© 2006 The Authors DOI: 10.1111/j.1466-822x.2006.00281.x
Journal compilation © 2006 Blackwell Publishing Ltd www.blackwellpublishing.com/geb 205
Global Ecology and Biogeography, (Global Ecol. Biogeogr.) (2007) 16, 205–219
RESEARCH
PAPER
ABSTRACT
Aim The biodiversity of geometrid moths (Lepidoptera) along a complete tropical
elevational gradient was studied for the first time. The patterns are described, and
the role of geometric constraints and environmental factors is explored.
Location The study was carried out along the Barva Transect (10° N, 84° W), a
complete elevational gradient ranging from 40 to 2730 m a.s.l. in Braulio Carrillo
National Park, Costa Rica, and adjacent areas.
Methods Moths were sampled manually in 2003 and 2004 at 12 rain forest sites using
light ‘towers’, each with two 15 W ultraviolet fluorescent tubes. We used abundancebased
rarefaction, statistical estimation of true richness (Chao 1), geographically
interpolated observed richness and Fisher’s alpha as measures of local diversity.
Results A total of 13,765 specimens representing 739 species were analysed. All
four measures showed a hump-shaped pattern with maxima between 500 and
2100 m elevation. The two subfamilies showed richness and diversity maxima at
either lower (Ennominae) or higher (Larentiinae) elevation than Geometridae as a
whole. Among the four environmental factors tested, relative humidity yielded the
highest correlation over the transect with the rarefaction-based richness estimates as
well as with estimated true species richness of Geometridae as a whole and of Larentiinae,
while rainfall explained the greatest variation of Ennominae richness. The
elevational pattern of moth richness was discordant with both temperature and with
tree species richness. A combination of all environmental factors in a stepwise
multiple regression produced high values of r2
in Geometridae. The potential effects
of geometric constraints (mid-domain effect, MDE) were investigated by comparing
them with observed, interpolated richness. Overall, models fitted very well for
Geometridae as a whole and for Ennominae, but less well for Larentiinae. Smallranged
species showed stronger deviations from model predictions than largeranged
species, and differed strikingly between the two subfamilies, suggesting that
environmental factors play a more pronounced role for small-ranged species. We
hypothesize that small-ranged species (at least of the Ennominae) may tend to be
host specialists, whereas large-ranged species tend to be polyphagous. Based on
interpolated ranges, mean elevational range for these moths was larger with increasing
elevation, in accordance with Rapoport’s elevational rule, although sampling effects
may have exaggerated this pattern. The underlying mechanism remains unknown
because Rapoport’s ‘rescue’ hypothesis could not explain the observed pattern.
Conclusions The results clearly show that moth diversity shows a hump-shaped
pattern. However, remarkable variation exists with regard to taxon and range size. Both
environmental and geometric factors are likely to contribute to the observed patterns.
Keywords
Barva Transect, Costa Rica, diversity, Geometridae, La Selva, Lepidoptera, MDE,
rain forest, Rapoport’s rule.
Blackwell Publishing Ltd
The role of environment and mid-domain
effect on moth species richness along a
tropical elevational gradient
Gunnar Brehm,1
†* Robert K. Colwell2
and Jürgen Kluge3
1
Department of Animal Ecology I, University of
Bayreuth, Universitätsstraße 30, D-95440
Bayreuth, Germany, 2
Department of Ecology
and Evolutionary Biology, University of
Connecticut, Storrs, Connecticut 06269-3043,
USA, and 3
Department of Systematic Botany,
University of Göttingen, Untere Karspüle 2,
37073 Göttingen, Germany
*Correspondence: Gunnar Brehm, Institut für
Spezielle Zoologie und Evolutionsbiologie mit
Phyletischem Museum, Erbertstraße 1,
07743 Jena, Germany.
E-mail: Gunnar_Brehm@yahoo.com
†Present address: Institut für Spezielle
Zoologie und Evolutionsbiologie mit
Phyletischem Museum, Erbertstraße 1, 07743
Jena, Germany.
G. Brehm et al.
© 2006 The Authors
206 Global Ecology and Biogeography, 16, 205–219 Journal compilation © 2006 Blackwell Publishing Ltd
INTRODUCTION
It is well known that the diversity of organisms on Earth is not
uniformly distributed along latitudinal and altitudinal gradients.
Species richness in most groups of organisms peaks at low
latitudes and decreases towards the poles (e.g. Gaston, 2000;
Willig et al., 2003). Along elevational gradients, hump-shaped
and monotonically decreasing patterns are most frequently
observed (Rahbek, 2005). However, the underlying mechanisms
are still poorly explored. Many factors (biotic, abiotic and historical)
have been discussed that may be responsible for elevational
patterns of species richness (for reviews see McCoy,1990; Rahbek,
1995, 2005). In recent years, geometric constraints have played
an increasing role in their interpretation (Colwell & Hurtt, 1994;
Willig & Lyons, 1998; Colwell & Lees, 2000; Colwell et al., 2004).
The random placement of species geographical ranges within a
bounded geographical domain produces a hump-shaped pattern
of species richness. The effect has been termed the mid-domain
effect (MDE), and the concept has been critically discussed by
various authors (e.g. Hawkins & Diniz-Filho, 2002; Zapata et al.,
2003, 2005; Colwell et al., 2004, 2005; Hawkins et al., 2005).
Although the MDE cannot provide biological explanations for
the causes of the distribution of individual taxa, it shows that
hump-shaped patterns (rather than uniform distributions) of
species richness within geographical domains are appropriate
null models.
Although many studies of species richness and diversity along
elevational gradients have been published, they show a strong
bias towards plants and the temperate zones. However, the global
majority of terrestrial organisms are tropical arthropods, and
knowledge of their richness patterns along altitudinal gradients
is still very poor. The reasons for this bias are: (1) accessible,
complete elevational gradients in intact habitats are scarce and
(2) arthropods present a methodological challenge because of
their often extreme richness in tropical regions and associated
difficulties in sorting and identification (Novotny & Basset, 2000;
Longino et al., 2002). In this paper we analyse a unique data set
that covers, at the species level, a complete tropical gradient
(fromnearsealeveltonearthetopof amountain)of averyspeciesrich
group of insects.
Geometrid moths have been used as a model group in a
number of recent ecological studies along habitat gradients in
various regions in the world (e.g. Beck et al., 2002; Brehm et al.,
2003a,b, 2005; Holloway & Intachat, 2003; Axmacher et al.,
2004). Previous studies found diversity maxima at medium
elevations (Indo-Australia: Holloway, 1987) or a diversity plateau
along a broad elevational range (Ecuador: Brehm et al., 2003b).
Although rich in species,provisional identification of geometrids
is usually possible and leads to reliable estimations of species
richness (Brehm et al., 2005). Until this study, no complete elevational
gradient in mature tropical forest habitats from sea level to
a mountaintop had been explored for this or any other group of
moths. The Barva Transect, which lies between near sea level and
2900 m a.s.l. on the Atlantic slope of the Cordillera Central in
Costa Rica, is the last remaining continuous gradient of mature
forest in Central America to cover such a wide elevational range
(Lieberman et al., 1996; Blake & Loiselle, 2000). Several studies
on the diversity of plants and animals have been conducted along
this gradient that allow a comparison of available data, and in
selected cases are adequate for regression analysis (see Materials
and Methods).
First, we investigated the pattern of geometrid species richness
along the Barva Transect. Secondly, we examined the explanatory
potential for elevational patterns of moth diversity of four environmental
factors for which data are available: humidity, rainfall,
temperature and tree diversity. In addition to such ecological
factors,stochastic MDE models were used to assess the explanatory
potential of geometric constraints on range location. To date,
elevational MDE models for arthropods have been explored only
for ants (Sanders, 2002) and butterflies (Lees et al., 1999), but
not previously for a species-rich group of moths. Thirdly, we
investigated ‘Rapoport’s elevational rule’, i.e. whether the average
range size of species increases with elevation as suggested by
Stevens (1992). So far, only two studies have examined evidence
for the rule along elevational gradients in insects (Fleishman
et al., 1998; Sanders, 2002), but no study has ever investigated a
complete elevational gradient. Our analyses were carried out for
all geometrid species recorded in this study, and, in addition, for
the subsets of species belonging to the two largest subfamilies:
Ennominae and Larentiinae.The focus of this paper is large-scale
richness patterns and their potential determinants. Analyses of
beta diversity, landscape diversity and seasonal patterns will be
published elsewhere (Brehm,in press; G.Brehm,unpublished data).
MATERIALS AND METHODS
Study area and elevational gradients
The study area is situated on the Atlantic slopes of Volcán Barva
within the Cordillera Central of Costa Rica. The Barva Transect
ranges from lowland tropical rain forests at c. 30 m to montane
rain forests near the summit of the dormant Volcán Barva at
2906 m a.s.l. Old growth forests dominate the vegetation, but old
secondary forests occur within the study area (Blake & Loiselle,
2000). The gradient spans over c. 35 km (Fig. 1). The 12 sampling
sites were situated at elevations between 40 and 2730 m a.s.l. in
Parque Nacional Braulio Carrillo (PNBC) or in adjacent areas,
including La Selva Biological Station (Fig. 1; elevations Table 1).
Virtually every environmental factor changes along an elevational
gradient as extensive as the one analysed in this study. Unfortunately,
quantitative data are available for only a few environmental
variables along the Barva Transect, and regression analysis was
restricted to those factors that might be expected to have an
influence on insect species richness or distributions. The four
factors analysed were: humidity, rainfall, temperature and tree
species diversity. Humidity and rainfall were expected to have
mainly indirect effects on moth richness via the vegetation, but
both factors are likely to play a significant role for moth flight
behaviour. Temperature was generally expected to be positively
correlated with moth species richness,because higher temperatures
usually allow faster larval development.As many geometrid species
feed on woody plants (Scoble, 1999; Brehm, 2002), a positive
Moth richness along an elevational gradient
© 2006 The Authors
Global Ecology and Biogeography, 16, 205–219 Journal compilation © 2006 Blackwell Publishing Ltd 207
correlation between tree and moth richness could reasonably be
expected. Original environmental data were extracted from
Kluge et al. (2006) and tree data from Lieberman et al. (1996).
Vegetation
In their botanical study along the Barva Transect, Lieberman
et al. (1996) found that woody species density (≥ 10 cm d.b.h.)
peaked at 300 m a.s.l. with 149 species ha−1
and decreased to 29
species ha−1
at 2600 m a.s.l. Rarefied to 400 stems to correct for
variation in stem density (Gotelli & Colwell, 2001), tree species
richness peaked at about 130 species at 300–500 m a.s.l., decreasing
to 25 species at 2600 m a.s.l. (Cardelús et al., 2006). To estimate
values at the elevations of the moth sampling sites, these rarefied
data for tree species richness were interpolated between adjacent
pairs of sites studied by Lieberman et al. (1996) (by fitting a
polynomial function to their data over the complete transect).
Tree height ranged between 38 and 47 m at 300 m a.s.l. and 22
and 28 m at 2600 m a.s.l. Among all plants investigated, dicot
trees dominated in all plots. Palms and lianas were relatively
common and species-rich at the lower sites but became rare or
absent at higher elevations (Chazdon, 1987), whereas tree ferns
and hemi-epiphytes mostly occurred at medium elevations.
Lieberman et al. (1996) found no discontinuities or evidence for
discrete vegetation zones along the gradient. Further quantitative
studies on particular plant taxa or life forms along the gradient
are still scarce, and would be required for the analysis for specific
moth taxa, e.g. the host plant relationship between moths of the
genus Eois and plants of the genus Piper. Studies on ferns (not
known to be used by neotropical geometrids) and epiphytic
plants have recently been carried out (Cardelús et al., 2006; Kluge
et al., 2006; Watkins et al., 2006) but botanical data from these
recent studies were not included in the analyses in this paper.
Climate
Mean annual rainfall at La Selva is c. 4000 mm (Sanford et al.,
1994). Rainfall increases orographically towards medium elevations
and might reach as much as 8000 mm year−1
at 700 m
a.s.l. (Lieberman et al., 1996). It decreases at higher elevations to
c. 3300 mm year−1
in the summit area (Sacramento; Hartshorn &
Peralta, 1988). Precipitation might be underestimated at high
elevations because of cloud-driven precipitation (McCain,
2004), but it is evident that rainfall, per se, shows a peak at
medium elevations. Figure 2(b) shows the rainfall values that
were used for our regression analyses, which were compiled from
the Instituto Meteorológico Nacional (IMN) in San José, Costa
Rica (Kluge et al., 2006) and from Heaney & Proctor (1990)
(2260 m a.s.l.).
Relative air humidity and temperatures were measured by
Kluge et al. (2006) at four points along the Barva Transect (40,
650, 1800 and 2800 m a.s.l.) from July 2002 to November 2003
using 27 Microdaq data loggers HoboPro RH/Temp. Data were
measured, whenever possible, in different vegetation types (zonal
forest, ravine forest, ridge forest) at each recording elevation.
Data loggers were positioned near tree trunks in order to avoid
direct sunlight. Data were recorded at intervals of 10 min at
heights of 50 cm, 200 cm and as high as possible (10–15 m) in the
inner canopy. Data were subsequently averaged to 1-h intervals.
By taking data from such a wide range of forest types and tree
heights, we sought to minimize possible biases in the humidity
data due to small-scale microclimatic differences. We chose daily
minimum values of humidity because maximum values were
always > 99%, and minimum humidity appeared to be more
likely to constrain moth flight physiology.
Figure 2(a) and (c) show the estimated mean daily minimum
values of air humidity and average mean temperature (linearly
interpolated between pairs of measurement sites). Kluge et al.
(2006) estimated the temperature lapse rate on the gradient to be
Figure 1 The Barva Transect study area in Costa Rica (Heredia
province). The 12 study sites at six elevational levels are situated at
elevations between 40 and 2730 m a.s.l. NP = National Park.
G. Brehm et al.
© 2006 The Authors
208 Global Ecology and Biogeography, 16, 205–219 Journal compilation © 2006 Blackwell Publishing Ltd
5.5 K per 1000 m. Data from Lieberman et al. (1996) as well as
temperature measurements during moth collection (G. Brehm,
unpublished data) correspond well with these data.
Other environmental factors
A number of other environmental factors change along the
elevational gradient. Quantitative data are available only for a
few, and not every factor is appropriate to explain insect diversity
patterns. For example, soil parameters such as total nitrogen and
total carbon increase with altitude in the gradient investigated
(Lieberman et al., 1996), but they are not expected to have a
direct impact on the species richness of moths.Although desirable,
the incorporation of available leaf nitrogen in the analysis
was hampered by the unavailability of such data along the
Barva transect as well as by current ignorance of the larval food
requirements of the vast majority of geometrid species.Ecosystem
productivity has often been discussed as a determinant of species
Table 1 Elevation of 12 sampling sites, number of individuals collected, recorded species numbers, and Fisher’s alpha of Geometridae and the
subfamilies Ennominae and Larentiinae
Elevation
(m)
Geometridae Ennominae Larentiinae
Individuals
Observed
species
Fisher’s alpha
± SD Individuals
Observed
species
Fisher’s alpha
± SD Individuals
Observed
species
Fisher’s alpha
± SD
40 424 88 33.8 ± 2.7 210 41 15.2 ± 1.7 21 13 14.6 ± 6.0
45 400 90 36.1 ± 2.9 195 45 18.3 ± 2.1 7 5 7.8 ± 6.2
525 397 147 84.4 ± 6.8 210 73 39.7 ± 4.4 41 24 24.3 ± 6.9
550 528 149 69.1 ± 4.8 243 62 26.9 ± 2.7 48 25 21.0 ± 5.2
1070 1224 212 74.0 ± 3.5 656 105 35.3 ± 2.3 408 70 24.3 ± 2.0
1115 1409 239 82.6 ± 3.7 793 121 39.8 ± 2.4 445 72 24.3 ± 1.9
1690 1372 230 79.0 ± 3.6 736 105 33.5 ± 2.1 563 98 34.3 ± 2.4
1710 1471 261 92.2 ± 4.0 888 122 38.3 ± 2.2 514 107 41.1 ± 2.9
2090 1647 247 80.6 ± 3.4 913 96 27.1 ± 1.6 635 117 42.1 ± 2.7
2140 1700 225 69.5 ± 2.9 1034 94 25.1 ± 1.5 588 108 38.8 ± 2.6
2725 1286 132 36.9 ± 1.9 613 53 13.9 ± 1.1 602 67 19.3 ± 1.4
2730 1907 133 32.5 ± 1.5 820 54 13.0 ± 0.9 1032 70 17.0 ± 1.1
13,765 739 167.1 ± 3.1 7311 325 69.7 ± 1.9 4904 248 55.1 ± 1.7
Figure 2 Environmental factors used for
regression analysis. Original data for (a)
humidity, (b) rainfall and (c) temperature
were extracted from Kluge et al. (2006).
Tree species data (d) were compiled from
Lieberman et al. (1996), and rarefied to
equalize the number of individual stems
sampled for richness. Data were interpolated
to estimate values at the elevations of the moth
sampling sites.
Moth richness along an elevational gradient
© 2006 The Authors
Global Ecology and Biogeography, 16, 205–219 Journal compilation © 2006 Blackwell Publishing Ltd 209
richness (species–energy theory; Rosenzweig, 1995; Whittaker
et al., 2001) but forest productivity estimates are not available for
the Barva Transect.
Moth sampling and identification
Moths were sampled manually using light traps (Fritz Weber,
Stuttgart, Germany) with two 15 W UVA tubes (Sylvania, blacklight
blue, F15W/T8/BLB and Sylvania blacklight F15W/BL350)
located in their centres (Brehm & Axmacher, 2006). The diameter
of the trap was c. 0.70 m and its height was c. 1.70 m. Sampling
was carried out from 18:30 to 21:30, to coincide with peak
activity of moths. Nights with bright moonlight were avoided
(Brehm & Axmacher, 2006). Twelve sites at six elevational
levels were selected (Fig. 1) and sampled at least twice in
the period April–June 2003, and at least twice in the period
February–March 2004.
Much greater numbers of moths per night were collected at
higher elevations than at lower sites (see Appendix S1 in Supplementary
Material), possibly due to reduced predation pressure
by ants, bats and birds. As a consequence, collecting the same
number of samples at each site would have produced highly
unequal numbers of individuals at each site, thereby biasing
observed richness (Gotelli & Colwell, 2001) as well as observed
elevational ranges (Colwell & Hurtt, 1994). In an effort to reduce
the disparity among sites in numbers of individuals collected,
additional sampling was carried out at the three lowest elevational
levels, resulting in a total of between four and eight samples from
each site. These samples yielded a minimum number of 397
specimens (site 3) and a maximum number of 1907 (site 12;
Table 1). In other words, despite double sampling effort at the
lowest sites, numbers of individuals at these sites were still less
than half the numbers at high-elevation sites.
Numbers of individuals and species collected per sampling
night may vary, for example depending on the weather and
moon conditions (e.g. Yela & Holyoak, 1997; Brehm & Axmacher,
2006). However, the applied sampling scheme very likely provides
reliable data, given the replication at each elevational level
and the subsequent careful statistical analysis of the results.
Appendix S1 in Supplementary Material provides a detailed
list of all sampling dates and the numbers of moths sampled on
each date. The moths were identified by the first author by comparison
with original type specimens or other reliably identified
material in several museums (see Acknowledgements). The
nomenclature follows Pitkin (2002) for the largest subfamily,
Ennominae, and otherwise follows Scoble’s (1999) catalogue.
A full species list is provided in Appendix S2 in Supplementary
Material.
Species richness and diversity measures
As discussed above, sample size (number of individual moths
collected) varied considerably among sites (Table 1) in spite of
efforts to equalize it. The number of species observed in
incompletely sampled communities increases nonlinearly with
sample size and is thus an unreliable measure of the true, local
species richness (Gotelli & Colwell, 2001). We used four distinct
approaches to cope with this problem. The first approach relied
on individual-based (classic) rarefaction (Gotelli & Colwell,
2001) to compare sites at comparable numbers of individuals
(which reduces the number of species to below the observed
richness for sites with more individuals in samples). We refer to
these estimates as rarefied richness. The second approach was to
use non-parametric statistical estimators of true local species
richness to reduce the bias of incomplete sampling (Chao, 2005).
We call these values estimated richness. The third approach was
based directly on the recorded species × site incidence data. On
the working assumption that each species’ range is continuous
along the transect, we indicated a species as being present at all
sampling sites between (and including) its lowest and highest
recorded presence, whether or not it was actually recorded at
intermediate sites. The resulting richness values for each elevation
are referred to here as interpolated richness. For the fourth
approach, to allow comparison with other studies that use it, we
computed Fisher’s alpha, a widely used index of diversity that is
relatively insensitive to undersampling (Magurran, 2004). Rarefied
richness values were calculated using a program developed
by Kenney & Krebs (2000), at the level of 397 individuals in
Geometridae, 195 individuals in Ennominae and 41 individuals in
Larentiinae. Fisher’s alpha and estimated richness (Chao 1) were
calculated using  7.50 (Colwell, 2005a), and interpolated
richness was computed from interpolated species ranges using
 4 (Colwell, 2005b).
Geometric constraints
We used  4 (Colwell, 2005b) to estimate predicted
richness under the assumption of random placement of interpolated
moth elevational ranges. Incomplete sampling routinely
underestimates range size (Colwell & Hurtt, 1994), with the most
extreme case being species recorded from only a single elevation
(including many species in our study) which thus have an
observed elevational range of 0 m. Following the strategy of
Cardelús et al. (2006), we adjusted for range underestimation by
adding 293 m to each end of each recorded range (for all ranges),
half the maximum elevational distance between any two adjacent
sampling elevations. For each species recorded at the lowest
sites (40 and 45 m a.s.l.), however, we added only 40 or 45 m to
the downhill end of the range (and 293 m to the uphill end).
Likewise, for each species recorded at the highest sites (2725 and
2730 m), we added only 171 or 176 m to the uphill end of the
range, because the top of the mountain lies at 2906 m a.s.l. Range
augmentation also adds realism in another way: without it,
single-site species would otherwise have been ‘lost’ between
sampling elevations during the randomization of the midpoints
and would thus have failed to contribute to expected patterns.
Because none of the augmented ranges extends to an adjacent
sampling elevation (considering the paired sampling sites in
Table 1 each as a single elevation, e.g. 525 m and 550 m), augmentation
has no effect on the pattern of interpolated richness.
All species ranges and midpoints are provided in Appendix S2 in
Supplementary Material.
G. Brehm et al.
© 2006 The Authors
210 Global Ecology and Biogeography, 16, 205–219 Journal compilation © 2006 Blackwell Publishing Ltd
 4 systematically reassigns the location of each of
the interpolated, augmented ranges within the domain (0–
2906 m) at random (sampling without replacement) and then
records the predicted richness at each of the 12 field sampling
points.After a specified number of randomizations (we specified
1000),  reports the mean richness and its 95%
confidence interval (CI) for each sampling point (inspired by
McCain, 2003).
Range interpolation assumes that a species observed at two
different elevational levels is present or at least potentially
present everywhere between these levels, i.e. it assumes continuous
species distributions, as commonly done in macroecology. As a
consequence, interpolated species richness is somewhat greater
than recorded species richness, particularly in the middle of
the gradient where interpolated species ranges are more likely
to overlap (Grytnes & Vetaas, 2002). To assess whether this
effect of interpolation might have produced artefactual results,
we regressed mean predicted species richness (generated by
) against: (1) rarefied species numbers, (2) extrapolated
species richness (Chao 1),as well as against (3) interpolated
species richness. To explore the effect of range size on MDE
predictions (Colwell et al., 2004), we split each of the three taxa
(Geometridae, Ennominae and Larentiinae) into two classes, the
50% of species with the largest elevational ranges and the 50%
of species with the smallest elevational ranges (Lees et al., 1999;
Jetz & Rahbek, 2002), and repeated the same analyses that we
conducted on the full geometrid data set.
To explore the potential of individual factors (humidity,
rainfall, temperature, tree species and geometric constraints) to
explain elevational patterns of species richness, we performed
simple ordinary least squares (OLS) regressions of species
richness (rarefied, estimated or interpolated) for each taxon and
range-size moiety against each of these factors. We report r2
for each regression. Because geographical data such as these are
generally spatially autocorrelated, data values for nearby points
are not statistically independent, so that degrees of freedom are
inflated and ordinary P-values for r2
are underestimated.To correct
for spatial autocorrelation in regression residuals, we calculated
for each regression the effective number of degrees of freedom
according to Dutilleul’s method (Dutilleul, 1993; Legendre et al.,
2002), and we report adjusted P-values (Padj for r2
) based on the
effective degrees of freedom.
We also computed multiple linear regressions to explore
multivariate explanations for elevational patterns of moth species
richness. We selected the best model from among all possible
combinations of simple variables, guided by Akaike information
criterion (AIC) statistics (Burnham & Anderson, 1998). For
comparison with the best model, we also computed an environmental
model that included all four environmental factors, as
well as a geometric constraints model that included only MDE
predictions. Regression residuals were examined for spatial
autocorrelation based on Moran’s I for distance-classes. However,
because of the limited sample size (12 sites, spatially
arranged in six pairs), it did not prove feasible to apply spatial
autoregressive analyses with five explanatory variables. Thus no
P-values are reported for the multiple regressions. All regression
analyses were performed using  (Rangel et al., 2006), freely
available at www.ecoevol.ufg.br/sam.
Rapoport’s rule
We used  software (Colwell, 2005b) to compute
mean elevational range size among the species of Geometridae,
Ennominae and Larentiinae occurring at each of the 60 elevations
for interpolated range data (following the method of Stevens,
1992).
RESULTS
Species richness and diversity along the elevational
gradient
A total of 13,765 geometrid moths were sampled at 12 sites along
the elevational gradient (Table 1). They were sorted to 739
morphospecies of which 388 (53%) were identified to named
species. A large proportion of the remaining taxa are likely to be
new to science and need to be described taxonomically in the
future. A full list is provided in Appendix S1 in Supplementary
Material. Of the specimens sampled, 7311 (325 species) were
assigned to the subfamily Ennominae, 4904 (248 species) to
the subfamily Larentiinae, and the remainder to four other
subfamilies (Desmobathrinae, Geometrinae, Oenochrominae,
Sterrhinae). Geometrid moth richness and diversity clearly
showed a hump-shaped pattern along the elevational gradient.
While recorded species richness, estimated species richness and
interpolated species richness show a broad peak of richness
between c. 1100 and 2100 m a.s.l. (Table 1 and Fig. 3a,b), rarefied
richness and Fisher’s alpha indicate a broader hump of similar
richness and diversity between c. 500 and 2100 m (Table 1 and
Fig. 3a). All measures clearly show that richness and diversity of
geometrid moths decreases towards both ends of the elevational
gradient.
The subfamily Ennominae (325 species) largely reflects the
pattern of the whole family. However, diversity of ennomine
moths tends to peak at lower elevations compared with the
geometrids as a whole (between c. 500 and 1700 m a.s.l.,depending
on the measure). In contrast, species richness and Fisher’s alpha
for the Larentiinae (248 species) peak at very high elevations
between c. 1700 and 2100 m a.s.l. (Table 1 and Fig. 3) but decline
in the summit area of Volcán Barva as well as at lower elevations
(1100 m and below).
Regression on environmental factors
The upper portion of Table 2 shows r2
values and associated
P-values (based on degrees of freedom adjusted for spatial autocorrelation)
for simple linear regressions between three measures
of species richness as a function of four environmental factors
plus geometric constraints (MDE model predictions) for the
three taxa considered (Geometridae and its subfamilies Ennominae
and Larentiinae). The lower portion of Table 2 reports the results
of model selection for multiple regressions of richness measures
Moth richness along an elevational gradient
© 2006 The Authors
Global Ecology and Biogeography, 16, 205–219 Journal compilation © 2006 Blackwell Publishing Ltd 211
Figure 3 (a) Estimated species richness (Chao 1, triangles) and rarefied species numbers (squares) of Geometridae and the subfamilies
Ennominae and Larentiinae along an elevational gradient in Costa Rica. (b) Predicted species richness under the assumption of random range
placement (MDE) (central line and upper and lower 95% CI limits) and interpolated,observed species richness (circles) of Geometridae and two
subfamilies along an elevational gradient in Costa Rica. The first row shows richness for all species, the second row shows richness for the 50%
of species with the largest ranges, and the third row shows richness calculated for the 50% of species with the smallest ranges. See Table 2 for a
regression analysis comparing interpolated richness and predicted richness. Note that no rarefied species numbers were calculated for the two
lowest sites for Larentiinae because the number of individuals was very low (7 and 21). r2
values indicate the variance in interpolated, observed
species richness that is explained by predicted species richness. Padj indicates the P-value for r2
, based on degrees of freedom adjusted to account
for spatial autocorrelation using Dutilleul’s (1993) method.
G.Brehmetal.
©2006TheAuthors
212GlobalEcologyandBiogeography,16,205–219Journalcompilation©2006BlackwellPublishingLtd
Table 2 Simple OLS and multiple regression analysis of (a) rarefied species richness, (b) estimated species richness (Chao 1), (c) interpolated species richness (all species, 50% of species with largest
ranges, 50% of species with smallest ranges) vs. four environmental factors and geometric constraints (MDE model predictions) for three geometrid taxa. Padj (in italics) is the P-value for r2
, based
on degrees of freedom adjusted to account for spatial autocorrelation using Dutilleul’s (1993) method. Model selection (best model) for multiple regressions (which do not account for spatial
autocorrelation) was based on minimizing AIC, with consideration of all possible models. For comparison with the best model, r2
and AIC are also shown for a model incorporating geometric
constraints only (geometric constraints model) and a model with all environmental factors included and geometric constraints excluded (environmental model). Beta is the standardized regression
slope for each factor in the best model. Numbers of specimens were rarefied to 397 in Geometridae, to 195 in Ennominae and to 41 in Larentiinae (see Fig. 2). Negative relationships are indicated
by (–). Bold faced entries indicate significant r2
(Padj) ≤ 0.05 for individual tests
Geometridae Ennominae Larentiinae
Rarefied Chao 1
Interpolated
Rarefied Chao 1
Interpolated
Rarefied Chao 1
Interpolated
All
Large
ranges
Small
ranges All
Large
ranges
Small
ranges All
Large
ranges
Small
ranges
Simple OLS regression
Humidity (r2
) 0.38 0.52 0.69 0.80 0.28 (–) 0.17 0.35 0.57 0.63 0.08 (–) 0.44 0.80 0.73 0.76 0.47
Padj 0.107 0.070 0.058 0.036 0.291 0.218 0.104 0.063 0.099 0.532 0.051 0.036 0.041 0.040 0.142
Rainfall (r2
) 0.30 0.20 0.07 0.00 (–) 0.58 0.55 0.26 0.18 0.15 (–) 0.91 0.03 (–) 0.01 (–) 0.08 (–) 0.10 (–) 0.03 (–)
Padj 0.077 0.206 0.442 0.910 0.019 0.021 0.121 0.218 0.306 0.002 0.739 0.746 0.418 0.280 0592
Temperature (r2
) 0.01 0.01 (–) 0.10 (–) 0.35 (–) 0.89 0.11 0.0 (–) 0.03 (–) 0.55 (–) 0.68 0.01 (–) 0.401 (–) 0.53 (–) 0.62 (–) 0.21 (–)
Padj 0.799 0.783 0.511 0.254 0.003 0.303 0.957 0.697 0.136 0.020 0.847 0.216 0.153 0.107 0.349
Tree species (r2
) 0.12 0.01 0.03 (–) 0.22 (–) 0.95 0.31 0.01 0.00 (–) 0.46 (–) 0.78 0.01 (–) 0.25 (–) 0.41 (–) 0.49 (–) 0.13 (–)
Padj 0.319 0.868 0.724 0.316 < 0.001 0.114 0.768 0.959 0.109 0.005 0.869 0.268 0.148 0.107 0.372
Geometric constraints (r2
) 0.76 0.87 0.95 0.78 0.00 0.59 0.69 0.96 0.45 0.02 0.26 0.69 0.53 0.51 0.52
Padj 0.003 0.001 < 0.001 0.016 0.942 0.012 0.009 < 0.001 0.119 0.773 0.212 0.027 0.090 0.108 0.044
Multiple regression
Geometric constraints model 0.76 0.87 0.95 0.78 0.00 0.58 0.69 0.96 0.45 0.02 0.26 0.69 0.53 0.51 0.52
(r2
, AIC) 68.48 90.50 78.96 100.29 92.69 57.71 85.39 59.46 96.33 96.21 31.19 89.07 93.04 87.67 60.71
Environmental model
(r2
, AIC)
0.94 0.84 0.99 0.99 0.94 0.88 0.73 0.99 0.99 > 0.99 0.86 0.93 0.99 0.99 0.94
70.89 112.87 78.34 85.28 76.64 62.95 103.56 65.04 68.88 38.85 44.61 90.72 64.48 57.52 54.47
Best model (r2
, AIC) 0.94 0.87 0.99 0.97 0.93 0.88 0.69 0.96 0.99 > 0.99 0.44 0.83 0.99 > 0.99 0.99
62.14 90.50 (*)78.3 62.85 60.60 47.27 85.39 59.46 52.66 38.85 28.42 86.28 61.04 31.61 32.57
Humidity (beta) 1.88 2.28 −0.455 −0.769 0.662 0.660 −0.276 −0.215
Rainfall (beta) −0.74 −0.616 −0.131 −0.626 1.239 −0.410 −0.169 −0.787
Temperature (beta) 1.90 3.25 −1.277 −0.854
Tree species (beta) −1.97 −0.353 0.965 0.542 −0.361 0.741 −0.445 −0.559
Geometric constraints (beta) 0.933 (*) 0.935 0.760 0.831 0.978 1.29 0.290 1.132 0.776 2.681
(*) An alternative model with only slightly higher AIC (78.96), includes only Geometric constraints (beta = 0.975)
Moth richness along an elevational gradient
© 2006 The Authors
Global Ecology and Biogeography, 16, 205–219 Journal compilation © 2006 Blackwell Publishing Ltd 213
on the environmental variables and MDE predictions. In each
case, the model with the lowest AIC is reported (best model), as
well as a model with all four environmental factors (environmental
model) and a model with MDE predictions only (geometric
constraints model). The results in Table 2 should be considered
exploratory, rather than definitive, for four reasons: (1) despite
the massive effort invested in the field work for this study, the
sample size for these regressions is nonetheless quite small
(n = 12 sites), especially for multiple regression with five explanatory
variables; (2) spatial non-independence (autocorrelation)
between sites has been accounted for in the simple regressions,
but not in the multiple regressions; (3) unmeasured environmental
variables may be important; and (4) there is evidence
of substantial collinearity among explanatory variables [e.g.
temperature and rain with tree species richness (r2
= 0.86,
r2
= 0.71, respectively)] that cannot be confidently resolved
(Graham, 2003) with such a small sample size.
With these caveats, the results for simple regressions (upper
section of Table 2) show that the strength of the relationships
depended upon: (1) the environmental factor (or geometric constraints),
but also on (2) the taxon investigated, (3) the particular
measure of species richness used, and (4) the size of the species
ranges. For the Ennominae (and for the Geometridae as a
whole), as predicted by the MDE theory (Colwell & Lees, 2000;
Colwell et al., 2004), environmental factors tended to be more
important in explaining the richness patterns of small-ranged
species, whereas geometric constraints (MDE) were more
explanatory for large-ranged species. For small ranges in the
Ennominae, rainfall, temperature and tree species richness all
showed evidence of a positive relationship with richness. In
contrast, for small-ranged Larentiinae, the environmental factors
considered were relatively less important, whereas humidity
showed a positive relationship with richness of large-ranged
species (a positive relationship with humidity was also evident
for the large-ranged Geometridae as a whole). Large-ranged
species showed results similar to those for all species (within a
taxon), to which they contribute disproportionately because they
occur at many sites (Jetz & Rahbek, 2002; Colwell et al., 2004).
The strength of the relationships differed considerably depending
upon the measure of species richness, but generally showed the
same tendencies.
The multivariate results (lower section of Table 2) were
generally supportive of these patterns. As often happens with
multivariate regressions when multicollinearity is present,
certain factors tended to ‘capture’ much of the variance shared
with other factors, making rigorous causal interpretations
difficult. Examination of Moran’s I for distance-class residuals
from the multiple regressions for the models selected using the
AIC revealed substantial spatial autocorrelation in some of the
models, calling for additional caution in interpreting these
results.
Regression with MDE model data
Figure 3(b) shows MDE model predictions generated with interpolated
ranges (augmented as detailed in the Materials and
Methods) and randomized range placement. Predicted species
richness approximated interpolated richness quite closely for
Geometridae and Ennominae (all species; 0.95 ≤ r2
≤ 0.96,
Padj < 0.001), but considerably less well for Larentiinae (all
species; r2
= 0.53, Padj = 0.090). Even though r2
values are very
high, overall, interpolated species richness does not fall within
the 95% confidence limits at most individual sites. Interpolated
richness in Geometridae and Ennominae is always higher than
predicted richness at the ends of the gradient, and tends to be
lower in the middle of the gradient (Fig. 3b). In Larentiinae, the
interpolated richness curve is shifted towards higher elevations
compared with the predicted richness curve. The pattern of all
species is mirrored by the large-ranged species in Geometridae,
which dominate the information in the pattern for all species
(Fig. 3b). However, a remarkable shift of the curves occurs
in Ennominae and Larentiinae. Large-ranged species of both
families are clearly underrepresented at the lower elevations,
compared to MDE expectations, whereas interpolated species
richness is disproportionately high at high elevations. On the
contrary, small-ranged ennomines are very species rich at low
elevations and strongly decline at high elevations. As a consequence,
the correlation between interpolated species richness and
the MDE model predictions drops to only r2
= 0.02 (Padj = 0.733)
for small-ranged Ennominae, contributing to a similar pattern in
small-ranged Geometridae as a whole (Fig. 3b). Interpolated
richness of small-ranged Larentiinae still peaks at 1700 m, but
shows higher values at the lowest sites than large-ranged species.
Rapoport’s elevational rule
Figure 4(a) shows that average geometrid species range sizes
increase with increasing elevation (r2
= 0.69). The relationship is
almost linear in the Ennominae (Fig. 4b, r2
= 0.82). However, the
curve of the Larentiinae is asymptotic, reflected by a lower value
of r2
(Fig. 4c, r2
= 0.39). We report no P-values for these results
because the data points in Stevens plots are not independent
(Rohde et al., 1993).
DISCUSSION
Richness patterns and environment
It was long believed that species richness in insects shows a
monotonic decline along elevational gradients. Peaks at midelevations
have sometimes been regarded as sampling artefacts
(Wolda, 1987; McCoy, 1990). However, the paradigm has changed,
and peaks at medium elevations are generally accepted as being
the rule rather than the exception (for a review see Rahbek,
2005). Such peaks often occur at relatively low elevations and
have sometimes been overlooked when sampling was restricted
to higher elevations (Rahbek, 1995; Lomolino, 2001; Kluge et al.,
2006). Examples include groups such as butterflies and ants,
which show a maximum diversity in tropical regions far below
1000 m a.s.l. (DeVries, 1994; Brühl et al., 1999; Fisher, 2002).
However, evidence is still scarce because only a very few insect
studies have investigated complete elevational gradients.
G. Brehm et al.
© 2006 The Authors
214 Global Ecology and Biogeography, 16, 205–219 Journal compilation © 2006 Blackwell Publishing Ltd
So far, only a few exceptions to an overall declining diversity of
insects at elevations higher than 1000 m have been documented.
Examples include arctiid and geometrid moths in Borneo
(Holloway, 1997; Schulze, 2000) and Ecuador (Brehm et al.,
2003b). Diversity patterns of these taxa have not previously been
investigated along a tropical elevational gradient ranging from
sea level to a mountain top. The present study confirms that
geometrid moths have a predominantly montane distribution
with exceptionally high species richness at elevations up to
2100 m a.s.l. Richness at the lowest elevations is markedly lower,
and also decreases towards higher elevations at the mountain
summit.
Regression analyses of selected environmental factors as
carried out in this study are generally hampered by the fact that
whereas moth richness may be explained statistically by certain
environmental factors there may well be no causal relationship.
Moreover, because some environmental factors are correlated
with each other, it is difficult to interpret positive correlations
with richness. Finally, the set of available environmental factors is
limited, especially in poorly investigated tropical regions (such as
the Barva Transect, which is rugged and difficult to access) where
suitable data (e.g. host plant relationships, host plant richness,
leaf nitrogen, etc.) do not yet exist. The available data appeared
nevertheless to be an appropriate first set of factors for an exploratory
analysis.
The results suggest, not surprisingly, that a combination of
environmental factors may affect the richness patterns of
geometrid moths, rather than a single factor. Humidity is
strongly correlated with richness of large-ranged Geometridae
and Larentiinae, whereas rainfall and temperature are strongly
correlated with small-ranged ennomine richness. Humidity and
rainfall can have direct or indirect effects, via the vegetation,
on herbivorous insects. It remains to be shown whether or not
ambient humidity levels could act as a physiological constraint to
certain moth species.
Geometrid moths have frequently been regarded as predominantly
being feeders of trees and shrubs (e.g. Scoble, 1999;
Brehm, 2002). Moth species richness in this study is positively
correlated with tree species diversity for small-ranged Geometridae
and Ennominae, but not for large-ranged species in these taxa,
nor for Larentiinae. These positive correlations are obscured
when species of all range sizes are considered together. In a
montane forest in Ecuador, Brehm et al. (2003b) found that tree
species richness declined along an elevational gradient spanning
c. 1600 m, but total diversity of geometrid moths remained at a
constantly high level up to elevations of c. 2700 m a.s.l. Hence, in
both areas, total species richness in this herbivorous taxon
remains high despite the declining availability of potential host
plants at higher elevations. Given the results of this study, it is
tempting to suggest that small-ranged species (at least of
Ennominae and Geometridae as a whole) may be host specialists,
whereas large-ranged species may tend to be polyphagous.
However, this conclusion remains speculative as long as no further
host plant data are available for Neotropical geometrids.
Temperature is obviously an important environmental factor,
especially for ectothermic organisms like insects. Brehm et al.
Figure 4 Stevens plots (mean moth range size for 60 elevational
bands) along an elevational gradient in Costa Rica: (a) Geometridae,
(b) Ennominae, (c) Larentiinae. Rapoport’s elevational rule is
supported because the mean range size increases with increasing
elevation, although the pattern of variation in the number of
individuals sampled may exaggerate this effect.
Moth richness along an elevational gradient
© 2006 The Authors
Global Ecology and Biogeography, 16, 205–219 Journal compilation © 2006 Blackwell Publishing Ltd 215
(2003b) and Brehm & Fiedler (2003) discussed possible physiological
adaptations of geometrid moths (particularly Larentiinae)
to low flight temperatures. Geometrids clearly demonstrate that
not all insects prefer the warmest climates, although richness of
small-ranged Ennominae and Geometridae as whole showed a
significantly positive correlation with temperature. The decline
of richness around 2700 m a.s.l. in Costa Rica could be an effect
of reduced habitat area rather than a negative effect of environmental
factors such as temperature (Fig. 1 gives an impression
how limited this elevational belt is in total area). Brehm et al.
(2003b, 2005) showed that diversity of geometrids at the same
elevation in Ecuador was still as high as it was at lower elevations.
Richness declined only at the highest elevations in the summit
area (3100 m a.s.l.) of the mountain range studied.
Richness patterns and geometric constraints
Considered alone, MDE models explained rarefied richness,
estimated richness and interpolated richness of all geometrid
moth species (all range sizes considered together) better than any
of the available single environmental factors (Table 2). Overall,
the models fitted the interpolated richness pattern very well for
Geometridae and Ennominae. However, the models consistently
predicted lower species richness at the ends of the gradient,
and often higher species richness at medium elevations than
observed in nature. A qualitatively similar pattern of underestimation
of richness by the model at the ends of the domain was
found by Kluge et al. (2006) and Watkins et al. (2006) for ferns,
and by Cardelús et al. (2006) for epiphytes along the same
elevational gradient. An explanation for the underestimation
of species richness at the limits of the domain could be the onedimensionality
of such models and the conservative assumptions
of the randomization algorithm, which requires each randomly
placed range to fit entirely within the domain without the
dynamic adjustment permitted in related, two-dimensional
models (Jetz & Rahbek, 2002). In reality, species richness at the
domain limits will probably never be zero (Grytnes & Vetaas,
2002; Connolly, 2005).
As for the higher predicted richness than interpolated richness
in the middle of the domain, the absolute magnitude of predicted
richness is influenced by the magnitude of range size
augmentations (see Materials and Methods); the greater the augmentation,
the higher the peak richness, because larger ranges
tend to be sampled at more sites. We assessed the sensitivity
of results to this effect by reducing the range size augmentation
to about half (150 m on each end) the values specified in the
Materials and Methods section. With this change, mid-domain
richness peaks were actually lower than interpolated richness,
but the relative magnitude (the‘shape’of the curves) was virtually
unchanged, and the r2
for fit was identical. Only a higher density
of elevational sampling sites and long-term sampling could
reduce the uncertainty regarding the appropriate degree of range
augmentation (which is necessitated by species that occur in only
one site) in future studies.
The separate analysis of the large subfamilies Ennominae and
Larentiinae shows that a good fit of the model for the whole
group is not necessarily reflected by all subordinate taxa. Richness
of larentiine moths is correlated more weakly with the model
richness predictions because the larentiine curve is ‘shifted’
towards higher elevations. Hence, richness of larentiines is likely
to be strongly driven by other than stochastic factors. Possible
causes for the richness of Larentiinae moths in high montane
habitats were discussed by Brehm & Fiedler (2003). Larentiines
such as winter moths could be particularly well adapted to cool
temperatures, allowing the colonization of montane forests.
Despite their potentially unfavourable climatic conditions, these
habitats could be advantageous in terms of reduced predation
pressure from ants, and insectivorous bats and birds.
Jetz & Rahbek (2002) and Colwell et al. (2004) argued that
large-ranged species were more likely to be constrained by geometry,
whereas small-ranged species were more likely to show the
effects of underlying environmental or historical drivers. Our data
strongly support this concept statistically for the Geometridae as
a whole and for the Ennominae, for which interpolated richness
of large-ranged species is much more strongly correlated
(r2
≥ 0.95) with MDE model predictions than interpolated
richness of small-ranged species, which is virtually uncorrelated
with MDE predictions (0.00 ≤ r2
≤ 0.02) (Fig. 3b). In fact, the
strong correlation between humidity and the richness of largeranged
(but not small-ranged) Geometridae and Larentiinae
(Ennominae was nearly significant, at Padj = 0.063) may actually
reflect the strong collinearity between humidity and MDE
predictions for large-ranged species, rather than an effect of
humidity itself on species richness.
In contrast, the strength of the correlation for Larentiinae was
largely independent of range size, though significant only for
small-ranged species and for estimated richness of all Larentiinae.
The patterns of deviation of interpolated richness curves from
model predictions, however, provide some of the most interesting
results. Both large-ranged ennomine and large-ranged
larentiine species showed a modest shift towards higher elevations
compared with MDE model predictions (Fig. 3b), whereas
the pattern of richness for small-ranged species in the two subfamilies
revealed a striking contrast between the two subfamilies:
small-ranged ennomine richness was strongly focused at
lower elevations whereas small-ranged larentiine richness was
concentrated at higher elevations. These contrasting patterns for
large-ranged vs. small-ranged species suggest that richness
patterns for these two rich groups are likely to have been driven
by completely different factors, depending upon range size, and
that the two subfamilies might have different biogeographical
origins — the ennomines at low to medium elevations and the
larentiines at high elevations.
Rapoport’s elevational rule
Rapoport’s rule (Stevens, 1989) states that species ranges become
larger with increasing latitude. Stevens (1992) extended the rule
to elevation as an explanation for monotonic decreases in species
richness with increasing elevation. So far, existing studies suggest
that the rule has no general applicability (Gaston et al., 1998),
but very little evidence exists for insects. The present study finds
G. Brehm et al.
© 2006 The Authors
216 Global Ecology and Biogeography, 16, 205–219 Journal compilation © 2006 Blackwell Publishing Ltd
a Rapoport effect in all three analysed taxa, based on interpolated
ranges (Fig. 4). When species richness varies among sites, and
equal numbers of individuals are sampled but the inventory is
incomplete at all sites, ranges tend to be underestimated at richer
sites, producing a spurious Rapoport effect (Colwell & Hurtt,
1994). Although, in our study, local richness was not correlated
with range size,the number of individuals collected (Table 2) was
rather strongly correlated with mean range size (Fig. 4) among
elevations. For this reason, the observed Rapoport effect is
probably exaggerated, and could even be entirely artefactual.
The same effect may shift curves for observed, large-ranged
species richness (Fig. 3b) to somewhat higher elevations than
equal sampling would produce. Individual-based rarefaction
of the species × elevations matrix to produce equal numbers of
individuals at each elevation might resolve these uncertainties,
but is beyond the scope of this paper.
As for the effect of area at different elevations, on a regional
scale, lowland areas dominate the Caribbean slopes of Costa
Rica as well as the study area, and area sharply declines with
increasing elevation (Kluge et al., 2006). Thus, the observed
hump-shaped richness pattern obviously cannot be positively
correlated with area.
To the degree that our Rapoport results are genuine, this study
confirms the findings of the only two existing insect studies on
Rapoport’s elevational rule that were carried out in North America
on butterflies (Fleishman et al., 1998) and ants (Sanders, 2002).
Both found the same combination of a mid-elevational richness
peak and Rapoport’s elevational rule, although no complete
gradients were investigated. The mechanism behind the rule is
still unknown, and different factors might be involved in different
groups of organisms. For example, the Rapoport ‘rescue’ hypothesis
(Stevens, 1992) suggests that low-elevation populations
are relatively intolerant of environmental stochasticity because
short-term climatic variability is positively correlated with
elevation. Hence, species that inhabit higher elevations must
have larger climatic tolerances and thus can be found along a
greater elevational range. Stevens (1992) used this hypothesis
to explain monotonically decreasing species richness patterns.
However, because species richness of geometrid moths is humpshaped,
the Rapoport ‘rescue’ hypothesis must be rejected.
Our results suggest that both stochastic and environmental
factors are likely to be responsible for the observed richness
peaks of geometrid moths at medium elevations. No single key
factor could be found to explain all patterns because marked
differences occurred between different taxa and between species
with different range sizes. An improved knowledge of the life
histories of species-rich tropical invertebrates as well as more
environmental data will be required in the future for a better
understanding of their richness patterns.
ACKNOWLEDGEMENTS
We thank three anonymous referees and Vojtech Novotny for
their constructive comments on previous drafts of this paper.
The Organization for Tropical Studies and the ALAS project
(Arthropods of La Selva) provided excellent working facilities at
La Selva Biological station and supported the work in the
adjacent Braulio Carrillo National Park. Martin Knopf, Vernon
Sota and in particular Edgar Corrales supported the field work.
The Instituto Nacional de Biodiversidad (Alvaro Herrera) and
the Ministerio del Ambiente y Energia, Sistema Nacional de
Areas de Conservacion helped to obtain research permits.Various
colleagues helped to access museum collections for the identification
of moths, particularly Martin R. Honey, Linda M. Pitkin,
Malcolm J. Scoble (Natural History Museum, London), John W.
Brown, Patricia Gentili-Poole (United States Museum of Natural
History) and Suzanne Rab-Green (American Museum of Natural
History). Falk Hüttmann calculated area–elevation relationships
for Costa Rica, and Claudia Raabe and Claudia Knake assisted
with spreading moths and data input. The Deutsche Forschungsgemeinschaft
and the National Science Foundation supported
this study (GB, BR 2280/1-1; RKC, DEB-0072702). Michael
Kessler and the Deutscher Akademischer Austauschdienst greatly
supported JK.
REFERENCES
Axmacher, J.C., Holtmann, G., Scheuermann, L., Brehm, G.,
Müller-Hohenstein, K. & Fiedler, K. (2004) Diversity of
geometrid moths (Lepidoptera: Geometridae) along an
Afrotropical elevational rainforest transect. Diversity and
Distributions, 10, 293–302.
Beck, J., Schulze, C.H., Linsenmair, K.E. & Fiedler, K. (2002)
From forest to farmland: diversity of geometrid moths along
two habitat gradients on Borneo. Journal of Tropical Ecology,
17, 33–51.
Blake, J.G. & Loiselle, B.A. (2000) Diversity of birds along an
elevational gradient in the Cordillera Central, Costa Rica. The
Auk, 117, 663–686.
Brehm, G. (2002) Diversity of geometrid moths in a montane rainforest
in Ecuador. PhD Dissertation, University of Bayreuth,
Bayreuth, Germany (http://opus.ub.uni-bayreuth.de/volltexte/
2003/20/).
Brehm, G. (in press) Diversity of geometrid moths in two
Neotropical montane rain forests. Forests in the mist. Science
for conserving and managing tropical montane cloud forests
(ed. by L.A. Bruijnzeel, J. Juvik, F.N. Scatena, L.S. Hamilton
and P. Bubb). University of Hawaii Press, Honolulu.
Brehm,G. & Axmacher,J.C. (2006) A comparison of manual and
automatic moth sampling methods (Lepidoptera: Arctiidae,
Geometridae) in a rain forest in Costa Rica. Environmental
Entomology, 35, 757–764.
Brehm, G. & Fiedler, K. (2003) Faunal composition of geometrid
moths changes with altitude in an Andean montane rain forest.
Journal of Biogeography, 30, 431–440.
Brehm, G., Homeier, J. & Fiedler, K. (2003a) Beta diversity of
geometrid moths (Lepidoptera: Geometridae) in an Andean
montane rainforest. Diversity and Distributions, 9, 351–
366.
Brehm, G., Süssenbach, D. & Fiedler, K. (2003b) Unique elevational
diversity patterns of geometrid moths in an Andean
montane rainforest. Ecography, 26, 356–366.
Moth richness along an elevational gradient
© 2006 The Authors
Global Ecology and Biogeography, 16, 205–219 Journal compilation © 2006 Blackwell Publishing Ltd 217
Brehm, G., Pitkin, L.M., Hilt, N. & Fiedler, K. (2005) Montane
Andean rain forests are a global diversity hotspot of geometrid
moths. Journal of Biogeography, 32, 1621–1627.
Brühl, C., Mohamed, M. & Linsenmair, K.E. (1999) Altitudinal
distribution of leaf litter ants along a transect in primary forest
on Mount Kinabalu, Sabah, Malaysia. Journal of Tropical Ecology,
15, 265–277.
Burnham, K.P. & Anderson, D.R. (1998) Model selection and
inference: an information-theoretic approach. Springer, New
York.
Cardelús, C., Colwell, R.K. & Watkins, J.E., Jr (2006) Vascular
epiphyte distribution patterns: explaining the mid-elevation
richness peak. Journal of Ecology, 94, 144–156.
Chao, A. (2005) Species richness estimation. Encyclopedia of
statistical sciences (ed. by N. Balakrishnan, C.B. Read and B.
Vidakovic), pp. 7909–7916. Wiley, New York.
Chazdon, R.L. (1987) The palm flora of Braulio Carrillo
National Park, Costa Rica. Brenesia, 28, 107–116.
Colwell, R.K. (2005a) EstimateS: statistical estimation of species
richness and shared species from samples, version 7.50 (http://
purl.oclc.org/estimates).
Colwell, R.K. (2005b) RangeModel: a Monte Carlo simulation tool
for visualizing and assessing geometric constraints on species
richness, version 4 (http://viceroy.eeb.uconn.edu/rangemodel).
Colwell, R.K. & Hurtt, G.C. (1994) Nonbiological gradients in
species richness and a spurious Rapoport effect. The American
Naturalist, 144, 570–595.
Colwell, R.K. & Lees, D.C. (2000) The mid-domain effect:
geometric constraints on the geography of species richness.
Trends in Ecology & Evolution, 15, 70–76.
Colwell, R.K., Rahbek, C. & Gotelli, N.J. (2004) The mid-domain
effect and species richness patterns: what have we learned so
far? The American Naturalist, 163, E1–E23.
Colwell, R.K., Rahbek, C. & Gotelli, N.J. (2005) The mid-domain
effect: there’s a baby in the bathwater. The American Naturalist,
166, E149–E154.
Connolly, S.R. (2005) Process-based models of species distributions
and the mid-domain effect. The American Naturalist,
166, 1–11.
DeVries, P.J. (1994) Patterns of butterfly diversity and promising
topics in natural history and ecology. La Selva: ecology and
natural history of a neotropical rain forest (ed. by A. McDade,
K.S. Bawa, H.A. Hespenheide and G.S. Hartshorn), pp. 187–
194. The University of Chicago Press, Chicago.
Dutilleul, P. (1993) Modifying the t test for assessing the correlation
between two spatial processes. Biometrics, 49, 305–
312.
Fisher, B.L. (2002) Ant diversity patterns along an elevational
gradient in the Reserve Speciale de Manongarivo, Madagascar.
Boissiera, 59, 311–328.
Fleishman, E., Austin, G.T. & Weiss, A.D. (1998) An empirical
test of Rapoport’s rule: elevational gradients in montane
butterfly communities. Ecology, 79, 2482–2493.
Gaston, K.J. (2000) Global patterns in biodiversity. Nature, 405,
220–226.
Gaston, K.J., Blackburn, T.M. & Spicer, J.I. (1998) Rapoport’s
rule: time for an epitaph? Trends in Ecology & Evolution, 13,
70–74.
Gotelli, N.J. & Colwell, R.K. (2001) Quantifying biodiversity:
procedures and pitfalls in the measurement and comparison of
species richness. Ecology Letters, 4, 379–391.
Graham,M.H. (2003) Confronting multicollinearity in ecological
multiple regression. Ecology, 84, 2809–2815.
Grytnes, J.A. & Vetaas, O.R. (2002) Species richness and altitude:
a comparison between null models and interpolated plant
species richness along the Himalayan altitudinal gradient,
Nepal. The American Naturalist, 159, 294–304.
Hartshorn, G.S. & Peralta, R. (1988) Preliminary description of
primary forests along the La Selva-Volcan Barva altitudinal
transect, Costa Rica. The tropical rainforests: diversity and
conservation (ed. by F. Almeda and C.M. Pringle), pp. 281–295.
California Academy of Sciences, San Francisco.
Hawkins, B.A. & Diniz-Filho, J.A.F. (2002) The mid-domain
effect cannot explain the diversity gradient of Nearctic birds.
Global Ecology and Biogeography, 11, 419–426.
Hawkins, B.A., Diniz-Filho, J.A.F. & Weis, A.E. (2005) The
mid-domain effect and diversity gradients: is there anything to
learn? The American Naturalist, 166, E140–E143.
Heaney, A. & Proctor, J. (1990) Preliminary studies on forest
structure on Volcán Barva, Costa Rica. Journal of Tropical
Ecology, 6, 307–320.
Holloway, J.D. (1987) Macrolepidoptera diversity in the IndoAustralian
tropics: geographic, biotopic and taxonomic variations.
Biological Journal of the Linnean Society, 30, 325–341.
Holloway, J.D. (1997) The moths of Borneo: family Geometridae,
subfamilies Sterrhinae and Larentiinae. Malayan Nature
Journal, 51, 1–242.
Holloway, J.D. & Intachat, J. (2003) Aspects of the diversity of
Geometridae (Lepidoptera) in Pasoh Forest Reserve. Pasoh:
ecology of a lowland rain forest in Southeast Asia (ed. by T.
Okuda, N. Manokaran, Y. Matsumoto, K. Niyama, S.C. Thomas
and P.S. Ashton), pp. 292–313. Springer, Tokyo.
Jetz, W. & Rahbek, C. (2002) Geographic range size and determinants
of avian species richness. Science, 297, 1548–1551.
Kenney, A.J. & Krebs, C.J. (2000) Programs for ecological methodology,
version 5.2. Vancouver, Canada.
Kluge, J., Kessler, M. & Dunn, R.R. (2006) What drives elevational
patterns of diversity? A test of geometric constraints, climate
and species pool effects for pteridophytes on an elevational
gradient in Costa Rica. Global Ecology and Biogeography, 15,
358–371.
Lees, D.C., Kremen, C. & Andriamampianina, L. (1999) A null
model for species richness gradients: bounded range overlap of
butterflies and other rainforest endemics in Madagascar.
Biological Journal of the Linnean Society, 67, 529–584.
Legendre, P., Dale, M.R.T., Fortin, M.-J., Gurevitch, M., Hohn,
M. & Myers, D. (2002) The consequences of spatial structure
for the design and analysis of ecological field surveys. Ecography,
25, 601–615.
Lieberman, D., Lieberman, M., Peralta, R. & Hartshorn, G.S. (1996)
Tropical forest structure and composition on a large-scale altitudinal
gradient in Costa Rica. Journal of Ecology, 84, 137–152.
G. Brehm et al.
© 2006 The Authors
218 Global Ecology and Biogeography, 16, 205–219 Journal compilation © 2006 Blackwell Publishing Ltd
Lomolino, M.V. (2001) Elevation gradients of species-density:
historical and prospective views. Global Ecology and Biogeography,
10, 3–13.
Longino, J.T., Coddington, J. & Colwell, R.K. (2002) The ant
fauna of a tropical rainforest: estimating species richness three
different ways. Ecology, 83, 689–702.
Magurran, A.E. (2004) Measuring biological diversity. Blackwell
Publishing, Oxford.
McCain, C.M. (2003) North American desert rodents: a test of
the mid-domain effect in species richness. Journal of Mammalogy,
84, 967–980.
McCain, C.M. (2004) The mid-domain effect applied to elevational
gradients: species richness of small mammals in Costa
Rica. Journal of Biogeography, 31, 19–31.
McCoy, E.D. (1990) The distribution of insects along elevational
gradients. Oikos, 58, 313–322.
Novotny, V. & Basset, Y. (2000) Rare species in communities of
tropical insect herbivores: pondering the mystery of singletons.
Oikos, 89, 564–572.
Pitkin, L.M. (2002) Neotropical ennomine moths: a review of the
genera (Lepidoptera: Geometridae). Zoological Journal of the
Linnean Society, 135, 121–401.
Rahbek, C. (1995) The elevational gradient of species richness: a
uniform pattern? Ecography, 18, 200–205.
Rahbek, C. (2005) The role of spatial scale and the perception of
large-scale species-richness patterns. Ecology Letters, 8, 224–
239.
Rangel, T.F.L.V.B., Diniz-Filho, J.A.F. & Bini, L.M. (2006)
Towards an integrated computational tool for spatial analysis
in macroecology and biogeography. Global Ecology and Biogeography,
15, 321–327.
Rohde, K., Heap, M. & Heap, D. (1993) Rapoport’s rule does not
apply to marine teleosts and cannot explain latitudinal gradients
in species richness. The American Naturalist, 142, 1–16.
Rosenzweig, M.L. (1995) Species diversity in space and time.
Cambridge University Press, Cambridge.
Sanders, N.J. (2002) Elevational gradients in ant species richness:
area, geometry, and Rapoport’s rule. Ecography, 25, 25–32.
Sanford, R.L. Jr, Paaby, P., Luvall, J.C. & Phillips, E. (1994)
Climate, geomorphology, and aquatic systems. La Selva: ecology
and natural history of a neotropical rain forest (ed. by A.
McDade, K.S. Bawa, H.A. Hespenheide and G.S. Hartshorn),
pp. 19–33. The University of Chicago Press, Chicago.
Schulze, C.H. (2000) Auswirkungen anthropogener Störungen
auf die Diversität von Herbivoren — Analyse von Nachtfalterzönosen
entlang von Habitatgradienten in Ost-Malaysia.
PhD Thesis, University of Bayreuth, Bayreuth, Germany.
Scoble, M.J. (ed.) (1999) Geometrid moths of the world — a
catalogue (Lepidoptera: Geometridae). CSIRO Publishing,
Collingwood, Australia.
Stevens, G.C. (1989) The latitudinal gradient in geographical
range: how so many species coexist in the tropics. The American
Naturalist, 133, 240–256.
Stevens, G.C. (1992) The elevational gradient in altitudinal
range: an extension of Rapoport’s latitudinal rule to altitude.
The American Naturalist, 140, 893–911.
Watkins, J.E., Jr, Cardelús, C., Colwell, R.K. & Moran, R.C.
(2006) Diversity and distribution of ferns along an elevational
gradient in Costa Rica. American Journal of Botany, 93, 73–83.
Whittaker, R.J., Willis, K.J. & Field, R. (2001) Scale and species
richness: towards a general, hierarchical theory of species
diversity. Journal of Biogeography, 28, 453–470.
Willig, M.R. & Lyons, K.S. (1998) An analytical model of latitudinal
gradients of species richness with an empirical test for
marsupials and bats in the New World. Oikos, 81, 93–98.
Willig, M.R., Kaufman, D.M. & Stevens, R.D. (2003) Latitudinal
gradients of biodiversity: pattern, process, scale and synthesis.
Annual Reviews in Ecology and Systematics, 34, 273–309.
Wolda, H. (1987) Altitude, habitat and tropical insect diversity.
Biological Journal of the Linnean Society, 30, 313–323.
Yela, J.L. & Holyoak, M. (1997) Effects of moonlight and meteorological
factors on light and bait trap catches of noctuid moths
(Lepidoptera: Noctuidae). Population Ecology, 26, 1283–1290.
Zapata, F.A., Gaston, K.J. & Chown, S.L. (2003) Mid-domain
models of species richness gradients: assumptions, methods
and evidence. Journal of Animal Ecology, 72, 677–690.
Zapata, F.A., Gaston, K.J. & Chown, S.L. (2005) The mid-domain
effect revisited. The American Naturalist, 166, E144–E148.
Editor: Pablo Marquet
BIOSKETCHES
Gunnar Brehm is an entomologist focusing on
biodiversity research of moth communities along
environmental gradients. He is working in tropical rain
forests in Ecuador and Costa Rica, and is also interested in
host plant relationships, in chemical ecology of moths and
butterflies associated with pyrrolizidine alkaloids, and in
conservation issues. He has a strong interest in insect
systematics and taxonomy, in particular of Neotropical
Geometridae and Arctiidae.
Robert K. Colwell is Board of Trustees Distinguished
Professor at the University of Connecticut, where he
works with biogeographical theory and models,
biodiversity informatics and statistics. His field studies
concern the ecology and evolution of species interactions,
particularly in the tropics.
Jürgen Kluge is a postdoctoral scientist in the
Department of Systematic Botany at the University of
Göttingen. He has research interests in the field of
diversity and biogeography with a special focus on
pteridophytes and bryophytes.
Moth richness along an elevational gradient
© 2006 The Authors
Global Ecology and Biogeography, 16, 205–219 Journal compilation © 2006 Blackwell Publishing Ltd 219
SUPPLEMENTARY MATERIAL
The following material is available for this article:
Appendix S1 Sampling dates and numbers of individuals
collected on each date at 12 sampling sites. Total numbers
sampled are provided in Table 1.
AppendixS2 Complete list of 739 geometrid species recorded at
12 rain forest sites along a complete elevational gradient in Costa
Rica (Barva transect).
This material is available as part of the online article from:
http://www.blackwell-synergy.com/doi/abs/10.1111/j.1466-
822x.2006.00281.x
(This link will take you to the article abstract).
Please note: Blackwell Publishing are not responsible for
the content or functionality of any supplementary materials
supplied by the authors. Any queries (other than missing
material) should be directed to the corresponding author of
the article.