The coincidence of rarity and richness and the potential signature of history in centres of endemism

12 Pages·2004·0.98 MB·English

Ecology Letters,(2004) 7:xxx–xxx doi:10.1111/j.1461-0248.2004.00678.x
REPORT
The coincidence of rarity and richness and the
potential signature of history in centres of endemism
Abstract
WalterJetz,1,2*CarstenRahbek3 We investigate the relative importance of stochastic and environmental/topographic
andRobertK.Colwell4 effects on the occurrence of avian centres of endemism, evaluating their potential
1EcologyandEvolutionary historical importance for broad-scale patterns in species richness across Sub-Saharan
BiologyDepartment,Princeton Africa. Because species-rich areas are more likely to be centres of endemism by chance
University,Princeton,NJ,USA alone,wetesttwonullmodels:Model 1calculatesexpectedpatternsofendemismusing
2BiologyDepartment,University
a random draw from the occurrence records of the continental assemblage, whereas
ofNewMexico,Albuquerque,
Model 2 additionally implements the potential role of geometric constraints. Since
NM,USA
Model 1 yields better quantitative predictions we use it to identify centres of endemism
3ZoologicalMuseum,University
controlled for richness. Altitudinal range and low seasonality emerge as core
ofCopenhagen,Copenhagen,
environmental predictors for these areas, which contain unusually high species richness
Denmark
4DepartmentofEcologyand comparedtootherpartsofsub-SaharanAfrica,evenwhencontrolledforenvironmental
EvolutionaryBiology,University differences. This result supports the idea that centres of endemism may represent areas
ofConnecticut,Storrs,CT,USA of special evolutionary history, probably as centres of diversification.
*Correspondenceandpresent
addressasofDecember2004: Keywords
DivisionofBiologicalSciences, Africa, birds, conservation, endemism, geographic range size, geometric constraints,
UniversityofCalifornia,San mid-domain effect, null model, random draw, species richness.
Diego,LaJolla,CA92093,USA.
E-mail:[email protected]
Ecology Letters (2004) 7: xxx–xxx
difficult to establish (e.g. Francis & Currie 2003). It follows
INTRODUCTION
thatgeographicanalysesofspeciesrichnessoftenexemplify
A growing number of regional and continental analyses adividebetweenecological(MacArthur1972;Endler1982a;
suggestthatalargeproportionofthegeographicvariationin Brown1995)andhistorical(Rosen1978;Nelson&Platnick
species richness can be explained by contemporary factors, 1981; Haffer 1982) approaches to the analyses of species
suchasproductivityandhabitatheterogeneity(Brown1995; distributionsthathasmarkedthepastfortyyearsofresearch
Rosenzweig1995;Currieet al.1999;Rahbek&Graves2001; in the interface of biogeography, ecology and evolution.
Jetz & Rahbek 2002; Francis & Currie 2003). However, Whilepalaeoecologicalevidenceallowsincreasingly more
there is a history behind species richness patterns, and accuratepredictionofpastvegetationpatterns,exactspatio-
despitetheprominentroleofpresent-dayfactorsthereisno temporal habitat dynamics and their specific effect on
doubtthathistoryisreflected,inoneformoranother,inthe animal distributions and gene flow remain obscure. How-
distribution of contemporary assemblages (Ricklefs & ever, coarse indices of climatic stability can be attained and
Schluter 1993; Cracraft 1994; Patterson 1999). Advocates yield interesting first insights about the direct effect of past
of the role of regional history, biogeographic barriers and climateonspeciesdistributions(Dynesius&Jansson2000).
the large-scale processes of allopatric speciation and Anotherpromisinganalyticalangleisgivenbytheevermore
extinction urge caution over the neglect of historical accurate and comprehensive phylogenies that allow phylo-
mechanisms such as past isolation dynamics, which tend geographic analyses at an increasingly large spatial and
to be ignored in analyses that focus on contemporary phylogenetic scale (Fjeldsa & Lovett 1997; Schneider et al.
patterns of species richness (Latham & Ricklefs 1993; 1999; Moritz et al. 2000).
Ricklefs et al. 1999; Ricklefs 2004). Meanwhile, other Across taxa, regions and scales, contemporary environ-
authorsholdthattheconsistentlystrongexplanatorypower ment models are repeatedly found to explain a very large
of contemporary climate suggests only a minor role for proportion of overall species richness, usually between
historicalprocesses for which direct evidence isnotoriously 60 and 85% (Schall & Pianka 1978; Currie 1991; Rahbek &
(cid:1)2004BlackwellPublishingLtd/CNRS
2 W.Jetz,C.RahbekandR.K.Colwell
Graves 2001; Jetz & Rahbek 2002; Francis & Currie 2003). number of species than other regions. Historical biogeog-
However, these statistically strong relationships mask two raphy has so far focused on the importance of areas of
important phenomena. First, outliers with much higher endemism in quantifying vicariance, but has only begun to
richness than predicted by contemporary environment specificallyaddresstheconfoundingissueofrandomeffects
models occur (Rahbek & Graves 2001; Jetz & Rahbek on area selection at a large scale (Mast & Nyffeler 2003).
2002) and often tend to be spatially clustered in areas that Beyond their role as indicators for testing biogeographic
havebeenpinpointedinthepasttocontainphylogenetically hypotheses, centres of endemism represent regions of high
orbiogeographicallyuniquespecies(Fjeldsa&Lovett1997; conservation concern (Stattersfield et al. 1998). With an
Stattersfield et al. 1998). Second, species with very small ever-increasing rate of extinction and lack of distributional
geographicrangestend toshowapatterninrichnessthatis information, knowledge about the potential large-scale
verydifferenttothatofallspeciestogether,theyareaffected predictability of centres of endemism from environmental
by different variables, and they are less predictable by factors reaches beyond traditional hypothesis testing.
contemporary environment (Jetz & Rahbek 2002). Both Whether centres of endemism – besides containing species
findings highlight locations where contemporary environ- thatarethreatenedaccordingtocurrentlyacceptedcriteria–
mental models fail, despite their strong explanatory power have a special role as areas of high past and possibly future
for overall species richness, and thus where historical evolutionary potential is a matter of particular importance
processes may prevail. We suggest that the occurrence of for large-scale conservation priority setting (Fjeldsa et al.
narrow-ranged species in so-called (cid:1)centres of endemism(cid:2) 1999; Crandall et al. 2000).
and underprediction of richness by environmental models Here we set out to address these issues by creating a
may be interrelated, each pointing to the geographic methodological bridge between those studies of species
occurrence of historical processes that affect both ende- richness that focus exclusively on contemporary environ-
micity and richness. mental correlates and those studies that attempt to infer
In the debate about historical interpretations of species historical process based on variance left conditionally
distributions, (cid:1)centres of endemism(cid:2) have repeatedly been unexplained by contemporary and stochastic models. Using
regardedasexemplifyingtheroleofhistoryincontemporary nullmodelstocontrolfortheconfoundingeffectofspecies
patterns of species distributions (Rosen 1978; Nelson & richness, we identify areas of endemism that cannot be
Platnick1981;Haffer1982;Prance1982).Onesuggestionis conditionally explained by contemporary environmental or
that these regions, if marked by primary endemics, are stochastic effects, suggesting a potential signature of
centresofcladeoriginandspeciationandshouldstilltestify historical processes on species richness. Specifically, we
to this special historical role by a very high overlap of ask(1)Howmanynarrow-rangedspecieswouldbeexpected
contemporary geographical ranges (Croizat et al. 1974; inanassemblagebasedsolelyontheoverallspeciesrichness
Terborgh 1992; Ricklefs & Schluter 1993). That is, overall of that assemblage, and how many and which centres of
species richness in such places should be higher than endemism remain when this effect is controlled for? (2) To
elsewhere, regardless of the particular endemic species what extent can the occurrence of these centres of
within them. endemism be explained from environment and topography
This view, and support for any specific historical alone? and (3) Do centres of endemism contain more
mechanismcouldpotentiallybechallengedifthegeographic species than other regions, even after controlling for any
distribution of centres of endemism largely follows sampling effects and accounting for potential differences in
contemporary factors (Endler 1982b; see also Francis & environment?
Currie 2003). Yet, even the latter interpretation may be
challengedifonecouldshowthatchancealonewasenough
to explain the observed pattern. The apparent local excess DATA AND METHODS
of narrow-ranged species that define centres of endemism
Distribution data
might simply represent the expected number of such
species, given locally higher richness, and such a simple The distributional data and grid used here are identical to
(cid:1)sampling effect(cid:2) would call any further historical or thatinJetz&Rahbek(2002),ascompiledbytheZoological
ecological inferences into question (Connor & Simberloff Museum, University of Copenhagen (Burgess et al. 1998).
1979; Gotelli & Graves 1996; Maurer 1999). It follows that This database consists of breeding distribution data for
any special (e.g. evolutionary historical) role of centres of all 1599 birds endemic to Sub-Saharan Africa across a
endemism can be evaluated only if the effect of species 1(cid:2) latitudinal–longitudinal grid of 1738 quadrats and con-
richness, per se, is properly accounted for. This requirement tains 366 853 species presence records(quadratscontaining
has so far left unresolved the issue of whether centres of £ 50% dry land were excluded). We defined species with
endemism do indeed contain an unexpectedly greater geographic range sizes £ 10 quadrats as (cid:1)narrow-ranged
(cid:1)2004BlackwellPublishingLtd/CNRS
Thecoincidenceofrarityandrichness 3
species(cid:2)(n ¼ 190 species, representing0.27% ofall quadrat Model 1
records)andthequadratsinwhichtheyoccuras(cid:1)Centresof Allelsebeingequal,agivenquadratismorelikelytocontain
Endemism(cid:2)(n ¼ 423).(cid:1)Areasofendemism(cid:2)aretraditionally species with a large than a small geographic range size (i.e.
defined as regions with at least two overlapping species numberofquadratrecords).Inordertoallowdifferencesin
restricted in range (Harold & Mooi 1994; Stattersfield et al. rangesizetobearconsequence onspeciesrepresentation in
1998; Hausdorf 2002). Our approach here is somewhat communities, the probability of a species(cid:2) inclusion should
different, in that we are interested in the potential special beproportionaltoitsgeographicrangesizeinrelationtothe
signature of the occurrence of narrow-ranged species as sum of all range sizes and the species richness N of the
A
such, disregarding their immediate relevance for vicariance assemblage. This representation in a local quadrat assem-
biogeography. For conservation studies a range size cut-off blage can be simulated by a random draw of species from
of 50 000 km2 is often used (e.g. Stattersfield et al. 1998), the observed list of species(cid:2) quadrat-records. Wide-ranged
but here we chose 10 quadrats (approximately species contribute more quadrats than narrow-ranged
110 000 km2) as a compromise between statistical power species, and are thus more likely to be sampled. Only
and the ability to identify unique regions (see also Fjeldsa sampling without replacement achieves an unbiased repre-
2003). sentation of species, and ensures that each species can be
represented in an assemblage at most once.
We performed Model 1 simulations using a custom-
Null model predictions
written C program. For each hypothetical quadrat species
Some quadrats may be more likely than others to include richness value N (between 1 and 1000), quadrat records
A
manynarrow-rangingspeciessimplybecauseofsample-size were drawn at random from the observed list of quadrat
effects(Connor&Simberloff1979;Gotelli&Graves1996; records(366 853overall)andthespeciesrepresentedbythat
Maurer 1999; Mast & Nyffeler 2003) or geometric record, if not already present in the quadrat, was added to
constraints (Colwell & Lees 2000; Jetz & Rahbek 2001; thelist ofspeciesuntilN distinctspecieshadbeendrawn.
A
Colwell et al. 2004; Pimm & Brown 2004). With regard to This procedure was repeated 1000 times for each value of
sample size, all other things being equal, one would expect N (yielding 1 million species lists, in all) and the average
A
species-rich quadrats to contain more species of all range number(and95%percentiles)ofnarrow-rangedspecieswas
size categories, including more narrow-ranged species, than calculated. This yielded a general relationship between
species-poor quadrats. Researchers have attempted to quadrat species richness N and expected number of
A
address this issue by deriving indices that down-weight the narrow-ranged species per quadrat, which we then applied
occurrenceofwiderangingspeciesand/ordividebyoverall to the empirical patterns.
richness (assuming a proportional relationship, e.g. Linder
2001), but none of these corrections control for richness Model 2
satisfactorily. With regard to geometric constraints, mid- Geometric constraints imposed by hard boundaries (such
domain models (Colwell & Lees 2000) predict that wide- as continental edge for terrestrial species) on species
rangingspeciesaremorelikelytooverlapintheinteriorofa richness are expected to (cid:1)force(cid:2) the occurrence of wide-
bounded domain (such as sub-Saharan Africa, bounded by ranged species towards the middle, while narrow-ranged
sea and desert) than nearer its edges for wholly non- species should be unaffected (Lees et al. 1999; Colwell &
biological reasons, thus producing a different pattern of Lees 2000; Jetz & Rahbek 2001; Colwell et al. 2004; Pimm
range-sizefrequenciesindifferentquadrats(Leeset al.1999; & Brown 2004). One potential prediction of this effect is
Jetz & Rahbek 2002). a higher overall richness of species in the middle of a
These issues have so far thwarted rigorous tests as to continent, driven by the higher number of wide-ranged
whether assemblages do in fact contain an unusual number species that tend to overlap there (e.g. Jetz & Rahbek
of endemics, given their overall richness, or conversely, 2002). However, a mixed scenario is possible, in which
whether putative centres of endemism are in fact more overall quadrat species richness is mostly driven by
species rich than other areas. Here we develop two null historical and environmental factors, but geometric
models that are intended to remove the (cid:1)noise(cid:2) (quadrats constraints may affect the range size composition of
that have high numbers of narrow-ranged species simply assemblages. If the middle of a continent is more likely to
because they have high richness) from the (cid:1)signal(cid:2) (areas contain wide-ranged species than the edge, quadrats of
with more narrow-ranged species than expected by chance equal overall species richness should contain more
given their level of richness). To select between the two narrow-ranged species near the edge than the middle.
models, we here assume that the null model that better fits Thus, compared with the assumption of Model 1 – that
the observed distribution quantitatively, and therefore the random draw from quadrat records applies equally to
removes the most (cid:1)noise,(cid:2) is to be preferred. all quadrats – consideration of geometric constraints
(cid:1)2004BlackwellPublishingLtd/CNRS
4 W.Jetz,C.RahbekandR.K.Colwell
predicts the presence of a relative (cid:1)excess(cid:2) of wide-ranging (AUC) of receiver–operating characteristic (ROC) plots, a
species in interior quadrats and a relative (cid:1)deficiency(cid:2) of threshold-independent measure of goodness-of-fit (Fielding
wide-ranging species in quadrats near the continental &Bell1997).FollowingSwets(1988)AUCvalues< 0.7are
edge. considered as poor, those > 0.7 as reasonable, and those
We modelled the location-specific predicted number of > 0.9asgood.Weuselogisticregressiontofurtherexamine
narrow-ranged species, given geometric constraints and null model predictions and to investigate the role of
observed overall species richness, as follows: first we used environmental predictors. In order to pre-select core
the two-dimensional (cid:1)spreading dye(cid:2) model presented by environmental variables for multiple regression within each
Jetz & Rahbek (2001) as implemented in GEOSPOD (Jetz type of predictor (Table 1) we identified pairs of variables
2001) and the observed list of range sizes to simulate for that showed high collinearity [abs(r ) > 0.5)] and only
S
each quadrat a list of species occurrences (with number of retained the variable with the higher explanatory power in
species per quadrat given by the geometric constraints one-predictor regressions. We used model simplification in
predictions), performing 100 runs. This resulted in a list order to find minimum adequate models for endemism
of 366 853 000 species occurrences across the 1738 status (logistic regression). Initially, all core variables were
1(cid:2)quadrats.Foreachquadratwethenperformedarandom included in the model. Subsequently, the predictor with the
draw (without replacement, i.e. each species was only lowest log-likelihood was excluded until all variables
sampled once) from the quadrat-specific list of species remaining in the model were significant at the 0.05 level.
occurrencesuntiltheactuallyobservedspeciesrichness,N , This analysis is confounded by spatial autocorrelation,
A
of that quadrat was reached, and repeated this procedure which can affect parameter estimates (Cliff & Ord 1981;
100 times. We then used this species list to calculate the Lennon 2000; Jetz & Rahbek 2002). Owing to a lack of
averagenumberofnarrow-rangedspeciespredictedforthis readily available methods for spatial logistic regression we
quadrat. did not control for spatial autocorrelationin this endemism
prediction analysis. The resulting likely spatial non-inde-
pendence of model residuals affects parameter estimates,
Predictor variables
measures of fitness and thus ranking of predictor variables.
Overall, we used 14 predictor variables to evaluate the The strength of this effect depends on the specific
effect of environmental and topographic conditions. These relationship between spatial autocorrelations of response
include eleven variables related to contemporary climate and predictor variable. Strong differences in the fit of
and derived features (e.g., net primary productivity, i.e. predictor variables with similar spatial autocorrelation are
NPP and NPP2, habitat heterogeneity, eight climatic likely to be robust to this issue.
variables, three of which reflect seasonality) and three We performed linear regression analysis on species
variables associated with altitude (mean and range) and richness of all 1599 bird species endemic to Africa over
area. Levels of overall richness may be affected by all 1738 quadrats of sub-Saharan Africa, similar to Jetz &
geometric constraints (mid-domain effect, Colwell & Lees Rahbek (2002), with an ad hoc model of all 15 environ-
2000; Jetz & Rahbek 2001; Colwell et al. 2004; Pimm & mental/topographic/geometric constraints predictor vari-
Brown 2004). Thus, for the analyses on overall species ables. We compared species richness predictions between
richness inside and outside centres of endemism we Center of Endemism quadrats and all others using a
included predictions (expected species richness values) traditional regression model. For statistical evaluation, we
from null model simulations using the observed range sizes enteredCenterofEndemismstatus(yes/no)asacategorical
and the assumption of fully continuous ranges (using the variablewiththeaforementioned15-predictorvariablesand
program GEOSPOD, Jetz 2001). Details on sources and compared its significance relative to other important
calculations of all 15 predictor variables are given in Jetz & predictors. We repeated the richness prediction analysis
Rahbek (2002). employing spatial regression techniques to evaluate and
control for spatial autocorrelation (Cliff & Ord 1981).
Spatialregressionseparatesthevariationacrossalatticeinto
Statistics
large-scale changes due to predictor effects and small-scale
We tested the performance of null models and environ- variation due to interactions with neighbours, which can be
mental variables in predicting presence and absence of modelled by iterative fitting of an autoregressive covariance
centresofendemismusingsensitivity andspecificity, with a modeltothedispersionmatrix.Here,weuseasimultaneous
cut-off of ‡ 1 narrow-ranged species for null model autoregressive model (SAR) and a King’s case (eight
predictions, and 50% presence–absence probability for immediate neighbours) neighbourhood structure and per-
environmental logistic regression predictions. Additionally formed the calculations in the S+ SPATIAL STATS Module
we calculated the accuracy measure (cid:1)area under the curve(cid:2) (Kaluzny et al. 1998).
(cid:1)2004BlackwellPublishingLtd/CNRS
Thecoincidenceofrarityandrichness 5
Table1 Environmental/topographic logis-
Minimum adequate
tic regression one-predictor and minimum
One-predictor models model
adequatemodelresultsfortheoccurrenceof
Model1 Centres of Endemism
Category Predictor Dir Chi-square P Dir Chi-square P
Altitude Meanaltitude + 18.80 ***
Altitudinalrange + 122.60 *** + 108.89 ***
Productivity NPP + 0.12
NPP2 + 0.11
Habitat NDVIclasses + 58.80 ***
heterogeneity TempRange ) 53.76 *** ) 52.43 ***
Seasonality NDVIseasonality ) 17.57 *** ) 14.08 ***
RainRange + 0.01
MaxTemp ) 62.39 ***
Precipitation + 0.08
Climatic factors AET + 0.14
PET ) 4.03 *
Radiation ) 2.04 + 7.05 **
Area Areadry land + 11.91 **
Habitat heterogeneity was estimated by counting the annual average of monthly classes of
NDVI (normalized difference vegetation index)perquadrat (0.05(cid:2)resolution).
Dir,directionoftherelationship; Chi-square, changein)2 log-likelihoodcomparedwitha
statistical modelwithout that predictor; NPP, netprimary productivity.
TempRangeandRainRangerefertoaverageannualrangeinmonthlymeantemperatureand
precipitation, respectively. NDVI seasonality refers to the intra-annual coefficient of vari-
ation in monthly mean quadrat NDVI. AET refers to actual evapotranspiration; PET to
potential evapotranspiration.
*P<0.05, **P<0.01, ***P<0.001.
the geographic pattern of Center of Endemism occurrence
RESULTS
(Fig. 2), but quadrat-by-quadrat Model 1 achieves a signi-
Model 1 predicts a non-linear increase of expected narrow- ficantly better fit than Model 2 (logistic regression, likeli-
ranged species with increasing quadrat species richness hood ratio test; v2 ¼ 52.30, P < 0.001). When model
(Fig. 1a). Any quadrat with over 285 species is expected to performance is tested on the number of narrow-ranged
contain at least one narrow-ranged species. Model 2 species among quadrats that contain them, both Model 1
similarly predicts a concave upward increase of narrow- and 2 are statistically significant predictors (Poisson regres-
ranged species richness with overall quadrat species rich- sion, likelihood ratio tests; Model 1: v2 ¼ 294.12,
ness, but with predicted richness levels modulated by the P < 0.001, Model 2: v2 ¼ 147.55, P < 0.001, n ¼ 423),
distance of quadrats to the continental edge (Fig. 1b). In but Model 1 yields a significantly stronger fit (v2 ¼ 146.57,
both models the maximum number of narrow-ranged P < 0.001). Comparing the logistic regression fits achieved
species predicted per quadrat is under four, which is in by the null models to those of 14 environmental/
stark contrast to the observed data with values as high as topographicvariableswefindthateventhemostsignificant
28 (Fig. 1c). predictor variable (habitat heterogeneity) has much smaller
Figure 2 illustrates the performance of null models in predictivepowerthanthenullmodelpredictions(likelihood
predicting the spatial occurrence of Centres of Endemism ratio test, Model 1: v2 ¼ 99.74, P < 0.001; Model 2: v2 ¼
(quadrats with at least one observed/predicted narrow- 78.34, P < 0.001).
ranged species). Both null models perform better in As Model 1 shows the better fit, we choose it over
correctly predicting absences than presences (high specific- Model 2asournullmodelinsubsequentanalysis.Thus,we
ity, Model 1: 0.86, Model 2: 0.85; low sensitivity, Model 1: set out to select those quadrats among the defined Centres
0.49, Model 2: 0.47) and have reasonable accuracy (AUC, ofEndemismthatcontainmorenarrow-rangedspeciesthan
Model 1: 0.77, Model 2: 0.75). They both show a highly expected by the chance effects of overall quadrat richness
significant logistic regression fit (likelihood ratio test, (as modelled by Model 1) alone. We use the upper 95%
Model 1: v2 ¼ 323.88, Model 2: v2 ¼ 271.59, P < 0.001 percentile predictions of Model 1 as lower cut-off to
in both cases, n ¼ 1738). Model 2 appears to better mimic identify these areas (Fig. 3a). Of the original 423 quadrats
(cid:1)2004BlackwellPublishingLtd/CNRS
6 W.Jetz,C.RahbekandR.K.Colwell
(a) Highlands, Eastern Zimbabwe mountains), as well as less
3 speciose areas such as the Central Somali Coast and North
s
s Somali Highlands, WesternAngola,Lesotho Highlands and
e
n
h Southeast Namibia.
c
s ri Weproceedtoevaluatethecontemporaryenvironmental/
e 2
ci topographic predictability of the occurrence of Model 1
e
p Centres of Endemism. We first perform single-predictor
s
ed logistic regressions with all 14 environmental/topographic
g
an 1 predictorvariablestoidentifyimportantvariables(Table 1).
w-r We find habitat heterogeneity, altitudinal range, maximum
o
arr temperature and annual range of temperature to be the
N statisticallymostpowerfulpredictorsofModel 1Centresof
0
Endemism.Inaminimumadequatemodelthataccountsfor
(b) collinearity and selects core predictors, a positive effect of
3 topographic heterogeneity (i.e. altitudinal range), a negative
s
es effect of two seasonality variables (seasonal temperature
n
ch range and variation in productivity) and, much less signifi-
s ri cant,apositiveeffectofsolarradiationemergeasimportant
e2
ci (Table 1). Of these, altitudinal range is by far the most
e
p
s significant. Using a 0.5 probability cut-off, this logistic
d
e regression model performs well in predicting absence, but
g
an1 not well for predicting presence (Fig. 3d, sensitivity: 0.10,
w-r specificity: 0.99). However, the cut-off independent good-
o
arr ness-of-fitisreasonable(AUC:0.92).Wellpredictedarethe
N
mountainous Centres of Endemism in East Africa and
0
CameroonandcoastalareasinAngolaandNorthSomali,but
(c) notnorthernNigeria,SomaliEastCoastandHighlands,and
ss25 theLesothohighlands.Ahighconcentrationofpredictedbut
e
n not (yet?) observed presences along the West Coast of
h
c
s ri20 CentralAfricadowntoSouthernNamibiaisnoticeable.
e Observed (unadjusted) Centres of Endemism are expec-
ci
pe15 ted to be more species rich than other regions due to the
s
d effect that species richness has on the probability of a
e
g quadratbeingaCenterofEndemism(seeabove).Observed
an10
w-r Centres ofEndemism contain on average 100 more species
o than other quadrats (U ¼ 129.09, P < 0.001). Yet, even
Narr 5 richness-controlled Model 1 Centres of Endemism harbour
onaverage69speciesmorespecies(U ¼ 45.02,P < 0.001).
0
0 100 200 300 400 500 This strong difference remains when the two to three (on
Species r ichness average,perquadrat)narrow-rangedspeciesaretakenoutof
the analysis (reduction in U marginal; P < 0.001 remains in
Figure1 Relationship between the observed overall species
all tests).This higher overall species richness inside Centres
richness and the predicted [(a) and (b)] or observed (c) narrow-
ofEndemism could simplybebecauseofdifferencesinthe
ranged species richness across 1(cid:2) quadrats of sub-Saharan Africa.
environmental conditions that correlate with species
Narrow-rangedspeciesarethosewithageographicrangesize£10
richness. That is, areas of endemism could simply be areas
quadrats. (a) Model 1 predictions (random draw of species from
continental list of quadrat occurrences). (b) Model 2 predictions that environmentally favour high species richness. To test
(randomdrawofspeciesfromGEOSPODsimulatedquadratspecific this hypothesis, we entered (cid:1)Endemism Status(cid:2) (whether a
listofquadratoccurrences).(c)Observeddata(notedifferentscale quadratisaCenterofEndemismornot)asabinaryvariable
ony-axis). in a full regression model with overall quadrat richness
minus narrow-ranged species richness as the dependent
with narrow-ranged species (Fig. 3b), 79 qualify under this variable and all 14 environmental/topographic predictor
criterion (Fig. 3c). These include species rich regions variables plus geometric constraints predictions as indepen-
(Cameroon Highlands, Albertine Rift Mountains, Kenya dent variables (Table 2). It emerges that Centres of
(cid:1)2004BlackwellPublishingLtd/CNRS
Thecoincidenceofrarityandrichness 7
(a) (b)
3
1.00
208
2.00
558
3.00
(c) (d)
1.00 1
2.00 13
3.00 27
Figure2 Nullmodelbasedataandpredictions.(a)Observedrichnessofall1599speciesendemictoAfrica.(b)Model1predictionsforthe
numberofnarrow-rangedspecies(rangesize £10,1(cid:2)quadrats)expectedacrossAfrica,giventheobservedrichnesspattern.Onlyquadrats
forwhichatleastonenarrow-rangedspeciesispredictedareplotted.(c)sameforModel2.(d)ObservedCentresofEndemismofAfricaand
theirnarrow-ranged species richness. Allare equalinterval classification.
Endemism are expected to contain relatively high numbers link between traditional environmental-correlate based
of species because of their environment alone, in particular analyses of species richness, null model focused investi-
because of their tendency to be located in productive areas gations and historical approaches that emphasize regional
(NPP is consistently the top predictor of quadrat richness). context.
However, both observed and Model 1 Centres of Ende- The notion that simple chance effects or constraints
mism, consistently contain even more species than the should be controlled for, or at least evaluated, is now an
environmental model predicts. In both cases, in the multi- established concept in community and broad-scale ecology
predictor model, Endemism Status came out as a highly (Connor&Simberloff1983;Gotelli&Graves1996;Gotelli
significant additionalvariableand asan important predictor 2001),althoughstilldebatedbysome.However,nullmodels
(Table 2). We repeated the analysis using spatial regression, still appear to play only a minor (and contentious, e.g.
which supported this result. Hawkins & Diniz 2002; Zapata et al. 2003; Colwell et al.
2004) role in biogeographic and historical analyses. In
analyses of local communities it has become clear that,
DISCUSSION
beyond environmental and historical explanation on the
Our study attempts a synthetic continental analysis of local scale, broader consideration is required of the
centres of endemism, seeking to investigate the interre- surrounding or source region in which thelocal assemblage
lated effects of species richness, environment and history. is embedded, as well as the degree to which local biotic
In methodology and approach it sets out to provide a composition actually differs from chance expectation. The
(cid:1)2004BlackwellPublishingLtd/CNRS
8 W.Jetz,C.RahbekandR.K.Colwell
(a) (b)
30
s
es 25
n
h
c
s ri 20
e
ci
e
p 15
s
d
e
g
n 10
a
w-r
arro 5
N c
0 ir
0 200 400 600 800
Species richness
(c) (d)
0.10
0.50
0.95
Figure3 Selection,distributionandenvironmentalpredictabilityofModel 1CentresofEndemism.(a)Richnessofnarrow-rangedspeciesin
aquadratinrelationtoitsoverallspeciesrichnessaspredictedbyModel1.Solidthickline:predictions[± 95%confidenceintervals(CI),thin
lines]ofModel 1.Opencircles:quadratswithnonarrow-rangedspecies.Crosses:quadratswithatleastonenarrow-rangedspecies,butless
than predictedby Model1 (£95% CI). Solid circles: Model1 Centres of Endemism,i.e. quadratswith more narrow-ranged species than
predictedbyModel 1(> 95%CI).(b)ObservedCentresofEndemismbeforeselection[allsymbolsin(a)].(c)Model1CentresofEndemism
[i.e.thesolidcirclesin(a)].(d)GeographicpredictionsofprobabilityofModel1CenterofEndemismoccurrenceaccordingtotheminimum
adequate environmental logistic regressionmodel (Table 2).Equal interval classification.
extent to which ecological characteristics of assemblages 1990). Therefore, the chance relationship between overall
deviate from random draw models has already provided andnarrow-rangedspeciesrichnessisnotlinear(Fig. 1).As
invaluable insights for many studies on island communities demonstrated by our simulation models, relatively more
(see Gotelli & Graves 1996 for review). The application of narrow-ranged species are expected in species rich assem-
this approachto local assemblages as part of regionalpools blages. Of course, by allocating species from the full
on the mainland may be similarly valuable (Maurer 1999; continental pool and thus sampling from the composite
Blackburn&Gaston2001),andthisisthefirststudytouse range-size frequency distribution across the continent, one
it on a continental scale. may neglect essential local processes or factors that
As typical for most taxa on broad scales (Gaston 2003), contribute to this frequency distribution. In the context of
the range size distribution of African birds is highly right- this analysis, the evolution and existence of extremely
skewed: there are many more narrow- than wide-ranged narrowrangeswithintheoverallrangesizedistributionmay
species(seealsoHall&Moreau1962;Pomeroy&Ssekabiira restonanexplanationoraconstraintthatdoesnotwarrant
(cid:1)2004BlackwellPublishingLtd/CNRS
Thecoincidenceofrarityandrichness 9
Table2 Select results of 15 predictors
Observed Centres ofEndemism (n¼ 423) Model1 Centres ofEndemism (n ¼79)
regression on overall minus narrow-ranged
speciesrichnessofbirdsacrosssub-Saharan
Rank Variable t P Rank Variable t P
Africa
GLM
1 NPP2 )12.72 *** 1 NPP2 )13.69 ***
2 RainRange 12.04 *** 2 NPP 11.99 ***
3 NPP 10.65 *** 3 RainRange 11.50 ***
4 Endemism Status 9.99 *** 10 EndemismStatus 4.10 ***
Moran’s I 0.67 *** Moran’s I 0.69 ***
SAR
1 NPP 9.75 *** 1 NPP 9.96 ***
2 Altitudinal range 8.87 *** 2 NDVIclasses 8.64 ***
3 NDVI classes 8.63 *** 3 NPP2 )8.10 ***
4 Endemism status 7.57 *** 6 Endemismstatus 4.86 ***
Moran’s I 0.06 *** Moran’s I )0.01 n.s.
Independent variables include the 14 topographic/environmental predictors and geometric
constraintspredictionsforoverallspeciesrichness(seeDataandMethods).Endemismstatus
is abinaryvariable thatindicates whetheraquadrat isaCentreofEndemismornot. Only
results of top three predictors are shown, together with the rank and results for the focal
variable, endemism status.
Moran’s I measures the spatial autocorrelation of the model residuals. For predictor terms
and symbolssee Table 1.
GLM, traditionallinear regression model; SAR,spatialautoregressive model.
random allocation across the continent. This critique can phical variation and occurrence of narrow homothermous
only be addressed by careful interpretation. Here we elevational bands and thus indicates the potential existence
propose that the close and highly confounding inter- of past and present barriers that facilitate speciation and
relationship between overall and narrow-ranged species beta-diversity (see also Janzen 1967; Vuilleumier 1969;
richnessbothjustifiesandrequiresanullmodelapproach.It Graves 1988; Rahbek 1997). Montane areas, in particular
confirms a potentially prominent role for random draw montane forests, have repeatedly been demonstrated to be
models in the study of continental biota. Our simplistic key areas for narrow-ranged species (Terborgh & Winter
criterion – best logistic regression fit – selected Model 1 as 1982; Stattersfield et al. 1998), also in Africa (Diamond &
best null model, but did not take into account spatial Hamilton 1980; Collar & Stuart 1988; Johnson et al. 1998;
autocorrelation effects and similarity in geographic pattern Linder 2001).
(Fig. 3) which may have favoured Model 2. Model 2 is Contemporary climate conditions are usually seen as
logically valid and may well have higher explanatory power important for processes maintaining species richness (e.g.
in other datasets. Currie et al. 1999), however they may also convey a strong
Because of its restriction to a standardized 1(cid:2) grid, our historical signature (e.g. Latham & Ricklefs 1993). In our
studyislimitedintheextenttowhichitallowsinterpretation study we identify low seasonality, best captured through
of exact distributional boundaries of narrow-ranged birds annual temperature range and obviously a measure of
(see e.g. Hall & Moreau 1962; Terborgh & Winter 1982; contemporary climate, as the second important predictor
De Klerk et al. 2002). Generally, we find that species with of centres of endemism. The role of low seasonality, as
narrow ranges show a very distinct pattern of occurrence such, in promoting occurrence of centres of endemism has
thatcanonlypartlybepredictedfromcontemporaryfactors. as yet not been quantitatively demonstrated. From an
Oneofthetwostrongpredictorsofthegeographiclocation ecological perspective the connection between very small
of centres of endemism is large altitudinal range within a range size and low seasonality may not be surprising.
quadrat (not altitude per se). Although altitudinal range has Seasonal environments are likely to limit tight habitat
sometimes been used as an estimate of habitat heterogen- specializations, which in turn may thwart small ranges
eity,topographicheterogeneitymeasuredasaltitudinalrange (Janzen 1967; Huey 1978; Stevens 1989). The significance
might better be viewed as a rough surrogate variable of seasonality may be related to the idea of ecoclimatic
reflecting historical opportunities for allopatric speciation stability furthering the persistence of relictual endemics
(e.g. Rahbek & Graves 2001; Jetz & Rahbek 2002). (Fjeldsa et al. 1997; Dynesius & Jansson 2000). One
Altitudinal separation within a quadrat measures topogra- interpretation of the role of eco-climatic stability was
(cid:1)2004BlackwellPublishingLtd/CNRS
10 W.Jetz,C.RahbekandR.K.Colwell
triggered by the observation that in African montane Foundation, German Academic Exchange Service and
forests the distributions of both phylogenetically old and German Research Foundation to WJ, and US-NSF grant
young species coincide (Fjeldsa & Lovett 1997), suggesting DEB-0072702 to RKC.
a causal link between speciation, endemism and long-term
stability (Fjeldsa et al. 1997).
REFERENCES
We may conclude that that centres of endemism are
concentrated in regions that offered unusually many Blackburn,T.M.&Gaston,K.J.(2001).Localavianassemblagesas
opportunities for past speciation, combined with stable random draws from regionalpools. Ecography,24, 50–58.
Brown, J.H. (1995). Macroecology. University of Chicago Press,
climates that allowed survival of narrow endemics despite
Chicago,IL.
their small geographic ranges. It has been argued that if
Burgess, N., Fjeldsa˚, J. & Rahbek, R. (1998). Mapping the dis-
centres of endemism really have acted as centres of
tributions of Afrotropical vertebrate groups.Species, 30,16–17.
speciation, as (cid:1)speciation pumps(cid:2), in the past (Terborgh Cliff, A.D. & Ord, J.K. (1981). Spatial Processes: Models & Applica-
1992), they are likely to contain more species than other tions.Pion, London.
regions today (Endler 1982b; Haffer 1982; Prance 1982). Collar, N.J. & Stuart, S.N. (1988). Key Forests for Threatened Birds in
Here we are able confirm this prediction and show that, in Africa. International Council for Bird Preservation, Cambridge,
UK.
Africa,centresofendemismindeeddocontainmorespecies
Colwell, R.K. & Lees, D.C. (2000). The mid-domain effect: geo-
than expected by chance or environment and topography
metric constraints on the geography of species richness. Trends
alone. However, we are unable to separate to what degree
Ecol.Evol.,15,70–76.
this pattern arises due to past differences in rates of Colwell,R.K.,Rahbek,C.&Gotelli,N.J.(2004).Themid-domain
speciation and extinction, or immigration and emigration, effect and species richness patterns: what have we learned so
and can therefore not directly test the (cid:1)speciation pump(cid:2) far?Am. Nat.,163, E1-E23.
hypothesis. Not only speciation but species persistence Connor, E.F. & Simberloff, D.S. (1979). The assembly of species
communities: chance orcompetition. Ecology,60,1132–1140.
(locally low extinction) may determine the occurrence of
Connor,E.F.&Simberloff,D.(1983).Neutralmodelsofspecies(cid:2)
centres of endemism (Mayr 1963). Knowledge of which
co-occurrencepatterns.In:EcologicalCommunities:ConceptualIssues
centres of endemism are primary (based on relictual,
andtheEvidence(edsStrong,D.R.,Simberloff,D.&Abele,L.G.).
autochthonous endemics) would yield stronger insights Princeton University Press, Princeton, NJ.
about historical mechanisms. The fact that clear patterns Cracraft, J. (1994). Species-diversity, biogeography, and the evo-
appear even in an analysis without that knowledge under- lutionof biotas. Am.Zool.,34,33–47.
lines the strength of a historical explanation and supports Crandall, K.A., Bininda-Emonds, O.R., Mace, G.M. & Wayne,
R.K.(2000).Consideringevolutionaryprocessesinconservation
the likely role of centres of endemism as past centres of
biology. TrendsEcol. Evol.,15, 290–295.
cladogenesis (Croizat et al. 1974; Ricklefs & Schluter 1993).
Croizat, L., Nelson, G. & Rosen, D.E. (1974). Centers of origin
We believe that our results on the distribution of centres
and related concepts. Syst.Zool.,23, 265–287.
ofendemismhelptoelucidatetheroleofhistoryinshaping Currie, D.J. (1991). Energy and large-scale patterns of animal
avian distributions in Africa. We have illustrated how an species andplant species richness.Am. Nat.,137, 27–49.
analysis with contemporary environmental predictors can Currie, D.J., Francis, A.P. & Kerr, J.T. (1999). Some general pro-
help to support historical interpretations, although better positionsaboutthestudyofspatialpatternsofspeciesrichness.
Ecoscience, 6,392–399.
palaeoclimatic and phylogenetic information is still badly
DeKlerk,H.M.,Crowe,T.M.,Fjeldsa,J.&Burgess,N.D.(2002).
needed. Further, our results demonstrate that not even
Biogeographical patterns of endemic terrestrial Afrotropical
advanced information on environmental variables and
birds. Diversity Distributions,8, 147–162.
modelling techniques are likely to be sufficient to delineate Diamond, A.W. & Hamilton, A.C. (1980). The distribution of
areas and species of prime conservation concern. Greater forestpasserinebirdsandQuaternaryclimaticchangeinAfrica.
recognition of the value of primary ecological surveys is J. Zool.,191, 379–402.
needed. Dynesius,M.&Jansson,R.(2000).Evolutionaryconsequencesof
changes in species(cid:2) geographical distributions driven by Milan-
kovitchclimateoscillations.Proc.NatlAcad.Sci.USA,97,9115–
ACKNOWLEDGEMENTS 9120.
Endler, J.A. (1982a). Pleistocene forest refuges: fact or fancy? In:
WethankPaulH.Harvey,RobFreckleton,KevinJ.Gaston, BiologicalDivsersificationintheTropics(ed.Prance,T.G.).Columbia
David J. Rogers, James H. Brown, Ethan P. White and University Press, New York, NY, pp. 641–657.
Claire Kremen for comments on earlier versions of the Endler, J.A. (1982b). Problems in distinguishing historical from
manuscriptandStuartPimmforprovidinghelpfulfeedback ecological factors inbiogeography.Am.Zool.,22, 441–452.
Fielding, A.H. & Bell, J.F. (1997). A review of methods for
on modelling procedures. This work was supported in part
the assessment of prediction errors in conservation presence/
by studentships and fellowships from the UK Natural
absence models.Environ. Conserv.,24, 38–49.
Environment Research Council, German Scholarship
(cid:1)2004BlackwellPublishingLtd/CNRS