Content uploaded by Matthew Van Dam
Author content
All content in this area was uploaded by Matthew Van Dam on Feb 13, 2020
Content may be subject to copyright.
SPECIAL
PAPER
Evaluating the influence of connectivity
and distance on biogeographical
patterns in the south-western deserts of
North America
Matthew H. Van Dam
1
* and Nicholas J. Matzke
2,3†
1
Department of Environmental Science,
Policy, and Management, University of
California, Berkeley, CA 94720-3140, USA,
2
Department of Integrative Biology, Center for
Theoretical Evolutionary Genomics, University
of California, Berkeley, CA 94720-3140, USA,
3
National Institute for Mathematical and
Biological Synthesis (NIMBioS), University of
Tennessee, 1122 Volunteer Blvd., Suite 106,
Knoxville, TN 37996-3410, USA
*Correspondence and current address:
Matthew H. Van Dam, Zoologische
Staatssammlung M€
unchen, M€
unchhausenstr.
21, 81247 M€
unchen, Germany.
E-mail: matthewhvandam@gmail.com
†
Current address: Discovery Early Career
Researcher Award (DECRA) Fellow, Division
of Ecology, Evolution, and Genetics, Research
School of Biology, The Australian National
University, Canberra, ACT 2601, Australia
ABSTRACT
Aim To examine the role of geological history, connectivity and distance in
shaping the biogeographical structure of North American desert clades that are
restricted to habitat islands (sand dunes and relictual aquatic habitats), using
statistical model choice on old and new probabilistic biogeographical models.
Location North America, Mojave, Sonoran and Chihuahuan Deserts.
Materials and methods Dated phylogenies were estimated for three field-
sampled insect clades (Trigonoscuta,Rhaphiomidas and sand treader crickets),
and five other literature-sampled clades (the snails Assiminea,Pyrgulopsis and
Tryonia; the desert fringe-toed lizard Uma; and the desert pupfish Cyprinodon).
BioGeoBEARS was used to statistically compare biogeographical models
assuming unconstrained or connectivity-constrained dispersal, with or without
founder-event speciation (jump dispersal) permitted. Finally, we introduce and
test a novel distance-based dispersal model (+x) where dispersal probability is
multiplied by distance to the power x.
Results We observed little concordance between biogeographical patterns and
timing of geological events. Model comparisons were decidedly in favour of
inclusion of founder-event speciation in the models for most taxa, with only a
small taxon, Uma, showing support for the model favouring vicariance. The
inclusion of a constrained-dispersal matrix was favoured by three of the eight
taxa examined (Cyprinodon, sand treader crickets, and Trigonoscuta). Surpris-
ingly, tests for distance influencing dispersal probability were mostly negative.
Main conclusions Our results do not show support for any one geological
event shaping the biogeographical patterns of these desert taxa. Instead, the
histories of desert dune and aquatic taxa are largely products of rare jump dis-
persal events, and can be considered island-like systems. Although results are
negative for the distance-based dispersal model, this in itself demonstrates the
superiority of explicit statistical model testing over a priori assumption of fixed
models in historical biogeography.
Keywords
BioGeoBEARS, Chihuahuan Desert, deserts, dispersal, historical biogeography,
Mojave Desert, Pleistocene, Sonoran Desert, vicariance
INTRODUCTION
The use of islands as natural laboratories to study evolution
dates back to Wallace and Darwin, and island systems have
been indispensable for understanding the processes generat-
ing biodiversity. Examples from the Galapagos, Caribbean
and Hawaiian archipelagos demonstrate the utility of islands
for the study of adaptation, community assembly and specia-
tion (Grant & Grant, 2002; Losos et al., 2003; Gillespie,
2004). In addition to true islands, habitat islands are also of
great interest to evolutionary biologists studying many of
the same questions (Wake, 1987; Masta, 2000; Knowles &
Carstens, 2007). Habitat islands are discrete patches of
habitat surrounded by a contrasting habitat (Whittaker &
ª2016 John Wiley & Sons Ltd http://wileyonlinelibrary.com/journal/jbi 1
doi:10.1111/jbi.12727
Journal of Biogeography (J. Biogeogr.) (2016)
Fern
andez-Palacios, 2007). Depending on the habitat and
taxon, the area surrounding habitat islands may be totally
uninhabitable (much like the ocean for terrestrial taxa), or
may be a less stringent constraint. Thus, the rules governing
dispersal–vicariance may not be the same between oceanic
islands and habitat islands.
Here, we examine the historical biogeography of taxa that
occupy sand dune habitat islands in the south-western
deserts of North America. These taxa are specialists on sand
dunes, and are never found more than 100 m outside dune
edges (Norris, 1958; Pierce, 1975; Hardy & Andrews, 1976;
Cazier, 1985). This highly specialized habitat preference,
combined with the isolated and disjunct distributions of
dunes, leads to the question of how these animals came to
occupy their current distributions. For example, Norris
(1958) first proposed that the fringe-toed lizard genus Uma
might have used the sandy river corridors and sand transport
pathways [the path aeolian (wind blown) sediments follow
from source to deposition] as a means of dispersal during
Pleistocene climate fluctuations, when sandy sediments
would periodically become available. Here we use phyloge-
netic dating and probabilistic biogeographical inference
methods to test this hypothesis for multiple clades.
Background: Geological hypotheses
We aim to test three geological hypotheses for their ability to
explain the distributions of dune taxa of south-western
deserts: (1) the uplift of the Sierra Madre Occidental, (2) the
formation of the Bouse Lakes, and (3) Pleistocene sand
transport pathways. The uplift of the Sierra Madre Occiden-
tal (c. 34–15 Ma) is postulated to be one of the major factors
responsible for dividing the Sonoran and Chihuahuan deserts
(Ferrari et al., 1999; Nieto-Samaniego et al., 1999; Riddle &
Hafner, 2006). A later-forming geological barrier was the
‘Bouse Embayment’ (Lucchitta, 1979; Lucchitta et al., 2001;
Turak, 2000; Spencer et al., 2013), which was created by the
proto-Colorado River when high rates of evaporation relative
to input formed a series of high-salinity palaeo-lakes (rather
than an actual embayment of the Gulf of Mexico) 4.83–
4.80 Ma (Spencer & Patchett, 1997; Poulson & John, 2003;
House et al., 2005, 2008; Roskowski et al., 2010; Spencer
et al., 2013). This will be referred to herein as the Bouse
Lakes Formation (BLFs). When the BLFs drained, the Color-
ado River was connected to the early Pliocene Gulf of Cali-
fornia by c. 4.80 Ma (Spencer et al., 2013). Phylogenetic
dating can yield some indication of the plausibility of these
two pre-Pleistocene geological explanations: if the estimated
date ranges of many nodes overlap with the date of a geolog-
ical event, it is possible that it caused approximately simulta-
neous speciation events; but if overlap is not observed, the
explanation is disconfirmed. If these geological events had an
effect on the biogeographical history, then we would expect
to see divergence between taxa east–west of the Mojave
section of the Colorado River during the formation of the
Bouse Lakes and divergence between sister taxa in the
Sonoran and Chihuahuan Deserts during the formation of
the Sierra Madre Occidental.
The third major event that might be important is the
repeated extension and retreat of lakes and rivers in the desert
south-west during Pleistocene glacial-interglacial cycles (Muhs
et al., 2003). Under this hypothesis, dune-restricted taxa fol-
lowed the sandy corridors of Pleistocene lakes and rivers to
achieve their current distributions. If the dune-restricted fauna
did in fact use river corridors as dispersal pathways, then they
should exhibit biogeographical patterns similar to those
observed for aquatic organisms of the south-western deserts.
This hypothesis is based on the premise that the river corridors
initially provided aquatic habitat connections required by fish
and other aquatic organisms, and as they began to dry,
exposed sandy sediments that were utilized by dune-restricted
taxa to traverse the same pathways. If the Pleistocene sand
transport pathways hypothesis is correct, then we would expect
to see speciation events dating to the Pleistocene, and geo-
graphical ranges that are well-explained by dispersal along
these pathways.
An alternative to these hypotheses is that geology and con-
nectivity are not important explainers of the geographical
range patterns of each clade, and long-distance ‘jump’ dis-
persal dominates instead. This hypothesis suggests that diver-
gence times will not correlate with geological events, and the
geographical range data will be better fit by unconstrained
models and those that make use of founder-event speciation.
To test these hypotheses, we estimated dated phylogenies
for four sand-dune-restricted clades, and also included four
aquatic clades that are found in desert regions adjacent to
many of our focal sand dunes. We then conducted maxi-
mum likelihood (ML) estimation of biogeographical history
under the dispersal-extinction-cladogenesis (DEC) model,
and modifications to this model that add founder-event spe-
ciation (DEC+J) and distance-dependent dispersal (DEC+x,
DEC+J+x). In addition, each model was run with and with-
out connectivity constraints. The ability of the models to
explain the data was assessed via standard procedures in sta-
tistical model choice.
Background: probabilistic biogeographical methods
Phylogenetic biogeography methods are reviewed in Lawing
& Matzke (2014). Although older methods remain popular –
such as construction of a historical narrative (Castoe et al.,
2009), or construction of total area cladograms from the area
cladograms of several taxa (e.g. Brooks et al., 2002) –the
field is clearly moving in the direction of statistical inference
with probabilistic, parametric models (Ree & Sanmart
ın,
2009). This tracks the dominant trends in genetics and phy-
logenetics. Although the term ‘phylogeography’ is sometimes
applied to historical biogeography, statistical phylogeography
(Knowles & Maddison, 2002) operates at the scale of popula-
tion genetics (Zink, 2002) and historical biogeography (Ron-
quist & Sanmart
ın, 2011) operates at the scale of
phylogenetics (Lawing & Matzke, 2014).
Journal of Biogeography
ª2016 John Wiley & Sons Ltd
2
M. H. Van Dam and N. J. Matzke
Parametric biogeography was initiated by the LAGRANGE
program (Ree et al., 2005; Ree & Smith, 2008) which used a
dispersal-extinction-cladogenesis (DEC) model. DEC has
proven widely popular and has been used in hundreds of
historical biogeography analyses. In DEC, ‘dispersal’ means
range expansion; and ‘extinction’ means local extirpation, or
range contraction. These processes are Markov processes
occurring anagenetically along the branches of a phylogeny,
and are controlled by the rate parameters dand e. These
parameters are free, and are estimated by ML. In DEC, geo-
graphical range can also change during cladogenesis through
vicariance or subset sympatry. Whereas dand eare esti-
mated, in DEC the cladogenesis model is fixed, with each
allowed event assigned equal probability conditional on the
ancestral range. Given the DEC model and an ML estimate
of dand e, the likelihood of the geographical range data can
be calculated, and the probabilities of ancestral ranges esti-
mated.
Although DEC pioneered the use of maximum likelihood
in historical biogeography, the presence of only a single
model meant that the full potential of the parametric
approach was not exploited. In particular, with only one
model available, it was impossible to make use of statistical
tools for model comparison and model selection. These tools
are ubiquitous in other areas of statistical inference, and
include the likelihood ratio test (LRT), Akaike’s information
criterion (AIC) and sample-size corrected AIC (AICc) (Burn-
ham & Anderson, 2002). Recently, the DEC model was
expanded to include founder-event speciation, also known as
‘jump dispersal’ (Fig. 2), via the DEC+J model (Matzke,
2013, 2014). The jparameter of DEC+J assigns a per-event
weight to founder-event speciation, a cladogenetic range-
inheritance event where one daughter lineage occupies an
area outside of the range of its ancestor. The two-parameter
DEC model is nested inside the three-parameter DEC+J
model, such that when j=0, DEC+J reduces to DEC. Statis-
tical model comparison often reveals that the data support
founder-event speciation as a process important for explana-
tion of the geographical range data at the tips of the phy-
logeny (Matzke, 2013, 2014; Harris et al., 2014; Voelker
et al., 2014), although traditional DEC cladogenesis processes
remain important.
The DEC+J model is included in the R package BioGeo-
BEARS (Matzke, 2013). However, the potential for using sta-
tistical model choice to test hypotheses in historical
biogeography is far from exhausted. BioGeoBEARS specifies
biogeography models as instances of a flexible supermodel.
This enables creation of new models with ease, which may
then be evaluated against the data via LRT, AIC, or AICc to
determine if the addition of free parameters is justified by
improvements in data likelihood.
Here we introduce a variant on DEC-type models wherein
dispersal probabilities are modified as a function of geo-
graphical distance between two areas. LAGRANGE allows
inclusion of a user-specified dispersal matrix, which alters
the transition matrix between states depending on the
relative probability of dispersal between areas. However, the
construction of the dispersal probability matrix was usually
somewhat subjective. Recent methods such as SHIBA (Webb
& Ree, 2012) and BayArea (Landis et al. 2013) allow for dis-
persal rate to be calculated as a function of distance. How-
ever, BayArea contains no model for range evolution at
cladogenesis, and SHIBA employs a simulation-based
approximate Bayesian approach, which is computationally
intensive, restricting analyses to a small number of taxa and
areas.
SHIBA and LAGRANGE could also specify constraints on
connectivity between areas, setting the probability of disper-
sal to 0 between unconnected areas. In BioGeoBEARS, both
connectivity and/or distance matrices can be included in the
biogeographical models, and these can be attached to any
cladogenetic models of interest (e.g. DEC or DEC+J). As the
likelihood of the geographical range data can be calculated
under each model, ML inference and standard model selec-
tion can be performed. To the best of our knowledge, this is
the first time a study has used formal model selection to test
the effects of both connectivity and distance. For a review of
the different biogeographical models, see Matzke (2013).
Natural history of focal taxa
Dune taxa
We selected four groups of taxa that are found throughout
portions of the North American deserts. All of the sand dune
taxa are ecologically restricted to this habitat type. Many
dune inhabiting species have specialized morphology for dig-
ging into and moving on the sand’s surface.
The fly genus Rhaphiomidas consists of 23 described spe-
cies and five subspecies all of which are endemic to the
deserts of North America (Van Dam, 2010). The adult flies
are active in spring and fall and feed on floral nectar. The
larvae are restricted to deep sandy soils and are believed to
be generalist predators. Several species are of threatened sta-
tus, for example, R. terminatus,R. trochilus and R. moapa
(Rogers & Van Dam, 2007). Rhaphiomidas terminatus termi-
natus is only known from 20 ha, in the middle of a golf
course on the Palos Verdes Peninsula (George & Mattoni,
2006).
The dune weevil Trigonoscuta (Coleoptera: Curculionidae),
represents 65 species and 90 subspecies (Pierce, 1975).
Trigonoscuta has a distribution which covers the Californian
coastal dunes as well as dunes in the Mojave and Sonoran
Deserts. In addition, each of the California Channel Islands
has endemic species. Trigonoscuta is highly restricted to sand
dunes, feeding on a variety of dune plants. All the members
of this genus are entirely flightless. Adults bury themselves
under the sand during the day and are nocturnally active.
Trigonoscuta are typically known from only one sand dune
system and most species are allopatric (Pierce, 1975). Some
authors have questioned the validity of the species that
Pierce described, especially the sympatric species as they
Journal of Biogeography
ª2016 John Wiley & Sons Ltd
3
Model choice in desert historical biogeography
show little if any external morphological variation (Ander-
son, 2002; Evans & Hogue, 2006).
The group commonly known as the sand treader crickets
(Rhaphidophoridae: Ceuthophilinae) is comprised of five
genera with 13 described species (and many undescribed).
All members of this family are entirely flightless. The name
‘sand treader’ refers to their enlarged tibial spines (sand-bas-
ket), which they use to dig into the sand to avoid desiccating
during the day. Their burrows are often a meter deep or
more in dry years, which requires them to occupy areas of
deep aeolian sand. In this study, we focus on a monophyletic
clade, all of which are dune-restricted, or are found in asso-
ciation with sandy soils/dunes in the case of Daihinia.
The fringe-toed lizards, genus Uma (Phrynosomatidae),
are entirely restricted to aeolian sand dunes. The genus is
adapted to life in sand dunes, possessing a specialized mor-
phology for dealing with this environment (e.g. fringes on
their toes which help them to move on sand) (Norris, 1958).
Many of these species are also threatened due to habitat loss
and other anthropogenic activities such as off-road vehicle
use, as are many dune arthropod species (Van Dam & Van
Dam, 2008).
Aquatic taxa
Pyrgulopsis (Hydrobiidae) is one of the most diverse aquatic
snails in North America, which consists of 134 species (Liu
et al., 2013). They have diversified extensively in the desert
south-west, and are found primarily in springs. Their special-
ization on specific water conditions, low dispersal ability and
highly restricted ranges (single springs in some cases) are
believed to be responsible for their high species diversity
(Hershler & Sada, 2002).
The genus Tryonia (Cochliopidae) has 32 described species
(Hershler et al., 2011). This aquatic snail is found primarily
in western North America but has one species in Florida and
one in Guatemala. Their young develop inside the female
genital duct (Hershler et al., 2011). They are found primarily
in springs but a few species occupy brackish water.
The snail genus Assiminea (Assimineidae) is found primar-
ily in brackish waters and has a world-wide distribution. The
North American species in the nitida complex are composed
of four described species (Hershler et al., 2007). One is
found on the coast of California in brackish waters, and a
second along the Gulf of Mexico. The other two are land-
locked and are found in Death Valley and the Chihuahuan
Desert. They have free-swimming larvae.
The North American pupfish Cyprinodon (Cyprinodonti-
dae) are members of a large egg-laying genus with c. 51 spe-
cies. Here we focus on the western clade exclusively (Echelle,
2008). They are found in waters of varying salinity and are
often highly restricted to small desert springs. Several species
are threatened due to their restricted ranges, and C. arcuatus
has been extinct since 1971. This species is morphologically
similar to C. macularius and C. eremus (Minckley et al.,
2002) and is not included here.
MATERIALS AND METHODS
Taxon sampling
For the three insect clades (Trigonoscuta,Rhaphiomidas and
the sand treader crickets), most specimens were collected
from 2006–2011 within the published ranges of these clades,
as well as from regions of Baja California and Sonora, Mex-
ico that were well outside the documented ranges for many
of these groups. The total number of sampled locations is
over 200 (for range maps of sand dune taxa sampled please
see Supporting Information). Specimens were preserved in
95% ethanol stored on ice in the field and then transferred
to a 20 °C freezer. Trigonoscuta outgroups were selected
from the putative sister taxa to Trigonoscuta (Pierce, 1975),
as well as other North American dune-restricted weevils
(Miloderes Casey, 1888). The ingroup taxa were selected from
each of the known populations described in Pierce (1975), as
well as localities from museum specimens (CAS, CDFA,
COB, EMEC, LACM, UCD, UCR). For Rhaphiomidas flies,
outgroups included members of four separate subfamilies of
Mydidae, Apioceridae and a single Asilidae. The outgroups
for the Ceuthophilinae subfamily of crickets were sampled
from the Pristoceuthophilini (Pristoceuthophilus), Argytini
(Argyrtes), and Macropathinae (Macropathini, Heteromal-
laus). For each location (isolated sand dunes), individual
specimens were initially treated as separate taxa in phyloge-
netic analysis so as not to bias the sampling by imposing
previous taxonomic concepts. A maximum of 12 individuals
were sampled per population, with an average of four for all
but the U.S. coastal species of Trigonoscuta. This sampling
regime was used to assess levels of incomplete lineage sorting
between samples, or instances of mitochondrial introgression.
Our Supporting Information, Appendix S1 includes: (1)
standard molecular lab procedures used for extraction, PCR,
sequencing and alignment; (2) references for sequences
derived from Genbank for non-insect taxa. For a list of genes
used see Table 1.
Phylogenetic dating analyses
We chose to partition sequences by codon position because
it has been demonstrated repeatedly that incorporating dif-
ferent rates of DNA evolution for each codon position out-
performs single-partition strategies (Brandley et al., 2005;
Fyler et al., 2005; Seago et al., 2011). Model selection was
performed in MrModeltest2 (Nylander, 2002). The models
for different partitions were selected using AIC and AICc.
For phylogenetic reconstruction, beast 1.7.5 (Drummond
et al., 2012) was used. Phylogenetic trees were dated using a
lognormally distributed relaxed clock model using the birth-
death tree prior (Drummond et al., 2006; see Table 1 for cal-
ibration priors). Each of the Markov chain Monte Carlo
(MCMC) analyses was run for sufficient generations to reach
stationarity, with trees and model parameters sampled from
the stationary posterior distribution. Stationarity was assessed
Journal of Biogeography
ª2016 John Wiley & Sons Ltd
4
M. H. Van Dam and N. J. Matzke
using the program Tracer 1.5.3 (Drummond, 2007). The
most challenging part of the analysis was derivation of dating
priors; we used (admittedly approximate) fossil and biogeo-
graphical calibration points; if none were available then an
informative prior on the relaxed clock rate was used
(Table 1). We decided against performing a gene-tree/species
tree analysis (Pamilo & Nei, 1988) mainly because of the
available data for our clades (Table 1). For 6/8 clades,
mtDNA (a single locus) was all that was available. For the
sand treaders crickets there were two available loci, COI
(mtDNA) and H3 (nuclear DNA). H3 is slowly evolving and
was sequenced only to provide improved resolution deep in
the tree. More loci are available for Rhaphiomidas flies, but
the nuclear genes chosen are slow-evolving and have sparse
sampling.
Biogeographical analyses
Biogeographical model fitting and ancestral range
estimation
For biogeographical analyses, we used the R package Bio-
GeoBEARS (Matzke, 2013). BioGeoBEARS requires as
inputs (1) a dated phylogeny, (2) a file of geographical
ranges indicating presence/absence of each species or coalesc-
ing population in each discrete area in the analysis and (3)
constraint matrices indicating connectivity and/or distance
between the discrete areas. The sampling localities were
grouped into the following discrete areas: Mojave River
watershed, Owens Valley River, Bristol Trough and Clarks
Path sand transport pathway, Parker Dunes, Colorado River
Dunes, Sonoran Desert, Chihuahuan Desert, Great Basin
Desert, Great Plains and Coastal Dunes (Fig. 1). We did not
have any species that occupied more than four areas, so we
allowed for a maximum of four areas at each node, therefore
a total of 562 possible states (geographical ranges) in the
state space. We could not include larger numbers of areas
due to computational limitations. Areas and distances
between the areas were defined in ArcGIS software (Fig. 1),
with distance measured in kilometres along the most likely
path of dispersal (for Pleistocene river connections) or
closest distance between the borders of two areas (for discon-
nected areas). These distances were used in the constrained-
distance-dependent dispersal matrix. The boundaries between
the sand transport pathways (Clarks, Bristol, Parker and
Mojave River) were set as defined by Muhs et al. (2003). The
Colorado River pathway was traced as the area adjacent to
the river, as this was one of the hypothesized dispersal corri-
dors of Norris (1958). For the distances and shapes of the
Great Basin, Sonoran and Chihuahuan deserts, we set the
perimeters of these areas to encompass the most peripheral
dunes of each region. The specimen-level phylogenies were
pruned so that a single operational taxonomic unit (OTU)
was left per species/coalescing monophyletic population. As
the branches are very short within the coalescing popula-
tions of the trees, it would make very little difference which
specimen was used, so we randomly selected one tip from
each population and used it as a representative of the popu-
lation. Pruning was done because DEC-type models, includ-
ing the DEC+J model (Fig. 2), assume that a lineage can
possibly inhabit more than one area. If a specimen-level tree
is used in a DEC analysis, biased results will be obtained if
the individual specimens themselves are used as OTUs,
because a specimen can only inhabit a single area. When all
OTUs inhabit single areas, the data will tend to prefer ‘+J’
models (Matzke, 2013, 2014). This is acceptable if each spe-
cies/monophyletic population really is restricted to a single
area, but not if this is due to the OTUs being specimens.
The geographical structure of gene trees within species/coa-
lescing populations also is interesting, but is not the topic of
this study. We implemented eight models in BioGeoBEARS.
The eight models are: (1) the DEC model, unconstrained,
(2) DEC+J unconstrained, (3) DEC model with dispersal
constrained to adjacent areas, (4) the DEC+J model with dis-
persal constrained to adjacent areas, (5) the DEC+x model,
where dispersal is limited to adjacent areas and modified by
distance taken to the power x, (6) the DEC+J+x model,
which adds jump dispersal, also modified by distance to the
power x, (7) DEC+x without adjacent area constraints and
(8) DEC+J+x without adjacent area constraints.
A common misconception about constraints in DEC/
DEC+J-type analyses is that constrained analyses will always
yield a lower maximum log-likelihood than unconstrained
analyses. Dispersal constraints can actually increase or
decrease the likelihood of the data, keeping in mind that the
ML estimates of dand ecan be different between constrained
and unconstrained models. In essence, a model would confer
the highest likelihood if it gave high probability to just the
dispersal events needed to explain the observed tip ranges,
and low probability to all of the events that would produce
unobserved ranges. When a constraints matrix rules out
unnecessary dispersal events, dcan be optimized to explain
just the ‘right’ events to explain the data. This is the case if
the data and the constraints match. For example, if many sis-
ter species in a large phylogeny are (A, B), and none are (A,
C), then a model that allows A,B dispersal but disallows
A,C dispersal might be favoured.
Modifying dispersal probability as a function of distance
The ‘+x’ models in BioGeoBEARS take the base dispersal
rates specified by the rest of the model, and modify them as
follows:
Rmod:
¼Rorig:
distancex(1)
Wmod:
¼Worig:
distancex(2)
where R
orig
is the original dispersal rate (for anagenetic range
expansion dispersal), R
mod
the modified dispersal rate, and
the distance is in kilometres. W
orig
and W
mod
are the origi-
nal and modified per-event weights for founder-event jump
dispersal at cladogenesis (under DEC-type models, for
range-change events at cladogenesis, each possible event is
Journal of Biogeography
ª2016 John Wiley & Sons Ltd
5
Model choice in desert historical biogeography
Table 1 Summary of sequence data used in the analysis. Loci count and type are given. Number of specimens sampled per species, plus specimen totals and number of species, are
provided. If species were pre-defined in the literature, they are listed as ‘predefined’. If they were discovered through examination in this study they are listed as ‘discovered’. The node/clock
prior column lists the node or clock priors set in beast used to calibrate the tree, followed by the references for each.
Taxon
mtDNA
loci Nuclear loci
Range no.
samples per
species (average)
and total
no. samples
(tips)
No.
species
Species
pre-defined
or discovered
MCMC
no. gens.
in beast Calibration node/clock prior Type and Reference
Rhaphiomidas CO1,
CO2,
16S
EF1-alpha,
PGD, SNF,
WG, CAD
1–34 (8.4)
(219 total)
26 Discovered 20,000,000 Higher Mydidae stem, Normal
(mean =120.0 SD =10.0)
Fossil from the Crato Formation,
Brazil (Albian) (Grimaldi, 1990)
Sand treader
crickets
CO1 H3 2–37 (9.5)
(227 ingroup,
420 total)
24 Discovered 40,000,000 Caelifera and Ensifera divergence
Normal(mean =250.0
SD =10.0); (Troglophilus,
Dolichopoda, Euhadenoecus,
Hadenoecus) Lognormal(mean =1.0
SD =1.0 offset =34.0),
Gryllidae Lognormal(mean =1.0
SD =1.0 offset =107.0).
includeStem =’true’ for all groups
Fossil, Permo-Triassic boundary
(Grimaldi & Engel, 2005). Fossil,
Baltic amber (Chopard, 1936;
Allegrucci et al., 2005). Fossil
(Heads & Leuzinger, 2011)
Trigonoscuta CO1 1–23 (5.5)
(138 reduced
to help
MCMC,
411 total)
25 Discovered 40,000,000 Gulf of California split parent node,
Normal(mean =6.0 SD =2.0)
Geological vicariance (Murphy &
Aguirre-L
eon, 2002; Oskin &
Stock, 2003)
Uma CytB,
ATPase
19–81 (36.8)
(184 total)
5 Pre-defined,
only five species
40,000,000 Uma stem Normal(mean =33.6
SD =5.0)
Fossil Phyrnosoma from Split
Rock Formation,clade origin
earlier (Robinson & Van
Devender, 1973). Split Rock
Formation overlies White River
formation of Oligocene age
(Love, 1961)
Cyprinodon CytB,
ND1
1–10 (3.8)
(87 total)
23 Pre-defined 40,000,000 Cyprinodon rubrofluviatilis Red river-Pecos
split parent node Normal(mean =4.5
SD =0.5); Cyprinodon
macularius-Cyprinodon eremus parent
node(Pinacate volcanic field eruptions
diverted the R
ıo Sonoyta from the Lower
Colorado River) Normal(mean =1.3
SD =0.5) (see Echelle, 2008)
Geological vicariance
(Echelle, 2008)
Journal of Biogeography
ª2016 John Wiley & Sons Ltd
6
M. H. Van Dam and N. J. Matzke
assigned a per-event weight, and the probability of the event
is this weight divided by the sum of the weights of all events;
(Matzke, 2014). When x=0, as in non-‘+x’ models, distance
has no effect on dispersal, and these models are nested
within distance-dependent models. When xis negative, dis-
persal probability drops off as distance increases. In the unli-
kely event that xwere positive, dispersal probability would
increase with distance. If x=2 (Fig. 3), the inverse square
law is being followed (often considered in studies of dispersal
in ecological studies; see Darlington, 1938 for a review and
critique of the law in dispersalist historical biogeography). It
should be noted that it is entirely possible that range expan-
sion and jump dispersal could follow different dispersal ker-
nels, and that dispersal kernels could have a shape more
complex than an exponential function. However, testing
these possibilities would require a more complex model, with
more free parameters, and possibly more data to estimate
them, and thus should be reserved for a future study. As it is
conceivable that the units on distance might be important
for successful search of the parameter space, we repeated the
ML analyses with the distances rescaled by dividing by the
maximum distance; likelihoods were identical so these results
are not shown.
Statistical model choice
As DEC-type models are nested within DEC+J-type models,
these models can be compared in pairwise fashion using the
LRT, a chi-squared test with one degree of freedom. For
comparing all of the models on a particular data set, nested
or not, AIC and AICc were used. AIC- and AICc-based
weights were calculated and used to calculated relative model
probabilities as percentages (Burnham & Anderson, 2002; see
Supporting Information for the full calculations), in order to
assess the support that geographical range data lend to each
model.
RESULTS
Sampling effort and missing taxa
The study area includes over 200 localities and encompasses
sand dune taxa from the Pacific Coast to the Great Plains of
North America. For Uma,Trigonoscuta and Assiminea, we
had complete sampling of species. For Rhaphiomidas we
sampled all but one of the described species, R. hoguei, which
is only known from one specimen occurring near Laredo,
Texas. Repeated attempts to sample this species in the field
were unsuccessful. This species is most similar to R. brevi-
rostris and could be another Sonoran-Chihuahuan Desert
divergence in Rhaphiomidas.
For the sand treader crickets we sampled all of the
described species except for two species of Daihiniodes from
the Chihuahuan Desert. This genus is most likely para- or
polyphyletic with Daihinibaenetes, and this group in general
is in need of revision. For all the other genera of sand
Table 1 Continued
Taxon
mtDNA
loci Nuclear loci
Range no.
samples per
species (average)
and total
no. samples
(tips)
No.
species
Species
pre-defined
or discovered
MCMC
no. gens.
in beast Calibration node/clock prior Type and Reference
Assiminea CO1,
16S
1–13 (3.5)
(42 total)
12 Pre-defined 20,000,000 Lognormal prior on clockrate
(mean =0.0183 SD =1.0)
Clock rate (Wilke, 2003)
Pyrgulopsis CO1,
ND1
1–37 (2)
(208 total)
100 Pre-defined 10,000,000 Pyrgulopsis owensensis,Pyrgulopsis
perturbata, parent node
Normal(mean =0.819, SD =0.015);
Pyrgulopsis stem Lognormal(mean =5.0
SD =1.0 offset =15.0)
Geological, Bishop Tuff
formation (Hershler & Liu,
2008). Fossil Pyrgulopsis
truckeensis (Firby, 1993;
Stewart & Perkins, 1999;
Hershler & Liu, 2004)
Tryonia CO1 1–14 (2.8)
(108 total)
38 Pre-defined 20,000,000 Tryonia stem Lognormal(mean =10.85
SD =1.0 offset =10)
Fossil (Hershler et al., 1999)
Journal of Biogeography
ª2016 John Wiley & Sons Ltd
7
Model choice in desert historical biogeography
treader crickets we had complete sampling. In the course of
this study many undescribed species from this group were
collected.
For species of the western Cyprinodon clade, previous
authors (Echelle, 2008) sampled all but one species, C. ar-
cuatus, which is believed to be extinct. For Pyrgulopsis we
included 100 of the 134 described species, drawing on
sequences from the literature, with almost complete repre-
sentation of the western species. In the Death Valley system
we lacked P. aardahli and P. nevadensis which have not
been collected since the late 1800s and are likely extinct
(Hershler, 1994). For Tryonia we have representatives from
29 of the 32 described species, as well as undescribed spe-
cies from the Chihuahuan desert region (Hershler et al.,
2011).
Molecular Data
Sequence data were obtained for total of 411 Trigonoscuta
individuals (855 bp mtDNA COI), 227 sand treader cricket
individuals [(1536 bp mtDNA COI), (353 bp nDNA H3)],
and 219 Rhaphiomidas individuals [(2904 bp mtDNA; COI,
COII, 16S), (3720 bp nDNA; EF1alpha, PGD, snf, Wg,
CAD)].
Phylogenetic dating analyses
The estimated ages of the study clades varied greatly. For
example, Rhaphiomidas diverged from the rest of the Mydi-
dae during the early Cretaceous (see Supporting Information
for trees). Other genera were much younger, such as Assimi-
nea, which is <10 million years old, with many divergences
dated to the Pleistocene. Graphics of the dated phylogenies
are available in Supporting Information.
Statistical comparison of biogeographical models
The results of ML inference on each data set and biogeo-
graphical model are given in Supporting Information, along
with the ML parameter estimates of d,eand where applica-
ble, jand x(in models that do not include jor x, these
parameters are fixed to 0). The data usually reject the null
hypothesis that the DEC-type model explained the data as
well as the DEC+J-type model. The exceptions were Uma,
Bristol Trough
Chihuahuan Desert
Clarks Pass
Colorado River Dunes
Great Basin
Mojave River Drainage
Owens Valley
Parker Dunes
Great Plains
Sonoran Desert
Dune Regions
(b)
(b)
(a)
sand dunes
Dispersal Paths
Figure 1 (a) The geographical areas used in
BioGeoBEARS analyses are shaded. Arrows
denote proposed dispersal paths and
distances. The connectivity shown by the
arrows was used to make the constraint
matrices. The distances are implemented in
the multiplier of dispersal probability
matrices. (b) Close-up of dispersal paths.
Journal of Biogeography
ª2016 John Wiley & Sons Ltd
8
M. H. Van Dam and N. J. Matzke
and constrained-dispersal analyses of Assiminea and
Trigonoscuta. With Assiminea and Trigonoscuta, addition of
the adjacency matrix dramatically reduced the log-likelihood
advantage of the DEC+J model over the DEC model, such
that DEC could not be statistically rejected. However, these
constrained models were much poorer explanations of the
data than the unconstrained models (see below), so failure to
distinguish the models may simply be a side-effect of the
general poor fit of the constrained models. In the case of
Uma, the data were so few that no power existed to distin-
guish the different models, and ML optimization failed in
one case, probably due to an extremely flat likelihood sur-
face.
Comparing all of the models with AIC and AICc (Table 2)
revealed that the constrained-dispersal matrix was preferred
over the unconstrained-dispersal matrix in two of the eight
groups, the Cyprinodon pupfish and the Ceuthophilinae
crickets. In one of the groups (Trigonoscuta), the constrained
and unconstrained matrices produced similar likelihoods,
separated only by 0.2 log-likelihood units. In Trigonoscuta,
the DEC+x+J model was a slight improvement (by 1.1 log-
likelihood units) relative to the DEC+J unconstrained model.
This means that DEC+x+J was the favoured model for
Trigonoscuta with AIC, but only ranks third under AICc,
which more strongly penalizes extra free parameters in small
datasets.
Four of the clades overwhelmingly favoured unconstrained
DEC+J models. These were Rhaphiomidas flies and the three
aquatic snail clades (Assiminea,Pyrgulopsis and Tryonia).
Unconstrained DEC was the poorest-performing model over-
all, never achieving even 1% relative model probability under
any data set, under AIC or AICc (except Uma, which had
the aforementioned power problems).
For the individual taxon-specific biogeographical recon-
structions, see Supporting Information.
Anagenetic range-change events
Cladogenetic range-change events
“Dispersal”
(range expansion) (d)
“Extinction”
(range contraction) (e)
AA
BA
A
B
Vicariance (v)
A B BA
Sympatry (subset) (s)
A B BA B
Sympatry (narrow) (y)
AA
A
Founder/jump dispersal (j)
AB
A
Dispersal-Extinction-Cladogenesis (DEC) model
+J
DEC
Figure 2 Cladogenetic and anagenetic
processes allowed in DEC and DEC+J
models.
0 200 400 600 800 1000
distance (km)
Multiplier on dispersal probability
1e−6
1e−5
1e−4
0.001
0.01
0.1
1x=−0.3
x=−1
x=−2
Figure 3 Relationship between the value of xand multiplier on
dispersal probability, given distance.
Journal of Biogeography
ª2016 John Wiley & Sons Ltd
9
Model choice in desert historical biogeography
Table 2 Summary of data likelihoods under each model, and results of statistical model choice. Note Uma was had too few taxa to calculate AICc, indicated by N/A. Unconstrained models
allowed dispersal between any of the areas and adjacent-only implemented the dispersal constraints, as seen in Fig. 1. The DEC+x models used distance between areas to implement dispersal
constraints.
DEC
unconstrained
DEC+J
unconstrained
DEC
adjacent-only
DEC+J
adjacent-only
DEC+x
adjacent-only
DEC+x+J
adjacent-only
DEC+x
unconstrained
DEC+x+J
unconstrained
No. free parameters (K) 2 3 23343 4
Log-likelihoods (LnL) n
Assiminea 12 26.5 19.8 26.7 25.9 26.7 25.9 24.7 18.9
Sand treaders 24 76.2 62.9 75.2 62.8 75.2 63.1 76.2 63.0
Cyprinodon 23 34.3 29.8 32.3 28.5 32.2 28.5 33.9 29.8
Pyrgulopsis 100 210.7 171.1 204.3 187.7 203.4 187.7 210.7 171.1
Rhaphiomidas 26 106.9 96.5 110.5 101.5 108.6 100.8 103.0 95.3
Trigonoscuta 25 59 54.1 56.1 54.3 54.9 53 58.5 52.5
Tryonia 38 104.5 88.7 105 95.5 104.7 95.5 104.4 87.6
Uma 59.2 9.2 9.3 9.3 8.1 9.3 8.5 9.1
AIC
Assiminea 57 45.6 57.4 57.8 59.4 59.8 55.4 45.8
Sand treaders 156.4 131.8 154.4 131.6 156.4 134.2 158.3 133.9
Cyprinodon 72.6 65.6 68.6 63 70.4 65 73.8 67.7
Pyrgulopsis 425.4 348.2 412.6 381.4 412.8 383.4 427.3 350.1
Rhaphiomidas 217.8 199 225 209 223.2 209.6 212.1 198.6
Trigonoscuta 122 114.2 116.2 114.6 115.8 114 123.0 113.0
Tryonia 213 183.4 214 197 215.4 199 214.8 183.2
Uma 22.4 24.4 22.6 24.6 22.2 26.6 23.0 26.1
Relative model probability (AIC)
Assiminea 0% 52% 0% 0% 0% 0% 0% 47%
Sand treaders 0% 36% 0% 40% 0% 11% 0% 13%
Cyprinodon 0% 15% 3% 54% 1% 20% 0% 5%
Pyrgulopsis 0% 73% 0% 0% 0% 0% 0% 27%
Rhaphiomidas 0% 45% 0% 0% 0% 0% 0% 55%
Trigonoscuta 0% 18% 7% 15% 8% 20% 0% 33%
Tryonia 0% 47% 0% 0% 0% 0% 0% 53%
Uma 21% 8% 19% 7% 23% 3% 16% 3%
AICc
Assiminea 58.3 48.6 58.7 60.8 62.4 65.5 58.4 51.5
Sand treaders 157.0 133.0 155.0 132.8 157.6 136.3 159.5 136.0
Cyprinodon 73.2 66.9 69.2 64.3 71.7 67.2 75.1 69.9
Pyrgulopsis 425.5 348.5 412.7 381.7 413.1 383.8 427.6 350.6
Rhaphiomidas 218.3 200.1 225.5 210.1 224.3 211.5 213.2 200.5
Trigonoscuta 122.5 115.3 116.7 115.7 116.9 116.0 124.1 115.0
Tryonia 213.3 184.1 214.3 197.7 216.1 200.2 215.5 184.4
Uma 28.4 48.4 28.6 48.6 46.2 N/A 47.0 N/A
Journal of Biogeography
ª2016 John Wiley & Sons Ltd
10
M. H. Van Dam and N. J. Matzke
DISCUSSION
Statistical model comparison reveals that our study clades fall
into three groups: (1) clades where the unconstrained DEC+J
model overwhelmingly dominates (Assiminea,Pyrgulopsis,
Tryonia and Rhaphiomidas), (2) clades with majority support
for models that constrain dispersal to occur only between
adjacent areas (Trigonoscuta, and Cyprinodon pupfish, Fig. 4)
and (3) a clade with too few data to confidently prefer any
model (Uma). If data were to be collected that allowed Uma
to be broken into a series of localized, monophyletic popula-
tions, it is possible that a model including jump dispersal
would be preferred, as Uma populations seem to have a bio-
geographical pattern similar to Trigonoscuta populations.
However, such data are not available at present.
Leaving Uma aside, all of the clades show support for
models that include founder-event speciation as a process.
This support is either moderate (Trigonoscuta) or very strong
(the rest). This confirms the results of Matzke (2014), which
showed that founder-event speciation is important in oceanic
island systems, and extends these results to habitat islands of
the desert south-west.
The fact that the inferred dating of speciation and biogeo-
graphical events shows little correlation with ancient geologi-
cal events (uplift of the Sierra Madre Occidental c. 34–
15 Ma, or BLF c. 4.80 Ma, see Table 3) accords with the
importance of founder-event speciation. If jump dispersal is
the primary means by which allopatric speciation happens,
there is not likely to be a strong geological correlation to
such events. While major geological events are the easiest
ones to the detect in the rock record (e.g. tectonic, mountain
uplift and major flooding events), and thus have played an
important role as explanations in the vicariance biogeogra-
phy tradition, our data show little support for the role of
ancient geological events in the diversification of the desert
biota under study. However, due to uncertainty in dating
analyses the assessment of the three major geological
hypotheses is put forward as a preliminary consideration.
The biogeographical statistical model choice results presented
here do not depend on the absolute dates, as none of the
models employ time-stratified constraints.
The young ages and very strong support for the uncon-
strained DEC+J model in snail taxa may be an indication of
the importance of bird-mediated dispersal. Other authors
have hypothesized this for snails (Hershler & Liu, 2008), and
there is some evidence for birds distributing snail larvae (van
Leeuwen et al., 2012). Tryonia have larvae that develop
inside the snail female’s genital tract (Hershler, 2001) unlike
the other genera of snails in this study, and this may help
explain why, with one exception, each species is confined to
a single locality, a pattern which strongly favours DEC+J
over DEC (Matzke, 2014). Tryonia porrecta is the only mem-
ber of this genus that is parthenogenic, which may explain
why this species has a relatively broad distribution (Hershler
et al., 2005). Mathematically, the low likelihood of the data
under the constrained-dispersal matrix is due to the fact that
Table 2 Continued
DEC
unconstrained
DEC+J
unconstrained
DEC
adjacent-only
DEC+J
adjacent-only
DEC+x
adjacent-only
DEC+x+J
adjacent-only
DEC+x
unconstrained
DEC+x+J
unconstrained
Relative model probabilities (AICc)
Assiminea 1% 80% 1% 0% 0% 0% 1% 18%
Sand treaders 0% 40% 0% 44% 0% 8% 0% 9%
Cyprinodon 1% 16% 5% 59% 1% 14% 0% 4%
Pyrgulopsis 0% 74% 0% 0% 0% 0% 0% 26%
Rhaphiomidas 0% 55% 0% 0% 0% 0% 0% 45%
Trigonoscuta 1% 21% 11% 17% 10% 15% 0% 25%
Tryonia 0% 54% 0% 0% 0% 0% 0% 46%
Uma N/A N/A N/A N/A N/A N/A N/A N/A
Journal of Biogeography
ª2016 John Wiley & Sons Ltd
11
Model choice in desert historical biogeography
Dune Region
Bristol Trough (A)
Chihuahuan Desert (B)
Clarks Pass (C)
Colorado River Dunes (D)
Pacific Coast (E)
Great Basin (F)
Mojave River Drainage (G)
Owens Valley (H)
Parker Dunes (I)
Great Plains (J)
Sonoran Desert (K)
Cyprinodon_rubrofluviatilis
Cyprinodon_bovinus
Cyprinodon_pecosensis
Cyprinodon_elegans
Cyprinodon_alvarezi
Cyprinodon_veronicae
Cyprinodon_eximius
Cyprinodon_pachycephalus
Cyprinodon_macrolepis
Cyprinodon_Rio_Aguanaval
Cyprinodon_atrorus
Cyprinodon_bifasciatus
Cyprinodon_nazas
Cyprinodon_meeki
Cyprinodon_salinus
Cyprinodon_diabolis
Cyprinodon_nevadensis
Cyprinodon_fontinalis
Cyprinodon_macularius
Cyprinodon_eremus
Cyprinodon_pisteri
Cyprinodon_albivelis
Cyprinodon_radiosus
6420
Millions of years ago
BFGH
B
B
B
BH
B
BH
B
B
B
B
B
B
B
BH
B
B
FG
F
B
K
B
J
B
B
B
B
B
B
B
B
B
B
B
B
B
G
F
FG
B
K
K
B
B
H
BH
B
B
B
BH
B
BH
B
B
B
B
B
B
B
H
B
B
F
F
B
K
B
B
J
B
B
B
B
B
B
B
B
B
B
B
B
B
B
G
G
F
K
K
B
Cyprinodon_rubrofluviatilis
Cyprinodon_bovinus
Cyprinodon_pecosensis
Cyprinodon_elegans
Cyprinodon_alvarezi
Cyprinodon_veronicae
Cyprinodon_eximius
Cyprinodon_pachycephalus
Cyprinodon_macrolepis
Cyprinodon_Rio_Aguanaval
Cyprinodon_atrorus
Cyprinodon_bifasciatus
Cyprinodon_nazas
Cyprinodon_meeki
Cyprinodon_salinus
Cyprinodon_diabolis
Cyprinodon_nevadensis
Cyprinodon_fontinalis
Cyprinodon_macularius
Cyprinodon_eremus
Cyprinodon_pisteri
Cyprinodon_albivelis
Cyprinodon_radiosus
6420
Millions of
y
ears a
g
o
J
B
B
B
B
B
B
B
B
B
B
B
B
B
G
F
FG
B
K
K
B
B
H
(a)
(b)
Figure 4 Ancestral state estimation for
Cyprinodon pupfish in BioGeoBEARS, using
the favoured DEC+J model, with
connectivity constraints. (a) Plot of the
single-most-probable state (geographical
range) at each node (just before speciation)
and post-split (just after speciation). (b) Pie
charts represent the probabilities of each
possible geographical range just before and
after each speciation event.
Journal of Biogeography
ª2016 John Wiley & Sons Ltd
12
M. H. Van Dam and N. J. Matzke
Table 3 Assessment of whether or not any dated biogeographical events match well with geological hypotheses. Columns list taxa, habitat type and hypothesized geological events. The
geological hypotheses were tested by seeing if speciation events occurred between specific areas, hypothesized to be influenced by the geology, during the time frame of the event. If
estimated dates were consistent with one of the geological hypotheses, the nodes and their ages are listed. If the expected biogeographical pattern between areas was not seen, then ‘no
speciation between areas’ is indicated. If the biogeographical pattern matched, but speciation did not occur in the correct time frame, then ‘no speciation between areas during time period’
is indicated.
Taxa Habitat type
Sierra Madre Occidental uplift [c. 34–15
Ma Sonoran-Chihuahuan Deserts (areas
K and B) (median, 95% credible interval)]
Bouse Lakes formation [4.83–4.80
Ma East and West sides of the
lower Colorado River (areas I
and A,C,D) (median, 95%
credible interval)]
2.58 Ma Quaternary glaciation sand transport
river corridors [(speciation between adjacent
areas and within area between adjacent
dunes/springs) (median, 95% credible interval)]
Rhaphiomidas Dune R. episcopus and R. pachyrhynchus (12.3, 4.0–22.8)
R. xanthos and R. painteri (12.8, 6.3–22.6)
No speciation between areas No speciation between areas during time period
Sand treaders Dune No speciation between areas No speciation between areas M. sp. Ibex Dunes and M. sp. Dumont
Dunes(2.5, 0.9–3.9); A. sp. Cadiz/Rice Dunes and
A. sp. Palen Dunes(2.9, 1.7–4.6)
Trigonoscuta Dune No speciation between areas No speciation between areas
during time period
T. sp. Ludlow Dunes and T. sp. Rice
Dunes group (0.9, 0.5–1.6)
Uma Dune U. scoparia,U. notata,U. inornata and U. exsul,
U. paraphygas (14.7, 7.6–22.4)
No speciation between areas U. inornata and U. notata (0.5, 0.2–0.7)
Cyprinodon Aquatic No speciation between areas during time period No speciation between areas C. albivelis and C. pisteri (0.2, 0.1–0.4); C. eremus and
C. macularius (0.5, 0.2–0.8); C. salinus and C. nevadensis,
C. diabolis (0.8, 0.5–1.2); C. nevadensis and C. diabolis
(0.5, 0.3–0.7); C. radiosus and rest of Western pupfish clade
(2.8, 1.9–3.7); C. sp. Rio Aguanava and C. bifasciatu,C. atrorus
(1.8, 1.1–2.6); C. pachycephalus and C. eximius (0.4, 0.2–0.6);
C. elegans and C. pecosensis,C. bovinus (2.7, 1.8–3.6); C. pecosensis
and C. bovinus (0.7, 0.4–1.0)
Pyrgulopsis Aquatic No speciation between areas during time period No speciation between area Many (45 speciation events between adjacent or within areas)
Assiminea Aquatic No speciation between area No speciation between area A. cienegensis and A. pecos (0.3, 0.005–1.9); A. sp. 20A and
A. infima clade (0.3, 0.005–1.4); A. infima and A. sp A25B
Inyo clade (0.1, 0.003–0.8); A. sp A25B and A. sp. 27A clade
(0.08, 0.002–0.5); A. sp 30B and A. sp 26A clade (0.1, 0.002–0.6)
Tryonia Aquatic No speciation between areas during time period No speciation between area Many (20 speciation events between adjacent or within areas)
Journal of Biogeography
ª2016 John Wiley & Sons Ltd
13
Model choice in desert historical biogeography
Dune Regions
Bristol Trough (A)
Chihuahuan Desert (B)
Clarks Pass (C)
Colorado River Dunes (D)
Pacific Coast (E)
Great Basin (F)
Mojave River Drainage (G)
Owens Valley (H)
Parker Dunes (I)
Great Plains (J)
Sonoran Desert (K)
(a)
(b)
Trigonoscuta_cf_coast_group
Trigonoscuta_San_Nicolas_Is
Trigonoscuta_Catalina_Island
Trigonoscuta_mainland_Island
Trigonoscuta_North_Channel_Is
Trigonoscuta_south_transverse
Trigonoscuta_north_transverse
Trigonoscuta_BCS_Pacific_Coast
Trigonoscuta_Gran_Desierto_sp
Planis_621m
Trigonoscuta_holtvillei_sp
Trigonoscuta_Gulf _of_CA
Trigonoscuta_Pacific_Baja_sp
Trigonoscuta_Pacific_Baja_sp
Trigonoscuta_sp_Mohawk_rainbo
w
Trigonoscuta_sleeperi_sp_group
Trigonoscuta_Cadiz
Trigonoscuta_Catfish
Trigonoscuta_BeaverDam
Trigonoscuta_Parker_sp_group
Trigonoscuta_Palen_sp_group
Trigonoscuta_Rice_group
Trigonoscuta_Ludlownonpeb
Trigonoscuta_Imperial_Valley_sp
Trigonoscuta_Kelso_sp_group
15 10 5 0
Millions of years ago
EK
EK
EK
E
E
E
E
K
K
K
K
K
K
K
ACK
A
AIK
A
D
ACIK
ACK
ACK
AD
K
K
E
E
E
E
E
E
K
K
K
K
K
K
K
K
AC
A
D
D
I
ACK
AD
A
K
GH
K
E
E
E
E
E
E
ACK
K
K
K
K
K
K
A
AIK
ACIK
D
D
ACK
K
AD
A
G
EK
EK
K
E
E
E
E
K
K
K
K
K
K
K
K
AC
A
A
D
I
ACK
ACK
AD
K
15 10 5 0
Millions of
y
ears a
g
o
K
E
E
E
E
E
E
K
K
K
K
K
K
K
K
AC
A
D
D
I
ACK
AD
A
K
GH
Trigonoscuta_cf_coast_group
Trigonoscuta_San_Nicolas_Is
Trigonoscuta_Catalina_Island
Trigonoscuta_mainland_Island
Trigonoscuta_North_Channel_Is
Trigonoscuta_south_transverse
Trigonoscuta_north_transverse
Trigonoscuta_BCS_Pacific_Coast
Trigonoscuta_Gran_Desierto_sp
Planis_621m
Trigonoscuta_holtvillei_sp
Trigonoscuta_Gulf _of_CA
Trigonoscuta_Pacific_Baja_sp
Trigonoscuta_Pacific_Baja_sp
Trigonoscuta_sp_Mohawk_rainbo
w
Trigonoscuta_sleeperi_sp_group
Trigonoscuta_Cadiz
Trigonoscuta_Catfish
Trigonoscuta_BeaverDam
Trigonoscuta_Parker_sp_group
Trigonoscuta_Palen_sp_group
Trigonoscuta_Rice_group
Trigonoscuta_Ludlownonpeb
Trigonoscuta_Imperial_Valley_sp
Trigonoscuta_Kelso_sp_group
Figure 5 Ancestral state estimation for
Trigonoscuta weevils in BioGeoBEARS,
using the favoured DEC+x+J model with no
connectivity constraints. (a) Plot of the
single-most-probable state (geographical
range) at each node (just before speciation)
and post-split (just after speciation). (b) Pie
charts represent the probabilities of each
possible geographical range just before and
after each speciation event.
Journal of Biogeography
ª2016 John Wiley & Sons Ltd
14
M. H. Van Dam and N. J. Matzke
snail clades often have sister taxa in areas that are not
allowed in the connectivity matrix, such as between the
Mojave River Drainage and Sonoran Desert. Pyrgulopsis also
has several sister taxa with distributions not allowed by the
connectivity matrix. The fact that Rhaphiomidas flies are also
well fit by unconstrained DEC+J suggests that dispersal
through flight is also important in this group.
In the case of Cyprinodon (pupfish), models including
the constrained-dispersal matrix gathered 79% of the total
probability based on AICc weights, with the DEC+J con-
strained model having 59% probability. This finding is con-
sistent with hypothesized riverine transport between areas.
The strong support for DEC+J as opposed to DEC in the
constrained models suggests that while new populations
were founded by dispersal during periods of river connec-
tivity, these were nevertheless rare events correlated with
cladogenesis. It must be admitted, however, that the vicari-
ance model available in DEC (and thus DEC+J) is actually
a quite crude equal-weights model, and thus does not
include anything increasing the probability of vicariance
events after a widespread habitat has been broken up; thus
future, more sophisticated probabilistic models of vicariance
might yield stronger statistical support for vicariance in
Cyprinodon.
Cyprinodon and Trigonoscuta species occupy Death Valley
and other northern extensions of the Mojave River Drainage
Basin, and they share pattern and timing in these basins.
Both taxa seem to have undergone local dispersal in the
Mojave River Basin during the Pleistocene, with long periods
of isolation indicated by deeper nodes. In addition, Trigonos-
cuta was the only lineage where the DEC+J+x model was one
of the best-fitting models (Fig. 5). This is likely due to their
poor dispersal ability (flightless with small slow moving legs),
which is reflected in biogeographical distributions where
close phylogenetic relatives tend to occupy nearby geographi-
cal areas. The fact that Cyprinodon shows little support for a
distance effect might indicate that rare, stochastic events –
perhaps major floods –were the main avenue of dispersal,
rather than slow, progressive colonization of a series of inter-
mediate locations.
It seems highly probable that dispersal probability and dis-
tance must be correlated to some extent in historical bio-
geography, so we were surprised that distance-based models
were often not supported in our study groups. There are sev-
eral possible explanations for this. Firstly, it might be the
case that distances may be too small to be a strong predictor
of dispersal in this sub-continental study region. Secondly,
connectivity of sandy washes during or after flooding may be
more important in some study groups, at which point dis-
tance becomes irrelevant. Thirdly, it is possible that the
heterogeneous nature of the intervening habitat plays a large
role in blocking dispersal, for example, mountain ranges pre-
venting dispersal even over short linear distances.
Extinction could be a partial explanation for some of the
observed disjunct distributions. However, this is very difficult
to test, due to the limitations of the fossil record for most of
these taxa. Extinction is a possible explanation particularly in
Cyprinodon, for example, if intermediate populations existed
between the Sonoran Desert and Owens Valley, became
extinct when ancient drainages such as those hypothesized to
link the Bristol Trough to the Colorado River dried up
(Crews & Gillespie, 2014), then extinction could help to
explain some of the disjunct distributions in this group.
CONCLUSIONS
Our results demonstrate that constructing new biogeographi-
cal models, and assessing their fit to the geographical range
data of different clades, can give insight into the biogeo-
graphical processes that are important in each group, and
also what life history traits may be influencing these distribu-
tions. Statistical model choice tools also indicate when cer-
tain predictors are not useful. In the case of these data and
models, the xparameter was rarely useful. This may be due
to the design of the study, which was focused on testing pre-
viously popular hypotheses for desert biogeography, based
on vicariance concepts, and was also focused within one
region of a continent. Distance is extremely likely to be an
important factor influencing dispersal probability, at least at
the coarsest geographical scales (dispersal between continents
and between remote oceanic islands). We believe that this
method has broad applicability to many biogeographical
regions other than just islands or habitat islands. For exam-
ple, defining biogeographical regions on a contiguous land-
scape often results in adjacent areas touching each other in
geographical space. Here, researchers could use various met-
rics for environmental distance between regions, and test
which have the strongest explanatory value.
Two main conclusions can be draw from this study. First,
the construction of spatially explicit models of connectivity
and distance is feasible using GIS, and the usefulness of these
models can be assessed with likelihood-based statistics. This
strategy is less problematic than past efforts which used sub-
jectively chosen dispersal rate modifiers to represent these
factors, which are then fixed in subsequent analysis, although
at the time it was the best option available (Clayton et al.,
2009; Condamine et al., 2013). Second, there is some biogeo-
graphical evidence supporting the sand/reverie transport
pathways for flightless dune taxa and the desert pupfish. The
other, more vagile taxa largely disperse through founder-
event speciation, with sand dune and aquatic habitats in the
deserts playing a role similar to oceanic islands.
ACKNOWLEDGEMENTS
M.H.V.D. was funded by the following: Desert Research
Fund, from The Community Foundation Serving Riverside
and San Bernardino Counties; Orthopterists’ Fund; UC
MEXUS Dissertation Research Grant; Theodore Roosevelt
Memorial Fund from the American Museum of Natural His-
tory; Anza-Borrego Foundation and Institute (ABFI) Ento-
mology; Point Reyes National Seashore National Park
Journal of Biogeography
ª2016 John Wiley & Sons Ltd
15
Model choice in desert historical biogeography
Service, U.S. Department of the Interior-Pacific Coast
Science and Learning Centre; Committee for Research and
Exploration of the National Geographic Society; Walker
Fund in Entomology UCB; Robert L. Usinger Memorial
Award UCB. N.J.M. was supported by NSF Award Number
0919124, and by the National Institute for Mathematical and
Biological Synthesis, an Institute sponsored by the National
Science Foundation, the U.S. Department of Homeland
Security, and the U.S. Department of Agriculture through
NSF Awards #EF-0832858 and #DBI-1300426, with addi-
tional support from The University of Tennessee, Knoxville.
N.J.M. was also funded by the Australian Research Council’s
Discovery Early Career Researcher Award #DE150101773,
and by The Australian National University. M.H.V.D. per-
formed sampling, sequencing, and dating; N.J.M. constructed
and ran the biogeography models; both conducted statistical
analysis and wrote the paper.
REFERENCES
Allegrucci, G., Todisco, V. & Sbordoni, V. (2005) Molecular
phylogeography of Dolichopoda cave crickets (Orthoptera,
Rhaphidophoridae): a scenario suggested by mitochondrial
DNA. Molecular Phylogenetics and Evolution,37, 153–164.
Anderson, R.S. (2002) Curculionidae. American Beetles, Vol.
2: Polyphaga. Scarabaeoidea through Curculionoidea (ed. by
R.H. Arnett, M.C. Thomas, P.E. Skelley and J.H. Frank),
pp. 722–815. CRC Press, Boca Raton, FL.
Brandley, M.C., Schmitz, A. & Reeder, T.W. (2005) Parti-
tioned Bayesian analyses, partition choice, and the phylo-
genetic relationships of scincid lizards. Systematic Biology.,
54, 373–390.
Brooks, D.R., Van Veller, M.G.P. & McLennan, D.A. (2002)
How to do BPA, really. Journal of Biogeography.,28, 345–
358.
Burnham, K.P. & Anderson, D.R. (2002) Model selection and
multimodel inference: a practical information-theoretic
approach. Springer, New York.
Casey, T. (1888) North American Rhynchophora. Annals of
the New York Academy of Sciences,4, 229–296.
Castoe, T.A., Daza, J.M., Smith, E.N., Sasa, M.M., Kuch, U.,
Campbell, J.A., Chippindale, P.T. & Parkinson, C.L.
(2009) Comparative phylogeography of pitvipers suggests
a consensus of ancient Middle American highland bio-
geography. Journal of Biogeography,36,88–103.
Cazier, M.A. (1985) A revision of the North American flies
belonging to the genus Rhaphiomidas (Diptera, Apioceri-
dae). Bulletin of the AMNH,182, 182–263.
Chopard, L. (1936) Orthopteres fossiles et subfossiles de
I’ambre et du copal. Annales de la Soci
et
e entomologique de
France,105, 375–386.
Clayton, J.W., Soltis, P.S. & Soltis, D.E. (2009) Recent long-
distance dispersal overshadows ancient biogeographical
patterns in a pantropical angiosperm family (Simarouba-
ceae, Sapindales). Systematic Biology,58, 395–410.
Condamine, F.L., Sperling, F.A.H. & Kergoat, G.J. (2013) Glo-
bal biogeographical pattern of swallowtail diversification
demonstrates alternative colonization routes in the North-
ern and Southern hemispheres. Journal of Biogeography,40,
9–23.
Crews, S.C. & Gillespie, R.G. (2014) Desert salt flats as oases
for the spider Saltonia incerta Banks (Araneae: Dic-
tynidae). Ecology and Evolution,4, 3861–3874.
Darlington, P.J., Jr (1938) The origin of the fauna of the
Greater Antilles, with discussion of dispersal of animals
over water and through the air. Quarterly Review of Biol-
ogy,13, 247–300.
Drummond, A.J. (2007) Tracer v1.5.3. Available at: http://
beast.bio.ed.ac.uk/Tracer.
Drummond, A.J., Ho, S.Y.W., Phillips, M.J. & Rambaut, A.
(2006) Relaxed phylogenetics and dating with confidence.
PLoS Biology,4, e88.
Drummond, A.J., Suchard, M.A., Xie, D. & Rambaut, A.
(2012) Bayesian phylogenetics with BEAUti and the
BEAST 1.7. Molecular Biology and Evolution,29, 1969–
1973.
Echelle, A.A. (2008) The western pupfish clade (Cyprinodon-
tidae: Cyprinodon): mtDNA divergence times and drainage
history.Late cenozoic drainage history of the Southwestern
Great Basin and Lower Colorado River Region: geologic and
biotic perspectives (ed. by L.M. Reheis, R. Hershler and
D.M. Miller). Special Papers- Geological Society of America,
439,27–38.
Evans, A.V. & Hogue, J.N. (2006) Field guide to beetles of
California. University of California Press, Berkeley, CA 334
pp.
Ferrari, L., L
opez-Mart
ınez, M., Aguirre-D
ıaz, G. & Car-
rasco-N
u~
nez, G. (1999) Space-time patterns of Cenozoic
arc volcanism in central Mexico: from the Sierra Madre
Occidental to the Mexican Volcanic Belt. Geology,27, 303–
306.
Firby, J.R. (1993) A new species of Vorticifex (Gastropoda:
Planorbidae) from late Cenozoic lake deposits. Nevada.
Journal of Paleontology,67, 370–374.
Fyler, C.A., Reeder, T.W., Berta, A., Antonelis, G., Aguilar,
A. & Androukaki, E. (2005) Historical biogeography and
phylogeny of monachine seals (Pinnipedia: Phocidae)
based on mitochondrial and nuclear DNA data. Journal of
Biogeography,32, 1267–1279.
George, J. & Mattoni, R. (2006) Rhaphiomidas terminatus ter-
minatus Cazier, 1985 (Diptera: Mydidae): notes on the
rediscovery and conservation biology of a presumed
extinct species. Pan-Pacific Entomologist,82,32–35.
Gillespie, R. (2004) Community assembly through adaptive
radiation in Hawaiian spiders. Science,303, 356–359.
Grant, P.R. & Grant, B.R. (2002) Unpredictable evolution in
a 30-year study of Darwin’s finches. Science,296, 707–711.
Grimaldi, D.A. (1990) Insects from the Santana Formation,
Lower Cretaceous, of Brazil. Bulletin of the American
Museum of Natural History,195,5–191.
Journal of Biogeography
ª2016 John Wiley & Sons Ltd
16
M. H. Van Dam and N. J. Matzke
Grimaldi, D.A. & Engel, M.S. (2005) Evolution of the insects.
Cambridge University Press, Cambridge.
Hardy, A. & Andrews, F. (1976) A final report to the Office
of Endangered Species on contract 14-16-0008-966.
Harris, R.B., Birks, S.M. & Leach
e, A.D. (2014) Incubator
birds: biogeographical origins and evolution of under-
ground nesting in megapodes (Galliformes: Megapodi-
idae). Journal of Biogeography,41, 2045–2056.
Heads, S.W. & Leuzinger, L. (2011) On the placement of the
Cretaceous orthopteran Brauckmannia groeningae from
Brazil, with notes on the relationships of Schizodactylidae
(Orthoptera: Ensifera). ZooKeys,77,17–30.
Hershler, R. (1994) A review of the North American fresh-
water snail genus Pyrgulopsis (Hydrobiidae). Smithsonian
Contributions to Zoology,554,1–115.
Hershler, R. (2001) Systematics of the North and Central
American aquatic snail genus Tryonia (Rissooidea: Hydro-
biidae). Smithsonian Contributions to Zoology,612,1–53.
Hershler, R. & Liu, H.-P. (2004) A molecular phylogeny of
aquatic gastropods provides a new perspective on biogeo-
graphic history of the Snake River region. Molecular Phylo-
genetics and Evolution,32, 927–937.
Hershler, R. & Liu, H.-P. (2008) Ancient vicariance and
recent dispersal of springsnails (Hydrobiidae: Pyrgulopsis)
in the Death Valley system, California-Nevada. Special
Papers- Geological Society of America,439,91–101.
Hershler, R. & Sada, D.W. (2002) Biogeography of Great
Basin aquatic snails of the genus Pyrgulopsis.Great basin
aquatic systems history (ed. by R. Hershler, D.B. Madsen
and D.R. Currey), pp. 255–276. Smithsonian Contributions
to the Earth Sciences,33. Smithsonian Institution, Wash-
ington D.C., WA.
Hershler, R., Liu, H.-P. & Mulvey, M. (1999) Phylogenetic
relationships within the aquatic snail genus Tryonia:
implications for biogeography of the North American
Southwest. Molecular Phylogenetics and Evolution,13,
377–391.
Hershler, R., Mulvey, M. & Liu, H.-P. (2005) Genetic varia-
tion in the Desert Springsnail (Tryonia porrecta): implica-
tions for reproductive mode and dispersal. Molecular
Ecology,14, 1755–1765.
Hershler, R., Liu, H.-P. & Lang, B.K. (2007) Genetic and
morphologic variation of the Pecos Assiminea, an endan-
gered mollusk of the Rio Grande region, United States
and Mexico (Caenogastropoda: Rissooidea: Assimineidae).
Hydrobiologia,579, 317–335.
Hershler, R., Liu, H.-P. & Landye, J.J. (2011) New species
and records of springsnails (Caenogastropoda: Cochliopi-
dae: Tryonia) from the Chihuahuan Desert (Mexico and
United States), an imperiled biodiversity hotspot. Zootaxa,
3001,1–32.
House, P.K., Pearthree, P.A., Howard, K.A., Bell, J.W., Per-
kins, M.E., Faulds, J.E. & Brock, A.L. (2005) Birth of the
lower Colorado River—Stratigraphic and geomorphic evi-
dence for its inception near the conjunction of Nevada,
Arizona, and California. Geological Society of America Field
Guide 6 (ed. by J.L. Pederson and C. Dehler), pp. 357–
387. Geological Society of America, Interior Western Uni-
ted States, Boulder, CO.
House, P.K., Pearthree, P.A. & Perkins, M.E. (2008) Strati-
graphic evidence for the role of lake spillover in the incep-
tion of the lower Colorado River in southern Nevada and
western Arizona. Late Cenozoic drainage history of the
Southwestern Great Basin and Lower Colorado River Basin:
geologic and biotic perspectives: Geological Society of America
Special Paper 439 (ed. by D.M. Miller, L.M. Reheis and R.
Hershler), pp. 335–353. Geological Society of America.
Knowles, L.L. & Carstens, B.C. (2007) Delimiting species
without monophyletic gene trees. Systematic Biology,56,
887–895.
Knowles, L.L. & Maddison, W.P. (2002) Statistical Phylo-
geography. Molecular Ecology,11, 2623–2635.
Landis, M.J., Matzke, N.J., Moore, B.R. & Huelsenbeck, J.P.
(2013) Bayesian analysis of biogeography when the num-
ber of areas is large. Systematic Biology,62, 789–804.
Lawing, A.M. & Matzke, N.J. (2014) Conservation paleobiol-
ogy needs phylogenetic methods. Ecography,37, 1109–1122.
van Leeuwen, C.H.A., van der Velde, G., van Lith, B. &
Klaassen, M. (2012) Experimental quantification of long
distance dispersal potential of aquatic snails in the gut of
migratory birds. PLoS ONE,7, e32292.
Liu, H.-P., Hershler, R., Lang, B. & Davies, J. (2013) Molecu-
lar evidence for cryptic species in a narrowly endemic
western North American springsnail (Pyrgulopsis gilae).
Conservation Genetics,14, 917–923.
Losos, J.B., Leal, M., Glor, R.E., de Queiroz, K., Hertz, P.E.,
Rodr
ıguez Schettino, L., Lara, A.C., Jackman, T.R. & Lar-
son, A. (2003) Niche lability in the evolution of a Carib-
bean lizard community. Nature,424, 542–545.
Love, J.D. (1961) Split Rock formation (Miocene) and
Moonstone formation (Pliocene) in Central Wyoming.
United States Geological Survey Bulletin,1121–1,1–39.
Lucchitta, I. (1979) Late Cenozoic uplift of the southwestern
Colorado Plateau and adjacent lower Colorado River
region. Tectonophysics,61,63–95.
Lucchitta, I., McDougall, K., Metzger, D.G., Morgan, P.,
Smith, G.R. & Chernoff, B. (2001) The Bouse formation
and post-Miocene uplift of the Colorado Plateau. The Col-
orado river: origin and evolution (ed. by R.A. Young and
E.E. Spamer). Grand Canyon, Arizona, Grand Canyon
Association Monograph 12, pp. 173–178.
Masta, S.E. (2000) Phylogeography of the jumping spider
Habronattus pugillis (araneae: salticidae): recent vicariance
of sky island populations? Evolution; International Journal
of Organic Evolution,54, 1699–1711.
Matzke, N.J. (2013) Probabilistic historical biogeography:
new models for founder-event speciation, imperfect detec-
tion, and fossils allow improved accuracy and model-test-
ing. Frontiers of Biogeography,5, 242–248.
Matzke, N.J. (2014) Model selection in historical biogeogra-
phy reveals that founder-event speciation is a crucial pro-
cess in island clades. Systematic Biology,63, 951–970.
Journal of Biogeography
ª2016 John Wiley & Sons Ltd
17
Model choice in desert historical biogeography
Minckley, W.L., Miller, R.R. & Norris, S.M. (2002) Three
new pupfish species, Cyprinodon (Teleostei, Cyprinodonti-
dae), from Chihuahua, Mexico, and Arizona, USA. Copeia,
2002, 687–705.
Muhs, D.R., Reynolds, R., Been, J. & Skipp, G. (2003) Eolian
sand transport pathways in the southwestern United
States: importance of the Colorado river and local sources.
Quaternary International,104,3–18.
Murphy, R.W. & Aguirre-L
eon, G. (2002) The nonavian rep-
tiles. A new island biogeography of the sea of Cort
es (ed. by
T.J. Case, M.L. Cody and E. Ezcurra), pp. 181–220.
Oxford University Press, New York.
Nieto-Samaniego, A.F., Ferrari, L., Alaniz-Alvarez, S.A.,
Labarthe-Hernandez, G. & Rosas-Elguera, J. (1999) Vari-
ation of Cenozoic extension and volcanism across the
southern Sierra Madre Occidental volcanic province,
Mexico. Geological Society of America Bulletin,111, 347–
363.
Norris, K.S. (1958) The evolution and systematics of the
iguanid genus Uma and its relation to the evolution of
other North American desert reptiles. Bulletin of the Amer-
ican Museum of Natural History,114, 251.
Nylander, J. (2002) MrModeltest 2.2. Program distributed by
the author. Evolutionary Biology Centre, Uppsala Univer-
sity, Uppsala, Sweden.
Oskin, M. & Stock, J. (2003) Marine incursion synchronous
with plate boundary localization in the Gulf of California.
Geology,31,23–26.
Pamilo, P. & Nei, M. (1988) Relationships between gene
trees and species trees. Molecular Biology and Evolution,5,
568–583.
Pierce, W.D. (1975) The sand dune weevils of the genus
Trigonoscuta, with a correlation of their anatomy to the geo-
logical history of our coast lines. Natural History Museum
of Los Angeles County, Los Angeles.
Poulson, S.R. & John, B.E. (2003) Stable isotope and trace
element geochemistry of the basal Bouse Formation car-
bonate, southwestern United States: Implications for the
Pliocene uplift history of the Colorado Plateau. Geological
Society of America Bulletin,115, 434–444.
Ree, R.H. & Sanmart
ın, I. (2009) Prospects and challenges
for parametric models in historical biogeographical infer-
ence. Journal of Biogeography,36, 1211–1220.
Ree, R.H. & Smith, S.A. (2008) Maximum likelihood
inference of geographic range evolution by dispersal,
local extinction, and cladogenesis. Systematic Biology,57,
4–14.
Ree, R.H., Moore, B.R., Webb, C.O. & Donoghue, M.J.
(2005) A likelihood framework for inferring the evolution
of geographic range on phylogenetic trees. Evolution,59,
2299–2311.
Riddle, B.R. & Hafner, D.J. (2006) A step-wise approach to
integrating phylogeographic and phylogenetic biogeo-
graphical perspectives on the history of a core North
American warm deserts biota. Journal of Arid Environ-
ments,66, 435–461.
Robinson, M.D. & Van Devender, T.R. (1973) Miocene
lizards from Wyoming and Nebraska. Copeia, 698–704.
Rogers, R. & Van Dam, M.H. (2007) Two new species of
Rhaphiomidas (Diptera: Mydidae). Zootaxa,68,61–68.
Ronquist, F. & Sanmart
ın, I. (2011) Phylogenetic methods in
biogeography. Annual Review of Ecology, Evolution, and
Systematics,42, 441–464.
Roskowski, J.A., Patchett, P.J., Spencer, J.E., Pearthree, P.A.,
Dettman, D.L., Faulds, J.E. & Reynolds, A.C. (2010) A late
Miocene-early Pliocene chain of lakes fed by the Colorado
River: evidence from Sr, C, and O isotopes of the Bouse
Formation and related units between Grand Canyon and
the Gulf of California. Geological Society of America Bul-
letin,122, 1625–1636.
Seago, A.E., Giorgi, J.A., Li, J. &
Slipi
nski, A. (2011) Phy-
logeny, classification and evolution of ladybird beetles
(Coleoptera: Coccinellidae) based on simultaneous analysis
of molecular and morphological data. Molecular Phyloge-
netics and Evolution,60, 137–151.
Spencer, J.E. & Patchett, P.J. (1997) Colorado Plateau uplift Sr
isotope evidence for a lacustrine origin for the upper Mio-
cene to Pliocene Bouse Formation, lower Colorado River
trough, and implications for timing of Colorado Plateau
uplift. Geological Society of America Bulletin,109, 767–778.
Spencer, J.E., Patchett, P.J., Pearthree, P.A., House, P.K.,
Sarna-Wojcicki, A.M., Wan, E., Roskowski, J.A. & Faulds,
J.E. (2013) Review and analysis of the age and origin of
the Pliocene Bouse Formation, lower Colorado River Val-
ley, southwestern USA. Geosphere,9, 444–459.
Stewart, J.H. & Perkins, M.E. (1999) Stratigraphy,
tephrochronology, and structure of part of the Miocene
Truckee Formation in the Trinity Range-Hot Springs
Mountains area, Churchill County, west-central Nevada.
USGS Open-file Report, 99–330. 19pp.
Turak, J. (2000) Revaluation of the Miocene/Pliocene deposi-
tional history of the Bouse Formation, Colorado River
trough, southern Basin and Range(CA, NV, and AZ).MS
Thesis, University of Wyoming, Laramie.
Van Dam, M.H. (2010) A new species and key for Rhaphio-
midas Osten Sacken (Diptera: Mydidae). Zootaxa,2622,
49–60.
Van Dam, A.R. & Van Dam, M.H. (2008) Impact of off-road
vehicle use on dune endemic Coleoptera. Annals of the
American Entomological Society,101, 411–417.
Voelker, G., Pe~
nalba, J.V., Huntley, J.W. & Bowie, R.C.K.
(2014) Diversification in an Afro-Asian songbird clade
(Erythropygia-Copsychus) reveals founder-event speciation
via trans-oceanic dispersals and a southern to northern
colonization pattern in Africa. Molecular Phylogenetics and
Evolution,73,97–105.
Wake, D.B. (1987) Adaptive radiation of salamanders in
Middle American cloud forests. Annals of the Missouri
Botanical Garden,72, 242–264.
Webb, C.O. & Ree, R. (2012) Biotic evolution and environ-
mental change in Southeast Asia. Cambridge University
Press, Cambridge.
Journal of Biogeography
ª2016 John Wiley & Sons Ltd
18
M. H. Van Dam and N. J. Matzke
Whittaker, R.J. & Fern
andez-Palacios, J.M. (2007) Island bio-
geography: ecology, evolution, and conservation. Oxford
University Press, Oxford.
Wilke, T. (2003) Salenthydrobia gen. nov. (Rissooidea:
Hydrobiidae): a potential relict of the Messinian salinity
crisis. Zoological Journal of the Linnean Society,137, 319–
336.
Zink, R.M. (2002) Methods in comparative phylogeography,
and their application to studying evolution in the North
American aridlands. Integrative and Comparative Biology,
42, 953–959.
SUPPORTING INFORMATION
Additional Supporting Information may be found in the
online version of this article:
Appendix S1 (1) Maps of sampling locations and dunes
(2) standard molecular lab procedures used for extraction,
PCR, sequencing, alignment and a table of primers used; (3)
references for sequences derived from Genbank of non-insect
taxa; (4) dispersal constraint matrices and BioGeoBEARS
results table; (5) results of beast analyses.
DATA ACCESSIBILITY
The outputs of all BioGeoBEARS analyses make for a mas-
sive PDF (>100 pages) not suitable for Supporting Informa-
tion. However, as the details of the range data and ancestral
range inferences may be of interest to some, we have perma-
nently uploaded the PDF to figshare: http://figshare.com/s/
53e25a28691511e5b8e906ec4bbcf141
BIOSKETCHES
Matthew H. Van Dam is interested in interaction of geo-
logical processes and natural history of an organism shaping
its biogeographical pattern; he is also interested in weevil sys-
tematics.
Nicholas J. Matzke is a Discovery Early Career Research
Award (DECRA) Fellow at The Australian National Univer-
sity, in the lab of Craig Moritz. He works on likelihood and
Bayesian methods in biogeography. He is also the author of
the R package BioGeoBEARS.
Editor: Brett Riddle
Journal of Biogeography
ª2016 John Wiley & Sons Ltd
19
Model choice in desert historical biogeography
