Habitat Use by Riparian and Upland Birds in Oldgrowth Coastal British Columbia Rainforest

2005·9 MB·English

by  Susan M Shirley

Wilson Bulletin 117(3):245-257, 2005
HABITAT USE BY RIPARIAN AND UPLAND BIRDS IN OLD-
GROWTH COASTAL BRITISH COLUMBIA RAINFOREST
SUSAN M. SHIRLEY^
—
ABSTRACT. The value of riparian habitats to birds differs among ecosystems. I tested whether riparian
habitat near large streams and rivers in the Pacific Northwest supported a higher abundance and diversity of
birds than adjacent upland forest. From 1996 to 1998, I surveyed breeding birds at four 9-ha sites in coastal
western hemlock forest on western Vancouver Island, British Columbia. Five species offorest generalists dom-
inated both riparian and upland bird communities: Winter Wren {Troglodytes troglodytes). Chestnut-backed
Chickadee (Poecile rufescens), American Robin (Turdus migratorius), Swainson’s Thrush (Catharus ustulatus),
and Pacific-slope Flycatcher (Empidonax difficilis). Species richness and total abundance were similar over the
riparian-to-upland gradient, whereas abundances ofriparian specialists and aerial foragers declinedwithdistance
from the river. To explore whether vegetation composition and structure explained bird distribution patterns, I
sampled three locations along both riparian and upland transects at each site. Riparian areas hadhigherdensities
of deciduous trees; conifer and snag densities were higher in upland areas. Salmonberry (Rubus spectahilis)
cover was marginally higher in riparian areas and blueberry {Vaccinium spp.) cover was higherin upland areas.
There was little effect of distance from the river on most bird species, but there were stronger associations of
birds with specific vegetation attributes. Tree and snag densities explained the most variation in abundance of
aerial foragers, and eight of nine individual species, whereas distance from the river and shrub cover were
important predictors of Hammond’s Flycatcher {Empidonax hammondii) abundance. Apart from riparian spe-
cialists and a few species with strong vegetation associations, bird assemblages in riparian and upland habitats
of this moist forest type were dominated by similar sets of generalist species. Received 1 December 2003,
accepted26April 2005.
Riparian habitats are influenced by both 1985, reviewed in Knopf and Samson 1994),
stream channel processes and the adjacent up- where bird diversity is higher in riparian and
land vegetation (Brinson et al. 1981, Naiman lower in upland areas. In contrast, studies of
et al. 1993). Topography, plant communities, mature, undisturbed stands in forests with
hydrologic regimes, and soil type typically greater annual precipitation (McGarigal and
distinguish riparian areas from upland areas. McComb 1992, Murray and Stauffer 1995,
Riparian habitats are heavily influenced by Wiebe and Martin 1998) have shown equal or
seasonal changes in water flow, and alluvial lower diversity in riparian habitats compared
soils in riparian habitats tend to be wetter than with upslope habitats; these studies (Mc-
soils in uplands. Riparian plant communities Garigal and McComb 1992, Murray and
have diverse vegetation structures, high edge: Stauffer 1995, Wiebe and Martin 1998), how-
area ratios, and are dominated by woody veg- ever, focused on riparian areas associated with
etation. These features are common to all ri- small (<5 m wide) mountain streams. Some
, parian habitats, but vary greatly depending on riparian areas show greater diversity near larg-
geographical location. Riparian ecosystems er streams and rivers (Knopf and Samson
' often support high bird diversity and abun- 1994, Lock and Naiman 1998), anti the avian
dance (Thomas et al. 1979, Knopfet al. 1988, use of riparian habitat relative to uplaiul hab-
Anthony et al. 1996) because oftheir complex itat along larger streams aiul rivers in the Pa-
vegetation structure (LaRue et al. 1995, Wicbe cific Northwest has not been well examinctl.
and Martin 1998), high plant diversity (Bull
1978, Raedeke 1988), and proximity to water. I studietl avian habitat use along larger
streams and rivers within continuous undis-
There is a strong bird diversity gradient
efrrnomanrdipaargiraincultotuurapllarnedgihoanbsitoatfstihne sUo.uSt.hw(ee.sgt.-, turMbeydffirosrtesotbjoefcttihveeIw^aicsiltico tNeostrtthhweeshty.pothesis
vStauHer and Best 1980, S/aro 1980, Knopf that bird species diversity and abundanee is
higher in old-growth riparian habitat associ-
' Dept, ofZoology, Univ. of British C'oliimbi:>. 6270 ated with large streams and ri\ers than in ad-
University Blvd., Vancouver, BC' V6I. IVS. ('anatla; jacent old-growth upland habitat. In the lii-
e-mail: Shirley /oology.ubc.ca cifie Northwest, riparian /ones tend to be
245
:
246 THE WILSON BULLETIN • Vol. 117, No. 3, September 2005
dominated by deciduous trees, whereas up- bedded within a mosaic of forest patches of
lands are dominated by conifers (McGarigal different ages across a landscape in which the
and McComb 1992, Pearson and Manuwal amount of primary forest varied from 50 to
2001). Disproportionate use of riparian habi- 70%. I classified rivers to stream order at a 1
tats by birds should be reflected in a decline 50,000-map scale based on branching follow-
in species richness and abundance with in- ing Kuehne (1962). The Nahmint and Taylor
creasing distance from the river, but associa- rivers (fourth order) and the Klanawa River
tion with riparian habitat may vary among (fifth order) ranged in width from 15 to 57 m.
species and guilds. Riparian specialists that Study sites were located in the Western Van-
rely on the stream or river as a food source couver Island ecoregion, within the moist-to-
and nest near streams should decline in abun- very-wet maritime biogeoclimatic subzones of
dance with increasing distance from the river. coastal western hemlock (Klinka et al. 1991,
Riparian forest edges along streams support a Nuszdorfer and Boettger 1994). The forest is
higher invertebrate biomass (Murakami and dominated by amabilis fir {Abies amabilis),
Nakano 2002), due to higher densities of western hemlock (Tsuga heterophylla), and
aquatic insects (Murakami and Nakano 2002), western red cedar {Thuja plicata). Red alder
and possibly greater primary productivity {Alnus rubra) and bigleafmaple {Acermacro-
(Ranney et al. 1981). Aerial foragers such as phyllum) occur at their greatest densities ad-
flycatchers may respond to emergent aquatic jacent to the rivers, but were also scattered
insects near streams and rivers (Gray 1993) throughout the forest in moister areas. The un-
and should occur at highest densities near the derstory is dense, highly stratified, and con-
water’s edge. Conversely, conifer specialists tains shrubs such as salmonberry {Rubusspec-
may increase with distance from the river due tabilis), red huckleberry {Vaccinium parvifo-
to increasing conifer densities (McGarigal and lium), salal {Gaultheria shallon), and devil’s
McComb 1992, Pearson and Manuwal 2001). club {Oplopanax horridus), with Alaskan
Second, I explored how variation in vege- {Vaccinium alaskense) and oval-leafblueberry
tation composition and structure from riparian {V. ovalifolium) predominating in the upland
to upland habitat explains distribution patterns areas. The climate is cool and wet in winter
of several species. Studies conducted in other and warm and dry in late summer (July-Sep-
temperate coniferous forests have revealed tember). Annual precipitation in the area av-
differences in vegetation structure and com- erages 3,100 mm, and daily temperatures av-
position between riparian and upland (Mc- erage 3.2° C in January and 15.6° C in July
Garigal and McComb 1992, Pearson and Ma- (Environment Canada Climate Data Services).
—
nuwal 2001). If the structural and species at- Vegetation sampling. I sampled vegeta-
tributes of riparian vegetation communities tion during 1995 within 20-m-radius circular
are the primary predictors of bird diversity plots (0.13 ha) using a procedure modified
and abundance, then use of riparian habitats from James and Shugart (1970). I accounted
should be related to the prevalence of these for the large size oftrees locally by increasing
structures relative to upland areas. Alterna- the plot radius (Mueller-Dombois and Ellen-
tively, bird diversity and abundance should berg 1974, Bryant et al. 1993). To improve
not differ between riparian and upland habi- the accuracy of visual estimation over a large
tats that are similar in vegetation structure and area, each plot was divided into four quadrats.
composition. I sampled vegetation in each ofthe four quad-
rats and then calculated means of the four
METHODS
quadrats for each variable. Plots were placed
Study area.—The study was conducted in at three stations 150 m apart along two 500-
m
three valleys on the west coast of Vancouver transects: one in riparian and one in upland
Island, British Columbia, between Ucluelet in habitat. Transects were oriented parallel to the
the north and Bamfield in the south (48° 5' N, river, 20 m (riparian) and 160 m (upland) from
125° 5' W). Four sites of continuous old- the river, for a total of six plots per site. Be-
growth forest were selected along the Nah- cause previous studies indicate that both flo-
mint {n = 2; 2 km apart), Taylor {n = 1), and ristics and structural attributes play roles in
Klanawa rivers {n = 1). The sites were em- avian habitat selection (MacArthur and Mac-
Shirley • HABITAT USE BY RIPARIAN AND UPLAND BIRDS 247
Arthur 1961, Holmes and Robinson 1981, 151-200 m) and calculated relative abundance
Robinson and Holmes 1984), I focused on 10 for each distance category as the abundance
variables representing broad measures ofveg- averaged over 3 years and four sites.
etation characteristics: density of deciduous I analyzed bird abundances by selected
and coniferous trees, snag density, volume of guilds and by individual species. I focused the
coarse woody debris (CWD), species richness, guild analysis on two guilds that I predicted
total percent coverofall shrubs and forbs, and would show a gradient in abundance from ri-
percent cover of the two dominant shrubs parian to upland: riparian specialists and aerial
(salmonberry and blueberry). Density (num- foragers. I assigned species to guilds after
ber/ha) was recorded for coniferous trees, de- Hatler et al. (1978), Ehrlich et al. (1988), and
ciduous trees, and snags within the entire 20- Campbell et al. (1990, 1997) (Appendix). For
m-radius plot. Trees <3 m in height and ferns individual species, I restricted my analysis to
were treated as shrubs. Richness and percent those with >5 observations in each year (nine
cover ofshrubs was sampled in a 10-m-radius species). When estimating species richness by
subplot nested within 20-m circular plots. site, I minimized the impacts of transient spe-
Richness and percent cover offorbs was sam- cies by excluding species that were likely mi-
pled using four l-m^ quadrats placed at the grants and species observed in only one cen-
center of shrub plots. CWD, defined as fallen sus session during each year (Willson and
logs >10 cm in diameter, was sampled at the Comet 1996).
—
point of intersection along the circumference Habitat associations. I used Akaike’s In-
of 20-m plots; I recorded diameter and length formation Criterion corrected for small sam-
to calculate volume (m^/ha) of CWD (Van ples (AICJ to select suitable models of asso-
Wagner 1968, Tho—mas et al. 1979). ciation between vegetation variables and avian
Bird sampling. Details of bird sampling abundance (Burnham and Anderson 2002). I
can be found in Shirley (2002). Briefly, birds used multiple regression and estimated the re-
were censused using a full-plot, area-search siduals to model species richness and abun-
method (Slater 1994). A 9-ha grid was estab- dance of avian guilds and individual species
lished at each site by running a 450-m line as a function of vegetation variables. Models
parallel and adjacent to the river’s edge and were based on a priori hypotheses of those
nine perpendicular lines extending 200 m vegetation variables that may be associated
from the river. Grid lines were set 50 m apart with a guild’s or species’ abundance. For each
and flagged at 25-m intervals. Censusing was model, I computed AIC^ and AAIC,.. Model
conducted at each site by at least two observ- likelihoods were standardized to sum to and
ers who walked the grid lines from 05:00 to expressed as Akaike weights (cd). The Ak1aike
10:00 (PST) on days without rain or high weight can be considered as the weight ofev-
winds. We censused birds at each site four idence supporting a given model as the best
times each breeding season from 1 May to 15 model; the higher the Akaike weight, the
July so that each site was censused once ap- stronger the model. To identify plausible mod-
proximately every 2 weeks. Birds of prey and els for each species or guild, 1 ranked the
I flyovers were not included in the censuses. I Akaike weights of models in a given set to
varied the order in which sites were sampled, produce evidence ratios (i.e., the weight ofthe
and three to four observers rotated among best model divided by that of a given model;
Ij sviotceaslanadndgrviidsulailneso.bTseorvaavtoiiodndsouwbelree-croeucnotridnegd, tBiuosrnehxapmresasndthAendelirkseolniho2o0d02)o.f Etvhiedesnecleectread-
I doinlastietenmuampbserthaantdwererleatilavteerabeuvnalduaantceedotfo bciarld- modDealtarelaantailvyesitso.—othVeergemtoadteilosn. attribute and
I species with respect to distance from the river. avian abundance data were tested for normal-
The numbers of observations over the four ity using the Shapiro-Wilks statistic (Shapiro
censuses in each year were averaged to pro- and Wilk 1965) before conducting paired t-
vide a mean number of species and individu- tests and ANOVAs. Homogeneity of varianc-
als per species for each site. 1 categorized ob- es for one-way ANOVA and repeated-mea-
servations into four distance categories from sures ANOVA were tested using the Levene
the river’s edge (0-50, 51-100, 101-150, and and Bartlett-Box /• tests, respectively (Norusis
248 THE WILSON BULLETIN • Vol. 117, No. 3, September2005
1994). Species abundance data that violated years. I compared species richness across the
these assumptions were either log (y + 1) or four 50-m intervals from the river’s edge and
rank transformed (Conover and Iman 1982). over time using two-way, repeated-measures
The a level of significance was set at 0.10 to ANOVA with year and distance from edge as
minimize the high biodiversity cost ofmaking fixed factors {n = 16).
a type II error in resource management deci-
RESULTS
sions (Toft and Shea 1983, Dayton 1998). I —
also define a level of “marginal significance” Vegetation. Four of 10 measu—res of veg-
as 0.15 > P > 0.10. All data were analyzed etation structure and composition density of
using SPSS for Windows 6.1.4 (Norusis coniferous and deciduous trees,—shrub-species
1994). richness, and blueberry cover differed be-
To compare riparian and upland means for tween riparian and upland habitats (Table 1).
each of the 10 vegetation variables, I used Effect sizes were medium for coniferous tree
paired Mests because riparian and upland hab- density and large for deciduous tree density,
itats were paired by site for each of the four reflecting substantial biological differences.
sites. Rather than correcting for multiple tests Riparian habitats had nearly five times the
using the standard Bonferroni method, which density of deciduous trees compared with up-
has several disadvantages when gauging the land areas, while upland areas had greater co-
effects of variables in ecological research nifer density and percent blueberry cover.
(Nakagawa 2004), I present effect sizes as rec- Snag density and salmonberry cover were
ommended by Hurlbert (1994). I evaluated the greater in upland and riparian areas, respec-
biological significance of the results using es- tively; effect sizes were large, but these dif-
tablished criteria (Cohen 1988) where a small ferences were only marginally significant due
effect size = 0.2, medium = 0.5, and large to a small sample size. CWD and forb cover
0 8 were not statistically different between habi-
. .
To compare abundances (all species com- tats, but effect sizes were medium and could
bined, two guilds, and nine individual species) indicate biological significance.
—
by distance category from the river and Avian abundance and diversity. During
among years, I used a two-way, repeated-mea- 1996-1998, I recorded 645 observations of36
sures ANOVA, with year and distance from species. For all sites combined, there were
river’s edge as fixed effects. Because the same >20 observations for 9 species, accounting
sites were censused over 3 years, I treated for 80% of all observations. The five most
year as a repeated variable in a model that abundant species were Winter Wren {Troglo-
specified polynomial contrasts to detect linear dytes troglodytes). Chestnut-backed Chicka-
or quadratic trends over time (Gurevitch and dee (Poecile rufescens), American Robin,
Chester 1986, von Ende 1993). Because there Swainson’s Thrush {Catharus ustulatus), and
were no significant year effects, data were Pacific-slope Flycatcher {Empidonax dijficil-
pooled across years forall comparisons except is). For all years and in each distance interval,
for American Robin (Turdus migratorius), assemblages were dominated by these five
which showed a significant distance-by-year species, wh—ich composed 53—58% oftotal ob-
interaction. I then tested for differences in servations with only minor variations in
pooled abundance (one-way ANOVA; all spe- their abundance rankings. Except for one for-
cies combined, two guilds, and eight species) est interior species (Pacific-slope Flycatcher),
acros—s four 50-m intervals from the river’s these species are forest generalists. Winter
edge with distance from edge as a fixed fac- Wren and Chestnut-backed Chickadee were
tor {n = 48). For American Robin, I per- the dominant species in upland sections and
formed the above analysis for each year sep- were replaced, in part, by American Robin
arately; however, results are presented for all and Swainson’s Thrush near the river. Ripar-
years together (Fig 1). Effect sizes forthe one- ian specialists generally occurred close to the
way ANOVAs were calculated using the Eta river and at low abundances, with the excep-
squared method (Levine and Hullett 2002). tion of Hammond’s Flycatcher {Empidonax
Because there was a significant year effect hammondii).
for species richness, I did not pool data across Because the dominant species were forest
Shirley • HABITAT USE BY RIPARIAN AND UPLAND BIRDS 249
Allspeciescombined
WinterWren Chestnut-backedChickadee AmericanRobin
abundance
Swainson'sThrush Pacific-slope Flycatcher Golden-crownedKinglet
Relative
VariedThrush Hammond'sFlycatcher HairyWoodpecker
3.5 3.5
3.0 3.0
2.5 2.5
2.0 2.0
1.5- 1.5-
1.0 1.0- n
0.5- 0.5- jn.
0.0. 00 [i.
0-50 51-100 101-150 151-200 0-50 51-100 101-150 151-200
Distance from river (m)
FIG. 1. Relative abundance (mean number ofob.servationsAsite/year) and .standard deviations (error bars) at
5()-m intervals from the river’s edge for (A) all species combined. (B) riparian specialists, (C) aerial foragers,
(D) Winter Wren, (E) Che.stnut-backed Chickadee, (F) American Robin, (G) Swainson's Thrush, (II) I’acitic-
slope Flycatcher, (I) Golden-crowned Kinglet, (I) Varied Thrush. (K) Hammond’s Flycatcher, and (L) Hairy
Woodpecker, 1996-1998, western Vancouver Island, British Columbia. Canada (;/ 48).
250 THE WILSON BULLETIN • Vol. 117, No. 3, September2005
TABLE 1. Vegetation characteristics of riparian and upland habitats, western Vancouver Island, British
Columbia, 1995. Of 10 variables, four (deciduous and coniferous tree densities, shrub species richness, and
percent cover of blueberry species) differed {P < 0.10) between riparian and upland habitats (n = 12 for all
tests).
Variable Riparianmean(SD) Uplandmean(SD) PairedMest p Effectsize®
Coniferous trees (no./ha) 247 (166) 356 (140) -1.83 0.095*’ -0.45
Deciduous trees (no./ha) 288 (102) 59 (37) 4.06 0.002*’ 0.79
Snags (no./ha) 72 (41) 134 (96) -1.77 0.10‘’ -0.88
Shrub cover (%) 27 (6) 25 (6) 0.80 0.44 0.06
Shrub richness (no. of species) 6 (2) 5 (2) 2.45 0.092*’ 0.17
Salmonberry cover (%) 8 (7) 3 (3) 1.65 0.13-’ 0.56
Blueberry cover (%) 2 (3) 5 (5) -2.37 0.098*’ -1.53
Forb cover (%) 44 (24) 57 (17) -0.80 0.48 -0.38
Forb richness (no. of species) 11 (3) 9 (4) 1.51 0.23 0.06
CWD (m3/ha) 2,512 (208) 3,549 (1,512) -0.94 0.37 -0.41
^^EBfofledc-tfasiczeed,vmaelauseusrdeednoaste(Ws]ign-ifimca2n)clem\a,twPhe<re0.W1]0.= meaninriparianhabitatandm2 = meaninuplandhabitat(afterHurlbert 1994).
‘^Marginallysignificant(0.15 >P>0.10).
generalists, total abundance did not differwith Species richness did not differwith distance
distance from the river (F344 = 1.26, P = from the river (F3 ,2 = 0.15, P = 0.93) and
0.30; Fig. lA). As expected, abundances of there was no significant interaction between
riparian specialists and aerial foragers de- distance and year (Fe,24 - 0.61, P - 0.62).
clined with distance from the river (Fig. IB, Species richness, however, differed among
C; riparian specialists: F344 = 7.98, P < years (F224 = 6.28, P — 0.028); 14% more
0.001; aerial foragers: F344 = 5.23, P = species were detected in 1997 than in 1996
0.027). Of the nine species for which I had and 1998 (Fig. 2). —
sufficient data for analysis, only two varied Habitat associations. For most bird spe-
significantly in abundance across the distance cies, associations with specific vegetation at-
intervals: Swainson’s Thrush (F344 = 2.85, P tributes were stronger than with distance from
= 0.10; Fig. IG) and Hammond’s Flycatcher the river. The best model for aerial forager
(F344 = 11.74, P = 0.001; Fig. IK) were more abundance was one showing a negative rela-
common near rivers (all other species: P ^ tionship with conifer density (Table 2). This
0.31, effect sizes < 0.07). model was strongly supported (co, = 0.850),
0-50 51-100 101-150 151-200
Distance from river (m)
FIG. 2. Means and standard deviations (error bars) of avian species richness at 50-m intervals from the
river’s edge, 1996-1998, western Vancouver Island, British Columbia, Canada.
Shirley • HABITAT USE BY RIPARIAN AND UPLAND BIRDS 251
TABLE 2. Habitat model selection using Akaike’s Information Criterion for abundance ofavian guilds and
species in undisturbed forest sites, western Vancouver Island, British Columbia, 1996-1998. Selected models
with the highest likelihood are shown. Secondary models are included if they were nearly equal to the first
model.
Guildorspecies Model^ AICc‘’ AAICe'= Kd CO^
Aerial foragers CT -5.089 0.000 3 0.850
Riparian specialists DI 27.940 0.000 3 0.244
SA 28.221 0.281 3 0.212
SH 28.444 0.504 3 0.190
SR 28.911 0.971 3 0.150
American Robin CT 27.795 0.000 3 0.405
Chestnut-backed Chickadee CT 25.831 0.000 3 0.387
SN 25.885 0.054 3 0.376
Golden-crowned Kinglet CT 16.730 0.000 3 0.334
Hammond’s Flycatcher DI 20.110 0.000 3 0.396
SH 20.165 0.554 3 0.385
Hairy Woodpecker CT 6.963 0.000 3 0.533
Pacific-slope Flycatcher CT 22.379 0.000 3 0.545
Swainson’s Thrush VA 21.764 0.000 3 0.266
DT 21.932 0.168 3 0.244
CT 22.065 0.301 3 0.229
Varied Thrush SN 9.212 0.000 3 0.457
Winter Wren CT 28.025 0.000 3 0.311
cov^erC,TS-Nd=enssniatygodefncsiotnyi,feSroRus=trsehers,ubDIspe=cideisstraicnhcneesfsr,omVArive=r,pDerTcen=tdbelnuseibteyrroyfsdpepc.idcuovoeurs.trees,SA =percentsalmonberrycover,SH=percentshrub
^Akaike’sInformationCriterioncorrectedforsmallsamplesizes.
AAICf= differencebetweenbestmodelandmodelwithminimumAIC,..
^Numberofparameters.
®0), = Akaikeweight.
being 14 times more likely than the second- dance was best predicted by two single-vari-
best model in the set. Abundance of riparian able models showing negative relationships
specialists was predicted equally by four sin- with conifer (o\ = 0.387) and snag densities
gle-variable models showing positive relation- (to, = 0.376); both models had almost equal
ships with salmonberry, percent shrub cover, support and were 3 times more likely than the
and shrub species richness, as well as a neg- third model in the set. Golden-crowned King-
ative relationship with distance from the river. let (Regulus satrapa) abundance was best pre-
Support for all four models, however, was dicted by a model showing a positive rela-
weak (all four oj, < 0.250) and the best model tionship with conifer density; however, this
was only 2.5 times more likely than the next- model was relatively weak (to, = 0.334) and
best model in the set. only 1.3 times more likely than the second
The abundances of four species (American model in the set. Hammond’s Flycatcher
Robin, Hairy Woodpecker \Picoides villosus]. abundance was best predicted by two single-
Pacific-slope Flycatcher, and Winter Wren) variable models representing a negative rela-
were best predicted by single-variable models tionship with distance from the river (to, =
showing negative relationships with conifer 0.396) and a positive relationship with percent
density (Table 2). Models for Hairy Wood- shrub cover (to, = 0.385). Both models had
pecker and Pacific-slope Flycatcher had mod- almost equal support and were 5 times as like-
erate support (to, = 0.533 and oj, = 0.545), ly as the third mt)tlel in the set. Swainson's
being 3-4 times more likely than the second- Thrush abundance was best predicted by three
best models in the sets. The models for Amer- single-variable mtHlels showing negative re-
piocratn R(too,bi=n a0n.d405WinatnedrtWo,re=n0h.a3d1 1w)eaankteirwseurpe- 0la.t2i6t6))nshainpds wdietnhsipteyrcoefntctb)lnuiefbeerroursy ctroeveesr((ttoo,, ==
only twice as likely as the sect)ntl nuKlcl in 0.244) anti a pt)sitive relatit)uship with density
the .set. Chestnut-backed Chickatlee abun- t)f tlecitlut)us trees (to, = 0.229). The models
252 THE WILSON BULLETIN • Vol. 117, No. 3, September 2005
had almost equal support, although support for 2001). McGarigal and McComb (1992) attri-
any one was quite weak. The best model for buted lower bird species richness in riparian
Varied Thrush {Ixoreus naevius) abundance as opposed to upland areas to the scarcity of
showed a positive relationship with snag den- conifers found along streams; however, in my
sity (o), = 0.457). This model had moderate study, conifers were not as scarce along ri-
support, being 4 times as likely as the second parian areas, perhaps accounting for the sim-
model in the set. ilarity in species richness between the two
habitats. The lack of a strong gradient in veg-
DISCUSSION
etation structure from riparian to upland is
Species diversity and abundance along the also reflected in the distribution of the most
—
riparian gradient. Contrary to my original common bird species. Abundances ofeight of
predictions, species abundance and diversity the most common bird species, as well as
ofbirds were similar along a distance gradient abundance of the aerial foraging guild, were
away from the river. Although species rich- associated most closely with densities of cer-
ness varied among years, total abundance re- tain canopy and understory species ratherthan
mained similar during the study. Other studies distance from the river. Complex topography,
in coniferous forests of the Pacific Northwest combined with consistently moist conditions,
have also found that riparian areas do not sup- provides suitable habitat for most species
port higher numbers of bird species or indi- across the riparian-upland gradient that I stud-
viduals (McGarigal and McComb 1992, Mur- ied, and it probably accounts for the lack of
ray and Stauffer 1995, Pearson and Manuwal strong riparian effects at the community level.
2001). In contrast, studies in more arid or ag- The large fourth- or fifth-order streams and
ricultural environments (Carothers et al. 1974, rivers in my study area contrast with the
Stevens et al. 1977, Wauer 1977, Stauffer and smaller, second-order streams that were thefo-
Best 1980) found large differences in diversity cus of some previous studies in northern for-
and abundance between riparian and upland ests (McGarigal and McComb 1992, Wiebe
habitats. McGarigal and McComb (1992) pro- and Martin 1998). In those studies, riparian
posed three hypotheses to account for the re- forests supported equal or fewer species and
gional difference in these patterns: (1) high individuals compared with surrounding up-
stream density and availability ofwater in up- lands (McGarigal and McComb 1992, Wiebe
land areas in the Pacific Northwest, (2) a less and Martin 1998). Studies of larger-order
pronounced microclimatic gradient (moderat- streams, however, have indicated that they
ed by maritime influences) in northwestern support denser, more complex riparian vege-
forests, and (3) a more subtle transriparian tation communities and greater avian density,
gradient in vegetation structure. Higher rain- species richness, and abundance (Knopf 1985,
fall and less variation in annual temperatures Lock and Naiman 1998). Lock and Naiman
on western Vancouver Island compared with (1998) found greater species richness and
Washington and Oregon may produce an even abundance along larger rivers where the ri-
less pronounced transriparian gradient. parian habitat contained a higher ratio of de-
In this study, riparian habitats had greater ciduous to coniferous vegetation; in my study,
densities ofdeciduous trees, and the understo- however, avian species richness and abun-
ry was dominated by salmonberry. In contrast, dance were similar across the riparian to up-
upland areas had higher densities of conifer- land gradient, even along larger streams. Most
ous trees, and blueberry species dominatedthe species used both riparian and uplandhabitats,
shrub understory; uplands also tended to have whereas only a few species specialized in ei-
greater snag densities. While low statistical ther habitat. In northwestern forests, these
power may have limited my ability to detect specialists represented a small fraction of the
statistically significant differences in some at- overall community. —
tributes, my results are consistent with those Habitat selection. Of the 36 species re-
of other studies that evaluated vegetation corded, five occurred only ne—ar the river. Four
structure across the transriparian gradient of of these riparian specialists Common Mer-
forests in the Pacific Northwest (McGarigal ganser {Mergus merganser), American Dipper
and McComb 1992, Pearson and Manuwal (Cinclus mexicanus). Belted Kingfisher {Cer-
Shirley • HABITAT USE BY RIPARIAN AND UPLAND BIRDS 253
yle alcyon),—and Spotted Sandpiper (Actitis aspen-coniferhabitat. Behavioral observations
macularius) depend on stream invertebrates by Sakai and Noon (1991) suggested that
and/or fish as food resources and they nest in some competition likely occurs between the
adjacent riparian vegetation or riverbanks Hammond’s and Pacific-slope flycatchers, but
(Enns et al. 1993, Campbell et al. 1997). The it does not result in competitive exclusion of
remaining species, Willow Flycatcher (Empi- one species by the other. Pacific-slope Fly-
donax traillii), rarely occurs in mature forest catcher abundance differed little along the
except in riparian areas. In the coastal western gradient ofdistance away from the river, a re-
hemlock zone. Willow Flycatchers more com- flection of its association with large conifer
monly occur in marshes and early succession- trees in both riparian and upland habitats. In-
al clearcuts (5-10 years of age) associated terestingly, studies farther south in the Pacific
with young red alder and willow trees (Enns Northwest (McGarigal and McComb 1992,
et al. 1993, Campbell et al. 1997). Pearson and Manuwal 2001) reported that Pa-
Five species occurred only at single sites in cific-slope Flycatchers are associated with ri-
upland sections of forest: Fox Sparrow {Pas- parian habitats, whereas Hammond Flycatch-
serella iliaca), Hutton’s Vireo (Vireo huttoni), ers are associated with upland habitats. The
Olive-sided Flycatcher (Contopus cooperi). reason for this difference is unclear, but may
Spotted Towhee (Pipilo maculatus), and Yel- relate to differences in species composition
low Warbler {Dendroica petechia). All of and size distribution of trees in the two forest
these species are rare in mature forests and habitats. For example, in the previous studies,
may select large patches of open, deciduous large-diameter conifers required by Ham-
vegetation in forest interiors. mond’s Flycatchers (Sakai and Noon 1991),
As predicted, I found that aerial foragers such as Douglas-fir (Pseudotsuga menziesii),
declined in abundance with increasing dis- occur in greater numbers in upland forests.
tance from the river. Aerial foragers include Several of the dominant species in my
Hammond’s Flycatcher, a species that occu- study, including Chestnut-backed Chickadee
pies a wide variety of habitats (Willson and and Golden-crowned Kinglet, showed associ-
Comet 1996). Throughout much ofthe Pacific ations with conifer or snag density. An excep-
Northwest, the species is an upslope specialist tion was the Swainson’s Thrush, which, al-
that is associated strongly with conifers at though a forest generalist, was more abundant
sites characterized by relatively open canopies in riparian habitat. This species is widespread
(Sakai and Noon 1991, McGarigal and Mc- on the west coast (Campbell et al. 1997) and
Comb 1992); however, farther north in Alaska often forages in salmonberry and devil’s club
it favors deciduous stands (Willson and Comet in riparian habitats. The positive association
1996), and in the forests of Vancouver Island with deciduous trees suggests that these struc-
(Waterhouse and Harestad 1999, Shirley tures or some other closely associated vege-
2002) and southeastern British Columbia tation may be an important influence on hab-
(Kinley and Newhouse 1997), this species is itat selection for this species. Furthermore,
largely restricted to mixed riparian forests that virtually all Swainson’s Thrush nests encoun-
include large deciduous trees and conifers. tered incidentally during surveys (// = 8) were
Whereas Hammond’s Flycatcher may use ri- found in salmonberry (SMS pers. obs.), sug-
parian habitat, it is sympatric with the Pacific- gesting a strong association with this shrub for
slope Flycatcher in old-growth forest (Camp- nesting habitat and/or food.
—
bell et al. 1997) and its distribution may re- Management implications. Patterns ofavi-
flect some habitat partitioning between the an diversity and abundance in riparian com-
two species. In southern Colorado, where munities often have been explained by dra-
Hammond’s and Cordilleran {Empidonax oc- matic gradients in microclimate and vegeta-
cidentalis) flycatchers co-occur, Hammond’s tion structure or composition (Carothers ct al.
Flycatcher densities were approximately one- 1974, Stevens et al. 1977, Dickson 1978, Sza-
half those of the Cordilleran Flycatcher (Bea- ro 1980). Where these gradients arc subtle, as
ver and Baldwin 1975). In these areas ofover- in forests ol the Pacific Northwest, the pat-
lap, Hammond’s Flycatcher inhabited aspen terns disappear (Wiebe and Martin 1998, Pear-
habitat, while the Cordilleran flycatcher used son and Manuwal 2001) or they may re-
254 THE WILSON BULLETIN • Vol. 117, No. 3, September2005
—
verse upland areas supporting greater diver- migration and other seasons (Harris 1984,
sity and abundance than riparian areas Wiebe and Martin 1998). For example, ripar-
(McGarigal and McComb 1992). Northwest ian habitats may provide critical habitat for
forests generally lack a strong microclimatic Neotropical migrants as they travel between
gradient from riparian to upland (Brosofske et their wintering and breeding grounds (Stevens
al. 1997), and ephemeral streams or ponds oc- et al. 1977, Finch 1991), and riparian habitat
cur in virtually every upland area. These fac- may be important for the survival and popu-
tors create a fine-scale habitat mosaic in which lation stability ofmigratory species during the
patches dominated by conifers are inter- breeding season. The diversity and density of
spersed with deciduous trees and shrubs that some migrants may be greater in riparian cor-
provide habitat for species more typical ofde- ridors because they are easy to follow and/or
ciduous-tree-dominated riparian areas. provide diverse foraging habitats (Wiens
Recent discussions of land management 1989, Wiebe and Martin 1998).
practices to preserve native biodiversity of
ACKNOWLEDGMENTS
forest species include using a landscape-level
approach that protects both riparian and up- I thank K. G. Beal, D. J. Huggard, L. G. Barrett-
land habitats to ensure connectivity across the Lennard, and two anonymous reviewers forcomments
landscape (McGarigal and McComb 1992, on earlier drafts ofthis manuscript, and K. T. Port, S.
Wiebe and Martin 1998). Maintaining connec- Frioud, S. L. Hicks, D. Lewis, R. Maraj, G. Matscha,
tivity may prevent isolation of remnant forest IF.aPlosouwt,haannkdGS..MWaetbsecrhafoarntdheCir.hMa.rdFwerogruksoinntfhoerftiheelid.r
patches (Fahrig and Merriam 1985, Saunders help in processing the vegetation data. I amgratefulto
and de Rebeira 1991, Gonzalez 2000); how- W. French and the engineering group atWeyerhaeuser
ever, the lack ofupland specialists in my study Canada fortheircooperation in locating field sites. Fi-
argues against placing too much emphasis on nancial support was provided by the Habitat Conser-
upland areasperse. Unmanagedriparian areas vation Trust Fund, Forest Renewal British Columbia,
not only provide habitat for those few species a University of British Columbia graduate fellowship
to SMS, and a Natural Sciences and Engineering Re-
that associate with specific features at the riv- search Council operating grant to J. N. M. Smith.
er’s edge, they also contain habitat elements
such as large conifers and snags important to LITERATURE CITED
many common forest species.
My study was limited in several ways that AnthSo.nKy.,NRe.lsGo.,n.G.19A9.6.GArveieann,aEb.unDd.anFcoersimnarni,paarinadn
should be considered in the development of zones ofthree forest types in the Cascade Moun-
land use plans. First, although the contrast be- tains, Oregon. Wilson Bulletin 108:280-291.
tween riparian and upland habitats is known Beaver, D. L. and R H. Baldwin. 1975. Ecological
overlap and the problem ofcompetition and sym-
to be subtle in the moist forests of the Pacific
Northwest (McGarigal and McComb 1992, Cpaotnrdyoirn7t7h:e1-W1e3s.tern and Hammond’s flycatchers.
Pearson and Manuwal 2001), differences in Brinson, M. M., B. L. Swift, R. C. Plantico, and J.
structure and composition ofvegetation in my S. Barclay. 1981. Riparian ecosystems: their
study may have been obscured by the small ecology and status. U.S. Department of the Inte-
sample size and resulting lack of statistical rior, Fish and Wildlife Service, Kearneysville,
West Virginia.
power. Statistically, several of the vegetation Brosofske, K. D., J. Chen, R. J. Naiman, and J. F.
characteristics that I measured were margin- Franklin. 1997. Harvesting effects on microcli-
ally or non-significant, but their medium to matic gradients from small streams to uplands in
large effect sizes indicate possible biological western Washington. Ecological Applications 7:
significance. Second, comparisons in this 1188-1200.
study were limited to measures ofavian abun- Bryan19t9,3.A.AvAi.a,nJ.coP.mmSuanviatride,sainndolRd.-gT.roMwctLhaaungdhlmiann.-
dance. Future work should focus on measures aged forests ofwestern Vancouver Island, British
of relative fitness or productivity in the two Columbia. Technical Report Series, no. 167. Ca-
habitats. Third, my study was conducted dur- nadianWildlifeService, Nanaimo, BritishColum-
ing the breeding season; work is also needed bia, Canada.
Bull, E. V. 1978. Specialized habitat requirements of
to assess avian distributions during other sea- birds: snag management, old growth, andriparian
sons, as the relative value of riparian and up- habitat. Pages 74-82 in Proceedings ofthe work-
land habitats may differ between periods of shop on nongame bird habitat management in the