Post on 09-Jul-2020
transcript
A necessarily complex model to explain the biogeography of Madagascar's amphibians and
reptiles
Jason L. Brown1,2, Alison Cameron3, Anne D. Yoder1, Miguel Vences4
1 Department of Biology, Duke University, Durham, NC 27708 Durham, NC, USA
2 Current address: Department of Biology, The City College of New York, NY, USA
3 School of Biological Sciences, Queen's University Belfast, 97 Lisburn Road, Belfast BT9 7BL, UK
4 Zoological Institute, Technical University of Braunschweig, Mendelssohnstr. 4, 38106 Braunschweig, Germany
Abstract
A fundamental limitation of biogeographic analyses are that pattern and process are inextricably
linked, and whereas we can observe pattern, we must infer process. Yet, such inferences are
often based on ad-hoc comparisons using a single spatial predictor such as climate, topography,
vegetation, or assumed barriers to dispersal without taking into account competing explanatory
factors. Here we present an alternative approach, using mixed-spatial models to measure the
predictive potential of combinations of spatially explicit hypotheses to explain observed
biodiversity patterns. In this study we compiled a comprehensive dataset of 8362 occurrence
records from 745 amphibian and reptile species from Madagascar. These data were used to
estimate species richness, corrected weighted endemism, and species turnover (based on
generalized dissimilarity modeling). We also created or incorporated, when previously available,
18 spatially explicit predictions of 12 major diversification and biogeography hypotheses, such
as: mid-domain, topographic heterogeneity, sanctuary, and climate-related factors. Our results
clearly demonstrate that mixed-models greatly improved our ability to explain the observed
amphibian and reptile biodiversity patterns. Hence, the observed biogeographic patterns were
1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
likely influenced by a combination of diversification processes rather than by a single
predominant mechanism. Further, selected genera of Malagasy amphibians and reptiles differed
in the major factors explaining their spatial patterns of richness and endemism. These differences
suggest that key factors in diversification are lineage specific and vary among major endemic
clades. Our study therefore emphasizes the importance of comprehensive analyses across
taxonomic, temporal, and spatial scales for understanding the complex diversification history of
Madagascar's biota. A "one-size-fits-all" model does not exist.
Keywords: Conditional Autoregressive Models, Orthogonally Transformed Beta Coefficients,
Generalized Dissimilarity Modelling, Species Distribution Modelling
The spatial distribution of biodiversity is at the core of biogeography, macroecology,
evolutionary biology, and conservation biology1,2. Biodiversity mapping indices are multi-
faceted concepts with the main components being local endemism, species richness, and species
turnover, of which the two latter correspond to alpha- and beta-diversity as used in community
ecology3,4. In different combinations, these components are invoked to identify biogeographic
regions5-7, prioritize geographic areas for conservation8, assess the effects of conservation
measures9, and/or delimit centers of speciation or extinction. These indices, however, are not
independent of one another. For instance, species turnover across an area is closely related to the
numbers of endemic species within each geographical unit or community, which in turn is often
used to estimate areas of endemism (AOE). These represent the coincident restrictedness of
taxa10-12 and are often used to identify unique geographic areas for biodiversity conservation or
2
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
biogeography studies13,14. Clearly, there is an inescapable circularity to these measures, and thus
also to the consequent inferences made regarding biogeographic processes.
Inferences of speciation mechanism fall prey to similar limitations. For example, it is
generally assumed that species formation and diversification of a range of co-distributed taxa
will be triggered or inhibited by similar barriers to gene flow, topographical and geological
settings, climatic conditions and shifts, and competition. Accordingly, it is the default
expectation that similar barriers (e.g., rivers, ecotones, climatic transitions) will lead to similar
patterns of species endemism, turnover, and richness; again, with the underlying assumption that
the observation of similar patterns among diverse species reveals a general causal mechanism of
diversification across all taxa. But there are additional processes by which species richness may
be generated. For example, climatic factors, environmental stability, land area, habitat
heterogeneity, paleogeography, and energy available all could be spatially correlated with
geographical barriers. Thus, any of these mechanisms might be indirectly, but not causally
related to diversification15. Patterns of endemism, on the other hand, are generally considered to
reflect a particular evolutionary history, with areas of endemism corresponding to centers of
diversification16 and often including some element of stochasticity. Consequently, it can be the
case that areas of high endemism often are also characterized by high species richness, though
the inverse is not necessarily true.
Species distribution models (SDMs) allow sophisticated calculations of centers of
historical habitat stability17. Yet, their spatial comparison with current patterns usually follows
narrative approaches and is similar to classical hypotheses of diversification mechanisms, with
no accounting for autocorrelation among the different explanatory variables. Based on either a
single explanatory variable or without employing statistics at all, often biogeography researchers
3
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
rely on ad-hoc comparisons with spatial distributions of single environmental factors such as
climate, topography, vegetation, or assumed barriers to dispersal. As a sign of progress, many
methodological advances are being developed to address the various problems described here.
For example, assessments of spatial biodiversity have typically used simple geographic measures
as the unit of analysis, such as the distribution range of individual species, though recent
methodological refinements include the inclusion of phylogenetic relationships among species
and their evolutionary age2,7. Moreover, carefully parameterized SDMs can generate accurate
estimates of distribution ranges18 and novel approaches are being developed to translate patterns
of species richness, endemism and turnover more objectively for determining those
biogeographic regions in greatest need for conservation and protection2,7,8,19-21. Despite this
progress in conceptual and statistical tools, biological explanation of these patterns is still in its
methodological infancy.
Here we aim to employ the latest techniques for sophisticated and improved statistical
methods for identifying the causal mechanisms that have determined the spatial distribution of
Madagascar's herpetofauna. Though the search for the drivers of biological diversification was
initially focused on the Neotropics, considerable attention has more recently been focused on
other areas such as the Australian wet tropics22 and Madagascar23. Madagascar is the world's
fourth largest island and hosts an extraordinary number of endemic flora and fauna. For example,
100% of the native species of amphibians and terrestrial mammals, 92% of reptiles, 44% of
birds, and >90% of flowering plants occur nowhere else24. This mega diverse micro-continent,
initially part of Gondwana, has been isolated from other continents since the Mesozoic. Its
current vertebrate fauna is a mix of only a few ancient Gondwanan clades and numerous
4
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
endemic radiations, originating from Cenozoic overseas colonizers arriving mainly from Africa25-
27.
The extraordinary levels of endemism at the level of entire clades in Madagascar, and
their long isolation from their sister lineages, provide a unique opportunity to study the
mechanisms driving divergence and diversification in situ28. Over the past decade, numerous
general mechanisms and models have been formulated to explain biodiversity distribution
patterns and species diversification in Madagascar, pertaining to environmental stability (or
instability), solar energy input, geographic vicariance triggered by topographic or habitat
complexity, intrinsic traits of organisms, or stochastic effects23,29-36. Evidence has supported
numerous hypotheses, though the evidence has typically been marshalled from limited or
phylogenetically-constrained taxa. Comprehensive statistical approaches comparing their relative
importance are rare37.
In this paper we apply an integrative approach to simultaneously test which of several
competing and complementary hypotheses are most strongly correlated with empirical
biodiversity patterns (Fig. 1). We first translate a total of 12 diversification mechanisms or
diversity models into explicit spatial representations. We then use diverse statistical approaches
to assess spatial concordance of these predictor variables with species richness, endemism and
turnover as calculated from original occurrence data of Madagascar's amphibians and reptiles,
with full species-level coverage. Our results best agree with the hypothesis that various
assemblages of species are under the influence of differing causal mechanisms. The clear
message is that the distribution of diverse organismal lineages will depend upon idiosyncratic
factors determined by their specific organismal life-history traits combined with stochastic
5
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
historical factors. Thus, any model that endeavors to explain island-wide patterns must
necessarily be complex.
RESULTS
To understand spatial distribution patterns in Madagascar's herpetofauna, we first
compared range sizes, and computed species richness and endemism from the modeled
distribution areas of amphibians and non-avian reptiles (hereafter reptiles). Mean range size (±
standard deviation) in our data set is smaller in amphibians than reptiles taking into account all
species (41,673 ± 55,413 km2 vs. 50,205 ± 84,078 km2; t= 3.981, p< 0.001; df=649.7) and after
excluding species known from only 1 or 2 localities (64,106 ± 57,532 km2 vs. 95,294 ± 87,495
km2; t= 4.511, p<0.001; df=427.4). Microendemics (species with distributions less than
1000km2) constitute 36.5% of all amphibian and 33.6 % of all reptile species in Madagascar
(difference not significant; Z =0.411, p=0.682).
Spatial patterns of species richness are quite similar between the two groups (Fig 2A &
E) and reach highest values in the eastern rainforest; in amphibians, richness peaks in the central
east, whereas in reptiles, it is more evenly distributed across the rainforest biome, with some
areas of high diversity also in the north, west, and southwest. Spatial patterns of endemism in
both groups (Fig 2B & F) reveal two centers in the north around the Tsaratanana Massif and in
the central east. Endemism values for reptiles are also high in southwestern Madagascar, the
most arid region of the island.
We applied Generalized Dissimilarity Modelling (GDM)38,39 to identify areas of
endemism on the basis of turnover patterns for non-avian reptiles and amphibians together. The
6
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
major AOE obtained in a 4-class categorization of the originally continuous GDM results (Figure
2C & H) largely mirrors the bioclimatic regions of Cornet (1974).
Our test includes a total of 12 predictor hypotheses, some of which focus on the
geographical pattern in which species diversity is distributed, but without making any clear
assumption about how the species originated (e.g., the mid-domain or topography heterogeneity
hypotheses). Others explicitly refer to mechanisms of diversification and make predictions about
how these processes affected the distribution of species diversity over geographical space (see
Supplementary Documents for detailed accounts). We divided all the hypotheses into two
categories: one in which predictions for continuous two-dimensional spatial richness and
endemism can be derived, and another in which nominal AOE predictions can be derived. The
first category includes (1) the Mid-domain Effect, (2) Topographic Heterogeneity, (3) Climatic
Refugia, (4) Museum (montane refugia), (5) Disturbance-Vicariance, (6) Climate Stability, (7)
Sanctuary and (8) Montane Species Pump. The second category includes (9) the River-Refuge
(large river model), (10) Riverine Barrier (minor and major rivers), (11) Climatic Gradient and
(12) Watershed. All these hypotheses were transformed into explicit spatial representations
(Supplementary Materials) and used as predictor variables for further analyses.
We calculated unbiased correlation of the continuous predictor and test variables
following the method of Dutilleul40, which reduces the degrees of freedom according to the level
of spatial autocorrelation between two variables (detailed results in Supplementary Materials
Table S3). We found that measures of both reptile and amphibian endemism significantly
correlated to the predictor hypotheses of Topographic Heterogeneity, Disturbance-Vicariance
and Museum (montane refugia). Reptile endemism (but not amphibian) is also correlated to
Sanctuary. Correlations to species richness were not tied to measures of endemisms. Whereas
7
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
reptile species richness is correlated to the Mid-domain Effect (distance) and Sanctuary
hypotheses, amphibian richness is correlated to the Sanctuary hypothesis as well as to the
Topographic Heterogeneity, Montane Species Pump, Disturbance-Vicariance, Museum, and
River-Refuge hypotheses.
In the univariate correlation analyses of nominal geospatial data (those related to AOE
predictions) we compared the biogeographic zonation of Madagascar as suggested by the GDM
analysis of amphibian and reptile distributions with zonations derived from five predictor
hypotheses. We found all predictor variables (corresponding to the hypotheses Riverine-major
and Riverine-minor, Gradient, River-refuge, and Watershed) to be significantly correlated to the
15-class GDM, and all but watershed with the 4-class GDM zonation (Table 1). Both GDM
classifications share the most overlap with the Riverine and Gradient hypotheses (between 40.9–
54.3% and 56.2–71.1%, respectively; Table 1).
Given the significant correlation of each of the spatial amphibian and reptile biodiversity
patterns with various predictor variables we used mixed conditional autoregressive spatial
models (CAR models) to test the influences of various predictors simultaneously. To avoid over-
parameterization we used AICc (corrected Akaike Information Criterion), an information-
theoretical approach, to compare models with different sets of predictors. We found that complex
models including most of the biogeography hypotheses (continuous predictor variables)
performed best, based on the lowest AICc values, and consequently used these for further
analysis. Detailed contributions of each predictor to the models of richness, endemism and GDM
zonation are summarized in Supplementary Materials Table S4. The top-five variables
contributed 49.4–75.9% to the models. For a more simplified graphical representation (Fig. 3),
we summarized the three Mid-domain Effect hypotheses (latitude, longitude, and distance), the
8
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
three principal components representing the Climate Gradient hypothesis, and three hypotheses
focused on topography (Topographic Heterogeneity, Disturbance-vicariance, Montane Species
Pump) were categorized together, respectively (Figs 3 & 4) . We found relevant influences of the
Mid-domain Effect especially on the GDM and the species richness and endemism of reptiles
(30.9%, 32.9% and 45.5%, respectively). However, it is important to point out that almost all the
Mid-domain correlation coefficients were negative. Thus, indicating that Mid-domain Effects do
not play a key role in determining spatial patterning. Climate Gradient effects influenced all the
models of biodiversity equally, contributing roughly a quarter to each (25.1–27.7%), though in
many cases the sign of the contribution varied. However in this case, a positive correlation was
not expected. The topography variables contributed positively to the richness and endemism
models of amphibians and reptiles, with joint influences of 9.1% and 22.4% on richness, and
6.5% and 17.3% on endemism. The two unique hypotheses, Sanctuary and Museum, each
contributed positively to all models, with Museum contributing between 7.1–17.1% (one of the
few hypothesis to contribute >5% and to be positively correlated to all biodiversity
measurements). The Sanctuary hypothesis also contributed positively to all hypotheses, though
to a lesser degree than the Museum hypothesis (which demonstrated little contribution to reptile
endemism).
To assess variation in biogeography patterns among major groups of the Malagasy
herpetofauna, we calculated mixed CAR models using the same methods for richness and
endemism of four exemplar sub-clades: the leaf chameleons (Brookesia), tree frogs (Boophis),
day geckos (Phelsuma) and Oplurus iguanas (including the monotypic iguana genus
Chalarodon). The top contributors to the models were drastically different for several of these
clades (Fig. 4). For instance, the topography variables had strong influences on Boophis richness,
9
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
with a joint contribution of 24.5%, but contributed much less to explaining the patterns of most
other groups. Further, the Sanctuary hypothesis had a strong influence on the Brookesia and
iguana models, though it contributed very little to the predictions of endemism in Boophis and
Phelsuma. Mid-domain Effects were apparent in most models but the sign on the correlation and
the contribution of each Mid-domain hypothesis varied considerably.
DISCUSSION
The results of this study clearly demonstrate that single-mechanism explanatory
hypotheses of spatial patterning in Madagascar's herpetofauna (and presumably, other Malagasy
vertebrates) are inadequate. Thus, we propose a novel method for examining and synthesizing
spatial parameters such as species richness, endemism, and community similarity. In this
framework, biogeographic hypotheses are explanatory variables. The resulting mixed-model
geospatial approach to biogeography analyses is both more robust, and more realistic. Our
approach has the potential to reduce bias and subjectivity in the search for prevalent factors
influencing the distribution of biodiversity, both in Madagascar and elsewhere. Currently,
researchers typically approach such questions by univariate and sometimes narrative analyses
that examine the fit of the observed patterns to only single explanatory models or mechanisms
(e.g. in Madagascar33,35,41) or compare a limited number of competing variables in univariate
approaches37. Such analyses are hampered, however, by spatial autocorrelation of biodiversity
patterns and predictor variables thereby inflating type-I errors in traditional statistical tests42,43.
Several solutions have been proposed for this problem. Some authors attempt to exclude spatial
10
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
autocorrelation from models44, whereas others incorporate spatial autocorrelation as a predictive
parameter in geospatial models45-47, as was applied in this study.
The results obtained here for some sub-clades are in agreement with previous analyses,
while others are not. For example, the high influence of the mid-domain effect on Boophis
treefrogs, one of the most species-rich frog genera in Madagascar, agrees with a previous
analysis performed by Colwell & Lees48 for all Malagasy amphibians (with a high representation
of Boophis). On the contrary, the negative contributions of the mid-domain effects on the
biodiversity patterns of the other genera in the analysis are obvious given that their centers of
richness and endemism are in either southern or northern Madagascar, but not in central parts of
the island. Previous studies postulated a high influence of topography on the diversification of
leaf chameleons (Brookesia),41,49 though this is not supported by our analysis. This latter example
exemplifies a dilemma of scale, inherent in all comparisons of spatial data sets. In fact, the
distribution of Brookesia is highly specific to certain mountain massifs in northern Madagascar
while the genus is largely absent from the equally topographically heterogeneous south-east.
This absence is probably due to its evolutionary history, with a diversification mainly in the
north and limited capacity for range expansion41. This historical distribution pattern probably
accounts for low influence of the topographic hypotheses on Madagascar-wide Brookesia
richness and endemism, while at a smaller spatial scale (northern Madagascar) these hypotheses
might well have a strong predictive value.
While patterns of richness and endemism of the Malagasy herpetofauna have been
analyzed several times for various purposes based on partial data sets8,35,37,41,48 the analysis of
turnover of species composition and the definition of biogeographic regions following from such
explicit analyses are still in their infancy. For reptiles, Angel's50 proposal of biogeographic
11
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
regions based on classical phytogeography, i.e., regions based on plant community
composition51, has usually been adopted52. Later, Schatz53 refined this zonation of Madagascar
based on explicit bioclimatic analyses, and Glaw & Vences54 proposed a detailed geographical
zonation based on the areas of endemism of Wilmé33. The GDM approach herein is the first
explicit analysis of a large herpetofaunal dataset to geographically delimit regions distinguished
by abrupt changes in their amphibian and reptile communities. This model turned out to agree
remarkably well with classical bioclimatic and phytogeographic zonations of Madagascar51,53,
strongly correlated to climatic explanatory variables (Fig. 3). Especially in the 4-classes GDM,
the regions almost perfectly correspond with those proposed by Schatz53 based on bioclimate,
i.e., eastern humid, central highland/montane, western arid, southwestern subarid zone. Although
the coincidence of the precise boundaries of these regions might be methodologically somewhat
biased, as we interpolated community distribution using climate variables in the analysis, the
model is still mainly based on real distributional information of species and thus provides
convincing evidence that amphibian and reptile communities strongly differ among the major
bioclimatic zones of Madagascar.
Several authors have suggested that the current distribution of biotic diversity in the
tropics resulted from a complex interplay of a variety of diversification mechanisms55,56. This
implies that no single hypothesis adequately explains the diversification of broad taxonomic
groups — our results support this assumption. Richness, endemism and turnover of large and
heterogeneous groups exemplified by the all-species amphibian and reptile data sets were in all
cases best explained by complex CAR models. These models have the advantage of
incorporating most or all of the originally included explanatory variables.
12
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
Several alternative explanations may account for this outcome. Patterns of biodiversity
may not be strongly correlated to any of the predictor mechanisms simply because none of them
provide the causal mechanism underlying the diversification processes. As another consideration,
spatial predictions of some of the biodiversity hypotheses may have been inaccurate, though we
took great care to avoid such mistakes. In any event, improvements in these methods may result
in different outcomes in future analyses.
Caveats aside, the results of this study almost certainly support a third explanation, that
different clades of organisms are each predominantly influenced by a different set of
diversification mechanisms. In turn, these are driven by intrinsic factors, such as morphological
or physiological constraints, or to extrinsic factors, such as an initial diversification in an area
characterized by a certain topography, climate, or biotic composition. This alternative is
supported by the observation that the patterns of several of the smaller subgroups in our analysis
were indeed best explained by opposing predominant variables, e.g., topographic heterogeneity
and museum (Boophis endemism) vs. climate stability and sanctuary (Brookesia endemism). An
overarching message is that the taxonomic scale of analysis is of extreme importance when
attempting to derive global explanations of biodiversity distribution patterns. Including too many
taxa will blur the existing differences among clades and lead to complex explanatory models,
whereas patterns within specific clades may be best explained by simple models.
The method proposed herein allows for a more objective quantification of the influences
of particular diversification mechanisms on biodiversity patterns, compared to traditional,
univariate approaches. Further developments of the method should especially focus on including
a phylogenetic dimension, and when appropriate (for predictor hypotheses), a temporal
component. Geospatial analyses of biodiversity pattern typically use species as equivalent and
13
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
independent data points, though in reality, they are entities with substantial variation in
parameters such as evolutionary age, dispersal capacity and population density, and with
different degrees of relatedness depending on their position in the tree of life. This multilayered
information can be included in various ways in the CAR/OTBC approach, e.g. by plotting
richness and endemism of evolutionary history rather than taxonomic identity, calculating
turnover only for sister species with adjacent ranges, or repeating the calculations for sets of
species defined by particular nodes on a phylogenetic tree. This latter approach— iterating the
analysis for successively more inclusive clades — appears particularly promising for identifying
those moments in evolutionary history wherein shifts in prevalent diversification mechanisms
have occurred. Given this perspective, we can begin to tease apart the diversification histories of
individual clades versus prevailing biogeoclimatic events that shape entire biotas.
MATERIALS AND METHODS
Biodiversity Estimates
Species Distribution Modeling
Species data consisted of 8362 occurrence records of 745 Malagasy amphibian and
reptile species (325 and 420 species, respectively). Species distribution models (SDMs) were
limited to species that had, at minimum, 3 unique occurrence points at the 30 arc-second spatial
resolution (ca. 1 km). The reduced dataset represented 453 species (consisting of 5440 training
points of 248 reptile and 205 amphibian species), with a mean of 12 training points per species
14
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
(max= 131). For 107 amphibian and 119 reptile species with only 1-2 occurrence records a 10km
buffer was applied to point localities in place of SDM. The species distribution models were
generated in MaxEnt v3.3.3e57 using the following parameters: random test percentage = 25,
regularization multiplier = 1, maximum number of background points = 10000, replicates = 10,
replicated run type = cross validate.
One limitation of presence-only data SDM methods is the effect of sample selection bias,
where some areas in the landscape are sampled more intensively than others58. MaxEnt requires
an unbiased sample. To account for sampling biases, we used a bias file representing a Gaussian
kernel-density of all species occurrence localities. The bias file upweighted presence-only data
points with fewer neighbors in the geographic landscape59. Species distributions were modeled
for the current climate using the 19 standard bioclimatic variables derived from interpolation of
climatic records between 1950 and 2000 from weather stations around the globe (Worldclim
1.460). Non-climatic variables geology, aspect, elevation, solar radiation, and slope were also
included61,62. All layers were projected to Africa Alber’s Equal-Area Cylindrical projection in
ArcMap at a resolution of 0.91 km2.
Correcting SDMs for Over-prediction
To limit over-prediction of SDMs, a problem common with modeling distributions of
Madagascar biota8,37, we clipped each SDM following the approach of Kremen et al.8. This
method produces models that represent suitable habitat within an area of known occurrence
(based on a buffered MCP), excluding suitable habitat greatly outside of observed range. The
size of the buffer was based on the area of the MCP. We used buffer distances of 20km, 40km,
and 80km, respectively, for three MCP area classes, 0-200km2, 200-1000 km2, and >1000 km2.
15
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
All corrected SDMs were proofed by taxonomic experts to ensure reliability; if a model did not
tightly match knowledge of areas where distributions were well documented, or if little prior
information existed regarding a species distribution, or taxonomy was convoluted, and because
of, its expected distribution could not be evaluated, the species was excluded from analyses (n=
71).
Range Sizes, Species Richness and Corrected Weighted Endemism
For descriptive range-size statistics, distribution range-sizes were sampled for all species
at 0.01 degrees2 from corrected SDMs (or buffered point data where applicable) and a student’s
t-test with unequal variance was performed between amphibian and reptile species. To assess
differences in the frequency of microendemics among the two groups, we converted all
distributions that were > or < than 1000 km2 to a value of 0 and 1, respectively. We then
calculated the mean frequency for both groups and ran a binomial test among both groups.
Species richness was calculated separately for amphibians and reptiles by summing the
respective corrected binary SDMs (based on a maximum training sensitivity plus specificity
threshold) and, for species with 1-2 occurrence records, buffered points in ArcGIS. This
provided a high resolution estimate of richness that is less affected by spatial scale and
incomplete sampling than traditional measurements based solely on occurrence records.
Measures of endemism are inherently dependent on spatial scale. We chose a grid scale
of 82 x 63 km, separating Madagascar into 24 latitudinal and 8 longitudinal rows, to reduce
problems associated with estimating endemism over too small or large areas12,35. Endemism was
measured as corrected weighted endemism (CWE), where the proportion of endemics are
inversely weighted by their range size (species with smaller ranges are weighted more than those
16
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
with large63) and this value divided by the local species richness12. CWE emphasizes areas that
have a high proportion of animals with restricted ranges, but not necessarily high species
richness. We calculated CWE separately for reptiles and amphibians using SMDtoolbox v1
(Brown in review).
Generalized Dissimilarity Modeling
Generalized Dissimilarity Modeling (GDM) is a statistical technique extended from
matrix regressions designed to accommodate nonlinear data commonly encountered in ecological
studies38. One use of GDM is to analyze and predict spatial patterns of turnover in community
composition across large areas. In short, a GDM is fitted to available biological data (the absence
or presence of species at each site and environmental and geographic data) then compositional
dissimilarity is predicted at unsampled localities throughout the landscape based on
environmental and geographic data in the model. The result is a matrix of predicted
compositional dissimilarities (PCD) between pairs of locations throughout the focal landscape.
To visualize the PCD, multidimensional scaling was applied, reducing the data to three
ordination axes, and in a GIS each axis was assigned a separate RGB color (red, green or blue).
Due to computation limitations associated with pairwise comparisons of large datasets,
we could not predict composition dissimilarities among all sites in our high resolution
Madagascar dataset. To address this, we randomly sampled 2500 points throughout Madagascar
from a ca. 10 km2 grid. We then measured the absence or presence of each of the 679 species at
each locality. We used the same high resolution environmental and geography data used in the
SDM. These 23 layers were reduced to nine vectors in a principal component analyses which
represented 99.4% of the variation of the original data. These data were sampled at the same
17
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
2500 localities. Both data (species presence and environmental data) were input into a
generalized dissimilarity model using the R package: GDM R distribution pack v1.1
(www.biomaps.net.au/gdm/GDM_R_Distribution_Pack_V1.1.zip). We then extrapolated the
GDM into the high resolution climate dataset by assigning ordination scores using k-nearest
neighbor classification (k=3, numeric Manhattan distance), calculating each ordination axes
independently38.
The continuous GDM was transformed into a model with four major classes, and each of
these was then classified separately into 3-5 minor classes. The numbers of major and minor
classes were based on hierarchical cluster analyses in in SPSS v1964 using a “bottom up”
approach. The number of classes equaled the number of dendrogram nodes with relative
distances (scaled from 0-1) at 0.71 and 0.63 for major and minor groups, respectively. The
distance cut off can be somewhat arbitrary, however in our data there were obvious
discontinuities (long dendrogram branches between nodes) at these two values. The resulting
classified models were interpolated into high resolution climate space using a k-nearest neighbor
classification as described above.
Biogeography hypotheses
We examined which specific spatial predictions for the three biodiversity patterns:
species richness, endemism and/or in areas of endemism (AOE- the coincident restrictedness of
taxa) in Madagascar could be derived from each of 12 biogeography hypotheses, and then
translated these predictions into spatial models in a GIS.
In a GIS, spatially explicit predictions of the three biodiversity patterns (species richness,
endemism or areas of endemism) were estimated for each biogeography hypothesis. For some of
18
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
the hypotheses not all three metrics of biodiversity were calculated due to lacking, or incomplete,
expectations (e.g. not all hypothesis make predictions about AOE). Because of these incomplete
biodiversity pattern predictions, comparisons among hypotheses are statistically non-trivial. This
is in part because few diversification hypotheses capture all facets of biodiversity (species
richness, endemism, AOE). Further, many estimates of biodiversity patterns rely on components
of climate or geography, thus some are based on the same data and are not entirely independent
of each other. Each hypothesis was generated at the spatial resolution of 30 arc-seconds
(matching the resolution of GDM and species richness estimates, later transformed to 0.91 km2).
For the endemism analyses, each biogeography hypothesis was upscaled to match resolution of
the endemism analyses by averaging all values encompassed in cell.
Spatial Statistics
The spatial predictions derived from the various biodiversity hypotheses resulted in either
continuous or nominal categorical data. Conducting statistical tests between data types is
nontrivial and, in some cases, not logical or impossible. Spatial data are represented in GIS by
two different formats: raster and vector. Geospatial raster data are composed of equal sized
squares, tessellated in a grid, with each cell representing a value (often continuous data), such as
elevation. Spatial vector data (commonly called ‘shapefiles’) can be represented by points, lines,
or polygons, such as: localities, roads and countries, respectively. Vector data are non-
topological and represent discrete features. They are often used to depict nominal data, where the
relationship of data categories to others is unknown or non-linear.
Raster data can be converted to vector data (and vice versa) and the data type (i.e.
nominal or continuous) may or may not change when converted. For example, in some cases
19
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
continuous data can be converted to ordered categories (ordinal data) when converted from raster
to polygon. However if the same data were converted back to a raster file, it would remain
categorical data due to data loss in the first conversion. Regardless of GIS data format, statistical
tests need be chosen according to the data types, however GIS data format remain equally
important, as often a single data format is required to perform a spatial statistic of interest (i.e.
software input limitations).
Analyses- Continuous Data
To assess a global measurement of correlation between continuous data, we calculated
Pearson correlations following the unbiased correlation method of Dutilleul40 and using the
software Spatial Analysis in Macroecology65.
Analyses- Nominal Categorical Data
Comparisons of nominal categorical spatial data (i.e. AOE predictions compared to
classified GDM) focused on the spatial distributions of the borders between the subunits. We
used the following methods to measure similarities and significance: (1) border overlap, and (2)
Pearson correlation coefficients (r) with Dutilleul’s spatial correlation (see above).
(1) Border overlap was calculated by sampling the landscape at 1 km resolution for the
presence of a border. If present, a point was placed. We then measured the spatial overlap of the
sampled borders of two landscapes. In all analyses, a 10 km buffer was applied to the overlap
calculation, and points datasets that overlap by 10km or less are were considered overlapping
boundaries. To account for differences in the level of subdivision of layers, overlap was
converted to a percentage and averaged for both layers being compared. Country outline was
excluded from all comparisons and thus, only intra-country boundaries were compared.
20
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
(2) To assess global correlation between two polygon shapefiles, each shapefile was
converted to a distance raster, measuring the closest distance from any point in the landscape to a
boundary. Using these layers we measured a Pearson correlation (unbiased correlation after
Dutilleul40), where high correlation coefficients represent two landscapes that have congruent
areas that are isolated from boundaries and others congruent areas that are adjacent to
boundaries. Each distance landscape was evenly sampled by 2000 points and correlations were
assessed on the values of these points.
Analyses- Mixed Models of Continuous Data
To determine the influence of each biogeography hypothesis in predicting the observed
biodiversity patterns, we integrated all continuous biogeography hypotheses into a single mixed
conditional autoregression model (CAR) using the software Spatial Analysis in Macroecology65.
To normalize the predictor variables, Box-Cox transformations (Box and Cox 1964) were
performed. The lambda parameter was estimated by maximizing the log-likelihood profile in R
package GeoR47. A Gabriel connection matrix was used to describe the spatial relationship
among sample points66. Using Gabriel networks, short connections between neighboring points,
is preferable (i.e. more conservative67) than using inverse-decaying distances because in most
empirical datasets the residual spatial autocorrelation tends to be stronger at smaller distance
classes68.
The main goal of our mixed spatial analyses were to determine the combination of
biogeography hypotheses that best predict the observed biodiversity patterns. If each explanatory
variable was incorporated natively, due to considerable multi-colinearity, often only a few
variables would end up contributing to a majority of the model. To estimate the true contribution
of each hypothesis in context of a mixed model (even if highly correlated to others), we
21
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
developed a novel approach that removes colinearity from the response variables (but in the
process explicit variable identity is temporarily lost). The transformed response variables are
then run in a CAR analysis and the resulting standardized model contributions are then
transformed back into original response variable identities; reflecting the relative contribution of
each in the model. This process is casually referenced here as Orthogonally Transformed Beta
Coefficients (OTBCs).
Orthogonally Transformed Beta Coefficients
Each biogeography hypothesis is standardized from zero to one. This ensured that the
component loadings reflected the relative contribution of each biogeography hypothesis. A
principal component analysis was performed on the standardized biogeography hypotheses using
a covariance matrix. All the resulting principal components (PCs) were extracted and then loaded
as explanatory variables in the CAR model. The CAR analyses were run iteratively, starting with
all PCs as response variables and then excluding each PC that did not contribute significantly to
the model (α = 0.05) until the final model included only PCs that contributed significantly to the
model. Because each PC represented a linearly uncorrelated variable, only the relevant,
independent data were incorporated into the final CAR model. The resulting standardized beta
coefficients (βj from the CAR analyses, Fig. 1 and Equation 1) were then multiplied by the value
of the corresponding component loadings (αij from the PCA, see Equation 1). The absolute value
of the product reflects the relative contributions of each biogeography hypothesis to each PC,
which are weighted by the PC’s contribution in the CAR model (herein termed the weighted
component loadings or WCLif , Equation 1). The weighted component loadings (WCLif, Equation
1) were then summed for each biogeography hypothesis across all PCs (Hi) and depict the
contributions of each hypothesis in the CAR model. The H i value was then converted to
22
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
percentages (HPi) to allow comparison among all CAR analyses. A positive or negative
correlation was determined for each biogeography hypothesis by running a separate CAR
analysis using the raw biogeography variables as a single response variable (all other parameters
were matched).
Equation 1
WCLij=|β j|×|α ij|H i=∑i
WCLij H All=∑ij
WCLij H Pi=( H i
Hall)∗100
Acknowledgments
We are grateful to numerous friends and colleagues who provided invaluable assistance
during fieldwork and previous discussions of Madagascar's biogeography, we would like to
particularly thank Franco Andreone, Parfait Bora, Christopher Blair, Lauren Chan, Sebastian
Gehring, Frank Glaw, Steve M. Goodman, Jörn Köhler, Peter Larsen, David C. Lees, Brice P.
Noonan, Maciej Pabijan, Ted Townsend, Krystal Tolley, Roger Daniel Randrianiaina
Fanomezana Ratsoavina, David R. Vieites, and Katharina C. Wollenberg. Fieldwork of MV was
funded by the Volkswagen Foundation. JLB was supported by the National Science Foundation
under Grant No. 0905905 and by Duke University start-up funds to ADY.
23
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
References
1 Kent, M. Biogeography and macroecology. Progress in Physical Geography 29, 256-264, doi:10.1191/0309133305pp447pr (2005).
2 Beck, J. et al. What's on the horizon for macroecology? Ecography 35, 673-683, doi:10.1111/j.1600-0587.2012.07364.x (2012).
3 Whittaker, R. H. Vegetation of the Siskiyou Mountains, Oregon and California. Ecol Monogr 30, 279-338, doi:10.2307/1943563 (1960).
4 Whittaker, R. H. Evolution and Measurement of Species Diversity. Taxon 21, 213-251, doi:10.2307/1218190 (1972).
5 Williams, P. H. Mapping variations in the strength and breadth of biogeographic transition zones using species turnover. Proceedings of the Royal Society B-Biological Sciences 263, 579-588, doi:10.1098/rspb.1996.0087 (1996).
6 Kreft, H. & Jetz, W. A framework for delineating biogeographical regions based on species distributions. Journal of Biogeography 37, 2029-2053, doi:10.1111/j.1365-2699.2010.02375.x (2010).
7 Holt, B. G. et al. An Update of Wallace's Zoogeographic Regions of the World. Science 339, 74-78, doi:10.1126/science.1228282 (2013).
8 Kremen, C. et al. Aligning conservation priorities across taxa in Madagascar with high-resolution planning tools. Science 320, 222-226, doi:10.1126/science.1155193 (2008).
9 Hoffmann, M. et al. The Impact of Conservation on the Status of the World's Vertebrates. Science 330, 1503-1509, doi:10.1126/science.1194442 (2010).
10 Platnick, N. I. On areas of endemism. Australian Systematic Botany 4, 2pp.-2pp. (1991).11 Harold, A. S. & Mooi, R. D. Areas of endemism: definition and recognition criteria.
Systematic Biology 43, 261-266, doi:10.2307/2413466 (1994).12 Crisp, M. D., Laffan, S., Linder, H. P. & Monro, A. Endemism in the Australian flora.
Journal of Biogeography 28, 183-198 (2001).13 Terborgh, J. & Winter, B. A method for siting parks and reserves with special reference
to Columbia and Ecuador. Biological Conservation 27, 45-58 (1983).14 Ackery, P. R. & Vane-Wright, R. I. Milkweed butterflies: Their cladistics and biology.
(British Museum of Natural History and Cornell University Press, 1984).15 Hawkins, B. A. et al. Energy, water, and broad-scale geographic patterns of species
richness. Ecology 84, 3105-3117, doi:10.1890/03-8006 (2003).16 Jetz, W., Rahbek, C. & Colwell, R. K. The coincidence of rarity and richness and the
potential signature of history in centres of endemism. Ecology Letters 7, 1180-1191, doi:10.1111/j.1461-0248.2004.00678.x (2004).
17 Carnaval, A. C., Hickerson, M. J., Haddad, C. F. B., Rodrigues, M. T. & Moritz, C. Stability Predicts Genetic Diversity in the Brazilian Atlantic Forest Hotspot. Science 323, 785-789, doi:10.1126/science.1166955 (2009).
18 Kozak, K. H., Graham, C. H. & Wiens, J. J. Integrating GIS-based environmental data into evolutionary biology. Trends in Ecology & Evolution 23, 141-148, doi:10.1016/j.tree.2008.02.001 (2008).
19 Lamoreux, J. F. et al. Global tests of biodiversity concordance and the importance of endemism. Nature 440, 212-214, doi:10.1038/nature04291 (2006).
24
525
526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567
20 Linder, H. P. et al. The partitioning of Africa: statistically defined biogeographical regions in sub-Saharan Africa. Journal of Biogeography 39, 1189-1205, doi:10.1111/j.1365-2699.2012.02728.x (2012).
21 Olivero, J., Márquez, A. L. & Real, R. Integrating fuzzy logic and statistics to improve the reliable delimitation of biogeographic regions and transition zones. Systematic biology 62, 1-21 (2013).
22 Graham, C. H., Moritz, C. & Williams, S. E. Habitat history improves prediction of biodiversity in rainforest fauna. Proceedings of the National Academy of Sciences of the United States of America 103, 632-636, doi:10.1073/pnas.0505754103 (2006).
23 Vences, M., Wollenberg, K. C., Vieites, D. R. & Lees, D. C. Madagascar as a model region of species diversification. Trends in Ecology & Evolution 24, 456-465, doi:10.1016/j.tree.2009.03.011 (2009).
24 Goodman, S. M. & Benstead, J. P. The natural history of Madagascar. (University of Chicago Press Chicago, 2003).
25 Yoder, A. D. & Nowak, M. D. Has vicariance or dispersal been the predominant biogeographic force in Madagascar? Only time will tell. Annual Review of Ecology, Evolution, and Systematics, 405-431 (2006).
26 Crottini, A. et al. Vertebrate time-tree elucidates the biogeographic pattern of a major biotic change around the K-T boundary in Madagascar. Proceedings of the National Academy of Sciences of the United States of America 109, 5358-5363, doi:10.1073/pnas.1112487109 (2012).
27 Samonds, K. E. et al. Spatial and temporal arrival patterns of Madagascar's vertebrate fauna explained by distance, ocean currents, and ancestor type. Proceedings of the National Academy of Sciences of the United States of America 109, 5352-5357, doi:10.1073/pnas.1113993109 (2012).
28 Yoder, A. D. et al. A multidimensional approach for detecting species patterns in Malagasy vertebrates. Proceedings of the National Academy of Sciences of the United States of America 102, 6587-6594, doi:10.1073/pnas.0502092102 (2005).
29 Pastorini, J., Thalmann, U. & Martin, R. D. A molecular approach to comparative phylogeography of extant Malagasy lemurs. Proceedings of the National Academy of Sciences of the United States of America 100, 5879-5884, doi:10.1073/pnas.1031673100 (2003).
30 Goodman, S. M. & Ganzhorn, J. U. Biogeography of lemurs in the humid forests of Madagascar: the role of elevational distribution and rivers. Journal of Biogeography 31, 47-55, doi:10.1111/j.1365-2699.2004.00953.x (2004).
31 Yoder, A. D. & Heckman, K. L. Mouse lemur phylogeography revises a model of ecogeographic constraint in Madagascar. (2006).
32 Dewar, R. E. & Richard, A. F. Evolution in the hypervariable environment of Madagascar. Proceedings of the National Academy of Sciences of the United States of America 104, 13723-13727, doi:10.1073/pnas.0704346104 (2007).
33 Wilme, L., Goodman, S. M. & Ganzhorn, J. U. Biogeographic evolution of Madagascar's microendemic biota. Science 312, 1063-1065, doi:10.1126/science.1122806 (2006).
34 Wollenberg, K. C., Vieites, D. R., Glaw, F. & Vences, M. Speciation in little: the role of range and body size in the diversification of Malagasy mantellid frogs. Bmc Evolutionary Biology 11, doi:10.1186/1471-2148-11-217 (2011).
25
568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612
35 Wollenberg, K. C. et al. Patterns of endemism and species richness in Malagasy cophyline frogs support a key role of mountainous areas for speciation. Evolution; international journal of organic evolution 62, 1890-1907, doi:10.1111/j.1558-5646.2008.00420.x (2008).
36 Pabijan, M., Wollenberg, K. C. & Vences, M. Small body size increases the regional differentiation of populations of tropical mantellid frogs (Anura: Mantellidae). Journal of evolutionary biology 25, 2310-2324, doi:10.1111/j.1420-9101.2012.02613.x (2012).
37 Pearson, R. G. & Raxworthy, C. J. The evolution of local endemism in madagascar: watershed versus climatic gradient hypotheses evaluated by null biogeographic models. Evolution; international journal of organic evolution 63, 959-967, doi:10.1111/j.1558-5646.2008.00596.x (2009).
38 Ferrier, S., Manion, G., Elith, J. & Richardson, K. Using generalized dissimilarity modelling to analyse and predict patterns of beta diversity in regional biodiversity assessment. Diversity and Distributions 13, 252-264, doi:10.1111/j.1472-4642.2007.00341.x (2007).
39 Allnutt, T. F. et al. A method for quantifying biodiversity loss and its application to a 50-year record of deforestation across Madagascar. Conservation Letters 1, 173-181, doi:10.1111/j.1755-263X.2008.00027.x (2008).
40 Dutilleul, P., Clifford, P., Richardson, S. & Hemon, D. Modifying the t test for assessing the correlation between two spatial processes. Biometrics, 305-314 (1993).
41 Townsend, T. M., Vieites, D. R., Glaw, F. & Vences, M. Testing Species-Level Diversification Hypotheses in Madagascar: The Case of Microendemic Brookesia Leaf Chameleons. Systematic Biology 58, 641-656, doi:10.1093/sysbio/syp073 (2009).
42 Kreft, H. & Jetz, W. Global patterns and determinants of vascular plant diversity. Proceedings of the National Academy of Sciences of the United States of America 104, 5925-5930, doi:10.1073/pnas.0608361104 (2007).
43 Hoeting, J. A. The importance of accounting for spatial and temporal correlation in analyses of ecological data. Ecological Applications 19, 574-577, doi:10.1890/08-0836.1 (2009).
44 Ohlemuller, R., Walker, S. & Wilson, J. B. Local vs regional factors as determinants of the invasibility of indigenous forest fragments by alien plant species. Oikos 112, 493-501, doi:10.1111/j.0030-1299.2006.13887.x (2006).
45 Bacaro, G. & Ricotta, C. A spatially explicit measure of beta diversity. Community Ecology 8, 41-46 (2007).
46 Bacaro, G. et al. Geostatistical modelling of regional bird species richness: exploring environmental proxies for conservation purpose. Biodiversity and Conservation 20, 1677-1694 (2011).
47 Diggle, P. J. & Ribeiro, P. J. J. Model-based Geostatistics. (Springer, 2007).48 Colwell, R. K. & Lees, D. C. The mid-domain effect: geometric constraints on the
geography of species richness. Trends in Ecology & Evolution 15, 70-76, doi:10.1016/s0169-5347(99)01767-x (2000).
49 Raxworthy, C. J. & Nussbaum, R. A. Systematics, speciation and biogeography of the dwarf chameleons (Brookesia; Reptilia, Squamata, Chamaeleontidae) of northern Madagascar. Journal of Zoology 235, 525-558 (1995).
50 Angel, F. Les Lézards de Madagascar. (Academie Malgache, 1942).
26
613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657
51 Humbert, H. Les territoires phytogéographiques de Madagascar. Année Biologique 31, 439–448 (1955).
52 Glaw, F. & Vences, M. Amphibians and Reptiles of Madagascar. (Vences, M. and Glaw Verlags, F. GbR., 1994).
53 Schatz, G. E. in Diversity and Endemism in Madagascar (eds W.R. Lourenço & S.M. Goodman) 1–9 (Société de Biogéographie, MNHN, ORSTOM, 2000).
54 Glaw, F. & Vences, M. Field Guide to the Amphibians and Reptiles of Madagascar. Third Edition edn, (Vences and Glaw Verlag, 2007).
55 Bush, M. B. Amazonian speciation: a necessarily complex model. Journal of Biogeography 21, 5-17, doi:10.2307/2845600 (1994).
56 Oneal, E., Otte, D. & Knowles, L. L. Testing for biogeographic mechanisms promoting divergence in Caribbean crickets (genus Amphiacusta). Journal of Biogeography 37, 530-540, doi:10.1111/j.1365-2699.2009.02231.x (2010).
57 Phillips, S. J., Anderson, R. P. & Schapire, R. E. Maximum entropy modeling of species geographic distributions. Ecological Modelling 190, 231-259, doi:10.1016/j.ecolmodel.2005.03.026 (2006).
58 Phillips, S. J. et al. Sample selection bias and presence-only distribution models: implications for background and pseudo-absence data. Ecological Applications 19, 181-197, doi:10.1890/07-2153.1 (2009).
59 Elith, J. et al. A statistical explanation of MaxEnt for ecologists. Diversity and Distributions 17, 43-57, doi:10.1111/j.1472-4642.2010.00725.x (2011).
60 Hijmans, R. J., Cameron, S. E., Parra, J. L., Jones, P. G. & Jarvis, A. Very high resolution interpolated climate surfaces for global land areas. International Journal of Climatology 25, 1965-1978, doi:10.1002/joc.1276 (2005).
61 Moat, J. & Du Puy, D. (ed Kew Royal Botanic Gardens) (1997).62 Jarvis, A., Reuter, H. I., Nelson, A. & Guevara, E. (CGIAR-CSI SRTM 90m Database,
2008).63 Williams, P. H. Some properties of rarity scores for site-quality assessment. British
Journal of Entomology and Natural History 13, 73-86 (2000).64 IBM SPSS Statistics for Windows v. 19.0 (IBM Corporation, Armonk, NY, 2010).65 Rangel, T. F., Diniz-Filho, J. A. F. & Bini, L. M. SAM: a comprehensive application for
Spatial Analysis in Macroecology. Ecography 33, 46-50, doi:10.1111/j.1600-0587.2009.06299.x (2010).
66 Legendre, P. & Legendre, L. Numerical ecology. Vol. 2nd (Elsevier, 1998).67 Griffith, D. A. in Practical handbook of spatial statistics (ed S. L. Arlinghaus) 65-82
(CRC Press, 1996).68 Bini, L. M. et al. Coefficient shifts in geographical ecology: an empirical evaluation of
spatial and non-spatial regression. Ecography 32, 193-204, doi:10.1111/j.1600-0587.2009.05717.x (2009).
27
658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697
698
699
700
Figure 1. Overview of work protocol and dataflow. Three types of original data were input
into the analyses: (1) biogeography hypotheses, (2) geography and climate data and (3) species
locality data. These data were used to predict distribution of species, and the distribution models
used to calculate biodiversity patterns (species richness, corrected weighted endemism,
turnover). We then tested for correlation of these biodiversity patterns with spatial predictions
28
701
702
703
704
705
706
derived from biogeography hypotheses, and used a multivariate model to simultaneously test the
influences of these hypotheses on the biodiversity patterns. *The response variables constituted
standardized principal components of the raw biogeography hypotheses ** The CAR models
were iterated until only response variables that contributed significantly the model were
included.
29
707
708
709
710
711
712
713
714
715
Figure 2. Observed Biodiversity Data. Reptile and amphibian species richness (A & E)
measures the number of species present. Endemism (B & F), based on a corrected weighted
calculation, reflects the proportion of unique species present within certain areas. The
generalized dissimiliarity model (GDM) analyzes compositional turnover of communities (here
jointly for amphibians and reptiles) and predicts dissimilarity throughout the landscape based on
an interpolation based on variation in climate and geographic data. The continuous GDM (G) can
then be classified into 4 major and 15 minor areas of endemism (H & C).
30
716
717
718
719
720
721
722
723
Figure 3. Explanatory contribution of continuous biogeography hypotheses to a conditional
autoregressive spatial model of each observed biodiversity measurements. Only hypotheses
contributing >5% are shown; see Supplementary Table S4 for complete and detailed data). Mid-
domain: I. latitude, II. longitude, III. distance. Climate-Gradient: I. PC1, II. PC2, III. PC3.
Climate-Stability: I. Precipitation Stability, II. Climate Stability (temperature and precipitation).
Topography: I. Topographic Heterogeneity, II. Disturbance-vicariance, III. Montane Species
Pump. An asterisk marks hypotheses that contributed negatively to the mixed CAR model.
31
724
725
726
727
728
729
730
731
732
733
734
735
736
737
Figure 4. Contribution of continuous biogeography hypotheses to a conditional
autoregressive spatial model of species richness and endemism for four focal groups. Only
hypotheses contributing >5% are shown; see Supplementary Table S4 for complete and detailed
data). Mid-domain: I. latitude, II. longitude, III. distance. Climate-Gradient: I. PC1, II. PC2, III.
PC3. Climate-Stability: I. Precipitation Stability, II. Climate Stability (temperature and
precipitation). Topography: I. Topographic Heterogeneity, II. Disturbance-vicariance, III.
Montane Species Pump. An asterisk marks hypotheses that contributed negatively to the mixed
32
738
739
740
741
742
743
744
745
746
CAR model. (1) Brookesia chameleons (number of species = 27, number of original distribution
points = 178). (2) Boophis treefrogs (n of spp.= 77, n of pts.=460) (3) Phelsuma day geckos (n of
spp.= 28, n of pts.= 304) (4) oplurid iguanas (Oplurus plus the monotypic Chalarodon; n of
spp.= 7, n of pts.= 147). Subgroups and colors of pie charts as in Fig. 3. An asterisk marks
hypotheses that contributed negatively CAR. For all hypotheses (with exception of the climate-
gradient variables) a positive correlation was expected between biodiversity metrics.
33
747
748
749
750
751
752
753
754
755
756
Table 1. Correlations of nominal biodiversity hypotheses to Generalized Dissimilarity Models (upper: 15-class; lower: 4-class
transformation of the original model). The left three columns show calculations based on the percentage of overlapping cells between
boundaries. High mean values depict high levels of shared boundaries among GDM and AOE (derived from the respective
hypothesis). R-values reflect non-spatial Pearson product-moment correlation coefficients. To assess significance of raster data, we
used an unbiased correlation following the method of Dutilleul (1993) that reduces the degrees of freedom according to the level of
spatial autocorrelation between the two variables.
Hypothesis Percent overlapping cells
(10km buffer)
Correlation to Generalized
Dissimilarity Model
GDM 15
classesHypothesis Mean r F-stat df p
Riverine- Major 77.11% 35.31% 56.21% 0.427 13.984 62.537 <.001
Riverine- Minor 72.43% 49.41% 60.92% 0.503 28.599 84.415 <.001
Gradient 74.63% 67.51% 71.07% 0.403 24.974 128.641 <.001
Watershed 60.14% 25.92% 43.03% 0.229 5.887 106.824 0.017
34
757
758
759
760
761
762
763
764
765
GDM -4 100.00% 38.71% 69.35% 0.551 26.257 60.171 <.001
Riverine – Refuge 71.00% 20.69% 45.85% 0.440 11.450 47.693 0.001
GDM- 4
classesHypothesis Mean Pearson's r F-stat df p
Riverine- Major 41.6% 40.3% 40.9% 0.556 12.519 28.042 0.001
Riverine- Minor 39.4% 58.1% 48.7% 0.51 16.345 46.550 <.001
Gradient 39.6% 69.0% 54.3% 0.378 13.012 77.926 <.001
Watershed 33.2% 33.3% 33.2% 0.213 3.775 79.701 0.056
GDM -15 52.1% 100.0% 76.1% 0.551 26.257 60.171 <.001
Riverine – Refuge 30.3% 22.1% 26.2% 0.837 52.355 22.433 <.001
35
766
767
768
769
SUPPLEMENTARY MATERIALS
Hypotheses
Climate Stability Hypothesis (Fig S2.J)
Climate stability is thought to create greater climatic stratification across environmental
gradients (Dynesius and Jansson 2000, Jansson and Dynesius 2002). In stable climates, orbitally
forced species’ range dynamics (ORD) are low, allowing localized populations to persist, and
thus become highly specialized and differentiated. Like the Gradient Hypothesis (following), this
model also focuses on bioclimatic disparities, but additionally incorporates climate stability. This
hypothesis states that areas of climate stability, particularly those with climatic stratification,
should possess higher species richness and endemism.
To estimate this model for Madagascar, coliniarity was measured for all BIOCLIM layers
(Bio1–19) using a Pearson Coefficient. If R2 values exceeded 0.5, one of the layers was
excluded. We preferentially selected layers based on raw data (e.g. selecting mean annual
precipitation over seasonality). The following layers were excluded: Bio3, Bio7, Bio9, Bio13–
18. For each remaining BIOCLIM layer we calculated the standard deviation of each cell
throughout the four time periods for which climate data were available (0 kya, 6kya, 21 kya, 120
kya). The resulting standard deviation of each BIOCLIM layer was standardized to 1 to account
36
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
for different units in raw data. All standardized stability layers were summed to create the final
climate stability layer; with lower values representing higher climate stability through time.
Disturbance-Vicariance Hypothesis (Fig S2.I)
Under this model, the major factor contributing to diversification was temperature fluctuations
(Colinvaux 1993, Bush 1994, Haffer 1997), rather than fluctuations in precipitation and forest
fragmentation (as in the preceding Refuge and River hypotheses). This hypothesis states cyclic
fluctuations of temperature during the Quaternary caused reoccurring displacement of taxa
towards lower and higher altitudes (during cool and warm periods, respectively). Range
retractions of taxa into the highlands occasionally resulted in allopatric divergence into new
species. The gradual displacement of temperature specialized taxa would cause reoccurring
invasions and counter-invasions of heterogeneous landscapes by both montane and lowland taxa.
This hypothesis predicts species richness and endemism would be highest in areas of topographic
heterogeneity and temperature instability.
This hypothesis was generated by measuring colinearity between all BIOCLIM layers
corresponding to temperature (Bio1– Bio11, see Climate Stability above for details). The
following layers were excluded: Bio3, Bio7 and Bio11. For each remaining BIOCLIM layer we
calculated the standard deviation of each cell throughout the four time periods for which climate
data were available (0 kya, 6kya, 21 kya, 120 kya). This layer was standardized from 0 to 1 to
account for different units between layers. All standardized stability layers were summed to
create the final temperature stability layer with lower values representing higher temperature
stability. This layer was inverted and multiplied by a standardized version of the final
37
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
topographic heterogeneity layer (see above). Higher values represented areas with high
temperature instability through time and high topographic heterogeneity.
Gradient Hypothesis (Fig S1.E)
According to the Gradient Hypothesis, diversification of Malagasy taxa was driven by
bioclimatic disparities throughout the island (i.e. between the east and west), causing parapatry
of populations along environmental gradients. This hypothesis was first formulated by Endler
(1982), however more recently it was adapted and formulated in detail for Madagascar by
Vences et al. (2010). Vences et al. focused on the climatic stratification longitudinally between
the dry west and humid east of Madagascar, terming this species case the Ecogeographic
hypothesis. Under the Gradient hypothesis populations adapt to local ecotones, diverging from a
generalist ancestor (or one of broader ecological tolerance). Due to local specialization, gene
flow within ecological similar sites is higher than those distributed at ecologically different sites
across a gradient. The subdivision of populations create a scenario where drift or selection can
override gene flow among ecogeographic subpopulations (Fisher 1930) and daughter species
occupy separate, adjacent niches. The exact mechanism of speciation is controversial (e.g.
prezygotic isolation, behavioral isolation or reproductive barriers) and beyond the focus of this
study. This hypothesis invokes no barriers or mechanism of allopatry and predicts species
richness and endemism to be highest in areas of high bioclimatic stratification. This GIS
prediction was obtained from Pearson & Raxworthy (2009). In our continuous CAR-OBTC
analyses, this hypothesis was represented by the first 3 PCs from a PCA on the 19 current
BIOCLIM data for Madagascar (Hijmans et al. 2005).
38
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
Mid-domain Effect Hypothesis (Fig S2. A-C)
Initially described to explain latitudinal trends in species richness, this mathematical hypothesis
demonstrates that if species’ ranges are distributed randomly between northern and southern
geographic limits, the highest overlap of species ranges would be in the middle (Lees 1996, Lees
and Colwell 2007). For Madagascar, this hypothesis (in two dimensions) results in increased
overlap of species toward the center of country, over the Ankaratra highlands, resulting from the
sum of random overlapping species ranges. We explore four variants of this hypothesis: 1
dimension (the mid-domain of latitudinal, altitude or longitudinal gradients) and 2 dimensions, as
described above (the mid-domain of both latitudinal and longitudinal gradients). Note the mid-
domain of altitude is the same GIS calculation of the Museum hypothesis. Thus, throughout the
manuscript this hypothesis is referred to as the Museum hypothesis rather than the mid-domain
altitude hypothesis. In one dimension, the latitude hypothesis predicts that species richness
would be highest around 18° S, or in two dimensions, centered around 18° S and 46.5° E. The
mid-domain hypotheses are included as a null hypothesis for spatial variation in species richness;
these hypotheses invoke no barriers or ecological/ habitat heterogeneity and rely solely on the
random distribution of species in a defined geographic space.
Montane Species Pump Hypothesis (Fig S2.M)
This hypothesis predicts that montane regions have higher species richness because of their
topographic complexity and climatic zonation - both increase opportunities for allopatric and
parapatric speciation (e.g., Moritz et al. 2000; Rahbek and Graves 2001; Hall 2005; Fjeldsa°
and Rahbek 2006; Kozak and Wiens 2007). This hypothesis predicts that speciation should be
highest in areas of high topographic and climatic heterogeneity. Within those habitats, rates of
39
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
speciation should be highest in mid-elevations. This hypothesis predicts species richness and
endemism would be highest in areas of topographic and climatic heterogeneity.
To create a spatially explicit model of this hypothesis we calculated the mean standard
deviation of the first three climate PCs (based on Bioclim current data) at ca. 10 km2 square
neighborhood of each cell. Each was summed together and then the product was standardized
from 0-1. The resulting layer was then multiplied by a standardized (0-1) topographic variation
hypothesis. High values depict areas of high topographic and climatic heterogeneity.
Museum Hypothesis (Fig S2.D)
According to this hypothesis speciation occurred in montane habitats. In the Montane Museum
hypothesis, more species exist at intermediate elevation because these elevations were simply
occupied the longest, because of this, there has been more time for speciation and the
accumulation of species in these habitats when compared to habitats at lower and higher
elevations (Stebbins 1974; Stephens and Wiens 2003). This hypothesis predicts levels of
endemism and richness will be highest in the middle elevations. Note that in execution, the
prediction for this hypothesis is identical to a Mid-domain hypothesis of altitude. The GIS
representation of this hypothesis for Madagascar represents the median elevation (378m), where
the overlap of species should peak. This elevation was given a value of one and from this
elevation, values linearly transitioned to zero at the maximum and minimum elevations.
Paleogeographic Hypothesis
This hypothesis states that vicariate differentiation of Malagasy lineages is associated with
formation of geologic barriers to dispersal. Each hypothesis is specific to the focal
40
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
paleogeographic event. There are two major classes of geological events that have occurred since
the separation from India (ca. 65 MYA). The first, directly a result from separation from India, is
the orogeny of eastern escarpment. This hypothesis states that deep lineages, pre-65Mya, should
exist between the east and western species. A second, more local barrier is the volcanisms of
several regions: Tsaratanana, Manongarivo, Ampasindava (ca. 50 Mya), Ankaratra (phase 1 ca.
28 Mya, phase 2 ca.15 Mya, phase 3 ca. 2 Mya) and Ambre (ca. 2 Mya, Krause 2007). Clades
should exhibit breaks around volcanic activity, though given localization of these barriers; biota
in most cases should have been able to dispersal around, though perhaps experiencing reduced
gene flow. All paleogeographic hypotheses are difficult to test because more strongly than other
hypotheses, they are dependent on the location of ancestral populations and the timing of
geological events relative to a clade's diversification. Given the variation of ages of origins in
Madagascar's vertebrates, it is likely that some clades were affected by most of the major
paleogeographical events while others were only affected by some. Because of these factors, it is
unlikely that species exhibit congruent biogeographic patterns. Thus in this paper, we were
unable to test this hypothesis.
Refuge Hypothesis -(Fig S2.D) This hypothesis holds that episodic fragmentation of forests
resulted in isolated patches of wet forest and this caused vicariant differentiation between
adjacent patches. These transitions were driven by periodic changes (every 20-100 Ky) in
insolation associated with Milankovitch cycles (aberrations of the orbit of the earth around the
sun due to the slight asymmetrical shape of the earth). Recurrent changes in insolation caused
various dry periods followed by humid periods, particularly pronounced at tropical latitudes. The
evolutionary consequences of paleoecological changes in climate likely depend on the regional
41
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
topography, predominant weather patterns and the impact on the ecosystem (for example, a
slight reduction in rain may have different biological consequences in a spiny forest versus
rainforest). In Amazonia during the late Tertiary and Quaternary during dry climatic periods, it
has been argued that extensive humid forests survived due to subtle topographic variation that
facilitated rainfall gradients adjacent to the Andes, Guianan highlands, Rondonia and hilly areas
east of Pará (Haffer 1969; Vanzolinii1970, 1973; Brown et al., 1974; Prance, 1982, 1996).
In addition to changes in insolation associated with Milankovitch cycles, the gradual
drifting of Madagascar northward towards the equator likely led to habitat changes, presenting
another source of refugial isolation. Thus, it is plausible that refugia persisted and species with
narrow ecological niches became isolated and diverged into separate species. To create a
continuous spatial representation of this for Madagascar, first colinearity was measured for each
BIOCLIM precipitation layer (Bio12 – Bio19) using a Pearson Correlation. If R2 values
exceeded 0.5, one of the layers was excluded. This resulted in the exclusion of Bio13, Bio15,
Bio16 and Bio17. We preferentially selected layers based on raw data (e.g. selecting mean
precipitation over seasonality). For each remaining BIOCLIM layer, we calculated the standard
deviation of each cell throughout the four time periods for which climate data were available (0
kya, 6kya, 21 kya, 120 kya). This layer was standardized from 0 to 1 to account for different
units/decimal places in raw data between layers. All standardized stability layers were summed
to create the final precipitation stability layer with lower values representing higher precipitation
stability.
Riverine Barrier Hypothesis (Fig S1. A,B).
42
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
Malagasy rivers are thought to have acted as barriers separating populations, resulting in intra-
riverine species and subspecies. As the geology of Madagascar has remained relatively stable
since absolute isolation major rivers should have persisted (at least seasonally). Under this
hypothesis, we expect vicariate differentiation of Malagasy lineages associated with large
tributaries. There have been several criticisms of this hypothesis, such that rivers frequently
change course, causing land and its inhabitants to passively transfer across the barrier, rivers
cease to act as barriers due to the lack geographic separation at headwaters (Wallace 1852) or
temporal fluctuations of climate causing rivers to change in size (e.g. the sizes of any rivers were
dramatically reduced during the Pleistocene).
To create a spatial prediction for this hypothesis, we selected all major permanent rivers
that have headwaters above 1000m and created polygons from the lowland areas between major
rivers and 1000m contour line. A second calculation focused on major riverine units composed
of areas between rivers with headwaters above 2000m and created biogeographic units from
lowland areas between rivers and 1000m contour line.
River-Refuge Hypothesis (FigS1.C)
The River-Refuge hypothesis, initially described by Haffer (1992, 1993) for the Neotropics, was
more recently proposed as a model of Malagasy diversification (Craul et al. 2007). This
hypothesis combines aspects of the Riverine Barrier hypothesis and the Refuge hypothesis under
which it states that lowland vicariant speciation occurs in refugia separated by broad lowland
rivers and by considerable unsuitable terrain in the headwaters.
To estimate this model, we combined the Riverine and a binary refuge hypothesis. The
continuous refuge layer was converted to a binary model by converting the top quartile to 1 and
43
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
all other values to 0. To account for differences between precipitation stability in arid areas
(versus wet), we excluded areas with less than 50cm in any of the time periods. Areas above
1000m were also excluded. Regions of the Riverine layer and binary refuge layer were then
combined into smaller river-refuge subunits.
Sanctuary Hypothesis (Fig S2.L)
Past climate changes greatly altered the distributions of organisms through time; causing local
extinctions, bottlenecks, isolation, range expansion and contraction of populations. Sanctuaries
represent specific areas of habitat stability that have remained present through time, differing
from refugia (which do not invoke geospatial consistency) and species track suitable habitat
across geography (Recuero & García-París 2011). We estimated sanctuaries in Madagascar by
calculating SDMs for 453 species (a subset of the GDM dataset, using species with three or more
unique occurrence localities). The SDM was projected into three historic time periods: LGM,
120KYA, 6 KYA. All SDMs were converted to binary models using the maximum training
sensitivity plus specificity threshold. The maximum training sensitivity plus specificity
threshold: 0.080 (SD +-0.047), area: 0.316 (SD +-0.196), training omission: 0.005 (SD +- 0.017),
number of training samples (mean: 12.321, max 137, min 3, SD 14.414). For each species, all
four binary SDMs were summed. The resulting layer was reclassified so that values of 3 and
below were converted to zero and values of 4 were converted to 1. Under this classification,
areas where the species was present for all four time periods were considered ‘sanctuaries’. This
was repeated for all species and all sanctuaries were summed to estimate areas of higher species
richness and endemism.
44
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
Topographic Heterogeneity (Fig S2.E)
In several studies, the level of topographic variation has been observed to be positively
correlated to species richness patterns and centers of endemism (Kerr & Packer 1997; Rahbek
and Graves 20001; Jetz & Rahbek 2002; Jetz et al. 2004). To characterize this hypothesis for
Madagascar we measured the standard deviation of elevation at ca. 10 km2 of each pixel.
Watershed Hypothesis (Fig S1.D)
One of the more recent diversification hypotheses is the Watershed hypothesis (Wilmé et al.
2006). According to this model, climatic changes caused retraction of forests to the surrounding
major rivers. If the headwaters of adjacent rivers were at lower elevations, the intervening areas
between rivers (the watersheds) become arid and forests populations became isolated, serving as
areas of speciation. By contrast, if the headwaters of rivers were higher elevations, the watershed
served as areas of retreat and forest refugia remained connected among rivers. These watersheds
are expected to contain proportionally much lower diversity and endemism. This GIS prediction
was obtained from Wilmé et al. (2006).
Additional references.
Brown Jr., K.S., Sheppard, P.M., Turner, J.R.G. 1974. Quaternary refugia in tropical America:
evidence from race formation in Heliconius butterflies. Proc. R. Soc. Lond. B. 187: 369-
378.
Bush, M.B. 1994. Amazonian speciation: a necessarily complex model. Journal of
Biogeography 21::5–17.
Colinvaux, P.A. 1993. Pleistocene biogeography and diversity in tropical forests of South
America P. Goldblatt (Ed.), Biological Relationships between Africa and South America,
Yale University Press, New Haven, CT.
45
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
Craul, M., Zimmermann, E., Rasoloharijaona, S., Randrianambinina, B., Radespiel, U. 2007.
Unexpected species diversity of Malagasy primates (Lepilemur spp.) in the same
biogeographical zone: a morphological and molecular approach with the description of
two new species. BMC Evolutionary Biology 7:83.
Dynesius, M., Jansson, R. 2000. Evolutionary consequences of changes in species_
geographical distributions driven by Milankovitch climate oscillations. Proc. Natl Acad.
Sci. U.S.A. 97:9115–9120.
Endler, J. 1982. Pleistocene forest refuges: fact or fancy? In Prance, G.T. (Ed.). Biological
Diversification in the Tropics. New York: Columbia University Press, p. 179-200.
Fisher, R.A. 1930. The Genetical Theory of Natural Selection. Clarendon Press.
Goodman, S. M. & Benstead, J. P. The natural history of Madagascar. (University of Chicago
Press Chicago, 2003).
Fjeldsa°, J., Rahbek, C. 2006. Diversification of tanagers, a species-rich bird group, from the
lowlands to montane regions of South America. Integr. Comp. Biol. 46:72–81.
(doi:10.1093/icb/icj009)
Haffer, J. 1969. Speciation in Amazonian forest birds. Science 165: 131-137.
Haffer, J. 1997.Alternative models of vertebrate speciation in Amazonia: an overview
Biodiversity and Conservation, 6:451–476.
Haffer,, J. 1992. On the “river effect” in some forest birds of southern Amazonia. Bol. Mus.
Para. Emilio Goeldi, sér. Zool. 8:217-245.
Haffer, J. 1993. Time’s cycle and time’s arrow in the history of Amazonia. Biogeographica
69:15-45.
Hall, J. P. 2005. Montane speciation patterns in Ithomiola butterflies (Lepidoptera:
Rhiodinidae): are they consistently moving up in the world? Proc. R. Soc. B 272:2457–
2466. (doi:10.1098/rspb.2005.3254)
Jansson, R., Dynesius, M. 2002. The fate of clades in a world of recurrent climatic change:
Milankovitch oscillations and evolution. Annu. Rev. Ecol. Syst. 33:741–777.
46
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
Jetz, W., Rahbek, C. 2002. Geographic range size and determinants of avian species
richness. Science 297:1548–1551.
Jetz, W., Rahbek, C., Colwell, R.C. 2004. The coincidence of rarity and richness and the
potential signature of history in centers of endemism. Ecol. Lett. 7:1180–1191.
Kerr, J.T. Packer, L. 1997. Habitat heterogeneity determines mammalian species richness in
high energy environments. Nature. 385:252-254..
Kozak, K.H., Wiens, J.J. 2007. Climatic zonation drives latitudinal variation in speciation
mechanisms. Proc. R. Soc. B 274:2995-3003. doi: 10.1098/rspb.2007.1106
Krause, D. W. 2003. Late Cretaceous vertebrates of Madagascar: A window into Gondwanan
biogeography at the end of the Age of Dinosaurs. Pp. 40-47 in S. M. Goodman and J. P.
Benstead (eds.), The Natural History of Madagascar. University of Chicago Press,
Chicago.
Lees, D. C. 1996. The Périnet effect? Diversity gradients in an adaptive radiation of
butterflies in Madagascar (Satyrinae: Mycalesina) compared with other rainforest taxa,
Pages 479-490 in W. R. Lourenço, ed. Biogéographie de Madagascar. Paris, Editions de
l'ORSTOM.
Lees, D. C., Colwell R.K. 2007. A strong Madagascan rainforest MDE and no equatorward
increase in species richness: Re-analysis of 'The missing Madagascan mid-domain
effect', by Kerr J.T., Perring M. & Currie D.J (Ecology Letters 9:149-159, 2006). Ecology
Letters 10:E4-E8.
Moritz, C., Patton, J. L., Schneider, C. J., Smith, T. B. 2000. Diversification of rainforest
faunas: an integrated molecular approach. Annu. Rev. Ecol. Syst. 31:533–563. (doi:10.
1146/annurev.ecolsys.31.1.533)
Pearson, R.G., Raxworthy C.J. 2009. The evolution of local endemism in Madagascar:
watershed versus climatic gradient hypotheses evaluated by null biogeographic
models. Evolution 63:959–967.
Prance, G.T. 1982. Biological diversification in the tropics. New York: Columbia University.
Prance, G.T. 1996. Islands in Amazonia. Phil. Trans. R. Soc. Lond. B 351: 823-833.
47
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
Rahbek, C. 1997. The relationship among area, elevation, and regional species richness in
Neotropical birds. Am. Nat. 149:875–902.
Rahbek, C., Graves, G. R. 2001. Multiscale assessment of patterns of avian species richness.
Proc. Natl Acad. Sci. USA 98:4534–4539. (doi:10.1073/pnas.071034898)
Recuero, E., García-París, M. 2011. Evolutionary history of Lissotriton helveticus: Multilocus
assessment of ancestral vs. recent colonization of the Iberian Peninsula. Molecular
Phylogenetics and Evolution 60: 170-182.
Stebbins, G. L. 1974. Flowering plants: evolution above the species level. Harvard University
Press, Cambridge, Mass.
Stephens, P. R., Wiens, J.J. 2003. Explaining species richness from continents to
communities: the time-for-speciation effect in emydid turtles. American Naturalist
161:112–128.
Vanzolini, PE. 1970. Zoología sistemática, geografía e a origem das espécies. São Paulo:
Instituto Geográfico de São Paulo. 56 p
Vanzolini, P.E. 1973. Paleoclimates, relief, and species multiplication in equatorial forests. In
Meggers, B.J., Ayensu E.S.. Duckworth, W.D. (Eds.). Tropical forest ecosystems in Africa
and South America: A comparative review. Washington: Smithsonian Institution. p.
255-258.
Wallace, A.R. 1852. On the monkeys of the Amazon. London. Proc. Zool. Soc., 1852:107-110.
Wilmé, L., Goodman, S.M., Ganzhorn, J.U. 2006. Biogeographic evolution of Madagascars
microendemic biota. Science 312:1063– 1065.
48
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
Supplementary Figure S1. Nominal Biogeography hypotheses. These hypotheses are
comprised of non-order categories from which relationships among contained data are not known
49
1078
1079
1080
1081
1082
or non-linear. Four hypotheses fell into this category: The Riverine (major and minor), River-
Refuge, Watershed and Gradient. See table S1 for summary of each hypotheses.
50
1083
1084
1085
1086
1087
1088
51
1089
Supplementary Figure S2. Continuous Biogeography hypotheses. Nine hypotheses fell into
this category: Mid-domain (longitude, latitude and distance), Museum, Topographic
Heterogeneity, Gradient* (PC1-PC3), Disturbance-vicariance, Climate Stability, Precipitation
Stability, Sanctuary and Montane Species Pump (see table S1 for summary of each hypotheses).
Inlayed tables represent the percent contribution of each corresponding hypothesis in the CAR
model with the lowest AICc of each observed biodiversity measurements. *The values of the
three climate principal components are not necessary assumed to reflect a positive correlation to
endemism and species richness, however, we are assuming each reflects a prediction of a linear
correlation (either positive or negative). The inclusion of the three climate principal components
is the result of the power CAR models and the ability to include multiple explanatory variables.
Due to the statistical limitations associated with nominal hypotheses, if a hypothesis could be
depicted by continuous data (even if it required several variables) they were converted to this
format and incorporated into the CAR. For example, if nominal data represented classified
continuous data, we include the continuous data (such as the 3 PCs of climate data).
52
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
Supplementary Table S1. Major Biogeographic Predictions Relevant to Madagascar.
Hypothesis Description Key Factors EffectsPredictions for Reptiles
and AmphibiansGIS Prediction
Temporal
ScopeKey Citations
Climate Stability
(Fig S2.J)
Climate stability creates
greater climatic stratification
across environmental
gradients
Stable climate; both
seasonally and
through geologic
time; no barrier is
necessary
In stable climates, orbitally
forced species’ range dynamics
(ORD) are low, allowing
localized populations to persist,
and thus become highly
specialized and differentiated.
Higher levels of
endemism in areas of
climatic stability
Use GIS and
spatiotemporal explicit
climate data to estimate
climate stability; stable
areas should harbor
higher species richness
and endemism.
None Dynesius & Jansson,
2000; 2002
Disturbance-
vicariance
(Fig S2.L)
Allopatry results from
altitudinal range retractions
caused by temperature
fluctuations.
Temperature
fluctuations,
changes in CO2 and
habitat
heterogeneity
(usually associated
with changes in
altitude)
Decreased temperatures allow
cool adapted species to disperse
south. Cyclic fluctuations in
temperature cause populations to
habitat track attitudinally, when
temperatures are at their highest,
populations become isolated on
sky islands
Diversity with
monophyletic lineages are
associated with a single
region. Most common
ancestor of sister clades
date to Pleistocene.
Estimate areas of high
temperate fluctuations
adjacent areas of slope;
higher values reflect
higher species richness
Quaternary Colinvaux 1993,
Bush 1994, Haffer
1997; Raxworthy
Nussbaum 1995.
Gradient
(Fig S1.E)
Parapatry of populations due
to environmental gradients
Divergent selection
and an
environmental
gradient; no barrier
Parapatric speciation across
climate space
Sister taxa are found in
different habitats along an
environmental gradient
Cluster analyses of all
current climate data to
estimate areas of
None Endler 1982
53
1105
is necessary endemism.
Mid-domain
Effect
(Fig S2. A-C)
Species’ ranges are
distributed randomly between
northern and southern
geographic limits, the highest
overlap of species ranges
would be in the middle.
Geographic space No mechanism invoked Species richness should be
highest in mid-domains
Richness is highest in the
mid-domain of latitude,
longitude and elevation.
None Lees 1996, Lees and
Colwell 2007
Montane Species
Pump
(Fig S2. M)
Topographic complexity and
climatic zonation of
mountains increase
opportunities for allopatric
and parapatric speciation
Topographic and
climatic
heterogeneity
Allopatric and parapatric
speciation across elevations
Sister taxa share common
ancestry with montane
ancestor. Extant montane
species display higher
levels of intraspecific
genetic variation.
Estimate areas of
topographic and climatic
heterogeneity- high
values reflect centers of
high endemism and
species richness
None Moritz et al. 2000;
Rahbek and Graves
2001; Hall 2005;
Fjeldsa° and Rahbek
2006; Kozak and
Wiens 2007, Roy et
al 1997, Fjeldsa et al
1999
Museum
(Fig S2.D)
More species occur at
intermediate elevations
simply because these
elevations were occupied the
longest and there has been
more time for speciation and
the accumulation of species in
Extended time
occupying in mid-
elevations
Increased differentiation at mid-
elevations
More species occur at
intermediate elevations
because these elevations
were occupied the longest
Use GIS to calculate the
median elevation of
Madagascar which
should possess the
highest species richness
and endemism
None Stephens and Wiens
2003
54
these habitats relative to those
at lower and higher elevations
Paleogeographica
l
Vicariant differentiation of
Malagasy lineages is
associated with formation of
geologic barriers to dispersal.
Each hypothesis is specific to
the focal paleogeographic
event.
Geological changes
resulting in vicariant
events
Vicariant differentiation across
barriers
Distinct east and west
lineages associated with
the central mountains, and
between the
southern/central/northern
massifs, tsingys.
Not Calculated Specific to
each
geological
event
Goodman &
Benstead 2003
Refuge
(Fig S2. K)
Allopatry due to retraction of
wet habitats
Repeated cycles of
drastic fluctuation in
precipitation
Episodic fragmentation of
forests resulting in isolated
patches of wet forest causing
vicariant differentiation between
adjacent patches
Evolutionary lineages
associated with refugia
(areas of continued
precipitation relative to
regional mosaic of
habitats)
Use GIS and
spatiotemporal explicit
climate data to estimate
stable wet habitats, these
areas reflect centers of
endemism
Cenozoic
(Tertiary
and
Quaternary
)
Haffer 1969m 1990,
1999, Endler (1982),
Brown (1987) Nores
(1999)
Riverine (Fig
S1. A, B)
Allopatry due to rivers acting
as barriers.
Permanent large
rivers
Vicariant differentiation of
Malagasy lineages associated
with large tributaries
Reciprocal monophyly of
clades on opposite sides of
river
Measured inter-riverine
areas which are areas of
endemism
None Wallace 1853,
Capparella 1991,
Patton et al. 1994,
Goodman &
Ganzhorn, 2004
River-Refuge Allopatry due the restriction
of wet habitats to lower
Reduced
precipitation,
Similar to Riverine Hypothesis;
fragmental faunal distributions
Sister taxa are found in
adjacent intra-riverine
Combine the Riverine
Hypothesis subunits and
Late
tertiary
Haffer 1992; 1993,
55
(Fig S1.C) elevations; higher elevation
habitats and intervening rivers
act as barriers
maintenance of
permanent rivers
into intra-riverine corridors,
isolation is associated with
increased aridity adversely
affecting habitat suitability at
headwater regions
corridors a binary precipitation
stability /low elevation
layer. Resulting areas
depict areas of high
predicted endemism.
(Post-
Miocene)
Craul et al 2007
Sanctuary
(Fig S2.L)
Extinction occurs more often
in instable habitats; thus
stable areas accumulate
species.
Climate fluctuations
through time across
heterogeneous
landscapes
Areas of niche stability
accumulate species through time.
Areas of niche stability
(specific aspects of
climate vs. overall climate
in Climate stability)
provide sanctuary for
species though time.
Use GIS and SDM to
estimate stable areas in
each species distribution
through time; areas of
highest stability should
harbor higher species
richness and endemism
None Recuero & García-
Paris 2011
Topographic
Heterogeneity
(Fig S1. D)
The level of topographic
variation has been observed to
be positively correlated to
species richness patterns and
centers of endemism
Topography No mechanism invoked Species richness should be
highest in areas of high
topographic heterogenety
Areas of high
topographic
heterogeneity harbor
higher species richness
None Kerr & Packer 1997;
Rahbek and Graves
20001; Jetz &
Rahbek 2002; Jetz et
al. 2004
Watershed
(Fig S1.D)
Allopatry results from
altitudinal range retractions
caused by simultaneous
decreases in temperature and
precipitation. Lower elevation
Repeated cycles of
drastic simultaneous
increases of both
temperature and
precipitation. Large
Dispersal during warm-humid
periods allows lowland species
to disperse attitudinally, across
headwater habitat (previously
too arid to occupy). Allopatry
Sister taxa are found in
adjacent intra-riverine
corridors. Most common
ancestor of sister clades
date to Pleistocene.
See Wilmé et al. 2006. Quaternary Wilmé et al 2006
56
rivers act as barriers. permanent rivers
and mountains
adjacent to
lowlands.
occurs as climate cools and the
species depends into lowlands
and lowland rivers prevent gene
flow between populations, now
occurring on both sides of river.
57
1106
Supplementary Table S3. Correlations of biodiversity hypotheses to observed biodiversity patterns. R-values reflect non-spatial
Pearson product-moment correlation coefficients. To assess significance of raster data, we used an unbiased correlation following the
method of Dutilleul (1993). This method reduced the degrees of freedom according to the level of spatial autocorrelation between the
two variables.
Hypothesis Reptile Amphibian
Correlation to Observed
Endemism R F-stat df p r F-stat df p
Mid-domain: Distance -0.148 0.539 24.037 0.470 -0.026 0.018 26.753 0.894
Topographic Heterogeneity 0.343 5.425 40.738 0.025* 0.662 21.664 27.724 <.001*
Refuge 0.036 0.04 32.328 0.848 0.154 0.552 22.722 0.465
Montane Species Pump 0.303 2.639 48.488 0.107 0.613 13.852 22.990 0.001*
Disturbance-vicariance 0.379 4.101 22.413 0.055 0.616 15.758 22.829 <.001*
Climate stability 0.241 0.62 14.580 0.444 0.356 2.033 15.124 0.174
Sanctuary 0.296 2.438 25.313 0.131 0.606 7.611 13.080 0.016*
Museum 0.285 4.971 56.429 0.030* 0.335 17.709 49.836 0.015*
58
1107
1108
1109
1110
1111
1112
River-Refuge (binary) 0.207 3.87 86.869 0.052 0.066 0.954
216.95
9 0.330
Correlation to Observed Richness R F-stat df p r F-stat df p
Mid-domain: Distance -0.480
10.26
5 34.353 0.003* -0.204 1.072 24.698 0.310
Topographic Heterogeneity 0.140 2.398 119.852 0.124 0.307 6.769 64.952 0.011*
Refuge -0.101 0.332 31.948 0.598 -0.093 0.216 24.605 0.646
Montane Species Pump 0.056 0.507 160.576 0.477 0.296 5.506 57.196 0.022*
Disturbance-vicariance 0.019 2.297 62.251 0.135 0.303 5.286 41.554 0.027*
Climate stability 0.233 1.506 26.191 0.231 0.210 0.723 15.672 0.408
Sanctuary 0.419
10.07
9 47.267 0.003* 0.818 34.069 16.885 <.001*
Museum 0.234 4.037 69.472 0.048* 0.250 4.481 66.989 0.038*
River-Refuge (binary) 0.108 2.378 201.187 0.125 0.234 4.94 85.022 0.029*
59
1113
1114
Supplementary Table S4. Mixed CAR spatial models of observed biodiversity data. A principal component analyses was performed
on the standardized biogeography hypotheses. All the resulting principal components (PCs) were extracted and then loaded as
explanatory variables. The CAR analyses were run iteratively, starting with all PCs as response variables and then excluding each PC
that did not contribute significantly to the model (α = 0.05) until the final model included only PCs that contributed significantly to the
model. The standardized beta coefficients (β) were then used to calculate contributions of each biogeography hypothesis (see methods
on OTBCs) in the final CAR analysis. To compare the contributions of each biogeography hypothesis among models of observed
biodiversity patterns (richness, endemism, GDM), β coefficents from each OTBC/CAR analyses were converted to the percentage of
contribution. *The mean of the 3 MDS vectors loadings were calculated and contributed as a single value to the total mean.
Percent Contribution to CAR Endemism Richness GDM Mean
contribution to
all observed
biodiversity
models*
Amphibia
n
Reptile Amphibian Reptile MDS-D1 MDS-D2 MDS-D3 Mean*
Mid-domain- Latitude 0.0% 27.7% 3.5% 3.6% 5.3% 20.7% 1.6% 9.2% 8.8%
Mid-domain- Longitude 8.1% 8.0% 6.7% 9.2% 10.9% 10.2% 8.1% 9.7% 8.4%
Mid-domain- Distance*** 15.1% 9.7% 7.7% 20.0% 14.4% 5.7% 15.8% 11.9% 12.9%
60
1115
1116
1117
1118
1119
1120
1121
1122
1123
Climate- PC1 15.4% 0.1% 11.4% 14.5% 15.1% 4.5% 13.8% 11.1% 10.5%
Climate- PC2 10.1% 12.7% 8.5% 11.0% 10.1% 11.7% 7.4% 9.7% 10.4%
Climate- PC3 2.2% 12.8% 4.6% 1.8% 4.4% 11.9% 3.9% 6.7% 5.6%
Refuge 1.4% 4.7% 6.9% 0.0% 3.8% 10.5% 0.0% 4.8% 3.5%
Climate Stability*** 4.7% 4.6% 7.4% 3.0% 0.0% 1.0% 3.8% 1.6% 4.2%
Topographic Heterogeneity 6.5% 1.7% 7.1% 3.7% 3.9% 0.5% 8.0% 4.2% 4.6%
Disturbance-vicariance 5.9% 4.8% 8.1% 3.1% 3.4% 2.2% 6.8% 4.1% 5.2%
Montane Species Pump 5.0% 0.0% 7.1% 2.3% 2.9% 0.0% 7.2% 3.4% 3.6%
Sanctuary 10.9% 0.3% 13.7% 12.6% 11.7% 3.1% 4.7% 6.5% 8.8%
Museum 14.8% 13.0% 7.1% 15.1% 14.3% 18.1% 19.0% 17.1% 13.4%
CAR Model Summary
Explained by Predictor
Variables: r2 (AICc)
0.482
(-372.3)
0.303
(-468.5)
0.624
(7861.9)
0.309
(21849.9)
0.869
(-4037.8)
0.435
(-1084.2)
0.659
(-3327.7)
Total Explained (Predictor and
Space): r2 (AICc)
0.646
(-426.0)
0.626
(-556.4)
0.849
(6349.3)
0.745
(19364.2)
0.954
(-6659.7)
0.942
(-6760.4)
0.868
(-5700.9)
Model significance: n, F, p-val 141, 42.5, 141, 9.7, 2501, 2501, 2501, 2501, 2501, NA
61
<0.001 <0.001 303.5
<0.001
278.1,
<0.001
1265.9,
<0.001
147.2,
<0.001
396.8,
<0.001
62
1124
1125
1126
1127
1128
1129
1130
Supplementary Table S5. Mixed CAR spatial models of focal subgroups. A principal component analyses was performed on the
standardized biogeography hypotheses. All the resulting principal components (PCs) were extracted and then loaded as explanatory
variables. The CAR analyses were run iteratively, starting with all PCs as response variables and then excluding each PC that did not
contribute significantly to the model (α = 0.05) until the final model included only PCs that contributed significantly to the model. The
standardized beta coefficients (β) were then used to calculate contributions of each biogeography hypothesis (see methods on OTBCs)
in the final CAR analysis. To compare the contributions of each biogeography hypothesis among models of observed biodiversity
patterns (richness, endemism, GDM), β coefficents from each OTBC/CAR analyses were converted to the percentage of contribution.
*The mean of the 3 MDS vectors loadings were calculated and contributed as a single value to the total mean.
Percent Contribution to CAR Endemism Richness
Boophis Brookesi
a
Oplurus* Phelsuma Boophis Brookesia Oplurus* Phelsuma
Mid-domain- Latitude 3.0% 4.0% 2.5% 17.0% 0.0% 5.1% 2.3% 4.2%
Mid-domain- Longitude 8.4% 9.9% 10.6% 11.8% 6.1% 8.3% 6.3% 10.2%
Mid-domain- Distance*** 16.0% 9.9% 6.3% 6.4% 10.6% 12.0% 11.0% 20.4%
Climate- PC1 12.2% 12.2% 13.5% 5.8% 14.4% 11.7% 15.9% 13.1%
Climate- PC2 6.0% 14.1% 13.2% 7.0% 13.0% 11.2% 8.9% 9.7%
Climate- PC3 5.2% 4.3% 2.8% 18.3% 1.6% 7.9% 0.0% 3.9%
63
1131
1132
1133
1134
1135
1136
1137
1138
Refuge 0.0% 10.5% 11.3% 2.3% 9.9% 4.5% 3.3% 0.0%
Climate Stability*** 3.6% 8.8% 11.2% 5.9% 8.9% 3.6% 5.8% 1.8%
Topographic Heterogeneity 9.0% 2.2% 1.0% 5.1% 2.8% 2.1% 6.5% 4.1%
Disturbance-vicariance 7.0% 4.7% 4.1% 7.2% 5.8% 2.5% 6.8% 2.7%
Montane Species Pump 8.4% 0.6% 0.0% 4.3% 2.7% 0.0% 4.5% 2.7%
Sanctuary 0.1% 18.9% 22.6% 0.0% 20.3% 17.2% 23.6% 7.9%
Museum 21.1% 0.0% 0.8% 8.8% 3.8% 13.9% 5.2% 19.3%
CAR Model Summary
Explained by Predictor
Variables: r2 (AICc)
0.251
(-252.2)
0.303
(-176.1)
0.432
(-317.9)
0.209
(-253.4)
0.569
(9710.5)
0.444
(5392.1)
0.178
(9789.1)
0.559
(6731.4)
Total Explained (Predictor and
Space): r2 (AICc)
0.524
(-316.0)
0.685
(-235.3)
0.756
(-436.8)
0.759
(-354.6)
0.863
(6851.0)
0.812
(2680.8)
0.504
(8530.6)
0.794
(4837.3)
Model significance: n, F, p-val 141, 46.6,
<0.001
141, 49.7,
<0.001
141, 20.5,
<0.001
141, 20.1,
<0.001
2501, 364.7,
<0.001
2501, 220.6,
<0.001
2501, 550.1,
<0.001
2501, 632.4,
<0.001
64
1139
1140