mm FORUM i!> a lighter channel of communication between readers and contnbutors, it aims to
stimulate discussion and debate, particularly by presenting new ideas and by suggesting
alternative interpretations to the more formal research papers published in ECOGRAPHY and
elsewhere A lighter prose is encouraged and no summary is required Contributions should be
concise and to the point, with a relatively short bibliography Formal research papers, however
short, will not be considered
The elevational gradient of species richness: a uniform pattern?
Carsten Rahbek, Centre for Tropical Biodiversity, Zoological Museum, University of Copenhagen,
Universitetsparken 15, DK-2100 Copenhagen 0, Denmark.
The decline m species nchtiess with increasing elevation
IS widely accepted as a general pattern (Table 1) In as
much as the elevational gradient is often claimed to
mirror the latitudinal gradient, spectes richness is assumed
to decrease monotonically (l e because of reduced
temperature and consequent decrease in producUvity)
Perhaps because such a relation is intuitive, biologists
have readily generalized the results of a few early studies
of tropical birds as supporting a general biogeographic
Table 1 Typical examples of statements atwut the relationship
between species nchness and elevation from books and papers
in major journals
"For all of these reasons, we expect the number of species to
decrease with altitude and, m fact, it does " (MacArthur 1972, p
107)
"In terrestrial habitats, vanation in species diversity along gradients
of elevauon and available soil moisture are [sic] almost as
general and striking as latitudinal vanation " (Brown and Gibson
1983, p 502)
"In terrestnal environments, a decrease in species with altitude
IS a phenomenon almost as widespread as a decrease with
latitude " (Begon et al 1990, p 805)
"Just as change of physical conditions with altitude resembles in
many respects the vanauon with latitude, so the decreasing
diversity of most organisms with increasing elevation mirrors in
most respects the latitudinal gradient of species nchness"
(Brown 1988, p 62)
"biologists have long recogmzed that elevational and latitudinal
species-nchness gradients mirror each other " (Stevens 1992,
p 899)
"In terrestnal ecosystems, diversity generally decreases with
increasing altitude there appear to be no substantiating data
for [the] 'mid-altitudinal bulge' as a general phenomenon"
(World Conservation Momtonng Centre 1992, pp 43,45)
"Decrease in the number of species with decreasing temperatures
at higher altitudes is as conspicuous as the decrease with
latitude (e g Brown and Gibson 1983), although exceptions
occur" (Rohde 1992, p 522)
pattern This has resulted in "citation inbreeding" Here, I
outline the supporting evidence for the generalization and
discuss the influence of samplmg regime and the often
Ignored influence of area I then present a quantitative
review of the data already present, although often ignored,
in the literature Altogether 97 papers (with 163
examples) have been reviewed. The analysis of these
empincal data support the view that species nchness
declines with elevation, but not the view that this decline
IS necessanly monotonic Some possible reasons for variation
in the exact shape of the relationship between
species nchness and elevation for different taxa and zoogeographic
areas are commented, but our understanding
of the relation between elevation and species nchness suU
appears to be immature
The empirical basis for the "general
pattern"
The generalization (Table 1) grew mamly from two studies
dealing with tropical birds one from the Peruvian
Andes (Terborgh 1977), and the other from New Guinea
(Kikkawa and Williams 1971) The textbook example
from New Guinea was published as a short note based on
compilation of the published distnbutional data (Kikkawa
and Williams 1971) It was conducted at a Ume
when the knowledge of the elevational distnbution of
New Guinean btrds was still somewhat rudimentary Another
more detailed study on birds of New Guinea has
also been cited as proof of the general pattern (Diamond
1972, cited in MacArthur 1972) However, Diamond's
New Guinea data actually show a small peak tn species
nchness at 1100 m with a marked decline m species
nchness only above this level
The second textbook example was based on a carefully
conducted field survey and cntical data analysis (Terborgh
1977). Based on mist-netting and opportunisuc
200 ECXX3RAPHY 18 2 (1995)
A 200 B 70
1000 2000 3000
ELEVATION (M)
4000 1000 2000 3000
ELEVATION (M)
4000
Fig I Species nchness of syntopic birds versus elevation on an Amazonian slope of the Andes in Peru Figure 1A is based on data
not standardized for area and sampling effort, whereas Fig IB is based on standardized samples of 300 mist-netted birds (data from
Terborgh 1977) I have fitted the lines by distance-weighted least-squares smoothing
field observattons at camps situated along ati elevational
gradient on the humid east slope of the Peruvian Andes,
Terborgh showed that species nchness declined monotonically
with elevation if the number of species is simply
plotted against elevation (Fig lA) However, when Terborgh
tned to eliminate the effect of sampling time by
standardizmg his mist-netting data, a different pattern
emerged (Fig IB) Terborgh explained the emerging
"hump-shaped" curve, which had a peak in species nchness
around 1400 m, as the result of a local "hot-spot" in
resources (Terborgh 1977) Although this unexpected
pattern is addressed through most of his discussion, only
the first non-standardized graph is usually cited in the
literature (e.g Brown 1988) Though the Ube of understorey
mist-netting data limited the scope of this study, it
serves to demonstrate the strong effect that sampling
effort can exert - especially in species-nch tropical forest
where most species occur at low densities
The importance of area
The effect of area on the relationship between spwcies
nchness and elevation has rarely been considered, although
we would expect area to have a significant impact
on the form of the elevational pattern, as the relationship
between area and species nchness seems as universal as
A 2500
2000
Din
o
ffl
1500
1000
500
B
ELEVATION (M) ELEVATION (M)
Fig 2 Species nchness of South Amencan tropical landbirds versus elevation Figure 2A is based on data not standardized for area,
whereas Fig 2B is based on data standardized for area set to 50000 km^ using equations for species-area curves (log S/log A
transformation) of each elevauonal zone (based on data from Rahbek unpubl) Area is set to 50000 km' because it is within the
range of the ongmal data upon which the equaUons are based, and is a size within the geographical regional scale (e g Wiens 1981,
Rosenzweig and Abramsky 1993)
ECOGRAPHY 18 2 (1995) 201
Table 2 Number of data sets (n = 163) found in the literature with data on the variation of species nchness with elevation
summanzed by biomes (NT = non-tropical biomes, T = tropical biomes) and scale (R = regional, L = local), listed for studies
conducted on mainland and island, respectively, and subdivided for whether the researcher(s) have attempted to standardize for the
effect of area and sampling regime and/or effort, only one of these factors, or none ("non-standardized")
Mainland
Area and sampling
Area
Sampling
Non-standardized
Island
Area and sampling
Area
Sampling
Non-standardized
Total
Invertebrates
NT(R/L)
0/2
0/0
4/1
9/11
1/0
0/0
0/0
0/5
14/19
T(R/L)
0/0
0/0
3/4
4/6
3/2
1/0
0/5
2/5
13/22
Vertebrates
NT(R/L)
0/6
0/0
0/6
3/0
0/0
0/0
0/0
0/0
3/12
T(R/L)
0/1
1/0
3/1
17/4
0/0
0/3
0/1
5/2
26/12
Plants
NT (R/L)
0/10
0/0
0/1
1/0
0/4
0/0
0/0
0/0
1/15
T(R/L)
0/6
4/3
0/0
8/3
0/0
0/0
0/0
2/0
14/12
Totals
NT (R/L)
0/18
0/0
4/8
13/11
1/4
0/0
0/0
0/5
18/46
T (R/L)
0/7
5/3
6/5
29/13
3/2
1/3
0/6
9/7
53/46
the latitudinal gradient To understand the relationship
between area and species nchness along elevational gradients,
especially on a regional scale, the effect of area
must be considered since areas of equal-sized elevational
belts may vary with elevation Thus, areas often decrease
with elevation because of generally steeper terrain toward
the highest peaks When landbird data from tropical
South Amenca, compiled at a regional scale using countries
as units, are standardized for area, the relationship
between species nchness and elevation gives a humpshaped
curve (Fig 2) However, area alone is unlikely to
explain any global pattern of species nchness, as close
couplings can be expected to exist between biological
diversification and habitat complexity (see also Rosenzweig
1992)
A quantitative review of the literature
The repeated citation of the same few studies provides a
false picture of the amount and diversity of data published
on the issue I have been able to find the surpnsmgly
high number of 97 papers, with 163 examples that
give data on the vanation in number of species with
elevation. It is highly probable that additional data exist,
as many data sets are published in httle known journals,
or in the "gray" literature.
Table 2 summanzes some charactenstics of each data
set (taxonomic group, region, scale and data treatment)
The mfluences of sampling regime/effort and the effect of
area are among the most influential biasing factors in
most field studies of species-nchness patterns, and, unfortunately,
equally difficult to eliminate successfully I
have thus only judged whether an attempt was made to
deal with these two factors, either in the design of the
survey or afterward, during the data analyses. Remaricably,
many of the pajjers reviewed do not give any details
on this subject These data sets are classified as "Nonstandardized"
together with those with no attempt to
standardize data (Table 2)
Unlike the traditional histoncal trend within most
fields of biological research, most data sets are from the
tropics (99 out of 163) The focus on the tropics is
presumably related to the circumstance that tropical elevational
gradients compnse a wider range of climatic
vanation than temjserate elevational gradients Out of 163
data sets, 68 are on invertebrates, 53 on vertebrates (including
36 on birds), and 42 on plants The majonty of
the data sets comes from mainland biota (122), whereas
41 data sets are fi-om islands
As shown, standardizing for sampling effort or effect
of area can have a significant influence on the emerging
shape of the relationship between sjjecies richness and
elevation In 87 of the 163 data sets, the data have not
been standardized for area or sampling effort (corresponding
figures for tropical and non-tropical biomes are
59% and 45%) Only in 35 (21 %) cases has a standardization
been attempted for area as well as sampling effort
(figures for tropical and non-tropical biomes are 12% and
36%, respectively)
Considenng the high mobility of birds compared to
other groups, the reliance by most reviewers on primanly
bird examples to illustrate a universal relationship seems
ill-founded, especially when the bulk of data in the hterature
actually denves from invertebrates and plants
Table 2 also provides an overview for which combmations
of, for example, taxonomic groups, scale, and region
we lack studies - especially those that consider the
effect of area and samphng regime on data.
Methodological problems and patterns
In many studies, a stated decline in species nchness with
202 ECOGRAPHY 18 2 (1995)
Table 3 The relationship between species nchness and elevation summarized by type of pattern Only the 90 data sets (of 163) that
provide data points spanning from below 500 m to above 1500 m are included (see text) The classification of each pattern is based
on a visual examination of bivanate plots NT= non-tropical biomes and T = tropical biomes
Scale Monotomcally
decreasing
NT T
Horizontal, then
decreasing
NT T
Hump-shaped
NT
Increasing
NT
Other
NT
Regional
Invertebrates
Vertebrates
Plants
Total regional
Local
Invertebrates
Vertebrates
Plants
7
6
10
23
10
2
1
Total local
Total
1 9
1 18
6
7
9
15
3
8
13
36
1
1
0
0
0
0
2
4
elevation was restncted to only a part of the elevational
gradient In other cases, mid-elevational data were lacking
Others used correlation tests on data sets that include
few stations from low- and mid-elevation, but many from
higher elevations, thereby biasing their findings toward a
strong negative correlation In such instances, the data are
inapplicable to support a monotonic relationship Conclusions
based on correlation tests sometimes ignore that
stations at mid-elevation actually have more species than
stations at low elevation
To analyze the general vanation of species richness
with elevation, a minimum requirement for any data set is
that It includes data spanning the entire gradient, albeit it
becomes lncreasmgly difficult to find appropnate gradients
with continuous natural habitat along the entire
gradient This is es{)ecially a problem with respect to
lowland stations, as lowlands and foothills often are the
most disturbed elevational zone(s) Of the 163 data sets,
47 do not include data from below 500 m. In the descriptive
analysis of the elevational pattern of species nchness,
I have only included data sets which are based on a
gradient from below 500 m to above 1500 m (see
Table 3) This hmits the analysis to 90 data sets of the
onginal 163 37 on invertebrates, 26 on vertebrates (lncludmg
19 on birds), and 27 on plants. As for the entire
data set, this subsample of data sets is biased toward the
tropics with 73 data sets compared to only 17 from
non-tropical biomes It also has more data sets from
mainland (n = 71) than islands (n = 19) Non-standardized
data sets are dominant (n = 49) In only 13 data sets
(14%) have attempts been made to take the effect of area
and sampling into account. The corresponding figure for
tropical data sets, which have been the main source of
generalizations on the elevational gradient, is only 7% (5
of 73 tropical data sets), in contrast to 47% of the nontropical
data sets
To conduct proper descnpuve statistical analyses of the
vanation of species nchness with elevation, the stations
(l e the data points) must be reasonably equally distnbuted
over the gradient, and the number of data points
sufficient to refiect any marked changes in habitats/
biomes over the analyzed gradient. Unfortunately these
two requirements for an optimal data set are rarely fulfilled
As It IS difficult to judge especially the latter
cntenon for most of the published data no attempt has
been made to select or exclude data on this basis.
Compansons of elevational patterns between taxa, latitudinal
climatic zones, biogeographic regions or mainland
versus islands could be misleading without correction
for the area effect and differences in sampling regime
Compansons of studies are also biased by
differences in the species included, and sometimes further
by limitations to Sjjecific trophic levels, guilds or
habitat These problems and the pronounced heterogeneity
of the quality of data sets make it difficult to
conduct proper cntical statistical tests for each data set of
the relationship between sp>ecies nchness and elevation
that are mutually comparable Still, disregarding these
biases, a rough companson based on a classification of
the pattern in each data set by visual examination of
bivanate plots serves to illustrate how ambiguous the
pattern is, both withm region, spatial scale and crude
taxonomic groupmgs (Table 3).
A decline in species nchness with elevation seems to
be a general trend Yet, a pattern where the speciesrichness
curve IS almost honzontal up to a certain elevaUon
before declining, or is hump-shaped, seems more
typical than a monotomc declme (Table 3)
ECOGRAPHY 18 2 (1995) 203
1000 2000 3000 4000 5000
ELEVATION (M)
Fig 3 Patterns of species nchness versus elevation may vary
within the same area for different taxa, here illustrated by New
Guinean bats (*) and rodents (ˇ) (data compiled <Tom Flannery
[1990]) Data are not standardized for area and sampling effort
Lines are fitted by distance-weighted least-squares smoothing
Just a reflection of the latitudinal
gradient?
The decrease in sjjecies nchness from the equator toward
the poles is one of the most universal biogeographic
patterns This pattern has been shown across an array of
taxa in aquatic as well as terrestnal ecosystems Exceptions
are few, and typically restncted to taxa with relatively
few species (Rohde 1992) At first glance, the
elevational gradient appears to share many climatic charactenstics
with the latitudinal gradient Thus, Stevens
(1992) claimed that ecologically restnctive climatic conditions
appear to increase with elevation as they do with
latitude Furthermore, the harsh climate and relatively
low species nchness at the equatorial treeline seem to
resemble the corresponding conditions found at northern
and southern temperate latitudinal zones The apparent
resemblance of climate at high elevauons and extreme
latitudes makes it intuitively tempting to expect the elevational
gradient of species nchness to just mirror of that
of the latitudinal gradient The impression of a mirrored
relationship is also indirectly imposed upon us by terminology
traditionally used, as we often divide tropical
mountains into tropical, subtropical, temperate and arctic
(or alpine) zones (eg Chapman 1917). Yet, detailed
analyses and compansons of vanation in biologically
significant climatic parameters between the latitudinal
and elevational gradient will reveal several important
differences For example, an important feature of climatic
vanation on the latitudinal gradient is the increased seasonality
towards the poles In contrast to this, mean temperature
remains fairly constant year round within the
same bands on a tropical elevational gradient, and variation
in temperature regime seems pnmanly to be diurnal
Interestingly, both types of vanation in temperature
regime are represented on temperate elevational
gradients One would expect this dissimilanty to have
different impact on fwpulation maintenance processes on
the two gradients and speciation
Actually, the existence of a "plateau" or a "hump" on a
curve companng species nchness with elevation should
not be regarded as unexpected considenng that - although
temperature declines with elevation, another lifesupport
factor, stable water supply, increases (at least to a
certain elevation) Most elevational gradients have a
more or less stable condensation zone (cloud zone) at a
certain level, especially conspicuous in the tropics, causing
favorable conditions for certain taxa (e.g ephiphytes
at mid-elevation, which in turn create microhabitats and
food for other taxa) As local climate can vary prominently
over a few kilometers or hundred meters (e g
between opfwsite slopes of the same mountain) in the
tropics, the exact location of such a "climatic optimum"
can vary considerable regionally and locally, causing
differences in the shape of the elevational gradient even
within the same taxa The latitudinal gradient does not
have such a "humidity peak" One could also add that
habitat fragmentation necessanly increases with elevation
but not necessarily with latitude.
Altogether, there seem to be no a pnon reasons to
believe that, climatically, the elevational gradient simply
mirrors the latitudinal gradient
A negative correlation between species richness and
elevation fits well with the general acceptance that the
lowland tropical rain forest has the nchest biota on Earth
(e g. MacArthur 1972, Erwin 1988). Recent research has
shown that this may not always be true on a regional scale
(South Amencan mammals [Mares 1992], and birds
[Rahbek unpubl ]). Histoncally, a monotonic decline in
Sjjecies nchness with elevation corresponds well with the
many theones suggesting mechanisms by which increased
energy availability often results in proliferation
of different species rather than larger populations of existing
species (e.g Hutchinson 1959, Pteston 1962, MacArthur
1972, Brown and Gibson 1983, Wnght 1983) We
now know that such an increase in species nchness with
productivity is not universal (e.g Rosenzweig 1971, Carson
and Barrett 1988) In fact, it has recently been suggested
that the relationship is hump-shaped (Rosenzweig
1992, Rosenzweig and Abramsky 1993)
Final remarks
Understanding elevational patterns must be based on well
collected qualitative data, and explanations must pnmarlly
be based on unravelling pnmary mechanisms, such as
204 ECOGRAPHY 18 2 (1995)
physical causes, including climatic factors, the narrow
width of the elevational gradient, and histoncal perturbations
that shape the available species pool Biotic mteractions
are secondary mechamsms that can be reflected in
the emergent patterns The vanation of sjjecies nchness
with elevation might be connected to the reduction of
temjjerature with elevation and the assumed corresfwndlng
reduction in productivity However, other factors
such as vanation in steepness, geological perturbations,
alterations of precipitation patterns, etc might also be
involved, probably with varying impact from case to
case For any correlation of species nchness with variables
measured over an elevational gradient, we need to
examine whether the correlation reflects a direct coupling,
or if It could be a result of interactions of several
other factors
The observation that high elevation suppwrts fewer
species than low elevation, which indirectly acted as a
catalyst for the belief in a general monotonic decline,
seems to be a general pattern (Table 3) Still, the pattern
of species nchness at low- and mid-elevations may differ
between taxa as well as within taxa between different
regions, and within the same region, at least on a regional
scale (see example in Fig 3) It is important to discnminate
between patterns refiecting recent diversification
and those reflecting long-term accumulation of species
(Fjeldsa 1994) The latter could well be an equilibrium,
provided that we compare areas that not only have similar
average conditions but also resemble each other in habitat
mosaicism and dynamism This could be a reason why
the position of humps or bend of curves vanes between
different sets of data Vanous taxa are also differently
affected by abiotic factors, such as, for example, humid-
ity
Further studies, including analyses of pnmary-level
processes that could influence the pattern, are needed to
reveal whether general patterns exit within biogeographic
regions, taxa, spatial scale, mainland versus islands, etc ,
or various combinations of these Appropriate null-models
should be considered before explaining elevational
gradient patterns as results of climatic, biological and
histoncal processes (Colwell and Hurtt 1994) Although a
difficult task, the development of testable hypotheses is
important to achieve significant progress within this field
and to contnbute to our general understanding of diversity
patterns This would be preferable to founding
new generalizations on simple accumulations of case
studies Much is still to be learned about this topic, for
now, we have to accept the unsatisfactory realization that
we do not know whether a general relationship exists
between species nchness and elevation, or whether an
universal explanation or model can be given
Acknowledgements -1 thank G Graves and J Fjeldsa for many
discussions and cnUcally reading the manuscnpt Also thanks to
J Lovett,J M CardosadaSilva,R Zusi,D ThurberandM E
Petersen for their cntical comments to the manuscnpt The study
was conducted dunng a stay at The National Museum of Natural
History, Smithsonian Instituuon supported by a Fulbnght grant
The complete list of papers which this Forum contnbution is
based on is available from the author
References
Begon, M , Harper, J L and Townsend, C R 1990 Ecology
Individuals, populations and communities (2nd ed) Blackwell.
Oxford
Brown, J H 1988 Species diversity - In Myers, A A and
Giller, P S (eds). Analytical biogeography - an integrated
approach to the study of animal and plant distnbution Chapman
and Hall, New York, pp 57-89
- and Gibson. A C 1983 Biogeography - C V Mosby, St
Louis
Carson, W P and Barrett, G W 1988 Succession in old-field
plant communities effects of contrasting types of nutrient
ennchment - Ecology 69 984-994
Chapman, F M 1917 The distnbution of bird-life in Colombia
a contnbution to a biological survey of South Amenca Bull
Amer Mus Nat Hist 36 1-729
Colwell, R K and Hurtt, G C 1994 Nonbiological gradients in
sfiecies richness and a spunous Rapoport effect - Am Nat
144 570-595
Diamond, J M 1972 Avifaunaof the eastern highlands of New
Guinea - Nuttall Omith Club 11 1^38
Erwin.T L 1988 The tropical forest canopy the heart of biotic'
diversity - In Wilson, E O (ed), Biodiversity National
Academy Press, Washington. D C . pp 123-129
FjeldsS, J 1994 Geographical patterns for relict and young
species of birds m Afnca and South Amenca and implications
for conservation pnonties - Biodiv Conserv 3 207-
226
Flanner>', T 1990 Mammals of New Guinea - Robert Brown
and Ass . Sydney
Hutchinson. G E 1959 Homage to Santa Rosalia, or "why are
there so many kinds of animals''" - A m Nat 93 145-159
Kikkawa. J and Williams, E E 1971 Altitude distnbution of
land birds in New Guinea - Search 2 64-65
MacArthur. R H 1972 Geographical ecology - Harper and
Rowe Publ , New York
Mares. M A 1992 Neotropical mammals and the myth of
Amazonian biodiversity - Science 25 976-979
Preston, F W 1962 The canonical distribution of commonness
and rarity - Ecology 43 185-215, 410-^32
Rohde, K 1992 Latitudinal gradients in species diversity the
search for the pnmary cause - Oikos 65 514-527
Rosenzweig, M L 1971 Paradox of ennchment destabilization
of exploitation ecosystem in ecological time - Science 171
385-387
- 1992 Species diversity gradients we know more and less
than we thought - J Mammal 73 715-730
- and Abramsky. Z 1993 How are diversity and productivity
related"' - In Ricklefs, R and Schluter, D (eds). Species
diversity in ecological communities Histoncal and geographical
perspectives Univ of Chicago Press, Chicago, pp
52-65
Stevens. G C 1992 The elevational gradient in altitudmal
range an extension of Rapoport's latitudinal rule to altitude
- A m Nat 140 893-911
Terborgh. J 1977 Bird species diversity on an Andean elevational
gradient - Ecology 58 1007-1019
Wiens, J A 1981 Scale problems in avian censusing -Stud
Avian Biol 6 513-521
World Conservation Monitonng Centre 1992 Global diversity
status of the Earth's living resources - Chapman and Hall,
London
Wnght, D H 1983 Species-energy theory an extension of
species-area theory - Oikos 41 496-506
ECOGRAPHY 18 2(199'>) 205