Tropical Rainforest Bird Community Structure in Relation To Altitude, Tree Species Composition, and Null Models in the Western Ghats, India

2005·7.2 MB·

by  T R Shankar Raman

Journal of the Bombay Natural History Society, 102 (2), May-Aug 2005 145-157
TROPICAL RAINFOREST BIRD COMMUNITY STRUCTURE IN RELATION
TO ALTITUDE, TREE SPECIES COMPOSITION, AND NULL MODELS IN THE
WESTERN GHATS, INDIA
1
T. R. ShankarRaman2-3,N. V.Joshi2-4andR. Sukumar2-5
'Accepted 2005
^Centre for Ecological Sciences, Indian Institute of Science, Bangalore 560 012, Karnataka. India.
’Email: [email protected]
4Email: [email protected]
’Email: [email protected]
Studiesofspeciesdistributiononelevationalgradientsareessentialtounderstandprinciplesofcommunityorganisation
as well as to conserve species in montane regions. This study examined the patterns ofspecies richness, abundance,
composition,rangesizes,anddistributionofrainforestbirdsat 14sitesalonganelevationalgradient(500-1400m)inthe
Kalakad-Mundanthurai Tiger Reserve (KMTR) ofthe Western Ghats, India. In contrast to theoretical possibilities,
resident bird species richness did not change significantly with elevation although the species composition changed
substantially (<10% similarity) between the lowest and highest elevation sites. Constancy in species richness was
possibly due to relative constancy in productivity and lack ofelevational trends in vegetation structure. Elevational
rangesizeofbirds,expected toincrease withelevation accordingtoRapoport’s rule, was foundtoshow acontrasting
inverseU-shapedpattern because species with narrowelevational distributions,includingendemics,occurredatboth
endsofthegradient(below 800inandabove 1200m). Birdspeciescomposition alsodidnotvaryrandomlyalongthe
gradientasassessedusingahierarchyofnullmodelsofcommunityassembly,fromcompletelyunconstrainedmodelsto
oneswithspeciesrichnessandrange-sizedistributionrestrictions.Instead,birdcommunitycompositionwassignificantly
correlatedwithelevationandtreespeciescompositionofsites,indicatingthe influenceofdeterministicfactorsonbird
community structure. Conservation of low- and high-elevation areas and maintenance of tree species composition
against habitat alteration are important forbirdconservation in the southern Western Ghats rainforests.
Keywords: structureandfloristics,mid-domaineffect,non-equilibriumdynamics,null models,communityassembly
INTRODUCTION apparentwhendataarecorrectedforsamplingeffortforlocal
species richness (Terborgh 1977; Patterson et al. 1996), or
SincethepioneeringworkofTerborgh(1971, 1977, 1985) area for regional species richness (Rahbek 1997; Bachman
and Terborgh and Weske (1975), research on altitudinal etal. 2004). Modelsthatevaluateconsequencesofgeometric
distribution patterns of birds has tended to focus on three constraints (hard upper and lower limits) on the altitudinal
—
major aspects variation in richness with altitude, range ranges of species also predict mid-altitude peaks called the
—
sizes, and species turnover rates and on one region with ‘mid-domain effect’ (Rahbek 1997; Colwell andLees 2000;
relatively well-documented information on bird altitudinal Veech2000).
distributions, the Neotropics (Rahbek 1997; Patterson etal. In contrast to species richness, community
1998;Stotz 1998;BlakeandLoiselle2000).Moststudieshave compositional change (turnover rates) at the regional level
shown ageneral pattern ofdecreasing species richness with mayincreasewithaltitude(Rahbek 1997),orshowpeaksand
increasing altitude, believed to mirror the well-recognised troughs corresponding to transition zones between lowland
latitudinal gradient in species richness (MacArthur 1972; andmontaneavifauna(Pattersonetal. 1996;Stotz 1998).Little
Pattersonetal. 1996, 1998; Bachmanetal. 2004).Monotonic isknownoffactorsinfluencingbirdcommunitycomposition
declineinspeciesrichnesswithaltitudemayalsooccurwhen at local levels in tropical rainforests. Along successional
altitudinal range sizes of species increase with altitude, as gradients, habitat structure and tree species composition
suggested by Stevens (1992) in an extension ofRapoport’s influencebirdcommunitystructure,wherein structurallyand
rule,amatterofintensedebate(Rapoport 1982;Gastonetal. floristicallysimilarsiteshave moresimilarbirdcommunities
1998;TaylorandGaines 1999). (Ramanetal. 1998).Alongaltitudinalgradients,paststudies
Other studies have suggested that species richness of local bird community structure have largely ignored the
may show a non-monotonic, hump-shaped pattern with relative influence ofdeterministic differences between sites
diversitypeakingatmid-altitudes(Janzen 1973; Rahbek 1997; in habitatattributesversustheeffectsofchance, null models,
Stotz 1998), as in diversity-productivity gradients oraltitudeperse.
(RosenzweigandAbramsky 1993).Thispatternmaybecome The present study explores variation in local bird
TROPICAL RAINFOREST BIRD COMMUNITY STRUCTURE IN WESTERN GHATS
communitystructure,speciescomposition,andturnoverrates Although 278 bird species have been recorded in and
along an altitudinal gradient in a tropical rainforest of the aroundKMTR,only84speciesoccurinrainforestsincluding
Western Ghats of India. Although recognised as a global 12endemicsand 13wintermigrants(Raman2001).Amajority
biodiversity hotspot(Myersetal. 2000)andanendemicbird ofspeciesbreedbetweenlateJanuaryandMay(AliandRipley
area(Stattersfieldetal. 1998),thisregionispeculiarinbeing 1983, TRSR pers. obs). The sites chosen for intensive bird
relativelydepauperateinitsrainforestbirdcommunitydueto communitysamplingwerelocatedatKannikatti(740m,8°37'
historicalisolationfromlargertractsofrainforestinnortheast Nand77° 16'E),Sengaltheri(1040m,8°31’Nand77°26E),
IndiaandsoutheastAsia(Danielsetal. 1992).Here,weexplore andKakachi(1220m,8°33’Nand77°24’E)inKMTR.These
how birdcommunity attributes vary with altitude in tropical areascontainrainforestsbetween500mand 1400maltitude,
rainforest. Specifically, weask: doesspeciesrichnessshowa contiguouswithrainforestselsewhereintheReserve.Fourteen
monotonic decline with altitude or a hump-shaped pattern sites located along this altitudinal gradient in relatively
duetothe ‘mid-domaineffect’?IstheresupportforRapoport’s undisturbed rainforest were selected randomly, within
rule?Doesaltitudinaldistanceanddegreeofchangeinhabitat logistical constraints, around three forest camps (four
influence and correlate with change in bird community accessedfromKannikatti,sixfromSengaltheri,andfourfrom
composition or are turnover rates estimated accurately by Kakachi) and locatedon topographic maps using landmarks
null models that simulate non-equilibrium dynamics? The andaglobalpositioningsystemreceiver(GPS).
results are used to assess the relative influence of
deterministic versus non-equilibrium factors on tropical Birdsurveys
rainforestbirdcommunity structure. The fixed-radius point count method was used to
survey bird populations in each site in a relatively uniform
MATERIALAND METHODS andefficientmannerduringthemainbreedingseason(Vemer
1985;Ralphetal. 1995).Pointcountsurveys(5minduration)
Studyarea werecarriedoutduringthefirstthreehoursaftersunrise(see
The 1600kmlongchainofhillscalledtheWesternGhats Raman2003fordetailsoffieldtechnique).Densitiesestimated
runs along the west coast ofthe Indian peninsulafrom 8° N by the fixed radius approach were used as they were highly
to21°N.Moistforests,includingtropicalwetevergreenforest, correlated to variable-radius point count estimates across
arefoundlargelysouthof 16° N(Pascal 1988),andcontaina species(Raman2003).
higher diversity ofendemic plant and animal taxa (Ali and Eachsampledsiterepresentedanareaofaround 12-15ha
Ripley 1983;Daniels 1992;Kumaretal. 2004).Thefieldwork lying at the designated altitude and was surrounded by
wascarriedout inthe Kalakad-MundanthuraiTigerReserve contiguous rainforests on at least three sides. Within each
(KMTR,895sq.km,8°25'to8°53’N,77° 101to77°35'E)inthe sitea600-700mlongtransectornarrowanimaltrailwasmarked
Agasthyamalai region at the southern end of the Western at 25 m intervals for the point count surveys. At each site,
Ghats. KMTR ranges between 50 m and 1700 m above msl 25 point count surveys (yielding 167-247 detections and an
with rainforest occurring chiefly above 500 m. This reserve estimated334-597individualbirdspersite)werecarriedout,
along with adjoining areas has one ofthe largest remaining excludingtheKodayarsitewhereonly 18countswerecarried
contiguoustracts(over400sq. km)ofrelatively undisturbed out. Sampling of all sites around Sengaltheri and Kakachi
tropicalrainforestintheWesternGhats(Rameshetal. 1997). was carried out mostly between February and May 1998,
The rainforestvegetation in KMTRhas beenclassifiedmid- during the peak breeding season when resident and migrant
elevation tropical wet evergreen rainforest of the Cullenia birds were present. For reasons of logistics, the four sites
exarillata-Mesuaferrea-Palaquium ellipticum type (Pascal aroundKannikattiweresampledonlyinMarch 1999.
1988; Ganesh et al. 1996; Parthasarathy 2001). Within the
rainforest, the mean monthly daytime temperature ranges Vegetationsampling
between 19 °C in January and 24 °C in April-May (at In each ofthe 14 sites, densities oftrees greater than
Sengaltheri, 1040m, range 15-31 °C). The totalrainfall was 30 cm girth at breast height (GBH at 1.3 m) were estimated
mm mm
2283 in 1998and2230 in 1999atSengaltheri.There using the point-centred quartermethod (PCQ, Krebs 1989).
are three seasons: (a) dry season (February to May), ThirteenPCQplots,withsuccessiveplotsspaced50mapart,
(b) southwestorsummermonsoon(JunetoSeptember), and were measured in each site, giving a sample of52 trees per
(c)northeastorwintermonsoon(OctobertoJanuary). KMTR site.Comparisonwithtwocompletelyenumerated 1 haplots
receivesoverhalfitsannualprecipitationduringthenortheast (>1800 stems each) in Sengaltheri showed that the density
monsoon. estimates fromthisPCQsample was<2%differentfromthe
146 J. Bombay Nat. Hist. Soc., 102 (2), May-Aug 2005
TROPICAL RAINFOREST BIRD COMMUNITY STRUCTURE IN WESTERN GHATS
morelaboriousenumeration.Alltreeswereidentifiedtospecies, for other vegetation and site variables: shrub, cane, and
orinafewcasestogenus(GambleandFischer1915-1935;Pascal cardamomdensities,leaflitterdepth,canopyheightandcover,
andRamesh 1997). Distancefromplotcentretothemiddleof andaltitude.Verticalstratificationwasmeasuredastheaverage
the bole and GBH were recorded foreach tree. Ateach ofthe numberofstratawithfoliageacrossthe25 pointssampledin
13PCQplots,2mradiusplotswerelaidtoenumerateshrubs.In eachsite.Thecoefficientofvariationofverticalstratification
addition,thenumberofcaneplantsandthepresenceorabsence was used as an index ofhorizontal heterogeneity (following
ofbamboowithin5mradiuswasrecorded. Ramanetal. 1998).Thetotalnumberoftreespeciesrecorded
Altitude,canopymeasuresandleaflittervariableswere in the PCQ plots was recorded as a measure oftree species
measuredat25points,evenlyspaced25 mapart,ineachsite. richness. Relationships between bird community and
Canopy height was measured using a rangefinder and vegetation variables were assessed using Kendall rank
percentage canopy cover using a spherical densiometer. correlations(SiegelandCastellan 1988).
Vertical stratification was assessed at these 25 points by Altitudinalrangesizeofeachspecieswasestimatedas
noting the presence or absence of foliage in the following thedifferencebetweenthemaximumandminimumaltitudeat
heightintervals(inmetres):0-1, 1-2,2-4,4-8,8-16, 16-24,24- which the species was recorded in the point count surveys.
32,and>32,directlyaboveandina0.5 mradiusaroundeach Thealtitudinalrangemidpointandaweightedrangemidpoint,
point(Ramanetal. 1998).Leaflitterdepthontheforestfloor estimated using the abundance ofeach species in each site,
was measured using a calibrated wooden probe. As ground werecalculatedforeachspecies.Similaritiesinbirdcommunity,
vegetation andlitterweredisturbedalongtrails, the samples foliage profile, and tree species composition between sites
weretaken 10mawayfromtrailsintotherainforestinterior. werecomputedusingtheMorisitaindexthatisleastsensitive
to sample size effects (Wolda 1981; Ramanetal. 1998). The
Dataanalysis matrixofpair-wisedissimilarities(100-Morisitaindexin%)in
The cumulative list of bird species recorded in each bird community composition was related to corresponding
site was used as a basic measure of bird species richness. matrices of similarity/distance in tree species composition,
However, because only 18 pointcounts were sampled at the foliageprofile,altitude,andgeographicdistancebetweensites
Kannikatti site, we obtained standardised estimates of bird using Mantel tests (Manly 1994). Geographic distance
community parameters for all sites for 17 sampled points. between sites was measured as the straight-line distance
UsingtheprogramEstimates(Colwell 1997), 100permutations betweensamplingsites.StatisticalsignificanceofManteltests
(17sampleswithoutreplacement)wereperformedtoestimate wasassessedthrough 10000randompermutations.Toexamine
the following parameters and their standard deviations: the independenteffectsofdifferentvariables, partial Mantel
(1)birdspeciesrichness,measuredasthecumulativenumber tests, derived from the Kendall tau approach, were used
of species; (2) bird abundance, estimated as the number of (Hemelrijk 1990). The tree and bird community similarity
individualbirds/ha;(3)Shannon-Weinerdiversity,calculated matrices were also used to ordinate study sites using
as-Z(p lnp),wherep istheproportion ofthei* speciesand multidimensionalscaling(Manly 1994).
the summation is across all species in the pooled samples; We used correlation and regression techniques to
(4) rarefaction richness, standardising for sampling effort assessstatisticalsignificanceofaltitudinaltrends.Linearand
usingColemanrarefactioncurves(Colwell 1997);(5)bootstrap non-linear (chiefly quadratic) regressions were applied to
species richness, an incidence-based estimator denoted as describe spatial trend data. We also used non-parametric
Shoo,-Sobi+ 1 ~Pk)m,wherep istheproportionofmsamples Kendall (tau) rank-order correlation coefficients to assess
with specieskandthe summationis across all S species in statisticalsignificance(SiegelandCastellan 1988).
the pooled samples; and(6)Chao 1 richness, an abundance-
basedestimatorofspeciesrichness(seeColwell 1997forthe Nullmodelanalyses
formulaandcorrectvarianceestimatorofthisindex). The mid-domain effect (Colwell and Lees 2000) was
Smaller scale point richness and abundance estimates assessed using the null model of Veech (2000). Species’
werealsoobtainedby averagingacrossreplicatepointcount altitudinal range widths were retained but located randomly
surveys in each site. The average number of species, onthe altitudinalgradientbetween500and 1400m(divided
m
detections,andindividualsperpointandtheirstandarderrors into 18zonesof50 width)in 1000simulationstoobtainan
were estimated for all species combined and for resident expected species richness curve along this gradient. The
speciesalone. Foreach site, treedensity andbasal areawere observed pattern of species richness was compared to the
calculatedusingthePCQmethod(Krebs 1989).Averagevalues simulated curve to obtain a mean displacement, D, where
acrossreplicate samplingpointsineach site werecalculated D = (Zd)/18, where d is the absolute difference in species
J. Bombay Nat. Hist. Soc., 102 (2), May-Aug 2005 147
P
TROPICAL RAINFOREST BIRD COMMUNITY STRUCTURE IN WESTERN GHATS
richness between thetwocurvesbeingcomparedforeachof in distribution and occurrence probabilities. Our range-
the 18zones(Veech2000).Afurther 1000simulationsofthe contiguity model incorporates, in addition, constraints on
data were used to determine the distribution ofD values in species’ altitudinal distributions and contiguity of ranges.
each run and to assess statistical significance. We retained in the contiguity model the number of
Null models were also usedto assess the likelihoodof intermediateabsencesor ‘holes’ inrangebutrandomisedtheir
obtaining, under the null model assumptions, estimates location. Even with these constraints, we checked and
equivalenttotwomeasuresoftheaveragesimilaritybetween established that over 5000 different matrices existed in
sites in birdcommunity composition. These measures were: simulations.Foreachoftheabovemodels,wesimulated5000
(i) average similarity betweenall possiblepairs ofsites(N= randomised matrices and the averages of the obtained
91 comparisons), and (ii) average similarity between distributionofsimilarityvalues(betweenall possiblepairsof
-
successive sites along the altitudinal gradient (N 13 sites and adjacent sites) were statistically compared to the
comparisons). The Jaccard index (Krebs 1989) was used to observedaveragesimilarity values.
measuresimilarity inbirdcommunitycomposition.Withthe
presence-absence matrix of species across sites, five null RESULTS
modelswereevaluated:
1.Unconstrainedmodel: matrixtotalconstant,rowand Birdspeciesrichness,diversity,andabundance
column totalsvaried; Across the 14 sites, we obtained over2900 detections
2.Saturationmodel:fixedcolumntotals(speciesrichness comprisingapproximately6600individualsofbirdsbelonging
at each site); to67 species. Ofthese,ninespecieswerelatitudinal (winter)
3.Distributionmodel: fixedrowtotals(numberofsites migrants and the remaining were resident species. Thirteen
each species occurred in); species were detected only once and afurtherseven species
4. Doubly constrained model: fixed row and column were seen less than five times overall. Total bird species
totalsbyaprocedureequivalenttotheKnight'stouralgorithm richness (cumulative list in each site) was significantly
(Sanderson2000); negatively correlated to altitude of the site (Kendall rank-
5. Range-contiguity model: this model simulated, in ordercorrelation,x=-0.52,n- 14,P=0.01,Table 1).Resident
additiontothedistributionconstraint,thecontiguityofranges bird species richness appeared to decline with altitude, but
ofspeciesalongthegradient. Simulationsshowedthatspecies this trend was not significant statistically ( = 0.08). The
occurrences were distributed more contiguously along the standardisedestimatesfor 17pointcountsobtainedover 100
gradientthanexpectedbyrandomplacement(P<0.001).Using runs (sampling with replacement) of species richness,
the observed occurrences in the species-by-site matrix, the Shannon-Weiner diversity, Coleman rarefaction richness,
relative probability of occurrence in any one of the 12 bootstraprichness,andtheChaol richnesswereallsignificantly
intermediate sites (excluding the 2 ‘edge' sites-the lowest negatively correlated toaltitude (P<0.05,Table 1). Mostof
andhighestsite) wasfoundtobeenhancedby afactorof9.0 thedecreaseoccurredbetween500mand900maltitude,after
if the species also occurred in sites immediately above and which there was little change in species richness. Bird
below,andbyafactorof4.5ifitoccurredonlyinoneadjacent abundance showed a converse positive association with
site either above orbelow. These enhancement factors were altitude that was marginally significant (x = 0.39, n = 14,
inrelation totheprobability ofoccurrence inasitewhen the P = 0.055, Table 1). The per point estimates of total and
species was absent fromboth adjacent sites on the gradient. resident bird species richness, detections, and abundance
For the edge sites, occurrence was enhanced by a factor of showed significant variation across the 14 sites (one-way
3.6 ifthe species occurred in the adjacent site in relation to ANOVA, Fj > 5.58. P < 0.004). Whereas the number of
33,9
occurrence when it was absent in the adjacent site. These detectionsperpointshowedweaknegativecorrelationswith
were then used to distribute occurrences in simulations to altitude (x=-0.41,P=0.04), thetrendsfortotal andresident
produce probabilistic contiguous distributions. species richness and detections with altitude were non-
Thesemodelsattempttosimulatetheeffectsofchance significant.Birdabundanceperpointincreasedwithaltitude,
and biological constraints on the assembly of species weakly and non-significantly in the case oftotal abundance
communities(FarnsworthandEllison 1996;Stoneetal. 1996; (x=0.38,P-0.06)andsignificantlyinthecaseofabundance
Sanderson2000). Incontrasttothe unconstrainedmodel,the ofresidentspecies(x=0.52,P=0.009).
saturation, distribution, and doubly-constrained models
implement constraints on the numberofspecies thatcan fill Vegetationandbirds:correlations
each siteand simulate theeffectsofinterspecific differences Vegetation variables were not significantly correlated
148 J. Bombay Nat. Hist. Soc., 102 (2), May-Aug 2005
—
)
TROPICAL RAINFOREST BIRD COMMUNITY STRUCTURE IN WESTERN GHATS
q
oCO oo 00 00 co 00 LO
00 Q. oo qrC-D OqCM OLLOO qO1 00r- OCCOM qCCOO
2 o d d d d d d d
"(O q
.c5c N CcLvDOi LTq-O’ ccqvoi CkqM Cc»Mb—1 CccvDo1i cqv1i CcqvDi
o
o 0
0 CM 0r- LO CO 00 CD
0 •*-" dLCVOJ dCLOO dq dq00 dCCMO dLCOO dqCM dqCM
o
O
"0c0 z LcOd CCOM C0O CqCrM-D TkL—O CdLOO kqCO ccvoi d010 CkqO dco ccTb~o kqCM cdCoM Okq
£CD CT”O O
D((CCDDD qT— 0CM0 CLMO CkCMD dCO CcLMoO cdCoO cqcvoi 0dCO CCkMO qd LcdoO dco qcCMo c00v0i
0 CT“O 0
kQ) LO LO CkD qCM qLO qCLMO qchv-i 0dCD qcb qk CLMO qcCvDi ccob q0k0
_l CO CM CM CM CM CM CM CM
C”\J
T O
o CO q dh- k0q0 0q0 dq kCD kCLMO dq cqo q CqD
CLOO CO C0V0J CCMO -r- CM cvi CM co cvi 0C6M cb
C“\J
T O
LcOd cT-o CrM- CLkMO CckOo Cq1M0 q0k0 cccvooi 0dh- qCCMD qCkO CqkO CcqvMi CC0MO0 CkD
C\l
altitude 0 CD krq- q CCkOD (qj) dc0o q dCrM^ q cqko O dCqD
with L(TC”MOO cCMo CO CCkqMO Tck~o C•q'MT 0kLO cCCvDDi hd0- cTTo~- q0d0 cccvooi ^kq CCqkMD TCd-D
X CC\OJ CCMD CCMO CM CCMD cvi C0M6 CCMD CM
correlations C~M
T 0
o CoCvDJ C7—O CKM CC0MM0 0qk0 CqCMM ^CJ—D ccCvoMi L0dO C0CM0D dCcDo ckLoO kco CcLMoO qrC-D
1
rank-order o q qkCD kO q q dcqo k00 dCqD co LqO CM d(°O0
LL CCOD CCMD CLMO CLMO T- cvi CM CCMD cvi CCMD
Kendall o 0 qk LLkOO O qqk cqo 00d0 LO cqko 0d0 0qk0 q 0q
LU o CO CCMO CM CCMO cvi CCMO co C0M6 cvi
<d
q q q
showing q q qk CO qk00 C0M d0 q kq kq CkqD q 0q0
Q qCO co CCMD CT-O CcvMi cb ccvoi co ccvoi cvi
CO
(00ED O LT-O CCOO 0CO CqCOO OCkO CqkM k0q c0cob qdLO CCkOM Ctk-O cqko qLcvOi CqLOO cqko
CO r--
CQ>Ol» q CqO q 0q qco qq q 0q0 qCM q cqo
c3 CO 'cvof CCOO C0O CcOo k CkM k cb d CkO k CkO k ccob cvi
E co
E 0 q
poo < LoOo 0q C1O0 cocoob CkD CcCvMDi qd CcCMbM qd CqkO CkCDM qqk Tcv-j qqd qcb
if LO
0 C0O
n 0 0
ra sC00c0((/zDD L"0E0CEcU/D5 C0c(DD 0-J0c>30CD 00C>D 0c00cCC0:DD CD
hO0o0Cco0C*oD 1"00CcE03cOo O1Q00CE—§=D>_. F1-0£0C055D X0d0CC0CcQ00DDD. C0“X"C0C000CO05DD i0-0CCO0C0fCDDL CQO i0-030c0Cc0fQJ OCO Cs§0c0cc00zO CQO O00c0E0000 CQO COO0CQC0cpDOD_ CQO -O000cCC00DD CQO
J. Bombay Nat. Hist. Soc., 102 (2), May-Aug 2005 149
TROPICAL RAINFOREST BIRD COMMUNITY STRUCTURE IN WESTERN GHATS
with altitude (Kendall correlations, ft > 0.05), except for detections per point was negatively correlated to leaf litter
bambooculmdensity(T=0.52,n= 14,P=0.01)mainlydueto depth(x=-0.47,P=0.037).
the occurrence ofbamboo at only two higher altitude sites.
The six standardised estimates of bird species richness. Rangedistributions,sizes,andspeciesrichness
Shannon-Weiner bird diversity, and the per-point total and The range size and distribution pattern across the 58
residentbirdspeciesrichnesswereallsignificantlynegatively residentbird species variedfrom speciesoccurring inone to
correlatedtooneofthevegetationvariables: leaf-litterdepth all 14sitesspanningtheentiregradient.Twenty-threespecies
(x<-0.45,n= 14,P<0.03 inallcases).Thesevariableswere (39.7%) were either restricted to lower altitudes or showed
also not correlated to any ofthe other vegetation variables. decliningtrendswithaltitudeinpopulationdensity (Kendall
Bird abundance (standardised and per-point) and the total correlations, P < 0.10, Appendix). Sixteen species (27.6%)
number of bird detections per point were not correlated to displayed a trend of increase in population density with
anyvegetationvariable.Theaveragenumberofresidentbird altitude(P<0.10)orwererestrictedlargelytohigheraltitude
sites(Appendix).Onlytwospeciesofwoodpeckers,Dinopium
javanenseandDryocopusjavensis, appearedtoberestricted
to mid-altitudes. The spatial distributions of the remaining
specieswerewidespreadandrelatively uniform 14species)
(
orcouldnotbedeterminedduetolowsamplesizes(3species).
Acrossthegradient,thenumberofresidentbirdspecies
whoserangesinterceptedeach50maltitudezonewasnearly
constantataround35species(Fig. la).Thiswassignificantly
different from the Veech (2000) null model that predicted a
mid-domainpeak(Fig. 1a,d=6.51,P<0.001).Therelationship
between altitudinal range size ofa species and the weighted
midpoint of its altitudinal range was also non-linear and
quadratic (Fig. lb, fitted curve: y = -0.0046x2 + 8.8633 x-
3496.8, R2 = 0.72). Similarly, in contrast to the expectation
under Rapoport’s rule, the mean altitudinal range ofspecies
ineachsiteshowedaquadraticrelationshipwiththealtitude
ofthe site (Fig. 1c, fitted curve: y= -0.0005x2+ 8.8839x+
216.99,.ft2=0.78).
Birdspeciesturnover
Weightedmidpointofaltitudinalrange(m) Speciesturnoverwasconsiderablealongthealtitudinal
gradient. Non-metric multidimensional scaling ordination
usingtheMorisitadissimilaritymatrixofthedataontreeand
bird species composition showed similar patterns (stress =
m
0.070and0.034,Fig. 2a,b, respectively). Sitesbelow900
tended to cluster together as did sites above 1 100 m, while
intermediatesiteswererelativelydissimilartotheothers(with
exceptions shown in tree species composition by site D at
843mandsiteMat 1341 m).
500 600 700 800 900 1000 1100 1200 1300 1400 1500 Partial Mantel tests showed that dissimilarity in bird
Altitude(m) community composition between sites was independently
Fig. 1: Bird species richness and range size relationships in positivelycorrelatedtobothaltitudinaldistancebetweensites
rainforestsofKMTR;(a)upperpanel:evaluationoftheVeech(2000) (T=0.60,ft<0.0001,Fig.3a)anddissimilarityintreespecies
—
null model of the mid-domain effect line shows best fit curve to composition(ft=0.22,ft=0.013,Fig.3b).Correlationswith
nullmodelestimatesagainstobservedcurve(dottedline);(b)middle both these variables were significant even when controlled
panel:altitudinalrangesizeof58residentbirdspeciesinrelationto for the effects of geographic distance (ft = 0.69 and 0.51,
theweightedmid-pointoftheiraltitudinalrange;and(c)lowerpanel:
meanaltitudinalrangeofspeciesateachofthe 14sampledsitesin ft < 0.0003). In contrast, geographic distance between sites
relationtoaltitudeofthesites. had noeffecton birdspeciescomposition, whentheeffectsof
150 J. Bombay Nat. Hist. Soc., 102 (2), May-Aug 2005
TROPICAL RAINFOREST BIRD COMMUNITY STRUCTURE IN WESTERN GHATS
modelwereclosetothatactuallyobserved(0.516vs.0.529),
butwere,nevertheless,significantlylower(P<0.0002).The
double-constraintmodelalsoperformedpoorlyinestimating
average similarity between adjacent sites (0.517 vs. 0.693).
The range-contiguity model estimated similarities between
adjacentsitesthatwerehigherandmuchclosertotheobserved
value (0.622 vs. 0.693). However, the model estimates were
significantly lower than the observed values for double-
constraintandrangecontiguitymodels(P<0.0002,Fig. 5).
Dimension1 DISCUSSION
Species richness
Speciesrichnessofresidentrainforestbirdsvariedlittle
despitesubstantialchangeincommunitycompositionacross
the altitudinal gradient, in contrast to expectation from the
four main models hitherto proposed (Rahbek 1997). The
generalmodel,propoundingamonotonousdeclineinspecies
richness with altitude in parallel with an assumed decline in
productivity,findsweaksupport,ifany,whenallbirdspecies
including latitudinal migrants areconsidered. The pattern of
range sizes observed in this study results in local species
Fig.2: Ordinationofstudysitesonthebasisofsimilarityin (a)tree richness patterns that depart significantly from expectations
species composition (upper panel) and (b) bird community
composition(lowerpanel)usingmultidimensionalscaling.Sitesare
namedAtoN inincreasingorderofaltitude.
altitudinal distance were controlled for(T=0.08, P=0.12),
indicating the primary importance ofaltitude. We found no
significant effect of foliage profile dissimilarity on bird
communitycomposition.
Species turnover between sites (average of 91 all-
possiblepairs)was0.529±0.139SDbytheJaccardsimilarity
indexand0.608 +0.251 SDby theMorisitasimilarityindex.
This was lower than average similarity in bird community
composition between adjacent sites (Jaccard index: 0.693 0 100 200 300 400 500 600 700 800 900
Altitudinaldistancebetweensites(m)
±0.107SD,Morisitaindex: 0.860±0.103 SD).Thepatternof
turnoverwas not monotonic or increasing with altitude, and
remained relatively low but for higher turnover (lower
m
similarity)around 1000 (Fig.4).
Nullmodelsofspeciesturnover
We assessed whether the five null models of bird
species turnover could accurately estimate the observed
Jaccard similarity between all possible pairs ofsites (0.529
±0.139SD)andbetweenadjacentpairsofsites(0.693±0.107
SD). The unconstrained model and saturation models gave 0 10 20 30 40 50 60 70 80 90 100
Treespeciesdissimilarity(Morisitaindex%)
similar results, producing similarity estimates that were
significantlylowerthantheobserved(P<0.0002,Fig.5).The pFiagi.rs3): iBnirrdeclaotmimonuntiotycodrirsesismpiolanrdiitnygbebtewteweeenns-istietse(9(1a)alall-tpiotsusdiibnlale
estimated average similarities between all possible pairs of distance (upper panel) and (b) tree species dissimilarity (lower
sitesbytherange-contiguitymodelandthedouble-constraint panel).
J. Bombay Nat. Hist. Soc., 102 (2), May-Aug 2005 151
TROPICAL RAINFOREST BIRD COMMUNITY STRUCTURE IN WESTERN GHATS
Altitude(m)
Unconstrained Saturation Distribution Double Range Observed
0.6 r constraint contiguity
0.5 - Fig.5:Between-sitesimilarities(averageJaccardindex)inrainforest
birdcommunitycompositionestimatedfromthefivenon-equilibrium
0.4 nullmodelscomparedwithobservedvaluesinthedataset.Vertical
(D linesare 1 SD.
c 0.3
t- extinction, (iv) the species pool contains species capable of
02
utilizing the entire range of resources and showing
0.1 compensatory shifts in abundance, and (v) the division rule
governing apportioning of resources across species results
0.0
500 600 700 800 900 1000 1100 1200 1300 1400 1500 in similarranked species-abundance distributions.
Altitude(m) Thereispartial supportforthesecriteriaforrainforest
Fig.4:Turnoverinrainforestbirdcommunitycompositionasafunction birdsinKMTR.Althoughproductivitycouldnotbedirectly
ofaltitudein KMTR. Intheupperpanel (a) similaritybetweeneach measured, the relative lack of clinal change in vegetation
siteandtheimmediatelyhighersiteisplottedagainstaltitudeofthe variablesacrossthealtitudinal gradientsuggeststhathabitat
lowersite. The lowerpanel (b) usesdatafrom 10 roughlyequally-
structureandresourceavailabilitydidnotvarysubstantially.
spacedsitesalongthegradientandplotsrateofchangeinsimilarity
between a site and the sites immediately lower and higher to it This is partly because a narrower altitudinal gradient was
againstaltitude ofthe central site. The rate ofchange ofsimilarity sampledinthisstudy(500-1400m)comparedtostudiesfrom
was calculated as 1/2[(d1/w,)+(d2/w2)], where d, and w1 are the theNeotropics(Terborgh 1971, 1977;Rahbek 1997). Overa
Morisita index dissimilarityand altitudinal distance between a site wideraltituderange,birddiversitymaydecline withaltitude
andthe lowersite, and d,andw thecorrespondingvaluesforthe
2 andcorrespondingreduction in foreststructuralcomplexity
highersite.
as in the Peruvian Andes (Terborgh 1977). In the present
study, bird species richness variables were uncorrelated to
under Rapoport's rule, hump-shaped relationships with any vegetation variables except for a negative, possibly
productivity(RosenzweigandAbramsky 1993),ormid-altitude spurious, correlation with leaf litter depth. The second
peaks arising from geometric constraints (the mid-domain condition of Brown et al. (2001), species turnover due to
effect,Colwell andLees2000). changesinabioticandbioticvariables,findsclearsupportin
Instead, the observed pattern of bird species richness the correlated variation in bird community composition in
isaspatial analogueoftherecenttheoretical modelproposed relationtoaltitudeandtreespeciescomposition.Theexistence
byBrownetai (2001),toexplaintheregulationoflocalspecies ofaspeciespoolthatisnearlytwiceaslarge(withatleast58
richness over time in changing environments. This model species)asinanyonelocalsamplingsite(around30species),
suggests that species composition may vary substantially adducessupportforthethirdcriterion.Finally,theoccurrence
over time, but species richness, as an emergent property of of a number of species that occurred across the entire
ecosystems, is often regulated within narrow limits. Five altitudinal gradient and the similar overall bird abundance
conditions are necessary and sufficient for this to occur and ranked species-abundance distributions (data not
(Brownetal. 2001 ): (i)productivity,orresourceavailability, presentedhere)suggestsupportforthefinalcriteria.Further
remainsrelativelyconstant, (ii)otherabioticorbioticfactors supportis, however,requiredregardingcompensatoryshifts
vary,causingturnoverinspeciescomposition,(iii)aregional in abundance ofbird species along the altitudinal gradient,
species pool provides a source ofcolonists tolocal sites that possibly in relation to occurrence of other bird species
are open systems with respect to species colonisation and (TerborghandWeske 1975).
152 J. Bombay Nat. Hist. Soc., 102 (2), May-Aug 2005
TROPICAL RAINFOREST BIRD COMMUNITY STRUCTURE IN WESTERN GHATS
Speciesrangesizesanddistribution demonstrated for communities of tropical rainforest trees
RangesizesofrainforestbirdsatKMTRdonotconform (Terborghetal. 1996;Pitmanetal. 2001)andbirds(Terborgh
totheexpectationsunderRapoport'srule,whichmayoperate andWeske 1975;Ramanetal. 1998).
only under specific ecological conditions including The significant effect of altitude and tree species
competition and the rescue effect (Taylorand Gaines 1999). compositionon birdcommunitycompositionandthe lackof
The failure ofRapoport’s rule and the mid-domain effect is influenceofgeographicdistance,suggeststhe inapplicability
consequentialtothepeculiardistributionpatternofrainforest ofnon-equilibriummodelsofrandomly-varyingdistribution-
birds in KMTR. A number of species, particularly those abundance patterns with spatial dependence (Terborghetal.
endemictotheWesternGhats, wererestrictedindistribution 1996).Thus,high-altitude sites inSengalthericlustered with
to lower(e.g., Ocycerosgriseus Cyornispallipes)orhigher high-altitudesitesfartheraway (Neterikal trail, Kakachi and
,
(Ficedula nigrorufa Brachypteryx major) altitudes Kodayar) than to virtually adjacent sites at lower altitudes.
,
(Appendix, Ali and Ripley 1983). Daniels (1992) has noted Bird community composition may be constrained by the
that a significant proportion ofendemics, among birds and altitude-specific environment including temperature,
angiosperins intheWestern Ghats, is restrictedtothehigher irradiance, and other biological factors known to vary with
hills(> 1000m). Inotherwords,birddistributionsweremainly altitude in tropical rainforests (Richards 1996). In addition,
of three types, restricted-range species of low and high tree species composition appears to be a key determinant of
altitudes,andwidelydistributedspecies,withnomid-altitude rainforest bird community composition in this study as in
species with small ranges. The low altitude species included otherstudiesfromsouth-westernIndia(RamanandSukumar
many that also use moist deciduous forest habitats (e.g., 2002),andnorth-easternIndia(Ramanetal. 1998).
Ocyceros griseus, Psittacula columboides, Pycnonotus Thepatternofspeciesturnoverwith altitude indicated
priocephalus), whereas high-altitude restricted species are theoccurrenceofdistinctivecommunitycompositionat low-
largely confined to wet evergreen rainforest (e.g., Ficedula altitudes(<900m)andhighaltitudes(>1100m)separatedby
nigrorufa, Brachypteryxmajor,Eumyiasalbicaudata,Aliand a transitional zone of high turnover. This paralleled the
Ripley 1983).Thispatternmaybeaconsequenceofhistorical observedrangesizedistributionsandplacementsofrainforest
factors that influenced the prevalence and distribution of birds in. KMTR. The occurrence of the transition zone at
rainforestovergeologicaltimescales. around 1000-1200 m altitude may be due to environmental
changes related to the formation of cloud- and mist-cover
Speciesturnover during the monsoon months in the southern Western Ghats
Aspectrumofopinionexistsonthefactorsinfluencing asinothertropicalmontanerainforestregions(Richards 1996).
the composition or assembly of species communities at
particular sites and their variation over space and time. At Nullmodelsofcommunityassembly
oneextreme,equilibrialmodelssuggestthatlocalcommunities In large-scale community studies, models with few or
areintegrated,repeatableunitswhosecomposition isstrictly no constraints that attempt to parsimoniously simulate the
regulated and predictable as aresult ofdeterministic factors effects of pure chance, almost invariably fail to explain
such as the varying environmental tolerances ofspecies and communitystructure(FarnsworthandEllison 996;Terborgh
1
competition(Clements 1916;MacArthur 1972;Pandolfi 1996; etal. 1996; Pitman etal. 2001). In this study, weevaluateda
Terborgh et al. 1996; Pitman et al. 2001). In contrast, non- hierarchyofnullmodelswithconstraintsthatattempttoinject
equilibrium approaches note that community composition varying degrees of biological realism. The results clearly
varies substantially over space and time, apparently due to indicatethatnon-equilibriummodelsthatdonotincorporate
stochasticorhistoricaleffectsofcolonization andextinction essential biological constraints fail to predict or reproduce
(Gleason 1926; Whittaker 1970; Strong etal. 1984; Hubbell the observed pattern of similarities in species composition
andFoster 1986;Brownetal. 2001).Afundamentaldistinction betweensites. Evenwithconstraintsonsiterichness,species’
betweennon-equilibriumandequilibriummodelsisthatunder range widths, and contiguity, the results suggest that
theformer,communitycompositionmaybeexpectedto ‘drift’ assumingspeciesaredistributedindependentlyofeachother
or vary continuously through space and time, whereas the can result in community similarities close to, but departing
latter predicts that spatially or temporally independent sites significantlyfrom,thoseactuallyobserved.Thus,Whittaker’s
with similar environmental conditions would have similar (1970)modelsoflocalcommunitystructureasaconsequence
communities(HubbellandFoster 1986;Terborghetal. 1996). of independent overlapping species distributions along
Using a space-for-time substitution approach, biological ecological gradients are only partly supported. The data
influences and deterministic structure have been suggests the possibility that some species have significantly
J. Bombay Nat. Hist. Soc., 102 (2), May-Aug 2005 153
:
TROPICAL RAINFOREST BIRD COMMUNITY STRUCTURE IN WESTERN GHATS
higher overlap or avoidance due to biological determinants T. MacArthur Foundation, USA. We thank the Tamil Nadu
such as joint distributions up to ecotones or exclusive ForestDepartmentforresearchpermissionsandV.K.Melkani
distributionsduetocompetitiveinteractions(Terborgh 1971, forhis interest and support in thefield. N.M. Ishwar, Divya
1985). Field research in the Western Ghats on competition Mudappa, and K. Vasudevan were helpful and inspiring
and on avian distributional ecology along with null model colleagues and we thank them and Ravi Chellam for
tests (e.g. Hofer et al. 19§9, 2000) may shed more light on unstintinglysharingthefieldstationresources. Divyahelped
these aspects. enormously in fieldwork, data collection, and discussions.
We are grateful to P. Jeganathan for assistance with
ACKNOWLEDGEMENTS vegetationdatacollectionandtoU. HoferandM.B. Krishna
for discussions. M. Jeyapandian, A. Silamban, Sashikumar
This research was funded by the Ministry of and many others are thanked for their assistance in the
Environment and Forests, India, and John D. and Catherine field.
REFERENCES
Ali, S. & S.D. Ripley (1983): Handbook of the Birds of India and Hofer, U.. L.F. Bersier & D. Borcard(2000): Ecotonesandgradient
Pakistan. Compact edition. Oxford University Press, Delhi. as determinants of herpetofaunal community structure in the
Bachman, S., W.J. Baker, N. Brummitt, J. Dransfield & J. Moat primary forest of Mount Kupe, Cameroon. J. Trop. Ecol 16:
(2004): Altitudinal gradients,areaandtropical islanddiversity: 517-533.
an example from the palms of New Guinea. Ecography 27 Hubbell, S.P. & R.B. Foster(1986): Biology, chance,and thehistory
299-310. and structure oftropical rain foresttree communities. Pp. 314-
Blake, J.G. & B.A. Loiselle (2000): Diversity of birds along an 329.In:Communityecology(Eds: Diamond.J.M.&T.J.Case),
altitudinal gradient in the Cordillera Central, Costa Rica. Auk HarperandRow,NewYork.
117: 663-686. Janzen,D. (1973):Sweepsamplesoftropicalfoliageinsects:effectsof
Brown, J.H., S.K.M. Ernest, J.M. Parody & J.P Haskell (2001): seasons, vegetation types, altitude, time ofday, and insularity.
Regulation of diversity: maintenance of species richness in Ecology 54: 687-708.
changing environments. Oecologia 126: 321-332. Krebs,C.J (1989):Ecologicalmethodology.HarperandRow,NewYork.
Clements,F.E.(1916):Plantsuccession:Ananalysisofthedevelopment Kumar,A., D. Mudappa& R. Pethiyagoda(2004):WesternGhatsand
of vegetation. Publication No. 242, Carnegie Institute, Sri Lanka. Pp. 152-157. In: Hotspots revisited: Earth's
Washington DC. biologically richest and mostendangeredterrestrial ecoregions
Colwell, R.K. (1997): Estimates: Statistical estimation of species (Eds: Mittermeier, R.A., P. R. Gil, M. Hoffmann, J. Pilgrim,
richnessandsharedspeciesfromsamples.Version6.01b. URL: T. Brooks, C.G. Mittermeier, J. Lamoureux & G. A. B. da
http://viceroy.eeb.uconn.edu/EstimateS. Fonseca). CEMEX, Mexico.
Colwell, R.K.& D.C.Lees(2000):Themid-domaineffect:geometric MacArthur, R.H. (1972): Geographical ecology: patterns in the
constraints on the geography ofspecies richness. Trends Ecol. distributionofspecies. HarperandRow, NewYork.
Evol. 15: 70-76. Manly, B.F.J. (1994): Multivariate statistical methods: a primer, 2nd
Daniels,R.J.R.(1992):Geographicaldistributionpatternsofamphibians edn. Chapman and Hall, London.
in the Western Ghats. India. J. Biogeog. 19: 521-529. Myers,N.,R.A.Mittermeier,C.G.Mhtermeier,G.A.B.daFonseca&
Daniels, R.J.R., N.V. Joshi & M. Gadgil (1992): On the relationship J. Kent (2000): Biodiversity hotspots for conservation
between bird and woody plant species diversity in the Uttara priorities. Nature 403: 853-858.
Kannada District of south India. Proc. Natl Acad. Sci. (USA) Pandolfi, J.M. (1996): Limited membership in Pleistocene reefcoral
89: 5311-5315. assemblages from the Huon Peninsula, Papua New Guinea:
Farnsworth.E.J. &A.M. Ellison(1996): Scale-dependentspatialand constancy during global change. Paleobiol. 22: 152-176.
temporal variabilityinbiogeographyofmangroverootepibiont Parthasarathy,N.(2001):Changesinforestcompositionandstructure
communities. Ecol. Monogr. 66: 45-66. in three sites of tropical evergreen forest around Sengaltheri,
Gamble, J.S. & C.E.C. Fischer (1915-1935): Flora ofthe Presidency Western Ghats. Cun: Sci. 80: 389-393.
ofMadras. Parts I-XI. Secretary ofState for India. London. Pascal. J.P. (1988): Wet evergreen forests of the Western Ghats of
Ganesh,T., R. Ganesan, M.S. Devy, P. Davidar& K.S. Bawa (1996): India: ecology, structure, floristic composition and succession.
Assessment of plant biodiversity at a mid-altitude evergreen Institut Fran^ais de Pondichery, Pondicherry, India.
forestofKalakad-MundanthuraiTigerReserve.WesternGhats. Pascal, J.P. & B.R. Ramesh (1997):Afieldkey tothetreesand lianas
India. Cun: Sci. 71: 379-392. of the evergreen forests of the Western Ghats (India). Institut
Gaston, K.J.. T.M. Blackburn & J.I. Spicer (1998): Rapoport's rule: Fran^ais de Pondichery, Pondicherry, India.
time for an epitaph? Trends Ecol. Evol. 13: 70-74. Patterson, B.D.,V. Pacheco& S. Solari (1996): Distributionofbats
Gleason,H.A.(1926):Theindividualisticconceptofplantassociation. alonganaltitudinalgradientintheAndesofsouth-easternPeru.
Bull. Torrey Bot. Club 53: 7-26. J. Zool. 240: 637-658.
Hemelrijk, C.K. (1990): Models of, and tests for, reciprocity, Patterson, B.D., D.F. Stotz, S. Solari & J.W. Fitzpatrick (1998):
unidirectionalityandothersocialinteractionpatternsatagroup Contrasting patterns of altitudinal zonation for birds and
level. Anim. Behav. 39: 1013-1029. mammals in the Andes of south-eastern Peru. J. Biogeog. 25:
Hofer, U„ L.F. Bersier & D. Borcard(1999): Spatialorganizationof 693-607.
aherpetofaunaonanaltitudinalgradientrevealedbynullmodel Pitman,N.C.A.,J.W.Terborgh,M.R.P. Silman.V.Nunez,D.A.Neill,
tests. Ecology 80: 976-988. C.E.Ceron,W.A. Palacios& M.Aulestia(2001): Dominance
154 J. Bombay Nat. Hist. Soc., 102 (2), May-Aug 2005