yamasaki, S.H., Messier, C., Bouthillier,
L. and Smith, C. 2004. Consequences of various landscape-scale
ecosystem strategies and fire cycles on age-class structure and harvest in
boreal forests. – Can. J. For. Res. 34:310-322.
Farnsworth, G.L., Pollock, K.H., Nichols, J.D., Simons, T.R., Hines, J.E. and
Sauer, J.R. 2002. A removal model for estimating detection probabilities
from point-count surveys. – Auk 119:414-425.
77
Folkard, N.F.G.and Smith, J.N.M. 1995. Evidence for bottom-up effects in the
boreal forest: Do passerine birds respond to large-scale experimental
fertilization? – Can. J. Zool. 73:2231-2237.
Forsman, J.T., Mönkkönen, M., Helle, P. and Inkeröinen, J. 1998. Heterospecific
attraction and food resources in migrants‘ breeding patch selection in
northern boreal forest. – Oecologia 115:278-286.
Franzreb, K.E. 1978. Tree species used by birds in logged and unlogged mixedconiferous forests. – Wilson Bull. 90:221-238.
Gray, S.L. 1995. A descriptive forest inventory of Canada‘s forest regions. –
Canadian Forest Service, Petawawa National Foresty Institute,
Information Report PI-X-122.
Gumpertz, M.L., Graham, J.M. and Ristaino, J.B. 1997. Autologistic model of
spatial pattern of Phytophthora epidemic in bell pepper: effects of soil
variables on disease presence. – J. Agric. Biol. Environ. Stat. 2:131-156.
Gunn, J.S., DesRochers, A., Villard, M.-A. and Ibarzabal, J. 2000. Playbacks of
mobbing calls of black-capped chickadees as a method to estimate
reproductive activity of forest birds. – J. Field Ornithol. 71: 472-483.
Haas, C.A. 1997. What characteristics of shelter belts are important to breeding
success and return rates of birds? – Am. Midl. Nat. 137:225-238.
Hagan, J.M., Vander Haegen, M.W. and Mckinley, P.S. 1996. The early effects of
forest fragmentation on birds. – Conserv. Biol. 10:188-202.
78
Hagan, J.M. and Meehan, A.L. 2002. The effectiveness of stand-level and
landscape-level variables for explaining bird occurrence in an industrial
forest. – For. Sci. 48:231-242.
Hames, R.S., Rosenberg, K.V., Lowe, J.D. and Dhondt, A.A. 2001. Site
reoccupation in fragmented landscapes: testing predictions of
metapopulation theory. – J. Anim. Ecol. 70:182-190.
Herms, D.A. and Mattson, W.J. 1992. The dilemma of plants: to grow or defend.
– Q. Rev. Biol. 67:283-335.
Howlett, J.S. and Stutchbury, B.J.M. 2003. Determinants of between-season
site, territory, and mate fidelity in hooded warblers (Wilsonia citrina). – Auk
120:457-465.
Hutto, R.L. 1998. Using landbirds as an indicator species group. – In: J. Marzluff
and Sallabanks, R. (eds.), Avian conservation and management, Island
Press, Washington, pp. 75-92.
Hunt, P.D. and Flaspohler, D.J. 1998. Yellow-rumped warbler. – In: A. Poole and
F. Gill (eds.), The Birds of North America. No. 376. Academy of Natural
Sciences: Philadelphia; The American Ornithologists' Union.
Imbeau, L., Mönkkönen, M. and Desrochers A. 2001. Long-term effects of
forestry on birds of the eastern Canadian boreal forests: a comparison
with Fennoscandia. – Conserv. Biol. 15:1151-1162.
Kareiva, P. and Marvier, M. 2003. Conserving biodiversity coldspots. – Am. Sci.
91:344-351.
Keitt, T.H., Bjørnstad, O.N., Dixon, P.M. and Citron-Pousty, S. 2002. Accounting
79
for spatial pattern when modelling organism-environment interactions. –
Ecography 25:616-625.
Klute, D.S., Lovallo, M.J. and Tzilkowski, W.M. 2002. Autologistic regression
modelling of American woodcock habitat use with spatially dependant
data. – In: Scott, J.M., Heglund, P.J., Morrison, M.L., Haulfer, J.B.,
Raphael, M.G., Wall, W.A. and Sampson, F.B. (eds.), Predicting species
occurrence issues of accuracy and scale. Island Press, Washington, pp.
335-355.
Korovin, G., Karpov, E., Islaev, A., Nefedjev, V., Efremov, D., Sedych, V.,
Sokolov, V., Schmidt, T., Blauberg, K., Ljusk Eriksson, O., Nilsson, S.,
Raile, G., Sallnäs, O. and Shvidenko, A. 1998. Siberia and Far East
Russia‘s future wood supply: an analysis. – International Institute for
Applied Systems Analysis, Laxenburg, Austria. Interim Report IR-987-001.
Lack, D. 1933. Habitat selection in birds. – J. Anim. Ecol 2:239-262.
Larsson S. and Danell, K. 2001. Science and the management of boreal forest
biodiversity. – Scand. J. For. Res. Suppl. 3:5-9
Legendre, P. and Legendre, L. 1998. Numerical ecology, 2nd English Edition. –
Developments in Environmental Modelling Vol. 20. Elsevier.
Lichstein, J. W., Simons, T.R. and Franzreb, K.E. 2002a. Landscape effects on
breeding songbird abundance in managed forests. – Ecol. Appl. 12:836857.
80
Lichstein, J. W., Simons, T.R. Shriner, S.A. and Franzreb, K.E. 2002b. Spatial
autocorrelation and autoregressive models in ecology. – Ecol. Monogr.
72:445-463.
MacArthur, R.H. and MacArthur, J.W. 1961. On bird species diversity. – Ecology
42:594-598.
Mathsoft 1999. S-plus 2000 for Windows.
Mazerolle, M.J. and Villard, M-.A. 1999. Patch characteristics and landscape
context as predictors of species presence and abundance: A review. –
Écoscience 6:117-124.
McCullagh, P. and Nelder, J.A. 1989. Generalized linear models. – Monographs
on statistics and applied probability 37, Chapman and Hall, New York.
McGarigal, K. and McComb, W 1995. Relationships between landscape
structure and breeding birds in the Oregon coast range. – Ecol. Monogr.
65:235-265.
Morse, D.H. 1989. American warblers: An ecological and behavioral perspective.
– Harvard University Press, Cambridge.
Nagelkerke, N.J.D. 1991. A note on the general definition of the coefficient of
determination. – Biometrika 78: 691-692.
Nilsson, S., Colberg, R., Hagler, R. and Woodbridge P. 1999. How sustainable
are North American wood supplies? – International Institute for Applied
Systems Analysis. Interim Report IR-99-003/January.
Nilsson, S.G., Hedin, J. and Niklasson, M. 2001. Biodiversity assessment in
boreal and nemoral forests. – Scand. J. For. Res. Suppl 3:10-26.
81
Niemelä, J., Larsson, S. and Simberloff, D. 2001. Concluding remarks–finding
ways to integrate timber production and biodiversity in Fennoscandian
forestry. – Scand. J. For. Res. Suppl 3:119-123.
Niemi, G., Hanowski, J., Helle, P., Howe, R., Mönkkönen, M., Venier, L. and
Welsh, D. 1998. Ecological sustainability of birds in boreal forests. –
Conserv. Ecol. [online] 2:17. URL:http://www.consecol.org/vol2/iss2/art17.
Parker, G.R., Kimball, D.G. and Dalzell, B. 1994. Bird communities breeding in
selected spruce and pine plantations in New Brunswick. – Can. Field-Nat.
108: 1-9.
Patuxent Wildlife Research Center. 2002. North American Bird Banding Manual.
Draft. - URL:http://www.pwrc.usgs/gov/BBL/manual.
Porneluzi, P.A. 2003. Prior breeding success affects return rates of territorial
male oven birds. – Condor 105:73-79.
Pulliam, H.R. 1988. Sources, sinks and population regulation. – Am. Nat.
132:652-661.
Pulliam, H.R. and Danielson, B.J. 1991. Sources, sinks, and habitat selection: a
landscape perspective on population dynamics. – Am. Nat. 137: S50-S66.
Raivio, S., Normark, E., Pettersson, B. and Salpakivi-Salomaa, P. 2001. Science
and management of boreal forest biodiversity–forest industries‘ views. –
Scand. J. For. Res. Suppl. 3:99-104.
Rodenhouse, N.L. Sherry, T.W. and Holmes, R.T. 1997. Site-dependent
regulation of population size: A new synthesis. – Ecology 78:2025-2042.
82
Roth, R.R. 1976. Spatial heterogeneity and bird species diversity. – Ecology
57:773-782.
Rowe, J.S. 1972. Forest regions of Canada. – Department of Environment,
Canadian Forest Service Publication No. 1300. Crown Copyright.
Saunders, D.A., Hobbs, R.J. and Margules, C.R. 1991. Biological consequences
of ecosystem fragmentation: A review. – Conserv. Biol. 5:18-32.
Schieck, J. 1997. Biased detection of bird vocalizations affects comparisons of
bird abundance among forested habitats. – Condor 99:179-180.
Schmiegelow, F.K.A. and Mönkkönen. 2002. Habitat loss and fragmentation in
dynamic landscapes: avian perspectives from the boreal forest. – Ecol.
App. 12-375-389.
Schneider, R.R., Stelfox, B.J., Boutin, S. and Wasel, S. 2003. Managing the
cumulative impacts of land uses in the Western Canadian Sedimentary
Basin: a modeling approach. – Conserv. Ecol. 7:8. [online] URL:
http//www.consecol.org/vol7/iss1/art8.
Schwab, F.E. and Sinclair, A.R.E. 1994. Biodiversity of diurnal breeding bird
communities related to succession in the dry Douglas-fir forests of
southeastern British Columbia. – Can. J. For. Res. 24:2043-2040.
Schwab, F.E., Simon, N.P.P. and Carroll, C.G. 2001. Breeding songbird
abundance related to secondary succession in the subarctic forests of
western Labrador. – Écoscience 8:1-7.
83
Seagle, S.W. and Sturtevant, B.R. 2005. Forest productivity predicts invertebrate
biomass and ovenbird (Seiurus aurocapillus) reproduction in Appalachian
landscapes. – Ecology 86:1531-1539.
Simon, N.P.P., Schwab, F.E. and Diamond A.W. 2000. Patterns of bird
abundance in relation to logging in western Labrador. – Can. J. For. Res.
30: 257-263.
Simon, N.P.P., Schwab, F.E. and Otto, R.D. 2002. Songbird abundance in clearcut and burned stands: a comparison of natural disturbance and forest
management. – Can. J. For. Res. 32:1343-1350.
Smith, W.P., Twedt, D.J., Cooper, R.J., Wiedenfield, D.A., Hamel, P.R. and Ford,
R.P. 1997. Sample size and allocation of effort in point count sampling of
birds in bottomland hardwood forests. – In: Ralph, C.J., Sauer, J.R. and
Droedge, S. (eds.), Monitoring bird populations by point counts. General
Technical Report PSW-GTR-149, pp 7-17.
Sutherland, W.J. 1997. From individual behaviour to population ecology. Oxford
Series in Ecology and Evolution. – Oxford University Press.
Trzcinski, K.M., Fahrig, L. and Merriam, G. 1999. Independent effects of forest
cover and fragmentation on the distribution of forest breeding birds. –
Ecol. Appl. 9:586-593.
Van Horne, B. 1983. Density as a misleading indicator of habitat quality. – J.
Wildl. Manag. 47:893-901.
84
Vaughn, I.P. and Ormerod, S.J. 2003. Improving the quality of distribution models
for conservation by addressing shortcomings in the field collection of
training data. – Conserv. Biol. 17: 1601-1611.
Venier, L.A. and Pearce, J.L. 2004. Birds and indicators of sustainable forest
management. – For. Chron. 80:61-66.
Vickery, P.D., Hunter, Jr., M.L. and Wells, J.V. 1992. Is density an indicator of
breeding success? – Auk 109:706-710.
Wade, T.G., Riitters, K.H., Wickham, J.D. and Jones, B.K. 2003. Distribution and
causes of global forest fragmentation. – Conserv. Ecol. 7:7. [online]
URL:http//www.consecol.org/vol7/iss2/art7.
White, G.C. 1983. Numerical estimation of survival rates from band-recovery
data. – J. Wildl. Manag. 47:716-728.
Wilton, W.C. 1965. The forests of Labrador. – Can. Dept. For., Dept. For. Publ.
No. 1066, Ottawa.
Zasada, J.C., Gordon, A.G., Slaughter, C.W. and Duchesne, L.C. 1997.
Ecological considerations for the sustainable management of the North
American boreal forests. – International Institute for Applied Systems
Analysis, Laxenburg, Austria. Interim Report IR-97-024.
85
Table 3.1. Descriptions and codes of variables used in bird-habitat models.
Code
Variable description
Stand level variables
blt<0.5
Broad-leaved tree crown cover < 0.5 m tall
blt>2
Broad-leaved tree crown cover > 2 m tall
blt0-2
Broad-leaved tree crown cover < 2 m tall
blt2-10
Broad-leaved tree crown cover > 2-10 m tall
blt>10
Broad-leaved tree crown cover > 10 m tall
con< 0.5
Conifer crown cover < 0.5 m tall
con0.5-2
Conifer crown cover > 0.5 - 2 m tall
con2-10
Conifer crown cover > 2-10 m tall
con>10
Conifer crown cover > 10 m tall
con>2
Conifer crown cover > 2 m tall
stem7-15
Number of stems >7 - 15 cm dbh per 400 m2
stem15-25
Number of stems >15 - 25 cm dbh per 400 m2
stem>25
Number of stems >25 cm dbh per 400 m2
snag7-15
Number of snags >7 - 15 cm dbh per 400 m2
snag>15
Number of snags >15 cm dbh per 400 m2
bot0.5
Tree crown cover < 0.5 m from ground
snagT
Total snags >7 cm dbh per 400 m2
Vol
Timber volume (m3) per 400 m2
RaS
Rivers and streams (m) within 100 m
Lake
Lake edges (m) within 100 m
W
Water edges (m) within 100 m (RaS + Lake)
Herbs
Herbaceous vegetation
Tall shrubs
Tall shrubs in the Ericaceae family
Landscape variables
RD500
Road length (m) within 500 m of point
RaS500
Rivers and streams (m) within 500 m of point
Lake500
Lake edges (m) within 500 m of point
HT500
Heavy timber (≥ 100 m3/ha) % within 500 m of point
OB500
Old burn (> 15 years old) % within 500 m of point
OC500
Old cut (> 15 years old) % within 500 m of point
YC500
Young cut (< 15 years old) % within 500 m of point
RD2000
Road length (m) within 2000 m of point
RaS2000
Rivers and streams (m) within 2000 m of point
Lake2000
Lake edges (m) within 2000 m of point
OC2000
Old cut (> 15 years old) within 2000 of point m
YC2000
Young cut (< 15 years old) within 2000 m of point
ES500
Early successional forest within 500 m of point (OB500 + OC500)
ES2000
Early successional forest within 2000 m of point (OB2000 + OC2000)
OB2000
Old burn (> 15 years old) % within 2000 m of point
W500
Water edges (m) within 500 m of point (RaS500 + Lake500)
W2000
Water edges (m) within 2000 m of point (RaS2000 + Lake2000)
Other variables
slope
Slope %
elevation
Elevation (m)
A
Spatial parameter; ρ in Autoregressive models, autocovariate in autologistic
models
86
Table 3.2. The effect of forest type on songbird detectability as determined by
SURVIV. Models with most support, i.e., lowest Akaike‘s Information Criterion
(AIC) are underlined. Model MFC distinguishes among forest type and
incorporated heterogeneity in detection probability within species, model MF
distinguishes among forest type, model MC incorporated heterogeneity in
detection probability within species, model M does not distinguish among forest
types or incorporate heterogeneity.
Speciesa
Codeb
*Yellow-rumped warbler, Dendroica coronata
*Ruby-crowned kinglet, Regulus calendula
*Dark-eyed junco, Junco hyemalis
*Fox sparrow, Passerella iliaca
YRWA
RCKI
DEJU
FOSP
MFC
7.95
12.43
9.17
16.98
Yellow-rumped warbler, Dendroica coronata
Ruby-crowned kinglet, Regulus calendula
Dark-eyed junco, Junco hyemalis
Fox sparrow, Passerella iliaca
Gray jay, Perisoreus canadensis
Pine Grosbeak, Pinicola enucleator
Black-backed woodpecker, Picoides arcticus
Hermit thrush, Catharus guttatus
Northern waterthrush, Seiurus noveboracensis
Pine siskin, Carduelis pinus
Three-toed woodpecker, Picoides tridactylus
Tennessee warbler, Vermivora peregrina
Swainson‘s thrush, Catharus ustulatus
Boreal chickadee, Poecile hudsonicus
White-throated sparrow, Zonotrichia albicollis
YRWA
RCKI
DEJU
FOSP
GRJA
PIGR
BBWO
HETH
NOWA
PISI
TTWO
TEWA
SWTH
BOCH
WTSP
3.69
0.96
3.63
4.69
3.54
3.52
5.13
4.42
5.28
2.81
4.14
2.23
4.90
4.81
6.08
a
AIC
MF
MC
24.50
0.00
10.77
0.00
4.36
1.12
7.01
2.00
M
18.50
3.17
0.00
0.00
18.22
12.45
3.80
1.74
1.03
0.00
1.75
1.50
1.28
1.89
0.76
6.39
1.89
1.99
2.08
16.25
11.11
2.17
0.00
0.00
0.30
0.00
0.00
0.00
0.00
0.00
4.43
0.00
0.00
0.00
0.00
0.00
0.00
2.01
1.46
2.18
2.00
1.90
2.00
0.54
1.99
0.00
1.02
0.89
2.63
Models identified with * distinguished among all 5 forest types where the
remainder distinguished only between non-commercial (LW, OS) and commercial
(LS, MS, HS).
b
Abbreviation of common names follows bird banding manual (Patuxent Wildlife
Research Center 2002).
87
Table 3.3. Best models (i.e., lowest AIC) and partial R2e due to local and
landscape variables as determined by logistic and autologistic regression for bird
species-years where R2e ≥ 0.15 for the global model. The full list of models
examined (global and nested for all species) and evidence for their support is in
Appendix 3.2. Predictor codes are in Table 3.1, species codes are in Table 3.2.
Model
PISI01 = - 5.10 - 4.30 * con2-10 + 0.05 * HT500
TEWA00 = - 5.18 + 0.09 * ES500 - 0.06 *RD500 + 1.00e-4 * RD2000 + 0.37 *A
BBWO01 = -2.97 - 0.52 * OB500 + 0.29 * EB2000
HETH00 = 0.38 + 4.01 * blt0-2 -26.02 * con>10 - 0.60 * snagT + 0.54 *
OB500
HETH01 = = - 0.52 - 10.60 * con>10 – 0.45 * snagT + 0.23 * OB500
NOWA00= -3.77 + 8.10 * blt02 + 9.26e-3 * W
NOWA01 = = -2.63 + 7.90e-3 * W
NOVA02 -= = -5.85 + 7.26 * blt02 + 5.38e-3 * W + 4.54e-4 W 500 + 17.01 * A
WTSP00 = -2.65 + 0.06 * ES500 - 2.57 *A
WTSP01 = = -3.20 - 8.93 * con>10 + 0.11 * ES500
88
R2e local
0.04
0.02
0.01
0.43
R2e land
0.14
0.30
0.15
0.06
0.18
0.18
0.11
0.02
0.00
0.08
≈ 0.00
0.00
0.00
0.14
0.16
0.19
Figure 3.1. Mean and standard error of vegetation features in relation to forest
type. Conifer, broad-leaved tree (BLT), and live crown (LC) ≤ 0.5 m abundance
is expressed as crown cover by height zone, stems and snags are numbers per
400 m2 by diameter class (cm), and volume is timber volume (m3) 400 m2. LW =
lichen woodland, OS = open spruce, LS = low volume spruce, MS = medium
volume spruce and HS = high volume spruce.
89
Figure 3.2. Species specific detection probabilities and standard errors over a 10
minute point count. Probabilities are estimates of the best model, i.e., lowest
AIC, as determined by the count removal method. Species codes are in Table
3.2.
90
Figure 3.3. Bird abundance and SE in relation to forest type. Abundance is the
mean number recorded across point count stations for YRWA, RCKI, and DEJU.
For the remaining species, abundance is the proportion of point counts for which
an individual was present. LW = lichen woodland, OS = open spruce, LS = low
volume spruce, MS = medium volume spruce and HS = high volume spruce.
Species codes in Table 3.2.
91
Figure 3.4. Reproductive activity and site occupancy (probability, SE) in relation
to forest type. LW = lichen woodland, OS = open spruce, LS = low volume
spruce, MS = medium volume spruce and HS = high volume spruce. Species
codes are in Table 3.2.
92
Appendix 3.1. Goodness of fit statistics for yearly models of bird abundance in
relation to forest type. Non-spatial models were least squares regression for
abundant species (YRWA, RCKI, DEJU) and logistic regression for less
abundant species. When spatial dependence was detected, conditional
autoregressive were used for abundant species and autologistic models for less
abundant species. See methods for details. Species codes are in Table 3.2.
Species
YRWA
Guilda
FG
PISI
FG
RCKI
FG
BOCH
FG
SWTH
FG
PIGR
FG
TEWA
FG
BBWO
SF
TTWO
SF
GRJA
GG
HETH
GG
FOSP
GG
DEJU
GG
NOWA
GG
Year
2000
2001
2002
2000
2001
2002
2000
2001
2002
2000
2001
2002
2000
2001
2002
2000
2001
2002
2000
2001
2002
2000
2001
2002
2000
2001
2002
2000
2001
2002
2000
2001
2002
2000
2001
2002
2000
2001
2002
2000
b
Neighbor (m)
2250
2750
500
3250
4000
750
3750
1500
250
c
R2e
0.03
0.04
0.11
0.09
0.20
0.06
0.08
0.11
0.05
0.02
0.05
0.02
0.08
0.03
0.06
0.03
0.06
0.12
0.22
0.08
0.03
0.02
0.05
0.04
0.13
0.05
0.03
0.02
0.05
0.02
0.46
0.26
0.11
0.07
0.10
0.13
0.05
0.02
0.09
0.13
93
R2 ρ
R2t
0.03
0.11
<0.01
0.11
0.16
0.12
0.16
0.14
0.18
0.11
0.01
0.03
0.29
0.29
0.47
0.04
0.48
0.08
WTSP
GG
2001
2002
2000
2001
2002
3750
4500
0.13
0.12
0.05
0.15
0.13
0.35
0.38
0.07
0.16
a
Predominant foraging guilds by species from Ehrlich et al. (1988). FG = foliage
feeder, SF = stem feeder, GG = ground feeder.
b
Neighbor is the distance at which space is accounted for in spatial models
c
R2e = variation explained by environment (non-spatial models), R2 ρ = Partial R2
of the spatial parameter ρ, the variance explained exclusively by space, R 2 t =
total variance explained by both environment and space.
94
Appendix 3.2. Goodness of fit statistics for yearly global models of bird
abundance in relation to vegetation characteristics. Non-spatial models were
least squares regression for abundant species (YRWA, RCKI, DEJU) and logistic
regression for less abundant species. Predictor codes are in Table 3.1, species
codes are in Table 3.2. When spatial dependence was detected, conditional
autoregressive models were used for abundant species and autologistic models
for less abundant species.
Modela
b
Variance explained
AIC

AIC
w
2
YRWA00 = con>10 + con2-10 + HT500
YRWA01 = con>10 + con2-10 + HT500
YRWA02 = con>10 + con2-10 + HT500
R e = 0.02
R2e = 0.01
R2e = 0.03; R2ρ =0.02; R2 t =0.05; N = 1250
PISI00 = con>10 + con2-10 + HT500
R2e = 0.04
PISI01= con>10 + con2-10 + HT500
- 5.10 - 4.30 * con2-10 + 0.05 * HT500
- 6.33 + 0-.05 * HT500
- 6.68+ 3.51 * con>10 + 0.05*HT500
- 5.51 + 2.56 * con>10 - 3.51 * con2-10 + 0.05 * HT500
R2e = 0.26
PISI02= con>10 + con2-10 + HT500
R2e = 0.03
RCKI00 = con>10 + con2-10 + HT500
R2e = 0.09; R2ρ =0.06; R2 t =0.11; N = 500
RCKI 01 = con>10 + con2-10 + HT500
R2e = 0.01; R2ρ =0.11; R2 t =0.11; N = 3250
RCKI 02 = con>10 + con2-10 + HT500
R2e = 0.07; R2ρ =0.01; R2 t =0.04; N = 4000
BOCH00 = con>10 + con2-10 + snagT + HT500
R2e = 0.06
BOCH01 = con>10 + con2-10 + snagT + HT500
R2e = 0.05
BOCH02 = con>10 + con2-10 + snagT + HT500
R2e = 0.01
SWTH00 = con>10 + con2-10 + blt0-2 + HT500
R2e = 0.08; R2ρ =0.06; R2 t =0.12; N = 1500
SWTH01 = con>10 + con2-10 + blt0-2 + HT500
R2e = 0.02; R2ρ =0.27; R2 t =0.29; N = 750
SWTH02 = con>10 + con2-10 + blt0-2 + HT500
R2e = 0.01; R2ρ =0.01; R2 t =0.15; N = 1250
PIGR00 = con>10 + con2-10 + HT500
R2e = 0.01
PIGR01 = con>10 + con2-10 + HT500
R2e = 0.05
PIGR02 = con>10 + con2-10 + HT500
R2e = 0.10
TEWA00 = blt02 + con>2m + ES500 + ES2000 + RD500 + RD2000
R2e = 0.35; R2ρ = 0.10; R2 t =0.35; N =750
97.14
97.69
97.89
98.26
0.00 0.33
0.55 0.25
0.75 0.23
1.12 0.19
= - 5.18 + 0.09 * ES500 - 0.06 *RD500 + 1.00e-4 * RD2000 + 0.37 *A
37.92 0.00 0.28
= - 6.05 + 0.08 * ES500 + 1.00e-4 * RD2000 + 1.53 *A
38.98 1.06 0.16
= - 4.25 - 0.07 *RD500 + 7.00e-5 * RD2000 + 0.14 *A
39.19 1.27 0.15
= - 5.52 + 5.44 * blt0-2 + 0.08 * ES500 - 0.05 *RD500 + 1.00e-4 * RD2000 - 0.24 *A
39.38 1.46 0.13
= - 5.08 + 9.00e-5 * RD2000 + 4.11 * A
39.97 2.06 0.10
95
= - 6.52 + 5.19 * blt0-2 + 1.89 * con>2 + 0.08 * ES500 - 0.05 *RD500 + 1.00e-4 * RD2000 - 0.40 *A
41.20 3.28 0.05
= - 6.55 + 5.12 * blt0-2 + 3.44 * con>2 - 0.06 *RD500 + 1.00e-4 * RD2000 + 1.57 *A
41.63 3.71 0.04
TEWA01 = blt02 + con>2m + ES500 + ES2000 + RD500 + RD2000
R2e = 0.13
TEWA02 = blt02 + con>2m + ES500 + ES2000 + RD500 + RD2000
R2e = 0.10
BBWO00 = snag7-15 + snag>15 + vol + OB500 + EB2000
BBWO01 = snag7-15 + snag>15 + vol + OB500 + EB2000
R2e = 0.04
R2e = 0.18
= -2.97 - 0.52 * OB500 + 0.29 * EB2000
96.50 0.00 0.51
= -2.53 - 0.30 * snag7-15 – 0.59 * OB500+ 0.30 * EB2000
97.65 1.15 0.29
= -2.65 - 0.36 * snag7-15 + 0.17 * snag>15 - 0.64 * OB500 + 0.29 * EB2000
BBWO02 = snag7-15 + snag>15 + vol + OB500 + EB2000
R2e = 0.09
99.39 2.89 0.12
TTWO00 = snag7-15 + snag>15 + vol + OB500 + EB2000
R2e = 0.03
TTWO01 = snag7-15 + snag>15 + vol + OB500 + EB2000
R2e = 0.08
TTWO02 = snag7-15 + snag>15 + vol + OB500 + EB2000
R2e = 0.10
HETH00 = blt0-2 + con>10 + snagT + OB500 + EB2000
R2e = 0.44
= 0.38 + 4.01 * blt0-2 -26.02 * con>10 - 0.60 * snagT + 0.54 * OB500
110.29 0.00 0.50
= 0.39 + 3.42 * blt0-2 -25.86 * con>10 - 0.62 * snagT + 0.49 * OB500 +0.30 * OB2000
112.16 1.87 0.20
= - 0.24 -30.02 * con>10 + 0.62 * OB500
112.24 1.95 0.19
= - 0.29 + 3.92 * blt0-2 -30.62 * con>10 + 0.60 * OB500
113.79 3.50 0.07
R2e = 0.23
HETH01 = blt0-2 + con>10 + snagT + OB500 + EB2000
= - 0.52 - 10.60 * con>10 – 0.45 * snagT + 0.23 * OB500
128.47 0.00 0.35
= - 0.45 - 6.17 * blt0-2 - 10.33 * con>10 – 0.44 * snagT + 0.25 * OB500
129.66 1.18 0.19
= - 0.42 - 10.21 * blt0-2 - 9.60 * con>10 – 0.60 * snagT + 0.08 * OB500 + 0.08 * EB2000
129.74 1.26 0.18
= - 0.50 - 4.2 * blt0-2 - 8.30 * con>10 - 0.49 * snagT
130.07 1.59 0.15
= - 0.50 - 4.2 * blt0-2 - 8.30 * con>10 - 0.49 * snagT
130.98 2.50 0.10
HETH02 = blt0-2 + con>10 + snagT + OB500 + EB2000
R2e = 0.08
FOSP00 = blt0-2 + con>2m + ES500 + ES2000 + RD500 + RD2000
R2e = 0.10; R2ρ = 0.28; R2 t =0.29; N =2000
FOSP01 = blt0-2 + con>2m + ES500 + ES2000 + RD500 + RD2000
R2e = 0.11; R2ρ = 0.14; R2 t =0.19; N =500
FOSP02 = blt0-2 + con>2m + ES500 + ES2000 + RD500 + RD2000
R2e = 0.08; R2ρ = 0.48; R2 t =0.50; N =4250
DEJU00 = con>10m + blt0-2 + HT500
R2e = 0.05; R2ρ = 0.04; R2 t =0.08; N =250
DEJU00 = con>10m + blt0-2 + HT500
R2e = 0.03
DEJU00 = con>10m + blt0-2 + HT500
R2e = 0.08
NOWA00 = blt0-2 + W + W 500 + W 2000
R2e = 0.15
-3
= -3.77 + 8.10 * blt02 + 9.26e * W
79.82 0.00 0.51
= -3.99 + 8.77 * blt02 + 7.91e-3 * W + 1.99e-4 W 500
81.46 1.64 0.22
= -3.20 + 8.04 * blt02 + 7.30e-3 * W + 3.88e-4 W 500 - 6.16e-5 * W 2000
81.82 2.00 0.19
-3
= -3.27 + 8.14e * W
83.78 3.96 0.07
R2e = 0.15
NOWA01 = blt0-2 + W + W 500 + W 2000
= -2.63 + 7.90e-3 * W
124.11 0.00 0.40
-3
= -2.81 + 3.90 * blt02 + 8.29e * W
124.50 0.39 0.33
= -2.95 + 4.35 * blt02 + 7.26e-3 * W + 1.44e-4 W 500
-3
-4
126.15 2.04 0.14
-5
= -2.39 + 3.69 * blt02 + 6.84e * W + 2.51e W 500 - 3.93e
* W 2000
R2e = 0.24; R2ρ = 0.36; R2 t =0.46; N =4000
NOWA02 = blt0-2 + W + W 500 + W 2000
-3
126.56 2.45 0.12
-4
= -5.85 + 7.26 * blt02 + 5.38e * W + 4.54e W 500 + 17.01 * A
96
72.92 0.00 0.24
= -5.34 + 5.17e-4 W 500 + 15.73 * A
73.04 0.12 0.23
= -5.73 + 7.05 * blt02 + 5.94e-4 W 500 16.22 * A
73.09 0.17 0.22
= -5.51 + 5.73 * blt02 + 7.75e-3 * W + 19.54 * A
-3
-4
73.53 0.61 0.18
-5
= -5.33 + 6.91 * blt02 + 5.25e * W + 5.54e W 500 - 3.98e
* W 2000 + 1.75 * A
74.36 1.44 0.12
WTSP00 = blt0-2 + con>2m + ES500 + ES2000 + RD500 + RD2000 R2e = 0.16; R2ρ = 0.12; R2 t =0.21; N =750
= -2.65 + 0.06 * ES500 - 2.57 *A
153.92 0.00 0.56
= -2.65 - 0.05 * blt0-2 + 0.06 * ES500 – 2.58 *A
155.92 2.00 0.21
= -2.47 + 0.79 * blt0-2 - 1.86 * con>10 + 0.07 * ES500 - 2.63 *A
157.17 3.25 0.11
= -2.28 + 0.66 * blt0-2 - 1.84 * con>10 + 0.07 * ES500 - 4.86 * RD500 - 2.53 *A
157.82 3.90 0.08
WTSP01 = blt0-2 + con>2m + ES500 + ES2000 + RD500 + RD2000 R2e = 0.27
= -3.20 - 8.93 * con>10 + 0.11 * ES500
56.56 0.00 0.39
= -3.35 + 7.17 * blt0-2 - 10.44 * con>10 + 0.11 * ES500
57.18 0.62 0.28
-4
= -2.28 + 7.03 * blt0-2 - 10.33 * con>10 + 0.11 * ES500- 3.83e * RD500
58.93 2.37 0.12
= -4.17 + 3.64 * blt0-2 + 0.80 * ES500
59.58 3.02 0.09
= -3.44 + 6.78 * blt0-2 - 10.44 * con>10 + 0.09 * ES500 - 0.05 * ES2000 + 5.18e-4 * RD500
59.74 3.18 0.08
2
WTSP02 = blt0-2 + con>2m + ES500 + ES2000 + RD500 + RD2000 R e = 0.11
a
Response variable is the species code (Table 3.3) with the year (00 = 2000, 01
= 2001, 02 = 2002).
b
R2e = variation explained by environment (non-spatial models), R2 ρ = Partial R2
of the spatial parameter ρ, the variance explained exclusively by space, R 2 t =
total variance explained by both environment and space. N = the distance at
which space is accounted for in spatial models.
97
Appendix 3.3. Correlograms of bird-vegetation model residuals where significant
autocorrelation was detected (global α=0.05) by species and years. Species
codes are in Table 3.2.
98
Appendix 3.4. Goodness of fit statistics for yearly models of bird reproductive
activity and site reoccupancy in relation to forest type. Non-spatial models were
logistic regression but when spatial dependence was detected, autologistic
models were used. See methods for details. Species codes are in Table 3.2.
Species
Year aNeighbor (m)
Reproductive activity
YRWA
2000
2001
2002
RCKI
2000
2001
2002
BOCH
2000
2001
2002
GRJA
2000
2001
2002
DEJU
2000
2001
2002
Site reoccupancy
YRWA
RCKI
SWTH
1250
GRJA
FOSP
4500
DEJU
b
R2e
R2ρ
R2t
0.24
0.25
0.38
0.38
0.03
0.01
0.07
0.03
0.01
0.04
0.06
0.05
0.07
0.01
0.01
0.02
0.05
0.03
0.05
0.10
0.12
0.10
0.04
0.11
0.07
a
Neighbor is the distance at which space is accounted for in spatial models
b
R2e = variation explained by environment (non-spatial models), R2 ρ = Partial R2
of the spatial parameter ρ, the variance explained exclusively by space, R 2 t =
total variance explained by both environment and space.
99
Appendix 3.5. Goodness of fit statistics for logistic regression models of bird site
reoccupation and reproductive activity in relation to vegetation characteristics.
When spatial dependence was detected, autologistic models were also used.
Predictor codes are in Table 3.1, species codes are in Table 3.2.
a
Model
Neighbor (m)
Reproductive activity
YRWA001 = con>10 + con2-10 + HT500
YRWA021 = con>10 + con2-10 + HT500
YRWA031 = con>10 + con2-10 + HT500
RCKI00 = con>10 + con2-10 + HT500
RCKI01 = con>10 + con2-10 + HT500
RCKI02 = con>10 + con2-10 + HT500
BOCH00 = con>10 + con2-10 + HT500+snagT
BOCH01 = con>10 + con2-10 + HT500+snagT
BOCH02 = con>10 + con2-10 + HT500+snagT
GRJA00 = con>10 + con2-10 + HT500
GRJA01 = con>10 + con2-10 + HT500
GRJA02 = con>10 + con2-10 + HT500
DEJU00 = con>10 + blt0-2 + HT500
DEJU01 = con>10 + blt0-2 + HT500
DEJU02 = con>10 + blt0-2 + HT500
b
R2e
R2ρ
R2t
0.05
0.03
0.02
0.01
0.09
0.02
0.13
0.02
0.08
0.09
≈ 0.00
0.05
0.05
0.01
0.01
a
Site reoccupancy
YRWA1 = con>10 + con2-10 + HT500
RCKI = con>10 + con2-10 + HT500
SWTH = con>10 + con2-10 + HT500 + blt0-2
GRJA = con>10 + con2-10 + HT500
FOSP = blt0-2 + con>10 + ES500 + ES2000 + RD500 + RD2000
DEJU = con>10 + blt0-2 + HT500
0.02
0.14
1250
0.02
0.24
0.24
0.02
4500
0.11
0.38
0.38
0.08
a
Response variable was binomial: + = site was occupied > 2 years for abundant species (YRWA,
RCKI, DEJU); + = site was occupied >1 years for less abundant (remaining) species.
100
Chapter 4 – The contribution of climate, arthropods and species interactions in
patterns of bird occupancy and habitat quality3.
3
I intend to submit this chapter to Oikos as Simon, N.P.P., and Diamond, A.W. The contribution
of climate, arthropods and species interactions in patterns of bird occupancy and habitat quality.
101
Abstract
We assessed the ability of 3 processes (microclimate, erratically varying
resources and weak competition) to explain weak habitat quality-vegetation
relationships for birds along a timber volume gradient near the boreal forest-taiga
boundary. There was no strong evidence that birds used microclimate factors in
selecting sites. Site occupation was inconsistent between years which may
suggest that birds did not choose ‗optimum‘ breeding sites deterministically, but
instead selected them stochastically when confronted with sites that were overall
satisfactory in quality. We attribute this, in part, to birds having difficulty
assessing resources within a site since arthropod biomass (indicating food
resources) showed little relationship to forest type, and varied between years and
within breeding seasons. Our results also indicate that bird abundances, and
thus their habitat selection patterns, are determined partly by densityindependent factors, perhaps climate. The coefficient of variation in yearly total
bird abundance was approximately 7 times higher than predicted by chance,
indicating bird community instability. The high yearly variance in total bird
abundance relative to the sum of the variances of all species signifies parallel
fluctuations in species densities indicative of low interspecific competition. Bird
community instability and broadening of species‘ niches, through low interspecific
competition, may also contribute to the weak bird-habitat relationships.
102
Introduction
Most habitat selection theories are based on the premise that individuals
select areas to maximize their fitness (Fretwell and Lucas 1969, Rosenzweig
1991, Sutherland 1997). For an individual to choose an area that increases its
fitness, that individual must know about the resources available within all areas
under consideration (Haila et al. 1996, Jonzén et al. 2004). Resource availability
depends upon the resources present within an area, the ability of the animal to
extract those resources, and the demands on those resources by intra- and
interspecific competitors (Fretwell and Lucas 1969, Cody 1981, Wiens 1985,
Morris 1988, Rosenzweig 1991). Thus, a thorough assessment of songbirdhabitat relationships requires studies of abundance-resource correlations to be
evaluated within the context of processes influencing selection (e.g., knowledge
of resources and competition).
Studies of songbird-habitat relationships can be grouped coarsely into 2
study extents, local and landscape (Diamond 1999, Drapeau et al. 2000,
Lichstein et al. 2002a). Most local-extent research directs measurements at
extents smaller than, or of similar magnitude to, the territory size; vegetation
variables (e.g., stem densities, stratification, and needle architecture) have been
used to predict songbird abundance, foraging and nest sites (Lack 1933,
MacArthur and MacArthur 1961, Roth 1976, Franzreb 1978, Morse 1989, Parrish
1995). Landscape variables are measured at larger extents (i.e., several
kilometers for songbirds) and include vegetation types surrounding the territory
103
and measures of spatial associations, e.g., isolation, patch-size and connectivity
(Saunders et al. 1991, Mazerolle and Villard 1999).
While vegetation variables are the most prominent factors included in
songbird-habitat studies, vegetation is only one component of their habitat. For
birds breeding near the extremity of their geographic range there may be
increased climatic influence in habitat selection, which may supersede birdvegetation relationships (Williams et al. 2003). Time constraints force northern
birds to assess breeding territories under partly snow-covered conditions (Orians
and Wittenburger 1991, Haila et al. 1996) raising the possibility that northern bird
populations use different cues (e.g. microclimate) to select territories than their
southern counterparts. Smith et al. (1998) suggested that early-arriving migrants
can offset food limitations by foraging in warmer microclimates that produce more
food. At large extents, climate influences songbird winter (Root 1988) and
summer geographic ranges (Venier et al. 1999, 2004), while at small extents,
microclimate determines nest placement (Walsberg 1981, With and Webb 1993).
At intermediate extents, weather influences feeding behaviour in winter and
spring (Wachob 1996, Smith et al. 1998). Similarly, elevation has been linked to
bird population trends and habitat selection (James et al. 1996, Lichstein et al.
2002a). This has lead to recommendations towards the inclusion of abiotic
factors such as climatic parameters and elevation in bird-habitat quality models
(James et al. 1996, Irwin 1998, Martin 2001).
Prospecting behaviour is one way that individuals can gain knowledge of
resource availability within a potential area to establish a territory (Wiens 1985,
104
Orians and Wittenburger 1991). When this level of knowledge is high, bird
habitat selection should closely track resource availability (Wiens 1985, Orians
and Wittenburger 1991, Haila et al. 1996). This implies that territories are
selected in a deterministic order where the optimal habitats are selected first and
most consistently. However, where birds cannot accurately assess resource
levels, optimal habitats are selected less frequently, producing apparently
stochastic patterns of occupation that are poorly correlated with resource levels
(Wiens 1985, Orians and Wittenburger 1991, Haila et al. 1996, Jonzén et al.
2004). Birds may have difficulty assessing resources i) in environments that vary
erratically (Wiens et al. 1985, Orians and Wittenburger 1991, Jonzén et al. 2004),
ii) where resources vary along subtle gradients (Haila et al. 1996) or iii) where the
short breeding season imposes time constraints or requires territories to be
selected under snow-covered conditions (Orians and Wittenburger 1991, Haila et
al. 1996).
The level of competition also influences the predictability of songbird site
occupation. With respect to intraspecific interactions, little or no competition
results in many higher quality sites will be unoccupied (Wiens 1985).
Conversely, extreme competition should result in saturation of the higher quality
sites, causing many poorer quality patches to be occupied (Rosenzweig 1991,
Rodenhouse et al. 1997). Thus, the clearest habitat relationships should occur
with intermediate levels of competition (Rosenzweig 1991) because all higher
quality sites and no poorer sites should be occupied (Morris 1988, Rosenzweig
1991, Rodenhouse et al. 1997). Conversely, increasing interspecific competition
105
results in a narrowing of an individual‘s realized niche to facilitate the coexistence
of ecologically similar species (Cody 1981, Wiens 1985, Rosenzweig 1991).
Narrowing of an individual‘s niche implies selection for a reduced range of
resources, hence clearer habitat relationships.
We evaluate 3 processes predicted to cause noisy bird-vegetation
patterns and assess their ability to explain previously documented weak habitat
quality-vegetation relationships for birds in central Labrador, Canada: i) birds
using microclimate factors in selecting breeding sites, ii) birds having difficulty
tracking resources and iii) reduced interspecific competition. In this region, most
species are near the northern extremity of their geographic range, potentially
increasing role of climatic cues in site occupancy. Birds may have difficulty
tracking resources because in northern environments, arthropod abundances
tend to vary erratically (Forsman et al. 1998). The harsh northern climate is also
believed to limit species numbers enough to reduce interspecific competition
(Järvinen 1979, Helle and Mönkkönen 1986, Morozov 1993), broadening species
niches (Rosenzweig 1991) and further weakening bird-habitat relationships.
These are evaluated within the context of 3 measures of songbird habitat quality:
bird abundance, site re-occupancy and breeding activity. These provide
different, but complimentary information about the quality of a particular habitat to
increase individual‘s fitness.
Study Area
The study area was within the perhumid high boreal ecoclimatic region
(Canada Committee on Ecological Land Classification 1989). The study was
106
conducted among mature (> 130 years) natural black spruce (Picea mariana)
stands within 50 km of Goose Bay, NL, Canada (53○ 20‘ N, 60○ 25‘ W). Sites
were dominated by black spruce and to a lesser extent balsam fir (Abies
balsamea). Feather moss (Pleurozium schreberi) occurs commonly on well
drained sites and Sphagnum spp. on poorly drained sites. Sites were distributed
among 5 forest types representing a gradient of timber volume: lichen woodland
(n = 50), open spruce (n = 47), low volume spruce (n=44), medium volume
spruce (n = 49) and high volume spruce (n = 30).
Methods
Birds were enumerated using 220 point counts of 10-minute duration,
divided into 3-, 2-, and 5-minute intervals in which all birds seen or heard were
recorded (Bibby et al. 1993, Smith et al. 1997). We used the maximum number
of males recorded within 100 m radius on a single visit as our index of
abundance because the number of birds recorded is the minimum number
present (Bibby et al. 1993). Individuals within 100 m that were flying over were
excluded. Observations were conducted between 04:00-11:00 AST from June
01 - July 31 in 2000-2002. Point count stations were distributed among the 5
stand types described above, were ≥ 250 m apart and ≥ 100 m from the nearest
edge. Two visits were made to each point count station and observers were
alternated between visits to reduce observer bias. To establish a consistent
protocol, all observers conducted simultaneous point counts with an experienced
observer before collecting data independently.
107
An index of breeding activity was used following Gunn et al. (2000), which
provides a reasonable indicator of reproductive success at coarse scales (Doran
et al. 2005). In all 3 years of the study, 5 minute play-backs of black-capped
chickadee (Poecile atricapillus) mobbing calls, following the point counts, were
used to attract songbirds to facilitate observations of songbird breeding activity.
We ranked breeding activity 0 = absent, 1= male only, 2 = pairs, 3 = adults
carrying nesting material, 4 = adults carrying food and 5 = the presence of
juveniles (adapted from Gunn et al. 2000). For additional data on reproductive
activity, after the completion of both point count rounds, we conducted 1 extra
visit to each station in 2001 and 2 extra visits in 2002 using only the mobbing
play-back.
Microclimate variables
We used the method of Bourque and Gullison (1998) and Bourque et al.
(2000) to calculate daily mean solar radiation and temperature during the
growing season (end of April – October) which roughly corresponds to the time
when migrants are present on the breeding grounds. Hourly predictions of
incoming solar radiation were based on sun-earth geometry through land-surface
attributes (elevation, slope, aspect, horizon angles, terrain configuration factor
and view factor) derived from a 1:50 000 digital elevation model of 50 m
resolution. Daily mean temperatures were estimated from the land-surface
attributes and the mean daily temperatures recorded at a reference site in Goose
Bay, NL, Canada. This technique has been successful in predicting temperature
estimates when compared with field data (Bourque and Gullison 1998) and
108
explaining variation in potential vegetation growth (Bourque et al. 2000). The
model produced rasters of solar radiation and temperature estimates with a
resolution of 50 m pixels. East facing slopes may be attractive because of higher
early morning temperatures, and snow tends to melt faster on south facing
slopes; thus, we included northness and eastness indices, adapted from Beers et
al. (1966) (Table 4.1) which produce high positive values for steep north and east
facing slopes and high negative values for steep south and western facing slopes
respectively.
Arthropod sampling
A subset of 24 point count stations (n = 5 for all forest types except n = 4
in HVS) were chosen on the basis of accessibility for insect sampling. At each
site, we haphazardly sampled one branch at each of 2, 4 and 6 m into the
canopy. We placed each branch in a bag prior to removal and the insects were
collected in the lab. We also placed water traps on the ground at each site; a
yellow bucket 33 cm high and 27 cm in diameter with approximately 10 cm water
and a tablespoon of detergent to remove surface tension. A 14.5 cm transparent
drift fence forming a cross to intercept insects from all directions was placed on
top of the bucket. Branch and water trap samples were collected weekly from
the first week of June to end of July, 2001-2002, corresponding to the bird survey
period. All samples were dried to constant weight and weighed to the nearest
mg.
Bird relationships with microclimate variables
109
To determine the importance of microclimate on our 3 measures of bird
habitat quality (abundance, reproductive activity and reoccupancy) we modelled
habitat quality-microclimate relationships while accounting for spatial
autocorrelation, using the approach of Keitt et al. (2002), Lichstein et al. (2002b)
and Klute et al. (2002) (see Chapter 3). Unfortunately, autoregressive models
with a Poisson response distribution are impractical because they can
incorporate only negatively correlated errors (Cressie 1993). As a compromise
between adhering to distribution assumptions and accounting for spatial pattern,
we used either Gaussian or logistic regression depending upon the species‘
abundance. Least-squared regression (on Log + 1 transformed data) and logistic
regression, was used for common (yellow-rumped warblers, dark-eyed juncos
and ruby-crowned kinglets) and uncommon species respectively, when the
model residuals were independent. When autocorrelation in the residuals was
detected, Gaussian conditional autoregressive models and autologistic
regression was used for common and uncommon species respectively.
We subjected all 4 microclimate measurements (solar radiation,
northness, eastness and temperature) to a Pearson‘s correlation. We excluded
solar radiation from bird-microclimate modelling as it was correlated with
northness (r > │0.9│). We constructed global models using all 3 microclimate
variables (northness, eastness and temperature) as predictors of our 3 habitat
quality measures. We used the R2 of Nagelkerke (1991) to evaluate the fit of the
global model. We constructed a series of nested models using all possible
combinations the 3 predictor variables and evaluated their importance through
110
Akaike‘s information criterion (AIC) (Burnham and Anderson 2002). If the R2 of
the global model ≥ 0.15, we felt the variance explained was too low to be
biologically relevant and thus did not evaluate nested models. Species that did
not occur on ≥ 5% of the point count stations in a single year were excluded from
the analyses.
Most of our reproductive activity observations were of pairs, so we
reduced the ordinal response to binomial categories (0 and 1 = no evidence, 2-4
= evidence of reproductive activity). Our site reoccupation data were counts
(number of years a site was occupied), but with a finite upper bound (i.e., 3)
which violates the assumption of a Poisson distribution (McCullagh and Nelder
1989). No transformation could normalize it so we converted site occupation to a
binomial variable. For common species most sites were occupied in 2 of the 3
years so we considered sites occupied 3 years as ―reoccupied‖. For less
common species 3 years of occupation was rare so we considered a site
―reoccupied‖ if it was occupied > 1 year. These reductions to binomial variables
also facilitated our spatial modelling framework.
Chapter 3 concluded that local and landscape vegetation was poor
predictor of habitat quality for most species in this study area. However,
vegetation was a weak to moderate predictor of habitat quality for some speciesyears (Tables 4.1, 4.2). To determine the influence of microclimate relative to
vegetation on habitat quality, we assessed whether the addition of microclimate
variables improved the vegetation models. For species-years were the R2 of
vegetation models was ≥ 0.15 we added climate models and assessed their
111
importance through backward elimination (i.e., the variable was discarded if its
removal reduced Akaike‘s information criterion, (AIC)).
Arthropod-forest type relationships
Arthropod biomass during particular periods of the breeding season or
heights in the canopy could be important for bird habitat quality. We analyzed
the effects of these factors, and forest type on total arthropod biomass, using
linear mixed effects models on log + 1 transformed data, conducted with the nlme
library of J. Pinheiro and D. Bates (1999) (URL:http://lib.stat.cmu.edu/S/; viewed
November 2003), implemented on S-plus 2000 (Mathsoft 1999). Site and year
were random effects because we wanted to extrapolate beyond these specific
years and sites (Pinheiro and Bates 2000); the remaining variables were fixed.
We analyzed arthropods from water traps as above but without branch height.
For each dataset, we constructed a global, linear mixed effects model using all
variables listed above and removed variables using backwards elimination (Table
3.2). Variance explained was computed on the global model only as this model
will fit the data best of all models confronted with data (Burnham and Anderson
2002). We evaluated model fit with a Pearson‘s correlation between the
observed and predicted values (Hosmer and Lemeshow 2000). We used
Akaike‘s Information Criterion (AIC) and AIC weights (Burnham and Anderson
2002) to evaluate the strength of evidence for each model.
Community stability and competition
Bird community stability was measured as the coefficient of variation in
total abundance between years (Järvinen 1979). We compared the coefficient of
112
variation with that expected from a Poisson distribution, to assess if the variability
in total abundance is greater than expected by chance (see Helle and
Mönkkönen 1986, Morozov 1993). We used the variance test of Schluter (1984)
and Järvinen (1979) to assess the level of competitive interactions between
species. We calculated the ratio (V) as variance of total bird abundance between
years ( S T 2 ) over the sum of the variance of individual species abundances
N
between years (  i2 ) with  i2  (1 / N ) ( X ij  t i ) 2 where X ij is the mean
i
density of species i in year j and ti is the mean abundance of species i. A ratio
significantly less than 1 indicates compensatory fluctuations are stronger than
parallel ones, suggesting competitive density interactions (Järvinen 1979,
Schluter 1984). A ratio greater than 1 indicates that species fluctuate in parallel.
The probability the ratio differed from 1 was determined with a Chi-square test
(Schluter 1984). We also used logistic regression to evaluate how site
occupancy in one year was related to site occupation of the previous year. A
strong positive relationship would indicate birds were selecting ‗optimum habitats‘
deterministically, whereas no relationship indicates a greater stochastic element
to site occupation and suggests birds were settling for ‗satisfactory habitats‘.
Results
Bird-microclimate relationships
Hermit thrush in 2000 and fox sparrow in 2002 were the only species
whose abundances were related to microclimate. The best hermit thrush model
showed a positive relationship with temperature. Other models with moderate
support show positive relationships with northness and eastness. Adding
113
temperature to the best vegetation model improved its fit (ΔAIC = 1.29). The
best model fox sparrow abundance showed an increase with temperature. Other
models with moderate support show positive relationships with northness and
eastness. Despite the poor vegetation model (Table 4.2) adding temperature
increased the AIC. The best model for fox sparrow site re-occupancy data was
an increase with eastness. Other models with moderate support show positive
relationships with northness and temperature. Adding temperature to the best
vegetation model improved its fit (ΔAIC = 3.50).
Arthropod Mass
Our models predicted arthropod biomass poorly, as the correlations
between observed arthropod mass and that predicted by our models were small
(0.38 and 0.46 for the branch and water trap data respectively). Our models
showed little support for our predicted increase in arthropod biomass with forest
productivity: our branch data overwhelming supported one model (Akaike
weight= 0.88), which excluded forest type (Table 4.4, Fig. 4.1). Arthropod
biomass declined linearly with branch height and there was less biomass in
2002. Although forest type was an important variable in our water trap models,
arthropod mass did not increase with productivity in 2001 (Fig. 4.2). In 2002,
arthropod biomass was consistently higher in high volume spruce but there was
little pattern in other forest types. There was a peak in biomass in week 4 of both
years (coincided with budburst) but the forest types with this peak varied
between years.
Community stability and competition
114
Bird species presence showed little relationship with site occupancy the
previous year, suggesting birds were not deterministically selecting for the
‗optimal sites‘. Fox sparrow presence in 2002 was weakly related to its
abundance in 2001 (R2 = 0.15) whereas all other species and years had R2 ≤
0.10. The coefficient of variation for total bird density across 3 years of 21.5% ±
8.9 (SE) is substantially higher than that predicted by a Poisson random variable
(2.8%). The ratio V = 4.45, is significantly higher than 1 (χ2 = 13.36, df =3, P =
0.004) indicating parallel fluctuations and little interspecific competition.
Discussion
Our hypothesis that birds use micoclimate cues to select territories did not
provide an adequate explanation for the weak bird-vegetation relationships
recorded in Chapter 3. Of our 45 abundance-microclimate models, only 2
showed a relationship with microclimate and these were weak to moderate.
Further, only hermit thrush abundance showed a microclimate association after
vegetation was controlled. Only 1 of our 6 reoccupancy models showed a
relationship with microclimate. However, this relationship persisted after
vegetation was controlled. None of our reproductive activity models showed a
relationship with microclimate.
We were unable to reject the hypothesis that the weak bird-vegetation
relationships were caused by birds having difficulty tracking resources. Our
results show that arthropod biomass (the primary food resource for most of our
breeding species) was weakly related to forest type and varied erratically within
the breeding season. Thus, arthropod biomass at the time a territory is selected,
115
does not indicate arthropod abundance throughout the breeding season. Where
arthropods vary fairly independently of vegetation features, strong preference for
particular vegetation features would not be advantageous for insectivorous birds.
Our results contrast with Seagle and Sturtevant (2005) but support Forsman et
al. (1998) who concluded that migrant birds in northern forests cannot use
arthropod abundance to determine habitat quality. Similarly, Orians and
Wittenburder (1991) partly attributed the poor correlation between female yellow
headed blackbird (Xanthocephalus xanthocephalus) densities and Odonates on
individual territories to the unpredictability of Odonate emergence.
In addition to the erratic variation in arthropod abundances, other factors
characteristic of our forests could contribute to birds having difficulty tracking
resources. Northern songbirds are obliged to acquire territories in the partially
snow-covered spring, contributing to their inability to assess resource levels
(Orians and Wittenburger 1991, Haila et al. 1996). Much of the variation in
vegetation in our forests occurs subtly along topographic gradients (Chapter 2),
unlike a patch-matrix-corridor system typical of frequent large-scale disturbances.
It seems likely that relative differences in resources would be more difficult to
assess in gradient systems than those where patches contrast more sharply
(Haila et al. 1996).
Our finding that sites occupied were inconsistent between years has 3
interpretations. 1) Birds were not deterministically selecting the highest quality
territories but instead were settling for ‗satisfactory conditions‘ rather than
seeking the ‗optimal conditions‘ as Haila et al. (1996) concluded. This result is
116
further suggestion that individuals‘ knowledge of resource availability is low and
hence deterministic selection of optimal conditions was not possible (Wiens
1985, Orians and Wittenburger 1991, Haila et al. 1996). 2) Birds were selecting
a resource that varied spatially between years. Arthropod biomass is the most
obvious bird resource that could vary spatially between years and our data show
this to be true. However, this variation is so great within the breeding season, it
is doubtful that birds can use early season arthropod biomass to assess food
resources throughout the breeding season, Nevertheless, they still could use
this ‗incomplete‘ information when making their decision (Orians and
Wittenburger 1991). However, because we have arthropod data for only a
subset of our sample area, we are unable to test this idea formally. 3) The
habitat quality of all sites was similar, i.e., there was no optimum to select. This
appears to be the case for species of which we has sufficient reproductive
activity and site occupancy data. If there were optimum sites but birds were
unable to select it, there should have been patterns with reproductive activity and
site occupancy data but not with abundance – we not find such a pattern.
Reduced competition, potentially driven by climate, could also contribute
to poor bird-habitat relationships. In northern Europe, harsh climates limit both
resident and migrant bird numbers, causing erratically-varying bird abundances
(Järvinen 1979, Helle and Mönkkönen 1986, Virkkala 1991a, 1991b, Morozov
1993, Hogstad et al. 2003). Järvinen (1979) reported a strong instability gradient
with latitude in Europe. Noon et al. (1985) found a similar, though weaker, trend
in North American coniferous forests. Our high coefficient of variation in bird
117
density indicates high annual variability similar to that of northern Europe
(Järvinen 1979). A variable and harsh climate would cause ―noisy‖ birdvegetation relationships because species abundances vary according to densityindependent factors (Wiens 1985, Helle and Mönkkönen 1986, Haila and
Järvinen 1990).
If climate reduces densities sufficiently to decrease interspecific
competition, birds could have broader niches (Morse 1974, Cody 1981, Martin
and Martin 2001) implying less specific habitat associations. The assumption
that harsh conditions reduce competition and favor the coexistence of
ecologically similar species has been challenged theoretically (Chesson and
Huntly 1997) and empirically (Mönkkönen 1990). However, Mönkkönen et al.
(2004) document positive associations between potentially competing species at
low densities (in northern areas) and negative associations at high densities.
Further, the prevalence of heterospecific attraction in northern areas suggests
competitive interactions are weak (Forsman et al. 1998, Mönkkönen et al. 1999).
Our parallel, rather than compensatory density fluctuations are evidence of weak
interspecific competition, a result similar to others in northern forests (Järvinen
1979, Helle and Mönkkönen 1986, Morozov 1993).
Studies concluding that bird habitat quality were weakly related to habitat
features, and suggest that birds show stochastic occupancy, will always be
subject to the following criticism: birds were in fact responding to an unmeasured
habitat feature. Though it is impossible to measure all factors potentially
influencing the species of interest, the vegetation variables used in the
118
regression models from Chapter 3 were based on the species‘ life history.
These, and similar variables have successfully predicted bird abundance in other
locations (Schwab and Sinclair 1994, Lichstein et al. 2002a, Drapeau et al.
2000). Further, their post-hoc exploratory data analysis (CART models) was a
thorough search using 42 uncorrelated variables representing local and
landscape vegetation. Both these procedures and the microclimate variables in
present study explained little variation in bird abundance so it is improbable that
key habitat features for all species within this study were missed. Further, since
species presence in a point-count station in one year was not related to its
presence at that site in the previous year, if birds were associated with an
unmeasured habitat feature, that feature must have varied spatially between
years. That is, birds could be deterministically selecting a habitat feature that
varied stochastically, perhaps arthropods.
Factors inherent in our data collection or analysis could also give the
appearance of poor bird-habitat relationships. Point counts have been criticized
because detections can vary with vegetation cover (Schieck 1997). We
assessed detectability bias in Chapter 3; except for pine grosbeak, there was no
evidence of bias, indicating that abundances derived from point counts reflect the
densities of the species. Other factors include data recorded at incorrect spatial
scales and failure to account for spatial autocorrelation (Vaughn and Ormerod
2003). It is doubtful that our data were recorded at the wrong scale since there is
abundant literature describing local and landscape bird-vegetation relationships
at 100 m and 500- 2000 m extents respectively (Hagan et al. 1996, Drapeau et
119
al. 2000, Lichstein et al. 2002a). There is no reason to expect that birds in our
study respond to different scales than birds elsewhere. We used 2 statistical
techniques, one of which incorporated autocorrelation, to evaluate birdvegetation relationships suggesting our results are not an artifact of the modeling
techniques or failing to incorporate autocorrelation. Other social factors (e.g.,
heterospecific attraction (Forsman et al. 1998, Mönkkönen et al. 1999), may also
influence habitat selection, but testing those is beyond the scope of this paper.
We based our conclusions of low competition on a variance ratio of
individual species abundances and total abundances (Schluter 1984).
Measuring competition directly can be difficult at any scale, particularly on a
spatial extent as large as ours, hence the popularity of this index for assessing
interspecific bird competition (e.g., (Järvinen 1979, Helle and Mönkkönen 1986,
Morozov 1993). A potential limitation with this index is that circumstances exist
in which a ratio >1 could indicate competition: all species fluctuate in unison to
identical, limiting resources (Schluter 1984). While there is overlap in resource
use between many of our species, many of them forage in very different areas
(e.g., tree canopy or ground) and some frequently forage on seeds while others
rarely do. So, it is improbable that all species fluctuate in unison to the same
resource, thus we are confident our result of V = 4.45 is indicative of low
competition.
Our results indicate that most species in our study area do not use
microclimate cues in selecting breeding sites. However, microclimate was weak
to moderately important for 2 species‘ habitat quality measures, even after
120
vegetation was controlled. This supports Irwin‘s (1998) appeal for researchers to
consider abiotic factors in bird-habitat relationships. Our results indicate that the
erratically varying arthropod abundances and low competition, the latter perhaps
induced through harsh climate could have caused the weak bird-habitat
relationships. If the population size is determined by density-independent factors
(i.e., climate), this implies that habitats may not be saturated and populations
may be more resistant to habitat loss.
Acknowledgements
Funding was provided by the Newfoundland and Labrador Department on
Natural Resources and Human Resources and Skills Development Canada.
Thanks are extended to those assisting in data collection, notably, B. Campbell,
J. Colbert, L. Elson, R. Flynn, M. Michelin, K. Mitchell, R. Neville, F. Phillips, B.
Rodrigues, F. Taylor. D. Goulding and D. Jennings assisted in GIS work and C.
Bourque assisted with, and provided a program to compute, abiotic variables.
Thanks are also extended to the Innu Nation. M. Betts, G. Forbes, J. Nocera, D.
Keppie and F. Phillips reviewed earlier drafts of the manuscript.
References
Beers, R.W., Dress, P.E. and Wensel, L.C. 1966. Aspect transformation in site
productivity research. – J. For. 64: 691-692.
Bibby, C.J., Burgess, N.D. and Hill, D.A. 1993. Bird census techniques. Academic Press, Toronto.
121
Bourque, C. P-.A. and Gullison, J.J. 1998. A technique to predict hourly potential
solar radiation and temperature for a mostly unmonitored area in the Cape
Breton Highlands. – Can. J. Soil Sci. 78:409-420.
Bourque, C.P.-A., Meng, F., Gullison, J.J. and Bridgeland, J. 2000. Biophysical
and potential vegetation growth surfaces for a small watershed in northern
Cape Breton Island, Nova Scotia, Canada. Can. J. For. Res. 30:11791195.
Burnham, K.P. and Anderson, D.R. 2002. Model selection and multimodel
inference. A practical information-theoretic approach, Second Edition. –
Springer, New York.
Canada Committee on Ecological Land Classification. 1989. Ecoclimatic Regions
of Canada: First approximation. – Environment Canada, Ecological Land
Classification Series No. 23. Canadian Wildlife Service.
Chesson, P. and Huntly, N. 1997. The role of harsh and fluctuating conditions in
the dynamics of ecological communities. – Am. Nat. 150: 519-553.
Cody, M.L. 1981. Habitat selection in birds: The roles of vegetation structure,
competitors, and productivity. – Bioscience 31:107-113.
Cressie, N.A. 1993. Statistics for spatial data. – John Wiley and Sons, New York.
Diamond, A.W. 1999. Concluding remarks: content versus context in forest bird
research. - In Diamond, A.W. and Nettleship, D.N. (eds.), Biology and
conservation of forest birds. Society of Canadian Ornithologists Special
Publication No. 1, pp. 139-143.
Doran, P.J., Gulezian, P.Z. and Betts, M.G. 2005. A test of the mobbing
122
plackback method for estimating bird reproductive success. – J. Field.
Ornithol. 76:227-233.
Drapeau, P., Leduc, A., Giroux, J-. F., Savard, J-. P.L., Bergeron, Y. and Vickery,
W. L. 2000. Landscape-scale disturbances and changes in bird
communities of boreal mixed-wood forests. – Ecol. Monogr. 70: 423-444.
Forsman, J.T., Mönkkönen, M., Helle, P. and Inkeröinen, J. 1998. Heterospecific
attraction and food resources in migrants‘ breeding patch selection in
northern boreal forest. – Oecologia 115:278-286.
Franzreb, K.E. 1978. Tree species used by birds in logged and unlogged mixedconiferous forests. – Wilson Bull. 90:221-238.
Fretwell, S.D. and Lucas, H.L. 1969. On territorial behaviour and other factors
influencing habitat distribution in birds. – Acta Biotheor. 19:16-36.
Gunn, J.S., DesRochers, A., Villard, M.-A. and Ibarzabal, J. 2000. Playbacks of
mobbing calls of black-capped chickadees as a method to estimate
reproductive activity of forest birds. – J. Field Ornithol. 71: 472-483.
Hagan, J.M., Vander Haegen, M.W. and Mckinley, P.S. 1996. The early effects of
forest fragmentation on birds. – Conserv. Biol. 10:188-202.
Haila, Y. and Järvinen. 1990. Northern conifer forests and their bird
assemblages. – In: Keast, A. (ed.), Biogeography and ecology of forest
bird communities. SPB Academic Publishing, The Hague, pp. 61-83.
Haila, Y., Nicholls, A.O., Hanski, I. and Raivio, S. 1996. Stochasticity in bird
habitat selection: year-to-year changes in territory locations in a boreal
forest bird assemblage. – Oikos 76:536-552.
123
Helle, P. and Mönkkönen, M. 1986. Annual fluctuations of land bird communities
in different successional stages of boreal forest. – Ann. Zool. Fenn.
23:269-280.
Hogstad, O., Selǻs, V. and Kobro, S. 2003. Explaining annual fluctuations in
breeding density of fieldfares Turdus pilaris – combined influences of
factors operating during breeding, migration and wintering. – J. Avian Biol.
34:350-354.
Hosmer, D.W. Jr. and Lemeshow, S. 2000. Applied logistic regression. – John
Wiley and Sons, New York.
Irwin, L.L. 1998. Abiotic influences on bird-habitat relationships. – In: J. Marzluff
and Sallabanks, R. (eds.), Avian conservation and management, Island
Press, Washington, pp. 209-217.
James, F.C., McCulloch, C.E. and Wiedenfeld, D.A., 1996. New approaches to
the analysis of population trends in landbirds. - Ecology 77:13-27.
Järvinen, O. 1979. Geographical gradients of stability in European land bird
communities. – Oecologia 38: 51-69.
Jonzén, N., Wilcox, C. and Possingham, H.P. 2004. Habitat selection and
population regulation in temporally fluctuating environments. – Am. Nat.
164:E104-E114.
Keitt, T.H., Bjørnstad, O.N., Dixon, P.M. and Citron-Pousty, S. 2002. Accounting
for spatial pattern when modelling organism-environment interactions. –
Ecography 25:616-625.
Klute, D.S., Lovallo, M.J. and Tzilkowski, W.M. 2002. Autologistic regression
124
modelling of American woodcock habitat use with spatially dependent
data. – In: Scott, J.M., Heglund, P.J., Morrison, M.L., Haulfer, J.B.,
Raphael, M.G., Wall, W.A. and Sampson, F.B. (eds.), Predicting species
occurrence issues of accuracy and scale. Island Press, Washington pp.
335-355.
Lack, D. 1933. Habitat selection in birds. – J. Anim. Ecol 2:239-262.
Lichstein, J. W., Simons, T.R. and Franzreb, K.E. 2002a. Landscape effects on
breeding songbird abundance in managed forests. – Ecol. Appl. 12:836857.
Lichstein, J. W., Simons, T.R., Shriner, S.A. and Franzreb, K.E. 2002b. Spatial
autocorrelation and autoregressive models in ecology. – Ecol. Monogr.
72:445-463.
MacArthur, R.H. and MacArthur, J.W. 1961. On bird species diversity. – Ecology
42:594-598.
Martin, T.E. 2001. Abiotic vs. biotic influences on habitat selection of coexisting
species: climate change impacts? – Ecology 82:175-188.
Martin, P.R. and Martin, T.E. 2001. Ecological and fitness consequences of
species coexistence: A removal experiment with wood warblers. – Ecology
82: 198-206.
Mathsoft 1999. S-plus 2000 for Windows.
Mazerolle, M.J. and Villard, M-.A. 1999. Patch characteristics and landscape
context as predictors of species presence and abundance: A review. –
Écoscience 6:117-124.
125
McCullagh, P. and Nelder, J.A. 1989. Generalized linear models. – Monographs
on statistics and applied probability 37. Chapman and Hall, New York.
Mönkkönen, M. 1990. Removal of territory holders causes influx of small-sized
intruders in passerine bird communities in northern Finland. – Oikos
57:218-288.
Mönkkönen, M., Forsman, J.T. and Thomson, R.L. 2004. Qualitative
geographical variation in interspecific interactions. – Ecography 27:112118.
Mönkkönen, M., Härdling, R., Forsman, J.T. and Tuomi, J. 1999. Evolution of
heterospecific attraction: using other species as cues in habitat selection.
– Evol. Ecol. 13: 91-104.
Morozov, N.S. 1993. Short-term fluctuations in a south-taiga bird assemblage:
support for an ―individualistic‖ view. – Ornis Fenn. 70:177-188.
Morris, D.W. 1988. Habitat-dependent population regulation and community
structure. – Evol. Ecol. 2:253-269.
Morse, D.H. 1974. Niche breadth as a function of social dominance. – Am. Nat.
108:818-830.
Morse, D.H. 1989. American warblers: An ecological and behavioral perspective.
– Harvard University Press, Cambridge.
Nagelkerke, N.J.D. 1991. A note on the general definition of the coefficient of
determination. – Biometrika 78: 691-692.
Noon, B.R., Dawson, D.K. and Kelly, J.P. 1985. A search for stability gradients in
North American breeding bird communities. – Auk 102:64-81.
126
Orians, G.H. and Wittenberger, J.F. 1991. Spatial and temporal scales in habitat
selection. – Am. Nat. 137:S29-S49.
Parrish, J.D. 1995. Effects of needle architecture on warbler habitat selection in
a coastal spruce forest. – Ecology 76:1813-1820.
Patuxent Wildlife Research Center, 2002. North American Bird Banding Manual.
Draft. - URL:http://www.pwrc.usgs/gov/BBL/manual.
Pinheiro, J.C. and Bates, D.M. 2000. Mixed-effects models in S and S-PLUS. –
Springer-Verlag, New York.
Rodenhouse, N.L. Sherry, T.W. and Holmes, R.T. 1997. Site-dependent
regulation of population size: A new synthesis. – Ecology 78:2025-2042.
Root, T. 1988. Energy constraints on avian distributions and abundances. –
Ecology 69:330-339.
Rosenzweig, M.L. 1991. Habitat selection and population interactions: The
search for mechanism. – Am. Nat. 137:S5-S28.
Roth, R.R. 1976. Spatial heterogeneity and bird species diversity. – Ecology
57:773-782.
Saunders, D.A., Hobbs, R.J. and Margules, C.R. 1991. Biological consequences
of ecosystem fragmentation: A review. – Conserv. Biol. 5:18-32.
Schieck, J. 1997. Biased detection of bird vocalizations affects comparisons of
bird abundance among forested habitats. – Condor 99:179-180.
Schluter, D. 1984. A variance test for detecting species associations, with some
example applications. – Ecology 65:998-1005.
127
Schwab, F.E. and Sinclair, A.R.E. 1994. Biodiversity of diurnal breeding bird
communities related to succession in the dry Douglas-fir forests of
southeastern British Columbia. – Can. J. For. Res. 24:2043-2040.
Seagle, S.W. and Sturtevant, B.R. 2005. Forest productivity predicts invertebrate
biomass and ovenbird (Seiurus aurocapillus) reproduction in Appalachian
landscapes. – Ecology 86:1531-1539.
Smith, W.P., Twedt, D.J., Cooper, R.J., Wiedenfield, D.A., Hamel, P.R. and Ford,
R.P. 1997. Sample size and allocation of effort in point count sampling of
birds in bottomland hardwood forests. – In: Ralph, C.J., Sauer, J.R. and
Droedge, S. (eds.), Monitoring bird populations by point counts. General
Technical Report PSW-GTR-149, pp 7-17.
Smith, R., Hamas, M., Dallman, M. and Ewert, D. 1998. Spatial variation in
foraging of the black-throated green warbler along the shoreline of
northern Lake Huron. – Condor 100:474-484.
Sutherland, W.J. 1997. From individual behaviour to population ecology. Oxford
Series in Ecology and Evolution. – Oxford University Press, Oxford.
Vaughan, I.P. and Ormerod, S.J. 2003. Improving the quality of distribution
models for conservation by addressing shortcomings in the field collection
of training data. – Conserv. Biol. 17: 1601-1611.
Venier, L.A., McKenney, D.W., Wang, Y. and McKee, J. 1999. Models of largescale breeding-bird distribution as a function of macro-climate in Ontario,
Canada. – J. Biogeogr. 26: 315-328.
128
Venier, L.A., Pearce, J.L., Mckee, J.E., McKenney, D.W. and Niemi, G.J. 2004.
Climate and satellite-derived land cover for predicting breeding bird
distribution in the Great Lakes Basin. – J. Biogeogr. 31:315-331.
Virkkala, R. 1991a. Population trends of forest birds in a Finnish Lapland
landscape of large habitat blocks: consequences of stochastic
environmental variation or regional habitat alteration? – Biol. Conserv.
56:223-240.
Virkkala, R. 1991b. Spatial and temporal variation in bird communities and
populations in north-boreal coniferous forests: a multiscale approach. –
Oikos 62:59-66.
Wachob, D.G. 1996. The effect of thermal microclimate on foraging site selection
by wintering mountain chickadees. – Condor 98:114-122.
Walsberg, G.E. 1981. Nest-site selection and the radiative environment of the
warbling vireo. – Condor 83:86-88.
Wiens, J.A. 1985. Habitat selection in variable environments: Shrub-steppe birds.
– In: Cody, M.L. (ed.), Habitat Selection in Birds. Academic Press, Inc.,
pp. 227-251.
With, K.A. and Webb, D.R. 1993. Microclimate of ground nests: the relative
importance of radiative cover and wind breaks for three grassland species.
– Condor 95: 401-413.
Williams, C.K., Ives, A.R. and Applegate, R.D. 2003. Population dynamics
across geographical ranges: time-series analyses of three small game
species. – Ecology 84:2654-2667.
129
Table 4.1. Descriptions and codes of variables used in bird-habitat models from
Chapter 3.
Code
Variable description
Stand level variables
blt0-2
Broad-leaved tree crown cover < 2 m tall
con2-10
Conifer crown cover > 2-10 m tall
con>10
Conifer crown cover > 10 m tall
snag7-15
Number of snags >7 - 15 cm dbh per 400 m2
snag>15
Number of snags >15 cm dbh per 400 m2
snagT
Total snags >7 cm dbh per 400 m2
Vol
Timber volume (m3) per 400 m2
W
Water edges (m) within 100 m (RaS + Lake)
Landscape variables
RD500
Road length (m) within 500 m of point
HT500
Heavy timber (≥ 100 m3/ha) % within 500 m of point
OB500
Old burn (> 15 years old) % within 500 m of point
RD2000
Road length (m) within 2000 m of point
ES500
Early successional forest within 500 m of point (OB500 + OC500)
ES2000
Early successional forest within 2000 m of point (OB2000 + OC2000)
OB2000
Old burn (> 15 years old) % within 2000 m of point
W500
Water edges (m) within 500 m of point (RaS500 + Lake500)
W2000
Water edges (m) within 2000 m of point (RaS2000 + Lake2000)
Birdsa
YRWA
Yellow-rumped warbler, Dendroica coronata
RCKI
Ruby-crowned kinglet, Regulus calendula
DEJU
Dark-eyed junco, Junco hyemalis
FOSP
Fox sparrow, Passerella iliaca
GRJA
Gray jay, Perisoreus canadensis
PIGR
Pine Grosbeak, Pinicola enucleator
BBWO
Black-backed woodpecker, Picoides arcticus
HETH
Hermit thrush, Catharus guttatus
NOWA
Northern waterthrush, Seiurus noveboracensis
PISI
Pine siskin, Carduelis pinus
TTWO
Three-toed woodpecker, Picoides tridactylus
TEWA
Tennessee warbler, Vermivora peregrina
SWTH
Swainson‘s thrush, Catharus ustulatus
BOCH
Boreal chickadee, Poecile hudsonicus
WTSP
White-throated sparrow, Zonotrichia albicollis
a
Abbreviation of common names follows bird banding manual (Patuxent Wildlife
Research Center 2002)
130
Table 4.2. Variance explained by environment for yearly models of bird
abundance, reproductive activity and site reoccupancy in relation to predicted
vegetation features (Chapter 3). Least squares regression was used for
abundant species (YRWA, RCKI, DEJU) and logistic regression for the
remaining, less abundant species. Variable codes in Table 4.1.
Model
Abundance
YRWA = con>10 + con2-10 + HT500
PISI = con>10 + con2-10 + HT500
RCKI = con>10 + con2-10 + HT500
BOCH = con>10 + con2-10 + snagT + HT500
SWTH = con>10 + con2-10 + blt0-2 + HT500
PIGR = con>10 + con2-10 + HT500
TEWA = blt02 + con>2m + ES500 + ES2000 + RD500 + RD2000
BBWO = snag7-15 + snag>15 + vol + OB500 + OB2000
TTWO = snag7-15 + snag>15 + vol + OB500 + OB2000
GRJA = con>10 + con2-10 + HT500
HETH = blt0-2 + con>10 + snagT + OB500 + OB2000
FOSP = blt0-2 + con>2m + ES500 + ES2000 + RD500 + RD2000
DEJU = con>10m + blt0-2 + HT500
NOWA = blt0-2 + W + W 500 + W 2000
WTSP = blt0-2 + con>2m + ES500 + ES2000 + RD500 + RD2000
Reoccupancy
YRWA = con>10 + con2-10 + HT500
RCKI = con>10 + con2-10 + HT500
SWTH = con>10 + con2-10 + blt0-2 + HT500
GRJA = con>10 + con2-10 + HT500
FOSP = blt0-2 + con>2m + ES500 + ES2000 + RD500 + RD2000
DEJU = con>10m + blt0-2 + HT500
Reproductive activity
YRWA = con>10 + con2-10 + HT500
RCKI = con>10 + con2-10 + HT500
BOCH = con>10 + con2-10 + snagT + HT500
GRJA = con>10 + con2-10 + HT500
DEJU = con>10m + blt0-2 + HT500
131
2000 2001 2002
0.02
0.04
0.09
0.06
0.08
0.01
0.35
0.04
0.03
0.01
0.26
0.01
0.05
0.02
0.05
0.13
0.18
0.08
0.03
0.03
0.07
0.01
0.01
0.01
0.10
0.09
0.10
0.44
0.10
0.05
0.15
0.16
0.23
0.11
0.03
0.15
0.27
0.08
0.08
0.08
0.24
0.11
0.02
0.14
0.02
0.02
0.11
0.08
0.05 0.03
0.01 0.09
0.13 0.02
0.09 ≈ 0.00
0.05 0.01
0.02
0.02
0.08
0.05
0.01
Table 4.3. Nested models of arthropod mass with associated Akaike‘s
Information Criteria (AIC), model weights calculated by linear mixed effects
models using maximum likelihood.
Model – Branch samples
AIC
-8504.00
Weight = Cover * Week + Branch + Branch:Cover + Year
Weight = Cover * Week + Branch + Year
-8516.16
Weight = Cover + Week + Branch + Year
-8538.44
Weight = Week + Branch + Year
-8545.60
Weight = Branch + Year
-8550.38
Weight = Week + Year
-8535.77
Weight = Week + Branch
-8538.41
Weight = Cover
-8526.61
Weight = Week
-8528.74
Weight = Branch
-8543.34
Weight = Year
-8540.65
Model – Water traps
Weight = Cover * Week + Year
1384.19
Weight = Cover + Week + Year
1371.46
Weight = Cover + Week
1374.37
Weight = Week + Year
1374.18
Weight = Cover + Year
1380.96
Weight = Cover
1383.24
Weight = Week
1377.11
Weight = Year
1383.22
132
46.38
34.23
11.94
4.76
0.00
14.62
12.00
23.78
21.65
7.05
9.73
Weight
7.50e-11
3.30e-08
2.22e-03
0.08
0.88
5.90e-04
2.21e-03
6.1e-06
1.8e-05
0.03
6.79e-03
12.73
0
2.91
2.72
9.5
11.8
5.65
11.76
1.10e-03
0.64
0.15
0.16
5.53e-03
1.77e-03
3.79e-02
1.79e-03
Table 4.4. The importance of microclimate for bird-habitat quality determined by
logistic and autologistic regression models. Nested microclimate models with
associated Akaike‘s Information Criteria (AIC) for species-years where the global
model explained ≥0.15 (see Appendix 4.1), species codes in Table 4.1. The AIC
values in the vegetation and microclimate models are the changes in AIC when
variables from the best microclimate model were included into the best
vegetation model.
Modela
ΔAICb
Nested microclimate
HETH2000 Re2 = 0.31
= -20.59 + 2.10*Temp
0.00
= -21.60 + 4.01*North + 2.20*Temp
0.87
= -20.60 + 0.48*East + 2.10*Temp
1.99
= -21.60 + 4.00*North + 0.29*East + 2.20*Temp
2.87
FOSP2002 Re2 = 0.26; Rρ2 = 0.57; Rt2 = 0.58; N = 1500
-3.90 + 0.17*Temp + 4.86*A
0.00
-2.37 + 6.66*East + 4.62*A
0.03
-2.52 + 1.63*North + 4.97*A
0.03
-3.32 + 6.61*East + 0.11*Temp + 4.69*A
2.00
-5.39 + 2.42* North + 0.32*Temp + 5.27*A
2.03
-2.45 +1.01*North + 6.50*East + 4.75*A
2.03
FOSPoccupancy Re2 = 0.22;Rρ2 = 0.52; Rt2 = 0.54; N = 3750
-2.80 + 7.44*East + 5.52*A
0.00
-3.07 + 3.71*North 7.06*East + 6.00*A
0.42
-3.48 + 7.43*East + 0.08*Temp + 5.56*A
1.97
-5.54 + 4.39*North + 6.91*East + 0.28*Temp + 6.22*A
2.05
-3.07 + 3.88*North + 6.07*A
3.23
Vegetation and microclimate modelsc
HETH2000
= -7.62 + 1.53*blt0-2 -19.37*con>10 - 0.54 * snagT + 0.48 * OB500 + 0.86*Temp
-1.29
FOSP2002
-4.43 + 0.02*ES500 + 0.23*Temp + 3.85*A
+1.11
FOSPoccupanc
-2.73 + 4.39*blt0-2 -2.16*Con>10 + 6.72*East + 5.63*A
-3.50
a
North = northness, East = eastness, Temp = mean daily temperature, A = autocovariate.
b
The sign (+/-) in the vegetation and microclimate models indicates whether adding the
microclimate variable reduced (-) or increased (+) the AIC relative to the vegetation only model.
c
Vegetation variables are described in Table 4.1.
133
Figure 4.1. Arthropod biomass (mean, SE) in relation to branch height.
134
Figure 4.2. Arthropod biomass (mean, SE) from water traps in relation to
collection week and forest type. LW = lichen woodland, OS = open spruce, LS =
low volume spruce, MS = medium volume spruce and HS = high volume spruce.
135
Appendix 4.1. Variance explained (R2) for global models of habitat quality
(abundance, reproductive activity and reoccupancy) predicted by microclimate
(northness, eastness and temperature). For common species (YRWA, RCKI,
DEJU), least squares regression was used when abundance was the response
variable. Logistic regression was used for all other models. Species codes are
in Table 4.1
Global model Re2
Species
2000 2001 2002
YRWA
0.02
0.04
0.03
PISI
0.06
0.21
0.04
RCKI
0.12
0.10
0.05
BOCH
0.03
0.08
0.01
SWTH
0.05
0.14
0.06
PIGR
0.07
0.01
0.10
TEWA
0.09
0.02
0.02
BBWO
0.02
0.05
0.05
TTWO
0.05
0.13
0.03
GRJA
0.03
0.03
0.01
0.31
HETH
0.11
0.07
0.26
FOSP
0.13
0.11
DEJU
0.05
0.01
0.07
NOWA
0.04
0.01
0.13
WTSP
0.02
0.12
0.05
Reproductive activity
YRWA
0.03 ≈0.00 0.08
RCKI
0.02
0.05
0.02
BOCH
0.02
0.01
0.09
GRJA
≈0.00 0.01
0.02
DEJU
0.02
0.01
0.01
Reoccupancy
YRWA
RCKI
SWTH
GRJA
FOSP
DEJU
All years
0.05
0.09
0.01
0.02
0.22
0.07
136
Chapter 5 – An evaluation of the environmental heterogeneity hypothesis and its
relation to site productivity4.
4
I intend to submit this chapter to Diversity and Distributions as Simon, N.P.P., and Diamond,
A.W. An evaluation of the environmental heterogeneity hypothesis and its relation to site
productivity
137
Abstract
We assessed the covariation between environmental heterogeneity and
productivity as a mechanism to produce productivity-diversity relationships in a
northern boreal bird community, Newfoundland and Labrador, Canada. We used
3 heterogeneity measures (foliage height diversity (FHD), horizontal diversity
(HD) and their interaction (FHD X HD)), 3 indices of productivity (timber volume,
mean daily temperature and solar radiation) quantified at 3 grain sizes (point
counts of 50 and 100 m radii and within the entire stand) to produce 27
heterogeneity-productivity models. Our models produced: 11 no patterns, 4
unimodal, 3 U-shaped, 2 linear declines, 4 curvilinear declines and 2 linear
increases and 1 curvilinear increase. Our only observed richness-productivity
patterns occurred at the stand grain and showed 3 different patterns: linear
decline, a linear increase and curvilinear decline with volume, temperature and
radiation respectively. Heterogeneity potentially explains the negative richnessvolume and positive richness-temperature relationships but not the richnessradiation relationship. We conclude that heterogeneity does not follow a
universal pattern with productivity, varying with scale and according to how
heterogeneity and productivity are quantified. We suggest that conflicting
richness-productivity patterns in the literature could result from different
heterogeneity-productivity relationships in areas and scales where richness is
driven by heterogeneity.
138
Introduction
Understanding which factors determine the variation in species richness
between areas is a central concern among community ecologists and is
fundamental in biodiversity conservation. Environmental heterogeneity and
productivity, the rate of energy flow through a system, are dominant factors that
have been demonstrated to affect species richness at particular scales (Currie
1991, Rosenzweig and Abramsky 1993, Tilman and Pacala 1993, BöhningGaese 1997, Atauri and de Lucio 2001, Mittelbach et al. 2001). Despite
extensive research on the relationship between productivity and species
richness, general patterns and mechanisms remain controversial (Rosenzweig
and Abramsky 1993, Böhning-Gaese 1997, Mittelbach et al. 2001). A unimodal
diversity-productivity relationship has been suggested to have the most
theoretical and empirical support (Rosenzweig and Abramsky 1993, Tilman and
Pacala 1993) but other relationships are common and have supporting theories
(Abrams 1988, 1995, Mittelbach et al. 2001, Chase and Leibold 2002). The
nature of the relationship depends partly upon the taxonomic group and the scale
of observation (Mittelbach et al. 2001, Chase and Leibold 2002). Further,
incomplete sampling of the productivity gradient could lead to false
interpretations, e.g., an increasing richness-productivity pattern could be the
increasing phase of a ‗true‘ unimodal relationship. However, Mittelbach et al.
(2001) and Abrams (1995) caution that evidence required to confirm nonunimodal relationships has been unacceptably stringent.
After reviewing various hypotheses, Rosenzweig and Abramsky (1993)
139
concluded the environmental heterogeneity hypothesis provided the best
explanation for the unimodal pattern: species richness increases with
environmental heterogeneity, which peaks at moderate productivity levels then
declines. This hypothesis was criticized by Abrams (1988, 1995) for the lack of
empirical support for a reduction in heterogeneity at high productivities. Models
proposed by Abrams (1988, 1995) indicate that both monotonic increasing and
unimodal relationships between heterogeneity and productivity are possible.
Abrams (1995) agreed that a unimodal relationship would most likely
occur in plants where understory light is limited at high productivities, as
suggested by Tilman and Pacala (1993). Rosenzweig and Abramsky (1993)
extended this pattern to animals: sites of low productivity support little vegetation
and are uniformly barren. As productivity increases, a variety of nutrient-light
combinations increases plant species and structural diversity. Beyond a certain
point, light limits understory plant growth, reducing plant species and structural
diversity and therefore animal diversity (Rosenzweig and Abramsky 1993,
Rosenzweig 1995). Heterogeneity, as a covariate of productivity, has been
proposed as driver of species richness for other taxa including small mammals
(Abramsky and Rosenzweig 1984, Rosenzweig 1995) and birds (Nilsson and
Nilsson 1978, Nilsson 1979). This apparent evidence for the environmental
heterogeneity hypothesis is weakened because heterogeneity was not quantified
in these studies; where heterogeneity has been quantified it is not always tightly
correlated with productivity and can explain richness better than productivity
(Böhning-Gaese 1997, Cueto and Lopez de Casenave 1999, Hurlbert 2004);
140
however see Currie (1991). The effect of heterogeneity in vegetation structure
on species richness is perhaps best studied in songbirds. Since Lack (1933) and
MacArthur and MacArthur (1961) first examined the issue, vertical and horizontal
diversity of vegetation structure have successfully explained variation in songbird
richness (Wiens 1974, Roth 1976, Rotenberry and Wiens 1980, Estades 1997).
If heterogeneity and diversity are correlated, then Rosenzweig and Abramsky‘s
(1993) heterogeneity-productivity curve should apply to songbirds.
Another source of confusion is that the nature of the covariance between
heterogeneity (particularly horizontal) and productivity could change with grain
size (i.e., size of the sample unit, sensu Legendre and Legendre 1998) (Abrams
1995). Low productivity sites tend to have sparser vegetation than more
productive sites. If low productivity sites (sparse vegetation) were embedded
within higher productivity areas (dense vegetation) measurements at small grain
sizes will probably reduce the variance in vegetation in productive areas. This is
because smaller sample units will likely sample either low productivity or high
productivity sites. Conversely, measurements made on broader grain sizes
would include both high and low productivity sites and overall, be classified as
moderate to high productivity. These samples would contain patches of sparse
vegetation along with dense vegetation, thus increasing the within-plot
heterogeneity of those areas (Abrams 1995). Further, since habitat selection can
occur over multiple scales, heterogeneity at landscape extents could influence
local species richness (Wiens 1989). For terrestrial species in forested areas,
landscape extent variables appear less influential in habitat selection than local
141
variables (McGarigal and McComb 1995, Trzcinski et al. 1999, Lichstein et al.
2002a), see however Drapeau et al. (2000). Yet, at extents of several
kilometres, landscape heterogeneity predicted species richness better than did
measures of productivity (Böhning-Gaese 1997, Kerr and Packer 1997, Atauri
and de Lucio 2001). Chase and Ryberg (2004) determined that landscape
connectivity was partly responsible for the scale-dependent productivity patterns
reported in Chase and Leibold (2002). An additional complication is that
productivity is rarely measured directly and the surrogates used may differ in the
strength in their correlation with productivity (Mittelbach et al. 2001). The
frequency of the surrogate used varies with the spatial extent (i.e., total size of
the study areas, sensu Legendre and Legendre 1998), e.g., evapotranspiration
and plant biomass are most commonly measured at extents > 4000 km and ≤
200 km respectively (Mittelbach et al. 2001). This raises the possibility that
scale-dependent patterns are partly dependent upon the choice of the
productivity surrogate. These potentially confounding factors imply that
examinations of productivity and heterogeneity as determinants of richness
should incorporate multiple grain sizes and extents and use several surrogate
measures if productivity is not measured directly.
Our objective is to assess the influence of heterogeneity as a mechanism
behind biodiversity-productivity relationships. Under the environmental
heterogeneity hypothesis, we predict: (1) the shape of the productivity-richness
curve is determined by the shape of the productivity-heterogeneity curve, (2) bird
species richness increases with heterogeneity (3) heterogeneity explains more of
142
the variation in species richness than productivity. We investigate these
relationships at 3 grain sizes as the nature of the covariance between
heterogeneity and productivity could vary with grain size. We also include
landscape diversity as this could influence local richness. We use timber
volume, mean daily solar radiation and temperature during growing season as
surrogates for productivity. Timber volume is related to plant biomass, a
presumed correlate of productivity, while radiation and temperature are thought
to reflect available energy (Wright et al. 1993, Mittelbach et al. 2001, Hurlbert
2004). Timber volume can indicate foraging substrate for canopy and stem
feeding species and can reduce foraging substrate for shrub dwelling species by
shading understory plants. Solar radiation and temperature influence insect
abundances and activity (i.e., insectivore food) (Smith et al. 1998) and nest
placement (Walsberg 1981, With and Webb 1993). All 3 measures have been
found to predict bird species richness (Nilsson 1979, Currie 1991)
Study area
We sampled 220 sites among mature (≥ 130 years) natural black spruce
(Picea mariana) stands within 40 km of Goose Bay, NL, Canada (53° 20‘ N, 60°
25‘ W). The study area is in the perhumid high boreal ecoclimatic region
(Canada Committee on Ecological Land Classification 1989). Sites are
dominated by black spruce and to a lesser extent balsam fir (Abies balsamea)
(Table 5.1). Feather moss (Pleurozium schreberi) commonly occurs on well sites
and Sphagnum spp. on poorly drained sites.
To ensure a range of productivity values, our sites were distributed across
143
5 forest cover types lichen woodland (LW), open spruce (OS), low volume spruce
(LS), medium volume spruce (MS) and high volume spruce (HS); reflecting a
gradient of timber volume chosen from forest inventory maps at 1:12 500 scale
(Newfoundland and Labrador Department of Natural Resources, unpublished).
While we used the timber volume from forest cover maps to choose our sites,
timber volume in our analyses were calculated using the vegetation
measurements conducted at individual sites described below. Our lowest
productivity sites were within the lowest productivity classification of the forest
inventory. Less than 3 percent of central Labrador contains stands of equal or
greater timber volumes than our highest productivity stands. Hence we have
sampled an adequate range of site productivity values to discern any major
patterns within our study area.
Methods
Bird survey
Birds were censused using 220 point count stations (Bibby et al. 1993,
Smith et al. 1997) conducted between 4:00-11:00 am AST from June 01 - July 31
between 2000 -2002. We attempted to have 50 point counts per forest type,
based on Smith et al. (1997) who suggested this level of sampling was sufficient
to discern any significant patterns. Availability, accessibility and the requirement
for sites to remain unlogged for the duration of the study limited our sampling
within each forest type as follows: LW = 50, OS = 47, LS = 44, MS = 49 , HS =
30. Nevertheless, the standard errors on our abundances for most species were
small relative to the forest type mean suggesting this sampling intensity was
144
sufficient. All songbirds seen or heard during a 10-minute interval were recorded
and noted as to whether they occurred within 50 or 100 m from the point count
station. Ten-minute point counts repeated twice per season have been found to
survey bird species richness adequately (Gutzwiller 1993, Sorace et al. 2000,
Siegel et al. 2001). Birds flying overhead that were obviously not interacting with
the forest structure (e.g., migratory flocks) were excluded from analyses. Two
visits were made to each point count station per year and observers were
alternated between visits to prevent observer bias. Point counts stations were 
250 m apart and  100 m from the nearest stand or road edge.
Vegetation survey
To estimate timber volume and heterogeneity, we sampled vegetation at
each point count station using 5 subplots (8 x 10 m), 1 at the centre and 1 at
each of the 4 cardinal directions approximately 60 m from the station center.
Within each subplot, we recorded the species and diameter at breast height
(dbh) of all trees  7 cm dbh. The crown cover of trees and tall shrubs were
estimated using Emlen (1967): within each subplot, the height of live crown was
recorded above 20 systematically placed points (n = 100 per point count station).
Points were established 1 m apart along 2 transects that were 4 m apart.
Temperature and Solar Radiation
We used the method of Bourque and Gullison (1998) to calculate daily
mean solar radiation and temperature during the growing season (end of April –
October) which roughly corresponds to the time when migrants are present on
the breeding grounds. Hourly predictions of incoming solar radiation were based
145
on sun-earth geometry through land-surface attributes (elevation, slope, aspect,
horizon angles, terrain configuration factor and view factor) derived from a 1:50
000 digital elevation model of 50 m resolution. Daily mean temperatures were
estimated from the land-surface attributes and the mean daily temperatures
recorded at a reference site in Goose Bay, NL, Canada. This technique has
been successful in predicting temperature estimates when compared with field
data (Bourque and Gullison 1998) and explaining variation in potential vegetation
growth (Bourque et al. 2000). The model produced rasters of solar radiation and
temperature estimates with a resolution of 50 m pixels. We averaged the solar
radiation and temperature estimates within buffers of 50 and 100 m radii around
each point count station for the 50 and 100 m grains respectively using
ARCVIEW 3.2 (Environmental Systems Research Institute 1999). Estimates for
the stand grain were the average of all 100 m point count estimates within the
stand.
Data Analysis
Data were analyzed at 3 grain sizes, 50 and 100 m radii within a stand
and at the entire stand (size ranged from 28 – 170 ha). We chose grains at 50
and 100 m radii (0.78 and 3.14 ha respectively) because they are common
songbird point count sizes and correspond roughly to territory sizes of our
common species [e.g., 0.45-0.8 ha for yellow-rumped warbler (Dendroica
coronata), Hunt and Flaspohler (1998) and 2.9 ha for ruby-crowned kinglet
(Regulus calendula), Ingold and Wallace (1994)]. We used forest stand as a
grain size as stands are commonly used as habitat patches in landscape ecology
146
studies.
The relationship between our productivity measurements were assessed
through correlation analysis. Our estimate of bird species richness for the 50 and
100 m grains, we used the Jackknife estimator (Mh) (Otis et al. 1978) conducted
on Specrich2, available online [http://www.mbr-pwrc.usgs.gov/software/
specrich2.html]. This approach uses the number of species accumulated across
our 6 site visits (2 visits per year across 3 years) to estimate richness while
accounting for differences in species detectability (Nichols et al. 1998). We used
the same approach to estimate richness at the stand grain using all 100 m point
counts that occurred within the stand as the Jackknife estimator does not require
equal sampling effort to estimate richness. Other factors (e.g. observer bias,
time of season or day) could influence richness estimates but we did not include
them because they tend to be less important than species detectability (Boulinier
et al. 1998). Further, using the approach of Farnsworth et. al (2002), we
determined that time of season or time of day did not influence species
detectability in our study area (Simon unpublished).
Vegetation from the center subplot was used to characterize timber
volume and heterogeneity at the 50 m grain, while all 5 subplots were used to
characterize the 100 m grain and all subplots (ranged from 20 – 95) within a
stand characterized the stand.
At each grain, we calculated timber volume using tree diameter data from
our vegetation plots and local volume tables (Newfoundland and Labrador
Department of Natural Resources, unpublished data). Structural diversity was
147
quantified as foliage height diversity (FHD) and horizontal diversity (HD). We
calculated FHD  -  pi loge pi (MacArthur and MacArthur 1961), where pi = the
i
proportion of foliage in 3 height zones: 0 - 0.5 m, > 0.5 - 10 m and > 10 m
determined by our Emlen points. These height zones roughly correspond to
vegetation layers that have successfully the abundances to bird communities in
similar regions (e.g., Schwab et al. 2000, Simon et al. 2002). We used the
spatial variation in total foliage crown cover > 0.5 m tall to calculate HD at all 3
scales (Wiens 1974). At the 50 m grain, the coefficient of variation was used as
an index of HD calculated on the proportion of points containing vegetation > 0.5
m tall. At the 100 m grain, all 5 subplots were used to calculate the coefficient of
variation (HD) while all subplots within a stand used to calculate the coefficient of
variation (HD).
For the 50 and 100 m grains, landscape composition was quantified at 3
spatial extents using circular buffers around each point count station of 500 m,
1000 m, and 2000 m radii. Because the long axis of a stand often approached
1000 m, buffers ≤ 1000 m would contain little information additional to
measurements within the stand. Thus, the landscape composition radii for the
stand grain were 2000 m, 3000 m, and 4000 m buffers centered from a point that
minimized the sum distance to all point count stations within the stand. Variables
were calculated with ARCVIEW 3.2 (Environmental Systems Research Institute
1999) using 1992 forest inventory data (Department of Natural Resources,
unpublished data) digitized from 1:12 500 scale aerial photographs. The
inventory was updated to include disturbances (fire, harvesting) since 1992. For
148
each buffer, we calculated the proportion occupied by heavily stocked (≥ 100 m3
timber/ ha) mature forest (> 130 years old), sparsely stocked mature timber (<
100 m3/ ha), recent clearcuts (< 15 years old), older clearcuts (≥ 15 years old),
and wildfires. Landscape diversity was based on these proportions and
calculated similar to foliage height diversity.
Some point count stations were within the same stand raising the
possibility of spatial autocorrelation between variables at the 50 and 100 m
scales, thus violating the assumption of independence (Legendre and Legendre
1998, Legendre et al. 2002). We verified that model residuals were not
autocorrelated using S-PLUS (Mathsoft 1999) code in Ecological Archives M072007-S1 from Lichstein et al. (2002b) which computes significance tests for
correlograms. The code uses 999 permutations to generate the null distribution
and the global significance was performed at α=0.05, Bonferroni corrected for 20
lags at 250 m intervals.
Bird species richness estimates were counts so we used generalized
linear models with a poisson distribution and a loge link to evaluate trends in bird
species richness (McCullagh and Nelder 1989). Other response variables used
least-squares regression. The information-theoretic approach was used to
determine whether variables were related to productivity, and if so, the form of
the relationship (Burnham and Anderson 2002). Two models were fitted for each
variable of interest: the global model included productivity and productivity
squared (productivity = volume, radiation or temperature) as predictors while the
reduced model included only productivity. Other models, e.g., power and
149
logarithmic functions have been used to describe diversity patterns, usually
species-area relationships (e.g., Fisher et al. 1943, Preston 1962). These
relationships are not commonly reported patterns for diversity – productivity
patterns (Mittelbach et al. 2001) (see however, Wright 1983), nor are they explicit
predictions of the environmental heterogeneity hypothesis. Therefore,
investigating these models is beyond the scope of this paper. Goodness-of-fit
was calculated only for the global model (Burnham and Anderson 2002) and was
assessed using R2 for linear regression models, and the R2 of Nagelkerke (1991)
for generalized linear models. If R2 < 0.15, we did not investigate further models
as we felt the variation explained was too low to be biologically significant.
Akaike‘s Information Criterion (AIC) was used to evaluate the relative support for
the full versus the reduced model (Burnham and Anderson 2002). Second-order
Akaike‘s information criterion (AICc) was used at the stand grain due to small
sample size (Burnham and Anderson 2002).
If the global model had the lowest AIC (c) we determined whether the
relationship was unimodal or U-shaped, or merely represented an asymptote
change (curvilinear), using the method of Mitchell-Olds and Shaw (1987). The
technique determines whether an unconstrained model with a particular
intermediate maximum (or minimum) provides a better fit than a model with a
nonintermediate maximum (or minimum). Fit was compared using AIC (c).
For structural diversity, we modeled species richness as a function of
FHD, HD and their interaction. Since we had no a priori hypotheses for the
relative importance of each component of structural diversity we examined all
150
possible nested models and report evidence for their support. Model support
was assessed by ΔAIC(c) and AIC(c) weights. The relative importance (RI) of
each variable was quantified by totaling the AIC(c) weight of all models in which
the variable of interest was included (Burnham and Anderson 2002). To assess
the influence of landscape diversity on species richnes we had to control for local
structural diversity. We added landscape diversity variables, in increasing radii,
to the best structural diversity model (i.e., lowest AIC(c)) for the respective grain
and we report the ΔAIC(c) relative to the structural diversity only model.
Results
The 3 productivity measures were not positively related to each other.
Volume was negatively correlated with temperature and the strength of this
relation increased with grain size. Solar radiation showed weak negative
relationships with volume at the 50 and 100 m grains and temperature at the
stand grain (Table 5.2).
Foliage height diversity followed a unimodal relationship with timber
volume at all three grain sizes (Table 5.3, Fig. 5.1). Horizontal diversity showed
weak, U-shaped relationships with volume across all grain sizes. The only
pattern found for FHD X HD was a weak linear decline at the stand grain.
Songbird species richness showed no relationship with volume at the 50 and 100
m grains and a weak decline at the stand grain.
Foliage height diversity declined in a curvilinear fashion with mean daily
temperature at the 100 m and stand grains while HD showed a weak unimodal
relationship with temperature at the 50 m grain and a linear increase at the stand
151
grain (Table 5.4, Fig. 5.2). FHD X HD increased in a curvilinear fashion with
mean daily temperature at the stand grain while richness showed a weak linear
increase with mean daily temperature. Mean daily solar radiation produced a
curvilinear decline with FHD X HD and species richness at the stand grain (Table
5.5, Fig. 5.3). There was a general increase in the R2 value with grain size
across all metrics.
Songbird species richness was related to structural diversity only at the
stand grain: richness increased with FHD and HD and was negatively related to
FHD X HD with all 3 variables of similar importance (Table 5.6). Landscape
diversity only improved the structural diversity models for the 50 m grain with
landscape diversity at 500 m being the more important landscape variable (Table
5.6).
Discussion
Our finding of a unimodal relationship between FHD and timber volume is
consistent with the environmental heterogeneity hypothesis and supports the
rationale of Rosenzweig and Abramsky (1993) that areas of moderate
productivity have a variety of nutrient-light combinations that produce the highest
structural diversity. Our FHD-volume pattern is congruent with studies which
propose the environmental heterogeneity hypothesis (reviewed in Rosenzweig
1995), but the lack of a unimodal richness-volume pattern conflicts with this
hypothesis. Although Abrams (1995) suggested the strength of heterogeneityproductivity relationship should increase with grain size, as ours does, our
interpretation of Abrams (1995) is that this would result in a monotonic increase.
152
Conversely, the U-shaped relationship between HD and volume across all grains
is a direct contradiction of the unimodal relationship proposed by Rosenzweig
and Abramsky (1993). In areas where species show a strong relationship with
HD (e.g., Wiens 1974, Roth 1976), this may explain U-shaped richnessproductivity relationships which account for 25% of studies across a variety of
taxa, yet no mechanism has been proposed to explain them (Mittelbach et al.
2001). The linear decline at the stand grain with FHD X HD is a direct
contradiction of the monotonic increase proposed by Abrams (1988, 1995) and
the natural vegetation biomass gradient of Nilsson (1979).
Temperature produced a variety of relationships with structural diversity
whereas radiation produced only curvilinear declines with FHD X HD. Radiation
and temperature are thought to reflect the energy available within a system and
are most commonly used at large spatial extents i.e., > 4000 km, Wright et al.
1993, Mittelbach et al. 2001, Hurlbert 2004). The most common richnessproductivity relationship with these variables is a monotonic increase (Wright et
al. 1993, Mittelbach et al. 2001, Hurlbert 2004). The hypothesis of energy
increasing food, supporting more individuals and therefore more species, dubbed
the ‗more individuals‘ hypothesis by Srivastava and Lawton (1998) was proposed
to explain this pattern. However, Srivastava and Lawton (1998) concluded the
‗more individuals hypothesis‘ was insufficient to explain richness. Heterogeneity
frequently explains a significant proportion of the variation in richness after
accounting for energy, and heterogeneity is often poorly correlated with available
energy (Böhning-Gaese 1997, Cueto and Lopez de Casenave 1999, Hurlbert
153
2004). Our results suggest that heterogeneity does not follow a universal pattern
with productivity, varying with the type of heterogeneity (i.e., vertial or
horizonatal), the measure of productivity and the scale of observation. In forests,
it seems likely that the relationship between the different productivity measures
and FHD would vary with the shade tolerance of the major tree species. A peak
in FHD could occur at higher volumes if the understory was dominated by more
shade tolerant species. Our sites are dominated by black spruce and balsam fir
which are relatively shade tolerant (Frank 1990, Viereck and Johnson 1990)
hence, the relatively high volume at which FHD peaks. The lack of a relationship
between FHD and solar radiation could also be due to the shade tolerance of our
major tree species.
Our only observed richness-productivity patterns occurred at the stand
grain and were relatively weak for timber volume and temperature. However, the
three different measures produced 3 different results; linear decline, a linear
increase and a curvilinear decline with volume, temperature and radiation
respectively. It is possible that a productivity-heterogeneity relationship produced
the first two of these patterns: FHD X HD and richness followed the same
pattern with volume and temperature and richness was positively associated with
HD. However, this was not the case with radiation. Species richness and the
only structural diversity measure showing a pattern (FHD X HD) showed a
curvilinear decline. However, FHD X HD was a negative predictor of species
richness and the richness-radiation curve was much stronger than the HD X
FHD-radiation curve, a result that conflicts with other research (e.g., Böhning-
154
Gaese (1997), Cueto and Lopez de Casenave (1999), Hurlbert (2004)).
Although the mechanisms behind our variable richness-productivity
patterns may be difficult to identify, it does raise the question of whether the
scale-dependent patterns described in the literature are a result of scaledependent processes or scale-dependent surrogate productivity measures. Our
3 productivity surrogates were either unrelated or negatively related to each
other, and the nature of the relationship varied with grain size. In our study,
volume was largely driven by soil moisture (Chapter 2) where solar radiation and
temperature were largely driven by slope/aspect and elevation respectively.
Indicies of moisture (e.g., drainage) were negatively related to elevation and
hence the negative relationship between volume and temperature (Chapter 2). If
productivity is not measured directly, researchers should use multiple surrogates
and consider other factors correlated with their surrogates that could have
produced the observed patterns.
Our strongest predictor of species richness was heterogeneity at the stand
scale with FHD, HD and their interaction of similar importance. Our estimate of
species richness was likely driven by uncommon species. Several of these,
(e.g., Picoides spp and Turdus migratorius ) have large territories while others
(e.g., Loxia leucoptera) are irruptive without firm territories and are wide ranging.
It may be measurements made at smaller grains do not characterize the
heterogeneity perceived by these species, hence the stronger richnessheterogeneity relationship at the stand grain. Further, a large proportion of the
richness estimates at the 50 and 100 m grains are small relative to that of the
155
stand, due to the much smaller areas surveyed. It may be that these grain sizes
are too small to reflect a large enough range in rare species for richnessheterogeneity patterns to be evident.
Acknowledgements
Funding was provided by the Newfoundland and Labrador Department on
Natural Resources and Human Resources and Skills Development Canada.
Thanks are extended to those assisting in data collection, notably, B. Campbell,
J. Colbert, L. Elson, R. Flynn, M. Michelin, K. Mitchell, R. Neville, F. Phillips, B.
Rodrigues, F. Taylor. D. Goulding and D. Jennings assisted in GIS work and C.
Bourque assisted with, and provided a program to compute abiotic variables.
Thanks are also extended to the Innu Nation. J. Nocera and F. Phillips reviewed
earlier drafts of the manuscript. This is ACWERN Publication No. UNB-XX.
References
Abrams, P.A. 1988. Resource productivity-consumer species diversity: Simple
models of competition in spatially heterogenous environments. – Ecology
69:1418-1433.
Abrams, P.A. 1995. Monotonic or unimodal diversity-productivity gradients: what
does competition theory predict? – Ecology 76:2019-2027.
Abramsky, Z. and Rosenzweig, M.L. 1984. Tilman‘s predicted productivitydiversity relationship shown by desert rodents. – Nature 309:150-151.
Atauri, J.A. and de Lucio, J.V. 2001. The role of landscape structure in species
richness distribution of birds, amphibians, reptiles and lepidopterans in
Mediterranean landscapes. – Landsc. Ecol. 16:147-159.
156
Bibby, C.J., Burgess, N.D. and Hill, D.A. 1993. Bird census techniques. –
Academic Press, Harcourt Brace and Company, Toronto.
Böhning-Gaese, K. 1997. Determinants of avian species richness at different
spatial scales. – J. Biogeogr. 24:49-60.
Boulinier, T., Nichols, J.D., Sauer, J.R., Hines J.E. and Pollock, K.H. 1998.
Estimating species richness: the importance of heterogeneity on species
detectability. – Ecology 79:1018-1028.
Bourque, C.P.-A. and Gullison, J.J. 1998. A technique to predict hourly potential
solar radiation and temperature for a mostly unmonitored area in the Cape
Breton Highlands. – Can. J. Soil Sci. 78:409-420.
Bourque, C.P.-A., Meng, F., Gullison, J.J. and Bridgeland, J. 2000. Biophysical
and potential vegetation growth surfaces for a small watershed in northern
Cape Breton Island, Nova Scotia, Canada. – Can. J. For. Res. 30:11791195.
Burnham, K.P. and Anderson, D.R. 2002. Model selection and multimodel
inference. A practical information-theoretic approach, 2nd Edition. –
Springer-Verlag, New York.
Canada Committee on Ecological Land Classification. 1989. Ecoclimatic Regions
of Canada: First approximation. – Environment Canada, Ecological Land
Classification Series No. 23. Canadian Wildlife Service.
Chase, J.M. and Leibold, M.A. 2002. Spatial scale dictates the productivitydiversity relationships. – Nature 415:427-430.
Chase, J.M. and Ryberg, W.A. 2004. Connectivity, scale-dependence, and the
157
productivity-diversity relationship. – Ecol. Lett. 7:676-683.
Cueto, V.R. and Lopez de Casenave, J. 1999. Determinants of bird species
richness: role of climate and vegetation structure at a regional scale. – J.
Biogeogr. 26: 487-492.
Currie, D.J. 1991. Energy and large-scale patterns of animal and plant-species
richness. – Am. Nat. 137:27-49.
Drapeau, P., Leduc, A., Giroux, J.- F., Savard, J.- P.L., Bergeron, Y. and Vickery,
W. L. 2000. Landscape-scale disturbances and changes in bird
communities of boreal mixed-wood forests. – Ecol. Monogr. 70:423-444.
Emlen, J.T. 1967. A rapid method for measuring arboreal canopy cover. –
Ecology 48:158-159.
Environmental Systems Research Institute (ESRI). 1999. ArcView GIS 3.2. –
ESRI.
Estades, C.F. 1997. Bird-habitat relationships in a vegetational gradient in the
andes of central Chile. – Condor 99:719-727.
Farnsworth, G.L., Pollock, K.H., Nichols, J.D., Simons, T.R., Hines, J.E. and
Sauer, J.R. 2002. A removal model for estimating detection probabilities from
point-count surveys. – Auk 119:414-425.
Fisher, R.A., Corbet, A.S. and Williams, C.B. 1943. The relationship between the
number of species and the number of individuals in a random sample of an
animal population. – J. Anim. Ecol. 12:42-58.
158
Frank, R.M. 1990. Balsam Fir. – In: Burns, R.M., and B.H. Honkala (tech. cords.),
Silvics of North America: Volume 1. Conifers, pp. 26-35. Agric. Handb.
654. USDA For. Serv., Washington, DC.
Gutzwiller, K.J. 1993. Refining the use of point counts for winter studies of
individual species. – Wilson Bull. 105:612-627.
Hunt, P.D. and Flaspohler, D.J. 1998. Yellow-rumped warbler (Dendroica
coronata). – In: Poole, A. and Gill, F. (eds.), The Birds of North America, No.
376. The Birds of North America, Inc., Philadelphia, PA.
Hurlbert, A.H. 2004. Species-energy relationships and habitat complexity in bird
communities. – Ecol. Lett. 7:714-720.
Ingold, J.L. and Wallace, G.E. 1994. Ruby-crowned kinglet (Regulus calendula).
– In: Poole, A. and Gill, F. (eds.), The Birds of North America, No. 119. The
Birds of North America, Inc., Philadelphia, PA.
Kerr, J.T. and Packer, L. 1997. Habitat heterogeneity and a determinant of
mammal species richness in high energy regions. – Nature 385:252-254.
Lack, D. 1933. Habitat selection in birds. – J. Anim. Ecol. 2:239-262.
Legendre, P., Dale, M.R.T., Fortin, M.-J., Gurevitch, J., Hohn, M. and Myers, D.
2002. The consequences of spatial structure for the design and analysis of
ecological field surveys. – Ecography 25:601-615.
Legendre, P. and Legendre, L. 1998. Numerical ecology, 2nd English Edition. –
Developments in Environmental Modelling Vol. 20. Elsevier, New York.
Lichstein, J. W., Simons, T.R. and Franzreb, K.E. 2002a. Landscape effects on
breeding songbird abundance in managed forests. – Ecol. Appl. 12:836-857.
159
Lichstein, J. W., Simons, T.R., Shriner, S.A. and Franzreb, K.E. 2002b. Spatial
autocorrelation and autoregressive models in ecology. – Ecol. Monogr.
72:445-463.
MacArthur, R.H. and MacArthur, J.W. 1961. On bird species diversity. – Ecology
42:594-598.
Mathsoft. 1999. – S-PLUS 2000 for Windows.
McCullagh, P. and Nelder, J.A. 1989. Generalized linear models. – Monographs
on statistics and applied probability 37. Chapman and Hall, New York.
McGarigal, K. and McComb, W. 1995. Relationships between landscape
structure and breeding birds in the Oregon coast range. – Ecol. Monogr.
65:235-265.
Mitchell-Olds, T. and Shaw, R.G. 1987. Regression analysis of natural selection:
statistical inference and biological interpretation. – Evolution 41:149-1161.
Mittelbach, G.G., Steiner, C.F., Scheiner, S.M., Gross, K.L., Reynolds, H.L.,
Waide, R.B., Willig, M.R., Dodson, S.I. and Gough, L. 2001. What is the
observed relationship between species richness and productivity. –
Ecology 82:2381-2396.
Nagelkerke, N.J.D. 1991. A note on the general definition of the coefficient of
determination. – Biometrika 78:691-692.
Nichols, J.D., Boulinier, T., Hines, J.E., Pollock, K.H. and Sauer, J.R. 1998.
Inference methods for spatial variation in species richness and community
composition when not all species are detected. – Conserv. Biol. 12:13901398.
160
Nilsson, S.G. and Nilsson, I.N. 1978. Breeding bird community densities and
species richness in lakes. – Oikos 31:214-221.
Nilsson, S.G. 1979. Density and species richness of some forest bird
communities in south Sweden. – Oikos 33:392-401.
Otis, D.L., Burnham, K.P., White, G.C. and Anderson D.R. 1978. Statistical
inference from capture data on closed animal populations. – Wildl.
Monogr. 62.
Preston, F.W. 1962. The canonical distribution of commonness and rarity: part I.
– Ecology 43: 185-215.
Rosenzweig, M.L. and Abramsky, Z. 1993. How are diversity and productivity
related? – In: Ricklefs, R.E. and Schluter, D. (eds.), Species diversity in
ecological communities. University of Chicago Press, IL, pp. 52-65.
Rosenzweig, M.L. 1995. Species diversity in space and time. – Cambridge
University Press, Cambridge.
Rotenberry, J.T. and Wiens, J.A. 1980. Habitat structure, patchiness, and avian
communities in North American steppe vegetation: A multivariate analysis.
– Ecology 1228-1250.
Roth, R.R. 1976. Spatial heterogeneity and bird species diversity. – Ecology
57:773-782.
Siegal, R.B., DeSante, D.F. and Nott, M.P. 2001. Using point counts to establish
conservation priorities: How many visits are optimal? – J. Field Ornithol.
72: 228-235.
161
Schwab, F.E., Simon, N.P.P. and Carroll, C.G. 2001. Breeding songbird
abundance related to secondary succession in the subarctic forests of
western Labrador. – Écoscience 8:1-7.
Simon, N.P.P., Schwab, F.E. and Otto, R.D. 2002. Songbird abundance in clearcut and burned stands: a comparison of natural disturbance and forest
management. – Can. J. For. Res. 32:1343-1350.
Smith, W.P., Twedt, D.J., Cooper, R.J., Wiedenfield, D.A., Hamel, P.R. and Ford,
R.P. 1997. Sample size and allocation of effort in point count sampling of
birds in bottomland hardwood forests. – In: Ralph, C.J., Sauer, J.R. and
Droedge, S. (eds.), Monitoring bird populations by point counts. General
Technical Report PSW-GTR-149, pp. 7-17.
Smith, R., Hamas, M., Dallman, M. and Ewert, D. 1998. Spatial variation in
foraging of the black-throated green warbler along the shoreline of
northern Lake Huron. – Condor 100:474-484.
Sorace, A., Gustin, M., Calvario, E., Ianniello, L, Sarocco, S. and Carere, C.
2000. Assessing bird communities by point counts: Repeated sessions
and their duration. – Acta Ornithol. 35: 197-202.
Srivastava, D.S. and Lawton, J.H. 1998. Why more productive sites have more
species: An experimental test of theory using tree-hole communities. –
Am. Nat. 152: 510-529.
Tilman, D. and Pacala, S. 1993. The maintenance of species richness in plant
communities. – In: Ricklefs, R.E. and Schluter, D. (eds.), Species diversity in
ecological communities. University of Chicago Press, IL, pp. 13-25.
162
Trzcinski, K.M., Fahrig, L. and Merriam, G. 1999. Independent effects of forest
cover and fragmentation on the distribution of forest breeding birds. – Ecol.
Appl. 9:586-593.
Viereck, L.A. and Johnston, W.F 1990. Black spruce. – In: Burns, R.M. and B.H.
Honkala (tech. cords.), Silvics of North America: Volume 1. Conifers,
Agric. Handb. 654. USDA For. Serv., Washington, DC, pp. 227-237.
Walsberg, G.E. 1981. Nest-site selection and the radiative environment of the
warbling vireo. – Condor 83:86-88.
Wiens, J.A. 1974. Habitat heterogeneity and avian community structure in North
American grasslands. – Am. Midl. Nat. 91:195-213.
Wiens, J.A. 1989. Spatial scaling in ecology. – Funct. Ecol. 3:385-397.
With, K.A. and Webb, D.R. 1993. Microclimate of ground nests: the relative
importance of radiative cover and wind breaks for three grassland species. –
Condor 95: 401-413.
Wright, D.H. 1983. Species-energy theory: an extension of species-area theory.
– Oikos 41: 496-506.
Wright, D.H., Currie, D.J. and Maurer, D.A. 1993. Energy supply and patterns of
species richness on local and regional scales. – In: Ricklefs, R.E. and
Schluter, D. (eds.), Species diversity in ecological communities. University of
Chicago Press, Chigaco, pp. 66-77.
163
Table 5.1. General description of dominant local vegetation (crown cover) and
landscape composition (% of circular buffer of varying radii) at 3 grain sizes.
50 m radius

Local vegetationa
BLT < 0.5m 0.46
BLT 0.5 - 2 3.89
BLT 2 - 10
1.57
BLT > 10
3.41
Con < 0.5
4.55
Con 0.5 - 2 19.27
Con 2 - 10 31.59
Con > 10
11.64
Landscape context
HT 500
52.75
HT 1000
50.54
HT 2000
49.13
HT 2000
HT 3000
HT 4000
ES 500
3.32
ES 1000
5.68
ES 2000
4.61
ES 2000
ES 3000
ES 4000
-
100 m radius
Stand
Min
Max

0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
15.00
105.00
75.00
75.00
35.00
80.00
100.00
95.00
0.47
3.06
1.11
0.15
5.45
19.67
31.71
12.96
0.00
0.00
0.00
0.00
0.00
1.00
9.00
0.00
8.00
41.00
34.00
26.00
29.00
52.00
68.00
55.00
0.00
1.88
0.61
0.00
4.89
17.89
31.39
10.05
0.00
0.00
0.00
0.00
1.00
5.00
16.00
0.20
0.00
8.00
3.00
0.00
10.00
30.00
47.00
23.00
0.00
3.45
5.75
0.00
0.00
0.00
-
99.50
95.60
87.64
46.24
56.81
36.29
-
52.94
50.77
49.29
3.32
5.68
4.61
-
0.00
3.45
5.75
0.00
0.00
0.00
-
99.50
95.60
87.64
46.24
56.81
36.29
-
48.62
47.16
45.92
4.50
5.12
4.00
6.34
10.31
13.13
0.00
0.70
0.57
82.28
72.00
64.04
25.04
18.48
10.99
Min
Max

Min
Max
a
BLT = broad-leaved trees (by height zone, m), Con = conifers (by height zone, m).
a
HT = Heavily stocked (>100 m3/ ha) mature forest (> 130 years old), ES = early successional
forest (< 15 years old).
164
Table 5.2. Scale-specific correlations (Pearsons) between the 3 productivity
measures, volume (V), solar radiation (SR), Temperature (T).
Grain size
50 m
V
SR
T
100 m
V
SR
T
Stand
V
SR
T
Productivity
V
1.000
-0.195
-0.393
SR
T
1.000
0.043
1.000
1.000
-0.190
-0.568
1.000
0.045
1.000
1.000
0.013 1.000
-0.812 -0.218
1.000
165
Table 5.3. The effects of volume on heterogeneity [foliage height diversity (FHD)
and horizontal diversity (HD) and their interaction (FHD:HD)] and species
richness as determined by least-squares and Poisson (species richness)
regression. Relationship determined by change in Akaike‘s Information Criteria (
AIC(c)), corrected for small sample size at the stand grain. The form of curvilinear
relationships (those including the squared term) was evaluated by the MitchellOlds and Shaw test (MOS).
Grain
Best Modela
Model
 AIC(c)
FHD
50 m
0.614 + 0.420 * vol - 0.133 * vol2
100 m 0.580 + 0.109 * vol -0.009 * vol2
Stand 0.500 + 0.159 * vol - 0.017 * vol2
HD
50 m
37.850 -7.754 * vol + 0.533 * vol2
100 m 51.683 -25.171 * vol + 7.989 * vol2
Stand 58.330 - 9.517 * vol + 0.818 * vol2
FHD:HD
50 m
100 m
Stand 34.381 -1.443 * vol
Richness
50 m
100 m
Stand 3.220 - 0.070 * vol
a
R2
MOS Relationship
 AIC(c)
9.939 0.23 31.93
47.149 0.45 5.35
7.827 0.69 0.75
Unimodal
Unimodal
Unimodal
8.691
5.602
1.29
0.20 37.86
0.18 35.54
0.69 7.40
U-shaped
U-shaped
U-shaped
3.36
0.02
0.05
0.23
None
None
L. decline
1.45
0.04
0.02
0.23
None
None
L. decline
Species richness is predicted by raising e to the power of the above equations.
166
Table 5.4. The effects of mean daily temperature (T) on heterogeneity [foliage
height diversity (FHD), horizontal diversity (HD) and their interaction (FHD:HD)]
and species richness as determined by least-squares and Poisson (species
richness) regression. Relationship determined by change in Akaike‘s Information
Criteria ( AIC(c)), corrected for small sample size at the stand grain. The form of
curvilinear relationships (those including the squared term) was evaluated by the
Mitchell-Olds and Shaw test (MOS).
Grain
size
Best Modela
Model R2 MOS Relationship
 AIC(c)
 AIC(c)
FHD
50 m
0.14
2
100 -4.892 + 1.396 * T - 0.085 * T
19.41 0.34 0.07
2
Stand -7.240 + 1.913 * T - 0.113 * T
3.73 0.53 2.41
HD
50 m -293.237 + 63.978 * T - 3.108 * T2 0.01 0.17 19.21
100
0.09
Stand 11.820 + 2.036 * T
2.21 0.48
FHD:HD
50 m
0.06
100
0.04
2
4.20 0.39 3.26
Stand -636.291 + 147.729 * T -8.144 * T
Richness
50 m
0.09
100
0.08
Stand 2.159 + 0.095 * T
0.01 0.27
a
None
C. decline
C. decline
Unimodal
None
L. increase
None
None
C. increase
None
None
L. increase
Species richness is predicted by raising e to the power of the above equations.
167
Table 5.5. The effects of mean daily solar radiation (SR) on species richness
and heterogeneity [foliage height diversity (FHD), horizontal diversity (HD) and
their interaction (FHD:HD)], as determined by Poisson (species richness) and
least-squares regression. Relationship determined by change in Akaike‘s
Information Criteria ( AIC(c)), corrected for small sample size at the stand grain.
The form of curvilinear relationships (those including the squared term) were
evaluated by the Mitchell-Olds and Shaw test (MOS).
Grain
size
Best Modela
Model
 AIC(c)
R2
MOS Relationship
 AIC(c)
FHD
0.03
0.01
0.04
None
None
None
0.01
0.02
0.14
None
None
None
0.80
<0.01
<0.01
0.23
0.24
None
None
C. decline
5.24
0.13
0.01
0.56
0.08
None
None
C. decline
50 m
100
Stand
HD
50 m
100
Stand
FHD:HD
50 m
100
2
Stand -397.567+72.652*SR-3.082*SR
Richness
50 m
100
Stand -28.50 + 5.569 * SR - 0.245 * SR2
a
Species richness is predicted by raising e to the power of the above equations.
168
Table 5.6. Poisson regressions of species richness predicted by structural (FHD,
HD) and landscape diversity. Relationships determined by change in Akaike‘s
Information Criteria ( AIC(c)), corrected for small sample size at the stand grain,
AIC(c) weight (w) and relative importance (RI). The AIC(c) values for the
landscape diversity models is the difference in AIC between the best structural
diversity model with the respective landscape diversity term added.
Grain
a
Models and Relative Importance (RI)
Structural diversity
50 m
100 m
Stand RI: FHD = 0.88, HD = 0.86, FHD:HD = 0.87
-9.305+13.406 *FHD+0.261*HD-0.278*FHD:HD
-0.213+2.374 *FHD+0.034*HD
Landscape diversity
50 m 0.796+1.292*FHD+0.026*HD-0.038*FHD:HD+0.439*H500
0.831+ 1.172*FHD+0.023*HD-0.034*FHD:HD+0.376*H2000
100
2.284-4.232E-4 *FHD+0.163*H500
2.303-6.137E-4 *FHD+0.086*H2000
Stand -6.208+10.217*FHD+8.565*HD-8.980*FHD:HD-0.340*H2000
-6.654+11.138*FHD+9.812*HD-10.142*FHD:HD-1.004*H4000
a
2
R
 AIC(c)
w
0.00
3.65
0.72
0.12
0.13
0.03
0.73
0.23 -18.00
0.18 -6.29
0.03 -1.08
0.01 +1.70
0.73 -0.55
0.77 -0.92
Species richness is predicted by raising e to the power of the above equations.
169
Figure 5.1. Relationships between timber volume, vegetation heterogeneity and
songbird species richness. Foliage height diversity = FHD, horizontal diversity
HD and their interaction = FHD X HD. Lines correspond to the particular models
with most support, i.e., lowest AIC(c), in Table 5.3. Where no relationship was
detected, lines represent the global model.
170
Figure 5.2. Relationships between mean daily temperature, vegetation
heterogeneity and songbird species richness. Foliage height diversity = FHD,
horizontal diversity HD and their interaction = FHD X HD. Lines correspond to
the particular models with most support, i.e., lowest AIC(c), in Table 5.4. Where
no relationship was detected, lines represent the global model.
171
Figure 5.3. Relationships between mean daily solar radiation, vegetation
heterogeneity and songbird species richness. Foliage height diversity = FHD,
horizontal diversity HD and their interaction = FHD X HD. Lines correspond to
the particular models with most support, i.e., lowest AIC(c), in Table 5.5. Where
no relationship was detected, lines represent the global model.
172
Chapter 6 – General discussion and conclusions.
Research has indicated that some food web structures that include
songbirds are controlled by bottom-up (resource availability) factors (Folkard and
Smith 1995, Seagle and Sturtevant 2005). Seagle and Sturtevant (2005)
determined that topography influenced nutrient input into detrital food webs. This
increased invertebrate biomass lead to higher Ovenbird (Seiurus aurocapillus)
abundance and reproductive output. Similarly, Folkard and Smith (1995)
suggested that changes in vegetation structure and increased invertebrate
densities caused increases in bird abundances following experimental
fertilization. Thus, it seemed reasonable not only to predict that changes in soil
and topography would influence vegetation structure in my study area, but that
these influences would extend to higher food chain levels. My results show only
the first of these two predictions to be true. Arthropods showed no relationship
with vegetation and factors relating to the ability of birds to assess resources,
and possibly limits imposed by climate were responsible for the failure of the
bottom-up control extending to songbirds.
Plants generally responded to environmental variables as reported by
other studies. Elevation was a dominant factor in predicting plant abundances,
however, this relationship did not result from elevation differences in temperature
or exposures as one would predict based on Rowe (1972). Rather, these
patterns were most likely due to edaphic factors which covaried with elevation
(Wilton 1959, Damman 1967, Siccama et al. 1970). Most birds in our study nest
or feed in trees and tall shrubs, thus these were the most likely plants to
173
influence bird habitat quality in our study area. High Picea mariana volumes
occurred on low to moderate slopes, but Abies balsamea volumes and the cover
of Alnus crispa and Betula papyrifera cover were more restricted to higher
elevations and steeper slopes. These results agree with Wilton (1959, 1965) and
Foster (1984) who concluded that productive forests were largely confined to
moist seepage routes on concave slopes at intermediate elevations where
continual water movement enhanced aeration and nutrient availability.
Despite the relatively strong relationship between plant abundances and
environmental variables, the variation in vegetation did not appear to influence
habitat quality (i.e., abundance, reoccupancy and breeding activity) for most
species. With the exception of hermit thrushes, relationships were either nonexistent or weak and inconsistent between years. Although forest types were
similar relative to large-scale disturbance induced change (i.e., they were
dominated by Picea mariana) there were structural differences. It seems
reasonable to expect canopy and stem feeders to be present in all forest types
but their densities should increase with tree height, crown cover, stem densities
and timber volume. These features varied by 300 – 500 % across forest types, a
level of variation that I expected to influence habitat selection for most of my bird
species. Where vegetation descriptions were reported in sufficient detail, I
compared my study with those conducted in warmer climates (DeGraaf and
Chadwick 1987, Parker et al. 1994, Schwab and Sinclair 1994). These studies
have a wide range of forest types, but I limited my comparisons to their
treatments that have a similar range of vegetation structure as mine. These
174
studies report greater differences in bird abundance across what appears to be a
range of vegetation structures similar to mine.
I evaluated 3 factors that predicted weak bird-vegetation relationships: i)
birds using climactic factors in selecting breeding sites, ii) birds having difficulty
tracking resources and iii) reduced interspecific competition. My bird-climate
models explained little variation in habitat quality, suggesting most birds do not
use climatic cues in selecting territories (across the scales we measured).
However, my results indicate the latter two predictions are, at least in part,
contributing to the weak bird-vegetation relationships.
I found little relationship between site reoccupancy and habitat variables,
and no consistency in sites occupied between years. This suggests that birds
were not deterministically selecting the highest quality sites but instead were
settling for ‗satisfactory conditions‘ rather than seeking the ‗optimal conditions‘ as
Haila et al. (1996) concluded. However, it is possible that the habitat quality of all
sites was similar, i.e., there was no optimum to select. Deterministically selecting
optimal conditions is not possible where individuals have poor knowledge of
resource availability (Wiens 1985, Orians and Wittenburger 1991, Haila et al.
1996). My results show that arthropod biomass (the primary breeding food
resource for most of our species) is poorly related to forest type and varies
erratically within the breeding season. Thus, arthropod biomass at the time a
territory is selected, does not indicate arthropod abundance throughout the
breeding season. Where arthropods vary fairly independently of vegetation
features, the latter may not be a good surrogate of food availability in birds. My
175
results contrast with Seagle and Sturtevant (2005) but support Forsman et al.
(1998) who concluded that migrant birds in northern forests cannot use arthropod
abundance to determine breeding site profitability.
Reduced competition, potentially driven by climate, could also contribute
to the poor bird-habitat relationships. Harsh northern climates limit bird numbers,
causing erratically varying bird abundances in northern Europe (Järvinen 1979,
Helle and Mönkkönen 1986, Virkkala 1991a, 1991b, Morozov 1993, Hogstad et
al. 2003). If climate reduces densities sufficiently to decrease interspecific
competition, birds could have broader niches (Morse 1974, Cody 1981, Martin
and Martin 2001) implying less specific habitat associations. The assumption
that harsh conditions reduce competition and favor the coexistence of
ecologically similar species has been challenged theoretically (Chesson and
Huntly 1997) and empirically (Mönkkönen 1990). However, Mönkkönen et al.
(2004) document positive associations between potentially competing species at
low densities (in northern areas) and negative associations at high densities
(more southern). Further, the prevalence of heterospecific attraction in northern
areas suggests competitive interactions are weak (Forsman et al. 1998,
Mönkkönen et al. 1999). My results indicate that bird abundances fluctuate in
parallel (i.e., all species tend to follow the same population trend) as opposed to
compensatory density fluctuations. This is evidence of weak interspecific
competition, a result similar to others in northern forests (Järvinen 1979, Helle
and Mönkkönen 1986, Morozov 1993).
176
Despite the general weak bird-vegetation relationships, where birdvegetation relationships existed there was a relatively strong landscape effect.
Landscape ecologists expect the influence of landscape content to be inversely
proportional to the amount of habitat on the landscape (Andrén 1994, Drolet et al.
1999, Fahrig 2003). Early successional forests were uncommon in our study
area, suggesting a landscape effect would be more likely for early successional
species. Consistent with this reasoning, three of our species whose abundances
were related to landscape composition variables, black-backed woodpecker,
Tennessee warbler, and white-throated sparrow, are characteristic of early
successional forests. Where vegetation was able to explain some variation in
abundance, partial R2 values indicate landscape variables were of similar or
greater importance than local variables in determining bird abundances, similar
to Drapeau et al. (2000) but contrary to other studies in forested areas (Hagan
and Meehan 2002, Lichstein et al. 2002).
My only observed richness-productivity patterns occurred at the stand
grain and were relatively weak for timber volume and temperature. However, the
three different measures produced 3 different results; linear decline, a linear
increase and a curvilinear decline with volume, temperature and radiation
respectively. It is possible that a productivity-heterogeneity relationship produced
the first two of these patterns: FHD X HD and richness followed the same
pattern with volume and temperature and richness was positively associated with
HD. However, this was not the case with radiation. Species richness and the
only structural diversity measure showing a pattern (FHD X HD) showed a
177
curvilinear decline. However, FHD X HD was a negative predictor of species
richness and the richness-radiation curve was much stronger than the HD X
FHD-radiation curve, a result that conflicts with other research (e.g., BöhningGaese (1997), Cueto and Lopez de Casenave (1999), Hurlbert (2004)).
Despite the infrequency of productive forests on central Labrador‘s
landscape, my results suggest that most vegetation and avifauna could tolerate
some reduction of these forests through logging. The lack of discrete vegetation
groupings suggests that eliminating a particular group on the landscape would
not likely cause major species extirpation. However, plant species often showed
very different abundances among groups, so elimination of a particular group
could drastically reduce the abundance of some species on the landscape. The
RDA and MRT showed soil and topographic features create different vegetation
assemblages. One obvious forest management-conservation conflict is that sites
with highest timber volumes are usually a high priority for logging and these sites
had the highest diversity of understory plants. However, this conservation issue
did not extend to the bird community as richness declined with timber volume.
My 3 measures of habitat quality did not show a strong positive increase with
productivity for any species suggesting that low productivity forests will support
most forest birds in this region. I have confidence in this result because these
different, but complimentary measures show similar patterns. Thus, reducing the
amount of high volume mature stands would not likely have long-term negative
impacts on birds in this landscape. This is partly because many species in this
system are characteristic of open areas or can tolerate canopy openings. Other
178
species typically associated with closed forests (e.g., boreal chickadees) showed
weak responses to vegetation. However, the relatively strong relationship
between pine siskins and the proportion of heavy timber within 500 m,
suggesting that the effects of cutting heavy timber could reduce pine siskin
numbers beyond the actual harvesting area.
References
Andrén, H. 1994. Effects of habitat fragmentation on birds and mammals in
landscape with different proportions of suitable habitat: a review. – Oikos
71:355-366.
Böhning-Gaese, K. 1997. Determinants of avian species richness at different
spatial scales. – J. Biogeogr. 24:49-60.
Chesson, P. and Huntly, N. 1997. The role of harsh and fluctuating conditions in
the dynamics of ecological communities. – Am. Nat. 150:519-553.
Cody, M.L. 1981. Habitat selection in birds: The roles of vegetation structure,
competitors, and productivity. – Bioscience 31:107-113.
Cueto, V.R. and Lopez de Casenave, J. 1999. Determinants of bird species
richness: role of climate and vegetation structure at a regional scale. – J.
Biogeogr. 26:487-492.
Damman, A.W.H. 1967. The forest vegetation of western Newfoundland and site
degradation associated with vegetation change. – Ph.D. thesis, University
of Michigan, MI.
DeGraaf, R.M. and Chadwick, N.L. 1987. Forest type, timber size class, and New
England breeding birds. – J. Wildl. Manag. 51:212-217.
179
Drapeau, P., Leduc, A., Giroux, J.- F., Savard, J.- P.L., Bergeron, Y. and Vickery,
W. L. 2000. Landscape-scale disturbances and changes in bird
communities of boreal mixed-wood forests. – Ecol. Monogr. 70:423-444.
Drolet, B., Desrochers, A. and Fortin, M.-J. 1999. Effects of landscape structure
on nesting songbird distribution in a harvested boreal forest. – Condor
101:699-704.
Fahrig, L. 2003. Effects of habitat fragmentation on biodiversity. – Annu. Rev.
Ecol. Evol. Syst. 34:487-515.
Folkard, N.F.G. and Smith, J.N.M. 1995. Evidence for bottom-up effects in the
boreal forest: Do passerine birds respond to large-scale experimental
fertilization? – Can. J. Zool. 73:2231-2237.
Forsman, J.T., Mönkkönen, M., Helle, P. and Inkeröinen, J. 1998. Heterospecific
attraction and food resources in migrants‘ breeding patch selection in
northern boreal forest. – Oecologica 115:278-286.
Foster, D.R. 1984. Phytosociological description of the forest vegetation of
southeastern Labrador. – Can. J. Bot. 62:899-906.
Hagan, J.M. and Meehan, A.L. 2002. The effectiveness of stand-level and
landscape-level variables for explaining bird occurrence in an industrial forest.
– For. Sci. 48:231-242.
Haila, Y., Nicholls, A.O., Hanski, I. and Raivio, S. 1996. Stochasticity in bird
habitat selection: year-to-year changes in territory locations in a boreal
forest bird assemblage. – Oikos 76:536-552.
Helle, P. and Mönkkönen, M. 1986. Annual fluctuations of land bird communities
180
in different successional stages of boreal forest. – Ann. Zool. Fenn.
23:269-280.
Hogstad, O., Selǻs, V. and Kobro, S. 2003. Explaining annual fluctuations in
breeding density of fieldfares Turdus pilaris – combined influences of
factors operating during breeding, migration and wintering. – J. Avian Biol.
34:350-354.
Hurlbert, A.H. 2004. Species-energy relationships and habitat complexity in bird
communities. – Ecol. Lett. 7:714-720.
Järvinen, O. 1979. Geographical gradients of stability in European land bird
communities. – Oecologia 38:51-69.
Lichstein, J. W., Simons, T.R. and Franzreb, K.E. 2002. Landscape effects on
breeding songbird abundance in managed forests. – Ecol. Appl. 12:836-857.
Martin, P.R. and Martin, T.E. 2001. Ecological and fitness consequences of
species coexistence: A removal experiment with wood warblers. – Ecology
82:198-206.
Mönkkönen, M. 1990. Removal of territory holders causes influx of small-sized
intruders in passerine bird communities in northern Finland. – Oikos
57:218-288.
Mönkkönen, M., Forsman, J.T. and Thomson, R.L. 2004. Qualitative
geographical variation in interspecific interactions. – Ecography 27:112118.
Morozov, N.S. 1993. Short-term fluctuations in a south-taiga bird assemblage:
support for an ―individualistic‖ view. – Ornis Fenn. 70:177-188.
181
Morse, D.H. 1974. Niche breadth as a function of social dominance. – Am. Nat.
108:818-830.
Orians, G.H. and Wittenberger, J.F. 1991. Spatial and temporal scales in habitat
selection. – Am. Nat. 137:S29-S49.
Parker, G.R., Kimball, D.G. and Dalzell, B. 1994. Bird communities breeding in
selected spruce and pine plantations in New Brunswick. – Can. Field-Nat.
108:1-9.
Rowe, J.S. 1972. Forest regions of Canada. – Department of Environment,
Canadian Forest Service Publication No. 1300. Crown Copyright.
Schwab, F.E. and Sinclair, A.R.E. 1994. Biodiversity of diurnal breeding bird
communities related to succession in the dry Douglas-fir forests of
southeastern British Columbia. – Can. J. For. Res. 24:2043-2040.
Seagle, S.W. and Sturtevant, B.R. 2005. Forest productivity predicts invertebrate
biomass and ovenbird (Seiurus aurocapillus) reproduction in Appalachian
landscapes. – Ecology 86:1531-1539.
Siccama, T.G., Bormann, F.H. and Likens, G.E. 1970. The Hubbard Brook
ecosystem study: productivity, nutrients, and phytosociology of the
herbaceous layer. – Ecol. Monogr. 40:389-402.
Virkkala, R. 1991a. Population trends of forest birds in a Finnish Lapland
landscape of large habitat blocks: consequences of stochastic
environmental variation of regional habitat alteration? – Biol. Conserv.
56:223-240.
182
Virkkala, R. 1991b. Spatial and temporal variation in bird communities and
populations in north-boreal coniferous forests: a multiscale approach. –
Oikos 62:59-66.
Wiens, J.A. 1985. Habitat selection in variable environments: Shrub-steppe birds.
– In: Cody, M.L. (ed.), Habitat Selection in Birds. Academic Press, Inc.,
pp. 227-251.
Wilton, W.C. 1959. Forest types of the Grand Lake and Northwestern Lake
Melville areas of Labrador. – Forest Research Division Technical Note No.
83. The Queen‘s Printer, Ottawa.
Wilton, W.C. 1965. The forests of Labrador. – Department of Forestry
Publication, No 1066. The Queens printer, Ottawa.
183
Curriculum Vitae
Candidate‘s full name:
Neal Phillip Perry Simon
Universities attended:
Labrador Community College (1991-1992)
Memorial University of Newfoundland (1992-1996 B.Sc. (Hons))
University of New Brunswick (1996-1998, M.Sc.F)
Publications:
Elson, L.T., Simon, N.P.P. In press. Plant abundances following clearcutting and
stripcutting in central Labrador. Northern Journal of Applied Forestry.
Schwab, F.E., Simon, N.P.P., Stryde, S. and Forbes, G. In press. Effect of postfire snag removal on breeding birds of western Labrador. Journal of Wildlife
Management
Schwab, F.E., Simon, N.P.P., Sinclair, A.R.E. In press. Breeding birds related to
vegetation structure in southeast British Columbia. Journal of Wildlife
Management.
Simon, N.P.P. and Schwab, F.E. 2005. Plant community structure following
wildfire in the subarctic forests of Labrador. Northern Journal of Applied
Forestry 22: 229-235.
Newbury, T.L. and Simon, N.P.P. 2005. The effects of clearcutting on snowshoe
hare (Lepus americanus) relative abundance in central Labrador. Forest
Ecology and Management 210: 131-142.
Schwab, F.E., Simon, N.P.P., Nash, S. 2005. Sex and age segregation of
wintering willow ptarmigan in Labrador. Northeastern Naturalist 13:113 -118.
Betts, M., Simon, N.P.P., Nocera, J.J. 2005. Point count summary statistics
differentially predict reproductive activity in bird-habitat relationship studies.
Journal of Ornithology 146: 151-159.
Simon, N.P.P., Schwab, F.E. 2005. The response of conifer and broad-leaved
trees and shrubs to wildfire and clearcut logging in the boreal forests of
central Labrador. Northern Journal of Applied Forestry 22:35-41.
Otto, R.D., Simon, N.P.P., Couturier, S. Schmelzer, I. 2003. Evaluation of
satellite collar sample size requirements for mitigation of low-level military
jet disturbance of the George River caribou herd. Rangifer special issue no
14: 297-302.
Simon, N.P.P., Diamond, A.W., Schwab, F.E. 2003. Do northern forest bird
communities show more ecological plasticity than southern forest bird
communities in eastern Canada? Écoscience 10:298-296.
Simon, N.P.P., Schwab, F.E., Otto, R.D. 2002. Songbird abundance in clear-cut
and burned stands: a comparison of natural disturbance and forest
management. Canadian Journal of Forest Research 32:1343-1350.
Simon, N.P.P., Stratton, C.B., Forbes, G.J., and Schwab, F.E. 2002. Similarity of
small mammal abundance in post-fire and clearcut forests. Forest Ecology
and Management 165: 163-172.
Schwab, F.E., Simon, N.P.P., Carroll, C.G. 2001. Breeding songbird abundance
in the subarctic forests of western Labrador. Écoscience 8:1-7.
Schwab, F.E, Pitoello, F.G. and Simon, N.P.P. 2001. Relative palatability of
green manure crops and carrots to white-tailed deer. Wildlife Society
Bulletin 29:317-321.
LeCoure, M.I., Schwab, F.E., Simon, N.P.P. and Diamond, A.W. 2000. Effects of
post-fire salvage logging on breeding birds in western Labrador. Northeast
Wildlife 55:39-46.
Simon, N.P.P., Schwab, F.E, and Diamond, A.W. 2000.Patterns of bird
abundance in relation to logging in western Labrador. Canadian Journal of
Forest Research 30: 257-263.
Simon, N.P.P., Schwab, F.E, LeCoure, M.I., Phillips, F.R. and Trimper, P.G.
1999. Effects of trapper access on marten population in central Labrador.
Northeast Wildlife 54:73-76.
Simon, N.P.P., Schwab, F.E, LeCoure, M.I., Phillips, F.R. 1999. Fall and winter
diet of Martens, Martes americana, in central Labrador related to small
mammal densities. Canadian Field-Naturalist 113: 678-680.
Simon, N.P.P., Schwab, F.E., Baggs, E.M. and Cowan, G.I. Mct. 1998.
Distribution of small mammals among successional and mature forest types
in western Labrador. Canadian Field-Naturalist. 112:441-445.
Conference Presentations:
2003 Evaluating the effects of landscape change on abundance, productivity,
and survival of forest birds (presented with M. Betts). ACWERN, Wolfville,
NS.
2003 Natural disturbance and forest management. Workshop on Natural
Disturbance Management. Sustainable Forest Management Network,
Corner Brook, NL.
2002 The relationship between forest productivity and songbird habitat quality.
ACWERN, Rocky Harbour, NL
2000 Songbird abundance and fecundity in relation to forest structure and
productivity. ACWERN, St. Andrews NB.
1998 Patterns of bird abundance in relation to logging in western Labrador/ Bird
plasticity in relation to latitude. Institute of Environmental Monitoring and
Research Seminar Series - Goose Bay, NL.
1997 The effects of logging on birds in western Labrador. ASWFB/ACWERN
conference - Alma, NB.
1996 The effects logging on birds in western Labrador. ACWERN, Bon Portage
Island, NS.
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement