Evolutionary causes and consequences of ungulate migration

Abstract

Ungulate migrations are crucial for maintaining abundant populations and functional ecosystems. However, little is known about how or why migratory behaviour evolved in ungulates. To investigate the evolutionary origins of ungulate migration, we employed phylogenetic path analysis using a comprehensive species-level phylogeny of mammals. We found that 95 of 207 extant ungulate species are at least partially migratory, with migratory behaviour originating independently in 17 lineages. The evolution of migratory behaviour is associated with reliance on grass forage and living at higher latitudes wherein seasonal resource waves are most prevalent. Indeed, originations coincide with mid-Miocene cooling and the subsequent rise of C4 grasslands. Also, evolving migratory behaviour supported the evolution of larger bodies, allowing ungulates to exploit new ecological space. Reconstructions of migratory behaviour further revealed that seven of ten recently extinct species were probably migratory, suggesting that contemporary migrations are important models for understanding the ecology of the past. The authors examine present and past drivers of ungulate migratory behaviour, finding that current migratory ungulates are larger, more grass-dependent and live at higher latitudes on average than non-migrants, and that migration probably emerged after the rise of C4 grasslands and increased seasonality towards the poles.

Full text

Articles
https://doi.org/10.1038/s41559-022-01749-4
1Department of Ecology and Evolutionary Biology, Princeton University, Princeton, NJ, USA. 2School of Life Sciences, Arizona State University, Tempe, AZ,
USA. 3Department of Ecology and Evolutionary Biology, Yale University, New Haven, CT, USA. 4Center for Biodiversity and Global Change, Yale University,
New Haven, CT, USA. 5Institute of Ecology and Evolution, University of Oregon, Eugene, OR, USA. 6Department of Fish and Wildlife Conservation, Virginia
Tech, Blacksburg, VA, USA. ✉e-mail: joeloa@princeton.edu
The scientific community and public imagination have long
been captivated by ungulate migrations. Migrations, like
those of wildebeest in the Serengeti, have been referred to
as one of the natural wonders of the world1 and continue to dem-
onstrate their value, both via ecotourism revenue to local econo-
mies2 and as the focus of critical ecological research3. By tracking
plant quality and quantity across space and time—a behaviour
known as ‘green wave surfing’ (hereafter referred to more gener-
ally as ‘resource tracking’3–5)—migratory ungulates can sustain
much larger populations than their resident counterparts6–8. Since
ungulates track spatio-temporally variable forage across landscapes
(that is, ‘resource waves’), they also serve as important vectors of
nutrients, seeds, spores and diseases along migration corridors
and between seasonal ranges9,10, thus linking ecosystem processes
across large spatial scales. However, despite their cultural, economic
and ecological importance, large gaps remain in our knowledge of
ungulate migrations8,11–13. Both resource tracking and avoidance of
predators, parasites and pathogens have been identified as proxi-
mate drivers of migration13,14 but scant evidence exists regarding
the evolutionary origins of this behaviour11,12. Classically, migratory
behaviour is thought to have evolved via natural selection on genetic
variation directly associated with a migratory phenotype11,15,16.
However, recent evidence suggests that ungulate migrations may be
a cultural phenomenon, wherein socially learned information about
spatio-temporal patterns of plant quality (that is, ‘resource waves’) is
transmitted across generations and improved on via asocial learning
within generations—a process known as cumulative cultural evo-
lution17. In either case, migratory behaviour is thought to emerge
from a combination of physiological, morphological and cogni-
tive traits11,18–20, suggesting that genetics at least partially underpin
the evolution and maintenance of migratory behaviour (that is,
‘migratory genes’ might be reinforced by cultural transmission of
migratory knowledge16,21). Altogether, understanding the role that
ungulate traits (for example, body size, digestion, metabolic physi-
ology) and environmental factors (for example, latitude, resource
waves) play in the evolution of migratory behaviour will bring clar-
ity to the mystery of why some ungulates migrate while others do
not.
Extant migratory ungulates are hypothesized to share a ‘migra-
tory syndrome’, a common suite of environmental, morphological
and behavioural characteristics that interact to form a migratory
phenotype11,19. Environmentally, migratory behaviour is prevalent
in seasonal environments where predictable resource waves are
present3,12,20. Because the seasonality of grass growth in the tropics
and subtropics tends to be more pronounced than that of trees22,23,
migration at lower latitudes is largely restricted to grazing ungu-
lates that depend primarily on grasses to meet their nutritional
needs20,24. In contrast, all plants in temperate and mountainous
regions are seasonally variable in their nutritional quality and quan-
tity (not just grasses)25, which drives the seasonal migrations of
browsers, mixed feeders and grazers alike20. Nevertheless, the most
consistent migrants are grazers even in these seasonal systems3,5,26.
Grass dependence may therefore be tied to the evolution of migra-
tion inside and outside the tropics (Fig. 1). Morphologically, larger
body size may also be a key component of a migratory syndrome
in ungulates27,28 (Fig. 1). Migratory mammals tend to be larger than
non-migratory taxa and larger species undertake longer migra-
tions28–30. Such allometry in migratory behaviour may stem from
the ability of large-bodied species to accumulate greater nutritional
reserves and thereby better tolerate the energetic demands of migra-
tion, reduced predation risk during their migratory journeys and
lower reliance on high-quality forage28,30–32. Thus, we hypothesize
that latitude, grass dependence and body size together may lead to a
migratory syndrome and that these characteristics have jointly con-
tributed to the evolution of migration (Fig. 1; see Supplementary
Notes for additional justification of hypotheses).
To test the relative support for hypothetical models of how
migratory behaviour evolved (Supplementary Fig. 1), we first esti-
mated the evolution of migration across a species-level ungulate
phylogeny and determined how the evolution of migration relates
Evolutionary causes and consequences of
ungulate migration
Joel O. Abraham 1 ✉ , Nathan S. Upham 2,3,4, Alejandro Damian-Serrano3,5 and Brett R. Jesmer 3,4,6
Ungulate migrations are crucial for maintaining abundant populations and functional ecosystems. However, little is known
about how or why migratory behaviour evolved in ungulates. To investigate the evolutionary origins of ungulate migration, we
employed phylogenetic path analysis using a comprehensive species-level phylogeny of mammals. We found that 95 of 207
extant ungulate species are at least partially migratory, with migratory behaviour originating independently in 17 lineages. The
evolution of migratory behaviour is associated with reliance on grass forage and living at higher latitudes wherein seasonal
resource waves are most prevalent. Indeed, originations coincide with mid-Miocene cooling and the subsequent rise of C4 grass-
lands. Also, evolving migratory behaviour supported the evolution of larger bodies, allowing ungulates to exploit new ecological
space. Reconstructions of migratory behaviour further revealed that seven of ten recently extinct species were probably migra-
tory, suggesting that contemporary migrations are important models for understanding the ecology of the past.
NATURE ECOLOGY & EVOLUTION | www.nature.com/natecolevol

Articles NATURE ECology & EvolUTIoN
to the evolution of other ungulate characteristics, namely species
mean adult body size, latitudinal centroid of species’ geographi-
cal range and degree of grass dependence. We then applied phy-
logenetic path analysis, a method for determining the underlying
causal structure in phylogenetically structured comparative data33,
to determine the evolutionary causes and consequences of ungulate
migration. Next, we used a global dataset of the normalized differ-
ence vegetation index (NDVI) to assess the role that resource waves
played in promoting the evolution of migration. Finally, we used
the relationships between migration and other ungulate charac-
teristics to reconstruct the migratory behaviour of recently extinct
ungulates. Overall, we found evidence that migratory behaviour in
ungulates evolved in response to relying on grass forage and living
at high latitudes, which in turn drove the evolution of large body
sizes, and that migration may have been more widespread histori-
cally than it is today.
Results
A migratory syndrome. Extant ungulates are highly variable with
regard to migratory behaviour, body mass, grass consumption and
latitude (Fig. 2). Compiling data from a range of literature sources,
we found that 95 of 207 (45.9%) extant ungulate species are at least
partially migratory. Ungulate body masses span more than 3 orders
of magnitude, from 2.78 to 2,950 kg. Likewise, ungulates range from
pure grazers, consuming entirely grass, to pure browsers, consuming
entirely trees and forbs, with yet others (mixed feeders) eating inter-
mediate amounts of both grass and trees. Furthermore, ungulates
can be found across latitudes, residing in the tropics through to the
Arctic (up to nearly 75° N).
Amid this ecological variation, and in accordance with the exis-
tence of a migratory syndrome, we recovered consistent differences
between the characteristics of migratory and non-migratory ungu-
lates. We found that migratory ungulates are larger, inhabit higF-
Browninanher latitudes and are more grass-dependent on average
than non-migratory ungulates (Fig. 2a–c and Supplementary
Table 1). Likewise, we found that larger ungulates tend to con-
sume more grass on average (Fig. 2d and Supplementary Table 1),
although migratory ungulates are still more grass-dependent than
non-migratory ungulates even accounting for differences in body
size. However, contrary to expectations (Supplementary Notes),
body size is not correlated with latitude across ungulates species
(Fig. 2e and Supplementary Table 1). As such, the larger body sizes
of migratory ungulates are not the direct result of their inhabiting
higher latitudes.
We found that migratory ungulates inhabit distinctly seasonal
environments compared to non-migratory species. Migratory
behaviour is most prevalent among taxa whose ranges include
highly seasonal resource waves (Extended Data Fig. 1). Specifically,
resource wave seasonality, rather than wave magnitude or resource
wave distance, best explains the observed interspecific variation
in migratory behaviour (Extended Data Fig. 1). When added to
the above model of migratory behaviour that includes latitude,
grass dependence and body mass as covariates, resource wave sea-
sonality loses its predictive power and becomes non-significant
(Supplementary Table 1). This suggests that the predictive power
of resource wave seasonality is obscured by its covariation with one
or more other predictors. Unsurprisingly, we found that resource
wave seasonality increases with both increased latitude but also with
increased grass dependence (Extended Data Fig. 1), indicating that
vegetation growth is more seasonal at higher latitudes and also that
ungulates inhabiting landscapes with more seasonal resource waves
are more grass-dependent (probably due to the inability of ungu-
lates to specialize exclusively on a particular plant functional type
in the face of seasonally variable plant availability20,34,35). Together,
these results suggest that inhabiting higher latitudes and relying on
grass for nutrition exposes ungulates to predictable spatio-temporal
variability in resource quality and quantity, ultimately making
migratory behaviour advantageous.
Dynamic evolution. Not only are extant ungulates ecologically
variable but ungulate characteristics have also varied dynamically
through evolutionary time (Fig. 3a-d). By estimating the evolution
of migration, grass consumption, body size and latitudinal range
centroid across the ungulate phylogeny (Supplementary Table 2 and
Supplementary Fig. 2), we found phylogenetic ev idence that the most
recent common ancestor (MRCA) of extant ungulates was most
probably a small-bodied mixed feeder living in the tropics to sub-
tropics, although with marginal statistical support (Supplementary
Table 3). Although the confidence intervals (CIs) on reconstructed
states are broad and encapsulate a variety of ecologically disparate
possibilities (Supplementary Table 3), these findings are consistent
with previous reconstructions of ungulate evolution34 and with fos-
sil evidence36,37. Taken together, these lines of evidence suggest that
some of the characteristics that define the present-day migratory
syndrome—large body sizes, grass dependence and living at high
latitudes—are derived relative to the MRCA. We also found that the
MRCA was more likely non-migratory, although only by a small
margin (Supplementary Table 3); therefore, this finding should be
interpreted as inconclusive, as is often the case when reconstructing
the evolution of binary traits with multiple independent transitions
across the tree. Although the state of the MRCA as non-migratory
is equivocal, we found 17 branches where transitions from
non-migratory to migratory are supported (defined as a shift in the
Body size
Latitude
Grass
dependence
Migration
H1: primed by the environment
Body size
Latitude
Grass
dependence
Migration
H2: driven by grass dependence
Body size
Latitude
Grass
dependence
Migration
H3: enabled by large body size
Fig. 1 | Hypothetical evolutionary models of migration. Migration may
evolve in direct response to a spatio-temporally fluctuating resource
environment, resulting both from living at high latitudes and being reliant
on seasonally variable grasses (H1); alternatively, migration might evolve
in large-bodied, grass-dependent ungulates as a consequence of their
particular need to track grass productivity across large spatial scales (H2)
or migration might evolve when ungulates get large enough to where
migration is energetically feasible (H3).
NATURE ECOLOGY & EVOLUTION | www.nature.com/natecolevol

Articles
NATURE ECology & EvolUTIoN
posterior probability of migratory behaviour from <0.5 to >0.5).
By the same token, we estimate 23 branches along which the loss of
migratory behaviour is supported (defined as a shift in the posterior
probability of migratory behaviour from >0.5 to <0.5). Altogether,
these findings suggest that migratory behaviour was highly labile
across ungulate evolution, with a complex history of independent,
and possibly convergent, gains and losses.
The evolution of migratory behaviour appears to have changed
the evolutionary trajectories of several other ungulate characteris-
tics. We modelled grass dependence, body size and latitude as con-
tinuous characters evolving within the discrete selective regimes
of being migratory or non-migratory, finding that multi-optimum
Ornstein–Uhlenbeck models are preferred almost every time (299
out of 300 iterations) for all characters (Supplementary Table 4).
This suggests that not only do characters associated with migra-
tory behaviour interact evolutionarily, but migratory behaviour is
also associated with distinct evolutionary optima for these charac-
teristics. Such results suggest that ungulate characteristics change
directionally in response to evolving migratory behaviour, although
the direction of causation cannot be ascertained from these results
alone.
Altogether, our results provide evidence for the existence of
a migratory syndrome within ungulates, characterized in part by
large body sizes, grass dependence and living at high latitudes.
Furthermore, this migratory syndrome appears to have evolved
multiple times independently over the course of ungulate evolution.
Finally, our results suggest that the advent of migratory behaviour
changed the adaptive landscape for other ungulate characteristics.
Causes and consequences. In accordance with the hypothesis that
environmental factors motivated the evolution of migratory behav-
iour in ungulates (H1; Fig. 1), we found directional phylogenetic
evidence that latitude and grass dependence underpinned the evo-
lution of migratory behaviour, which in turn drove body size evo-
lution. To do this, we used phylogenetic path analysis, a method
that tests claims of conditional independence implied by various
causal hypotheses to determine the most probable causal relation-
ship between phylogenetically distributed characters. By comparing
alternative models, we found that the most probable causal model
for the evolution of migration (the average of all causal models with
C statistic information criterion corrected for small sample sizes
(ΔCIC)c < 2) shows that two characteristics—inhabiting higher
latitudes and being highly dependent on grass—promoted the evo-
lution of migratory behaviour. Migratory behaviour, in turn, pro-
moted the evolution of large body sizes (Fig. 4a and Supplementary
Table 5).
To fur ther interrogate the hypothesis that res ource waves mediate
the relationship between latitude and grass dependence on migra-
tory behaviour, we tested additional path models that included
links between resource wave seasonality and both latitude and grass
dependence (Supplementary Fig. 3). The average causal model
(Fig. 4b) is structurally similar to the path model without data on
100
Body mass (kg)
1 10 1,000 10,000
0
20
40
60
Latitude (°)
e
Migratory
Non-migratory
P = 0.271
Migratory
Non-migratory
100
Body mass (kg)
1 10 1,000 10,000
Dietary grass fraction
0
0.25
0.50
1.00
0.75
d
P = 2.483 × 10–4
a
1
10
100
1,000
Body mass (kg)
10,000
Non-migratory Migratory
*
0
20
40
60
Non-migratory Migratory
*
b
Dietary grass fraction
0
0.25
0.50
1.00
0.75
Non-migratory Migratory
*
c
P = 1.584 × 10–4 P = 2.636 × 10–5
P = 2.254 × 10–5
Latitude (°)
Fig. 2 | Evolutionary correlations between ungulate characteristics. a–c, Using phylogenetic modelling, we found that migration is positively correlated
with body mass (a), latitude (b) and grass dependence (c), such that migratory ungulates tend to be larger, inhabit higher latitudes and consume more
grass than non-migratory ungulates (two-sided PGLM; n = 207 species). d,e, Across all extant ungulate species, grass dependence (d) is positively
correlated with body mass (two-sided PGLM; n = 207 species), such that larger ungulates tend to eat more grass on average, but latitude and body mass
(e) are not significantly correlated (two-sided PLM; n = 207 species). The colour gradients along the axes correspond to those in Fig. 3. The asterisks in a–c
and solid regression line in d denote a significant relationship (P < 0.05), whereas the dashed line in e denotes the lack of a clear relationship (P ≥ 0.05),
corrected for multiple comparisons. The white bands in a–c represent the median values, the coloured black and red bars represent the interquartile range
and the white whiskers extend to ±1.5× the interquartile range. The grey shaded regions in d,e represent the 95% CIs on the regression. Full model details
are available in Supplementary Table 1.
NATURE ECOLOGY & EVOLUTION | www.nature.com/natecolevol

Articles NATURE ECology & EvolUTIoN
resource wave seasonality (Fig. 4a): links are the same but additional
links arise between latitude and green wave seasonality, green wave
seasonality and migratory behaviour and green wave seasonal-
ity and grass dependence (Fig. 4). The most notable difference is
that green wave seasonality mediates some of the effects of latitude
on migratory behaviour (Fig. 4b). Altogether, this provides addi-
tional support for the hypothesis that latitude and grass dependence
exposed ungulates to seasonal green waves and thereby selected for
the evolution of migratory behaviour (Fig. 1). However, no causal
model including resource waves is well supported (Supplementary
Table 6). This suggests that all models we tested make claims of
independence that are violated given our data; this is somewhat
unsurprising given our aforementioned findings that green wave
seasonality covaries significantly with both latitude and seasonality.
Additionally, the origins of migratory behaviour may be tem-
porally correlated with the mid-Miocene cooling of the Earth (and
resultant increases in seasonality towards the poles38) and the conse-
quent rise of C4 grasslands39. Branches along which migration arose
overlap the time intervals when these two changes to the Earth
system occurred (Fig. 5). This suggests that these environmen-
tal changes may have contributed to the emergence of migratory
behaviour, further emphasizing the central roles that living at high
latitudes and relying on grass forage have played in the evolution of
migratory behaviour.
Hippopotamidae
Giraffidae
103.5
Body mass
100.4
0°˚ 75°˚
Latitude
0 1
P(migration)
Caprinae
Aepycerotinae
Tapiridae
Camelidae
Cervidae
Bovinae
Moschidae
Cephalophinae
Antilopinae
Hippotraginae
Alcelaphinae
Reduncinae
Equidae
Rhinocerotidae
Suidae
Tragulidae
Antilocapridae
a
c d
b
Dietary grass fraction
0 1
Fig. 3 | Ungulate character evolution. a–d, The evolution of migratory behaviour in ungulates was estimated using stochastic character mapping, whereas
the evolution of dietary grass fraction (b), latitude (c) and body mass (d) in ungulates were each reconstructed using Ornstein–Uhlenbeck models of
character evolution (n = 207 species). Branch colors represent reconstructed character values and color gradients correspond to those in Fig. 2. In a,
branch colors correspond to P(migration), or the posterior probability (computed as the relative frequency across stochastic maps) of migration along the
branch, where red indicates high posterior probability of migration. Ungulate families and Bovidae subfamilies74 are denoted around the perimeter of the
phylogenies, along with the silhouette of a representative from each group. The colour gradients correspond to those in Fig. 2.
NATURE ECOLOGY & EVOLUTION | www.nature.com/natecolevol

Articles
NATURE ECology & EvolUTIoN
Finally, we found evidence that now-extinct ungulates may have
been disproportionately migratory. We reconstructed the migratory
phenotype of ten recently extinct ungulates using phylogenetic impu-
tation and found that seven out of the ten extinct taxa are supported
as being migratory (Fig. 6 and Supplementary Table 7). Migration is
significantly more prevalent among these extinct taxa compared to
extant ungulates (phylogenetic generalized linear model (PGLM);
n = 217, z = 3.007, P < 0.001; Supplementary Table 1); the proportion
of extinct ungulates that were migratory was 1.52× that of extant taxa
(70.0% for extinct taxa compared to 45.9% for extant taxa). However,
this result is based on only ten extinct taxa that could be adequately
placed in the ungulate phylogeny from existing genetic data and
should therefore be interpreted with caution.
Discussion
Ungulate migrations are important for maintaining both robust
population sizes and ecosystem dynamics6,7,10, yet little is known
about the ultimate drivers of migration and what the emergence
of migratory behaviour has meant for ungulate evolution11,12,16. We
used phylogenetic path analysis to evaluate the coevolution between
migratory behaviour and ungulate characteristics, finding that: (1)
migratory ungulates exhibit a migratory syndrome, tending to be
larger, depending more on grass and inhabiting higher latitudes
than their non-migratory counterparts; (2) migratory behaviour
appears to have arisen 17 times independently across the ungulate
phylogeny, contemporaneously with an increasingly seasonal cli-
mate and the subsequent spread of C4 grasslands; and (3) migra-
tory behaviour most likely evolved in response to selective pressures
associated with being grass-dependent and living at high latitudes
(or other highly seasonal environments), in turn enabling the evo-
lution of large body sizes. Our work provides a causal explanation
for the origin of migratory behaviour in ungulates and consequent
evolution of large body sizes in grazing mammals.
These results illuminate the critical role that migratory behav-
iour has played in ungulate evolution. The evolution of migratory
behaviour appears to have been driven, at least in part, by living at
high latitudes and depending on grass for nutrition (Fig. 4). Both
characteristics likely exposed ungulates to substantial resource vari-
ability; vegetation at high latitudes is highly variable across seasons
and grass is both fast-growing and responsive to environmental
variation relative to other plant functional groups22,25,31,40. In sup-
port of this hypothesis, most of the probable gains of migratory
behaviour that we estimated are temporally coincident with two
dramatic changes in the ecology of the planet: global cooling in the
mid-Miocene38 and the subsequent rise of C4 grasslands39 (Fig. 5).
Both of these changes drastically altered patterns of terrestrial
resource availability and applied new selective pressures on the
foraging ecology of ungulates35,37. Therefore, migratory behaviour
likely evolved as a strategy to cope with this increasingly variable but
also highly predictable vegetation growth (that is, resource waves).
Recent work has similarly demonstrated that many (although not
all) extant migratory ungulates track resource waves3,5,8. Thus, the
environmental contexts that historically selected for migratory
behaviour probably resemble those that continue to make this an
adaptive strategy for nearly half of the ungulate species today.
We found evidence that these global shifts in climate and vegeta-
tion triggered the evolution of migratory behaviour multiple times
across the ungulate phylogeny. Two compatible mechanisms may
have contributed to the many independent origins of migratory
behaviour (Fig. 5). First, ungulates and other migratory taxa use
spatial memory to form cognitive maps that enable them to track
resource waves across large spatial scales8,41–43, which suggests that
the ancestor of modern ungulates likely also possessed the cogni-
tive capacity to remember and integrate spatial information at large
scales41. This ability may have been subsequently co-opted by differ-
ent lineages for the purposes of migration in response to local selec-
tion pressures. Second, cultural evolution may have facilitated the
evolution of migratory behaviour, following evidence from contem-
porary migrations that knowledge of when and where to migrate
results from the cumulative cultural transmission of social and aso-
cial information about spatial patterns of plant phenology17. Cultural
evolution can exert particularly strong selection on behaviour since
culture can allow rapid diffusion of a particular behaviour through
a population, accelerating its genetic fixation16,21,44. Nevertheless,
some ungulate species appear unable to learn migratory behaviour,
even under extreme conditions (like severe drought31). Thus, while
the repeated evolution of migratory behaviour may have been facili-
tated by social learning and cultural evolution, our results indicate
that other physiological, morphological and ecological characteris-
tics likely constrained which species did and did not evolve migra-
tory behaviour.
Our results suggest that the evolution of migratory behaviour
precipitated the evolution of large body size in ungulates (Fig. 4).
This finding is consistent with the Behavioral Drive hypothesis,
which proposes that behaviour is not simply a product of morphol-
ogy but rather a powerful selective force that shapes evolutionary
trajectories, capable of initiating evolutionary shifts in morphology,
physiology or ecology44,45. Accordingly, increases in body size after
the emergence of migratory behaviour may have been the result
of selection pressures to mitigate the costs of migrating. Although
Latitude
1.00
0.41
Grass
dependence
Migration
Body size
0.01
0.02
1.08
(0.26 to 0.56)
(–0.04 to 0.05)
(–0.08 to 0.12)
(0.63 to 1.36)
(0.69 to 1.48)
n = 207
a b
n = 189
Latitude
Grass
dependence
Migration
Body size
Green wave
seasonality
0.13
(0.02 to 0.24)
–0.02
(–0.19 to 0.15)
0.44
(0.29 to 0.59)
0.91
(0.55 to 1.27)
0.42
(0.29 to 0.55)
–0.01
(–0.10 to
0.07)
0.01
(–0.05 to 0.05)
Fig. 4 | The causes and consequences of evolving migratory behaviour in ungulates. a, The average causal path model of migratory evolution
demonstrates that migration evolved in response to living at high latitudes and being dependent on grass (n = 207 species). b, Causal path models
incorporating green wave seasonality (NDVI semi-variance) suggest that green wave seasonality mediates some of the effects of latitude and grass
dependence on migration (n = 189 species). The arrows are coloured by whether or not they are significant; the links for which the CIs of the regression
coefficients overlap zero are depicted in grey since they cannot be taken to be significant. The numbers below the arrows represent the strength of the
effects, along with the corresponding 95% CI (via bootstrapping).
NATURE ECOLOGY & EVOLUTION | www.nature.com/natecolevol

Articles NATURE ECology & EvolUTIoN
long-distance migration is energetically intensive30, larger organ-
isms can move more efficiently and freely, such that large body
sizes may reduce the energetic costs associated with migrating26,30,46.
Additionally, evolving migratory behaviour may have freed ungu-
lates from resource limitation by providing them access to a larger
forage pool, thereby allowing them to evolutionarily explore a
broader phenotypic space and exploit unoccupied niches31,40,46.
Regardless of the mechanism, phylogenetic evidence suggests
that migration changed the adaptive landscape for ungulate body
size and this may have been the case for other mammal lineages
also47. Some of the largest extant mammals are migratory: savanna
elephants migrate seasonally in response to forage green-up47 and
blue whales, which share a common ancestor with artiodactyl ungu-
lates48, track resource waves in a manner similar to their terrestrial
relatives49. Thus, migratory behaviour may have played a key role in
the evolution of large body sizes in mammals more generally.
Our results suggest that migratory mammal species may
have been more numerous in the Earth’s past. Given that extant
large-bodied grazing species tend to be migratory (Fig. 2) and that
many such large grazing ungulate species roamed high latitude
environments before the Pleistocene megafaunal extinctions35,50–52,
it follows that many of these extinct megafauna likely also exhibited
migratory behaviour. Our results directly support this hypothesis,
with seven out of ten ungulate species that went extinct within the
past 1 Ma reconstructed as migratory (Fig. 6). As such, landscapes
were probably more spatially connected before the Pleistocene
extinctions, with migratory Pleistocene megafauna conveying
nutrients, seeds, spores and diseases across vast distances much as
they do today9,10. Indeed, the legacies of these lost migrations likely
continue to inform the ecology of modern ecosystems via persistent
effects on soil properties, fire regimes and plant communities35,53.
Hence, contemporary ecosystem dynamics may be somewhat
anachronistic54,55, informed by a past where migrations were more
widespread. The few remaining ecosystems with intact migrations
are therefore critical for understanding how these lost migrations
continue to influence the dynamics of ecosystems today.
We also speculate that the disruption of migrations may have
played a key role in the progression of the megafaunal extinctions
in North America, Europe and Asia and the ongoing loss of ungu-
late migrations56. Expansion of humans out of Africa in tandem
with changing environment conditions are chiefly implicated in
megafaunal extinctions50,57,58 but the precise mechanisms underly-
ing these extinctions are unclear28. Migratory behaviour is currently
under severe threat from global change11,13 and many large-scale
migrations have either already collapsed or are now imperilled by
intensifying anthropogenic pressures from land use change, over-
hunting and the construction of physical barriers11,56,59. If similar
drivers (a changing climate and human impacts) also caused the
collapse of migratory behaviours during the Pleistocene28,50, trig-
gering associated population declines7,24, then migratory species
would have become more vulnerable to stochastic events, ulti-
mately leading to extinction. Our findings that migratory ungu-
lates generally occur at higher (especially northern) latitudes and
are larger-bodied than non-migratory species (Fig. 2) may thereby
account for the size-biased nature of the Pleistocene extinctions as
well as their severity outside Africa50,51. The Pleistocene megafaunal
extinctions and subsequent decline of ungulate diversity may thus
serve as an analogue for contemporary and future loss of migratory
Cooling of the
temperate zone
middle Miocene
(15 Ma)
Rise of C4
grasslands
late Miocene and
Pliocene (3–8 Ma)
60 40 30 20 10 0
0
1
2
3
4
5
Relative surface temperature
(δ18O (‰) of foraminifera shells)
Time (Ma)
/ /
/ /
/ /
/ /
0 1
P(migration)
Fig. 5 | The Earth system context for the evolution of migratory behaviour. The evolutionary origins of migratory behaviour are temporally aggregated
and are coincident with the onset of mid-Miocene cooling as well as the global proliferation of C4 grasslands. The 17 branches along which the origins of
migration are most likely are marked with grey circles along the branches of the phylogeny; the timing of these origins is depicted by the semi-transparent
red circles along the x axis. The Cenozoic record of Earth surface temperature changes is derived from the δ18O (‰) of foraminifera shells and is
reproduced from Zachos et al.38 and Edwards et al.39. The blue line represents the mean temperature over the last circa 60 Ma and the grey shaded region
represents the variation around the mean.
NATURE ECOLOGY & EVOLUTION | www.nature.com/natecolevol

Articles
NATURE ECology & EvolUTIoN
behaviour if ongoing trends of habitat fragmentation and degrada-
tion are not reversed.
Conclusions
Resource waves associated with mid-Miocene cooling and the
spread of C4 grasslands created widespread selective pressures that
helped drive the repeated evolution of migratory behaviour in
high-latitude, grass-dependent ungulates (Figs. 4 and 5). The wide-
spread evolution of migratory behaviour across ungulate lineages
was likely facilitated by a suite of cognitive or physiological preadap-
tations and possibly also cultural evolution. New migratory behav-
iour, in turn, resulted in the selection for larger body sizes (Fig. 4),
which perhaps mitigated the energetic costs associated with migra-
tory behaviour and leveraged the additional resources accessed by
migrating. Dependence on migration for sustaining their popula-
tions may have exposed migratory ungulates to an increased extinc-
tion risk in the face of a changing Pleistocene climate and expanding
human impacts, subsequently contributing to the extinction of
many large-bodied grazing taxa (Fig. 6). By extension, we suggest
that the Pleistocene megafaunal extinctions are both an analogue
for the present and a warning for the future of ungulate species as
threats to migrations continue.
Methods
Incidences of migratory behaviour. To determine the incidence of migratory
behaviour in ungulates, we rst made an operational list of all ungulate species to
be included in our analyses. To do this, we used a recently constructed species-level
mammal phylogeny48, focusing all analyses on the node-dated DNA-only
consensus tree (maximum clade credibility of 10,000 trees in the credible set).
We pruned the whole mammal tree down to just ungulates (species in the orders
Perissodactyla and Artiodactyla but excluding Cetacea). erefore, our list of
ungulates included 207 extant and 10 extinct species for which DNA sequence
information was available (see Supplementary Dataset 1 and Supplementary Table
7 for the complete list of ungulate species and references consulted).
We then sought to determine which of these species were migratory. To curate
a list of migratory behaviour in ungulates, we first compiled published syntheses
of migratory species and performed an exhaustive literature review, searching
Web of Science and Google Scholar for any records of migratory behaviour for
each ungulate species. For the purposes of this study, we reduced migration to
a binary characteristic; ungulates were considered migratory if any population
exhibited seasonal round-trip movements between discrete areas and/or if they
were explicitly described as migratory in the published literature13,59; therefore, our
categorization of migratory ungulates includes elevational and latitudinal migrants.
We coded species as migratory if there was any record of the species having ever
exhibited migratory behaviour in the past or present.
Covariates of migration. Next, we gathered data on the three ungulate
characteristics we hypothesized to be relevant to the evolution of migration:
body size; latitude; and grass dependence. Species mean adult body masses were
assembled for all ungulate species from various mass datasets60–62, which are
themselves compilations from the primary literature. Body mass values were
log-transformed for all analyses.
To summarize the latitudinal niche of each ungulate species, we calculated
the latitudinal centroid of species’ geographical ranges. For extant species, expert
geographical range maps were downloaded from the International Union for the
Conservation of Nature63; the mean latitude and longitude were calculated. For
extinct ungulate species, the latitudinal centroids of their ranges were estimated
based on known fossil localities (Supplementary Table 7).
Our final hypothesized driver of migration was ungulate grass dependence.
Therefore, we performed a targeted literature search to determine the grass
dependence of each species, defined in this study as the mean dietary grass fraction
over the duration of each given study. As above, we searched Web of Science and
Google Scholar for published studies that reported ungulate diet composition. For
some understudied ungulates (29 out of 207 extant species), quantitative dietary
data were not available. Thus, the dietary grass fraction for these understudied
species was estimated from available qualitative information on their diets. Dietary
data were even sparser for extinct ungulates and entirely lacking for many taxa.
When diet data were absent, we used the degree of hypsodonty to estimate diet (for
example, see Toljagić et al.64).
Resource seasonality. To quantify resource waves across the ranges of globally
distributed ungulate species, we used metrics derived from spatial semi-variance
and semi-variograms of the NDVI (8 × 8 km, 16-day composites, 816 composites
spanning 34 years (1982–2015)) data housed in the Global Inventory Modelling
and Mapping Studies database65. For each 16-day composite, we calculated the
semi-variance among pairs of locations (NDVI pixels) across spatial scales ranging
from 5 to 100 km. We used the maximum semi-variance (that is, the ‘sill’; excluding
the last 1/4 of each semi-variogram) to determine the magnitude of resource waves,
and the distance lag of the peak semi-variance (that is, the ‘range’) to represent the
1
0
P(migration) Hippidion
principale
Coelodonta
antiquitatis
Rucervus
schomburgki
Hippidion
saldiasi
Equus
ovodovi
Bos
primigenius
Gazella
saudiya
Myotragus
balearicus
Hippotragus leucophaeus
Megaloceros giganteus
Fig. 6 | Reconstructed migratory behaviour in extinct ungulates. The pie charts represent phylogenetically imputed values of migration and reflect the
likelihood that ten recently extinct ungulate species were migratory. These recently extinct ungulates are disproportionately migratory relative to extant
taxa (two-sided PGLM; n = 217 species, z = 3.007, P < 0.001). The likelihood of migratory behaviour in these taxa was imputed from available data on grass
dependence, body mass and latitudinal range centroid from the fossil record (see Supplementary Table 7 for further details). Images of extinct ungulates
are courtesy of R. Uchytel and D. Boh and are reproduced with permission.
NATURE ECOLOGY & EVOLUTION | www.nature.com/natecolevol

Articles NATURE ECology & EvolUTIoN
distance over which the resource wave travelled (Extended Data Fig. 2). We also
estimated seasonal variation in resource wave strength by calculating the difference
between maximum semi-variance throughout the annual cycle over the 34-year
time series (Extended Data Fig. 2). By doing so, we identified which species ranges
possessed the seasonal resource waves that would make migration a viable strategy.
Note that the NDVI semi-variance data could only be derived for 189 of the 207
species in our dataset because the scale of semi-variance data was too coarse to be
relevant for ungulates with small species ranges, such as small-island endemics.
Data analysis. Data were analysed in R v.3.6.1 (ref. 66). All phylogenetic
analyses used the consensus tree as described above. First, to test whether these
characteristics are heritable across the ungulate phylogeny, we calculated multiple
indices of phylogenetic signal for all characters using the packages phytools
v.0.7-70 and adephylo v.1.1-1167,68. Then, to determine the manner in which these
characters evolved, we fitted evolutionarily explicit and non-evolutionary models
of character change across the phylogeny (white noise, star Brownian motion
(BM), BM, early burst, Ornstein–Uhlenbeck) for each character using the package
geiger v.2.0.7 and compared Akaike information criterion (AICc) support values to
select the best-fitting model69 (Supplementary Table 2).
Next, to evaluate how each of these characteristics changed over the course
of ungulate evolution, we estimated ancestral character states across the tree
from the species tip data (Supplementary Table 3). For continuous characters,
we used maximum likelihood estimations implemented in phytools67, employing
the evolutionary model of character change with the lowest AICc score based on
the above model selection (Supplementary Table 2). Therefore, the best-fitting
Ornstein–Uhlenbeck models from the above model selection were used to estimate
grass dependence, body mass and latitude across the ungulate phylogeny (Fig. 3).
To estimate migration (a binary character), we performed stochastic character
mapping in phytools with 1,000 simulations67.
To evaluate whether these characteristics coevolved, we used 100 stochastic
character maps of migration as maps of different selective regimes on the tree
and evaluated whether migration resulted in different evolutionary optima for
each character. Using the package OUwie v.2.670, we fitted Ornstein–Uhlenbeck
models with multiple optima and rates of evolution matched to the estimated
migration regimes (Ornstein–Uhlenbeckmv, Ornstein–Uhlenbeckma, and Ornstein–
Uhlenbeckmva), a single optimum Ornstein–Uhlenbeck model, a multi-rate BM
model (BMs) and a single-rate BM null model, following the analyses in Cressler
et al.71. As above, we compared their corrected AICc support values to select the
best-fitting model (Supplementary Table 4).
Then, we used phylogenetic models to estimate the evolutionary correlations
between characteristics with the phylolm v.2.6 package72 (Supplementary Table
1). First, we used a binomial PGLM to determine if grass dependence and body
mass are correlated across ungulates. We used a phylogenetic linear model (PLM)
to evaluate whether body mass is related to latitude. Finally, we tested whether
migration is related to body mass, grass dependence and latitude also using a
binomial PGLM.
To investigate whether the presence of resource waves predicted migratory
behaviour, we again used PGLMs to estimate the relationships between resource
wave metrics and migratory behaviour (Supplementary Table 1). We tested how
well each of the three resource wave metrics we calculated (that is, green wave sill,
green wave range and green wave seasonality) predicted migration by constructing
separate PGLMs for each metric, again using binomial distributions (Extended
Data Fig. 1). Because green wave seasonality was determined to significantly
predict migration, we then modelled how green wave seasonality predicted
migratory behaviour in concert with latitude, grass dependence and body mass
with a binomial PGLM. Finally, we modelled the relationship between green
wave seasonality and grass dependence and latitude, employing separate PLMs
(Extended Data Fig. 1).
Next, to evaluate the directionality of these relationships (that is, whether
migration is the cause or consequence of inferred relationships), we performed
phylogenetic path analysis73. Based on the plausible relationships between the
characteristics outlined above, we defined a list of probable candidate path models
(Supplementary Fig. 1). We compared the support for these different candidate
models using the CICc with the package phylopath v.1.1.273 (Supplementary Fig. 4).
All models with a ΔCICc < 2 were weighted and averaged (with full averaging) to
yield the average path model (Fig. 4a).
We sought to determine whether resource wave metrics mediated the causal
relationships between environmental predictors and migration. Because the
seasonality of the green wave was identified to be a significant predictor of
migration, we defined another set of candidate models that included green wave
seasonality as an additional independent variable (Supplementary Fig. 3). As above,
we compared support for the candidate models using the CICc (Supplementary
Fig. 5)
and computed the weighted average of all models with a ΔCICc < 2 to yield the
average path model (Fig. 4b). This analysis included only the 189 taxa for which we
could calculate the NDVI semi-variance data.
Finally, to illuminate migration’s role in the ecology of Earth’s past, we
performed phylogenetic imputation with the phytools package67 to reconstruct
the migratory phenotype of ten extinct ungulates included in our phylogeny from
data on body mass, grass dependence and latitude (Fig. 6). After reconstructing the
migratory behaviour of these extinct species, we compared the imputed migration
phenotypes of extinct species with observed migration among extant species using a
PGLM (with phylolm72, as above) to evaluate if the prevalence of migration differed
significantly between extinct and extant ungulates (Supplementary table 1).
Reporting Summary. Further information on research design is available in the
Nature Research Reporting Summary linked to this article.
Data availability
All data generated and analysed during this study are included in Supplementary
Dataset 1 and are also available in tabular form from the Dryad Data Repository
(https://datadryad.org/stash/dataset/doi:10.5061/dryad.g79cnp5rj).
Received: 19 October 2021; Accepted: 22 March 2022;
Published: xx xx xxxx
References
1. Dobson, A. P. et al. Road will ruin Serengeti. Nature 467, 272–273
(2010).
2. Larsen, F. et al. Wildebeest migration drives tourism demand in the Serengeti.
Biol. Conserv. 248, 108688 (2020).
3. Aikens, E. O. et al. e greenscape shapes surng of resource waves in a large
migratory herbivore. Ecol. Lett. 20, 741–750 (2017).
4. Bischof, R. et al. A migratory northern ungulate in the pursuit of spring:
jumping or surng the green wave? Am. Nat. 180, 407–424 (2012).
5. Merkle, J. A. et al. Large herbivores surf waves of green-up during spring.
Proc. Biol. Sci. 283, 20160456 (2016).
6. Fryxell, J. M., Greever, J. & Sinclair, A. R. E. Why are migratory ungulates so
abundant? Am. Nat.131, 781–798 (1988).
7. Staver, A. C. & Hempson, G. P. Seasonal dietary changes increase the
abundances of savanna herbivore species. Sci. Adv. 6, eabd2848 (2020).
8. Kauman, M. J. et al. Causes, consequences, and conservation of ungulate
migration. Annu. Rev. Ecol. Evol. Syst. 52, 453–478 (2021).
9. Lundberg, J. & Moberg, F. Mobile link organisms and ecosystem functioning:
implications for ecosystem resilience and management. Ecosystems 6,
0087–0098 (2003).
10. Bauer, S. & Hoye, B. J. Migratory animals couple biodiversity and ecosystem
functioning worldwide. Science 344, 1242552 (2014).
11. Bolger, D. T., Newmark, W. D., Morrison, T. A., & Doak, D. F. e need for
integrative approaches to understand and conserve migratory ungulates. Ecol.
Lett. 11, 63–77 (2007).
12. Fryxell, J. M. & Holt, R. D. Environmental change and the evolution of
migration. Ecology 94, 1274–1279 (2013).
13. Shaw, A. K. Drivers of animal migration and implications in changing
environments. Evol. Ecol. 30, 991–1007 (2016).
14. Hebblewhite, M. & Merrill, E. H. Trade-os between predation risk and
forage dier between migrant strategies in a migratory ungulate. Ecology 90,
3445–3454 (2009).
15. Nelson, M. E. Development of migratory behavior in northern white-tailed
deer. Can. J. Zool. 76, 426–432 (1998).
16. Berg, J. E., Hebblewhite, M., St. Clair, C. C. & Merrill, E. H. Prevalence and
mechanisms of partial migration in ungulates. Front. Ecol. Evol. 7, 325 (2019).
17. Jesmer, B. R. et al. Is ungulate migration culturally transmitted? Evidence of
social learning from translocated animals. Science 361, 1023–1025 (2018).
18. Sih, A., Bell, A. & Johnson, J. C. Behavioral syndromes: an ecological and
evolutionary overview. Trends Ecol. Evol. 19, 372–378 (2004).
19. Found, R. & St. Clair, C. C. Behavioural syndromes predict loss of migration
in wild elk. Anim. Behav. 115, 35–46 (2016).
20. Abraham, J. O., Hempson, G. P., Faith, J. T. & Staver, A. C.Seasonal strategies
dier between tropical and extratropical herbivores. J. Anim. Ecol. 91,
681–692 (2022).
21. Whitehead, H., Laland, K. N., Rendell, L., orogood, R. & Whiten, A.
e reach of gene–culture coevolution in animals. Nat. Commun. 10, 2405
(2019).
22. Scanlon, T. M., Caylor, K. K., Manfreda, S., Levin, S. A. & Rodriguez-Iturbe,
I. Dynamic response of grass cover to rainfall variability: implications for the
function and persistence of savanna ecosystems. Adv. Water Res. 28, 291–302
(2005).
23. Staver, A. C., Wigley-Coetsee, C. & Botha, J. Grazer movements exacerbate
grass declines during drought in an African savanna. J. Ecol. 107, 1482–1491
(2019).
24. Fryxell, J. M. & Sinclair, A. R. Causes and consequences of migration by large
herbivores. Trends Ecol. Evol. 3, 237–241 (1988).
25. Running, S. W. et al. A continuous satellite-derived measure of global
terrestrial primary production. Bioscience 54, 547–560 (2004).
26. Langvatn, R., Albon, S. D., Burkey, T. & Clutton-Brock, T. H. Climate, plant
phenology and variation in age of rst reproduction in a temperate herbivore.
J. Anim. Ecol. 65, 653–670 (1996).
NATURE ECOLOGY & EVOLUTION | www.nature.com/natecolevol

Articles
NATURE ECology & EvolUTIoN
27. Webber, Q. M. R. & McGuire, L. P. Heterothermy, body size, and locomotion
as ecological predictors of migration in mammals. Mamm. Rev. 52, 82–95
(2022).
28. Mann, D. H., Groves, P., Gaglioti, B. V. & Shapiro, B. A. Climate-driven
ecological stability as a globally shared cause of Late Quaternary megafaunal
extinctions: the Plaids and Stripes Hypothesis. Biol. Rev. Camb. Philos. Soc.
94, 328–352 (2018).
29. Jarman, P. J. e social organisation of antelope in relation to their ecology.
Behaviour 48, 215–267 (1974).
30. Hein, A. M., Hou, C. & Gillooly, J. F. Energetic and biomechanical constraints
on animal migration distance. Ecol. Lett. 15, 104–110 (2012).
31. Abraham, J. O., Hempson, G. P. & Staver, A. C. Drought-response strategies
of savanna herbivores. Ecol. Evol. 9, 7047–7056 (2019).
32. Owen-Smith, R. N. Megaherbivores: the Inuence of Very Large Body Size on
Ecology (Cambridge Univ. Press, 1988).
33. Gonzalez-Voyer, A. & von Hardenberg, A. in Modern Phylogenetic
Comparative Methods and eir Application in Evolutionary Biology: Concepts
and Practice (ed. Garamszegi, L. Z.) 201–229 (Springer, 2014).
34. Pérez-Barbería, F. J., Gordon, I. J. & Nores, C. Evolutionary transitions
among feeding styles and habitats in ungulates. Evol. Ecol. Res. 3, 221–230
(2001).
35. Staver, A. C., Abraham, J. O., Hempson, G. P., Karp, A. T. & Faith, J. T. e
past, present, and future of herbivore impacts on savanna vegetation. J. Ecol.
109, 2804–2822 (2021).
36. Janis, C. M. in e Ecology of Browsing and Grazing (eds Gordon, I. J. &
Prins, H. H. T.) 21–45 (Springer, 2008).
37. Janis, C. M. Tertiary mammal evolution in the context of changing climates,
vegetation, and tectonic events. Annu. Rev. Ecol. Syst. 24, 467–500 (1993).
38. Zachos, J., Pagani, M., Sloan, L., omas, E. & Billups, K. Trends, rhythms,
and aberrations in global climate 65 Ma to present. Science 292, 686–693
(2001).
39. Edwards, E. J. et al. e origins of C4 grasslands: integrating evolutionary and
ecosystem. Science 328, 587–591 (2010).
40. Bhat, U., Kempes, C. P. & Yeakel, J. D. Scaling the risk landscape drives
optimal life-history strategies and the evolution of grazing. Proc. Natl Acad.
Sci. USA 117, 1580–1586 (2020).
41. Fagan, W. F. et al. Spatial memory and animal movement. Ecol. Lett. 16,
1316–1329 (2013).
42. Merkle, J. A. et al. Spatial memory shapes migration and its benets: evidence
from a large herbivore. Ecol. Lett. 22, 1797–1805 (2019).
43. Mueller, T., O’Hara, R. B., Converse, S. J., Urbanek, R. P. & Fagan, W. F.
Social learning of migratory performance. Science 341, 999–1002 (2013).
44. Wcislo, W. T. Behavioral environments and evolutionary change. Annu. Rev.
Ecol. Syst. 20, 137–169 (1989).
45. Wyles, J. S., Kunkel, J. G. & Wilson, A. C. Birds, behavior, and anatomical
evolution. Proc. Natl Acad. Sci. USA 80, 4394–4397 (1983).
46. Yeakel, J. D., Kempes, C. P. & Redner, S. Dynamics of starvation and recovery
predict extinction risk and both Damuth’s law and Cope’s rule. Nat. Commun.
9, 657 (2018).
47. Purdon, A., Mole, M. A., Chase, M. J. & van Aarde, R. J. Partial migration in
savanna elephant populations distributed across southern Africa. Sci. Rep. 8,
11331 (2018).
48. Upham, N. S., Esselstyn, J. A. & Jetz, W. Inferring the mammal tree:
species-level sets of phylogenies for questions in ecology, evolution, and
conservation. PLoS Biol. 17, e3000494 (2019).
49. Abrahms, B. et al. Memory and resource tracking drive blue whale
migrations. Proc. Natl Acad. Sci. USA 116, 5582–5587 (2019).
50. Barnosky, A. D., Koch, P. L., Feranec, R. S., Wing, S. L. & Shabel, A. B.
Assessing the causes of late Pleistocene extinctions on the continents. Science
306, 70–75 (2004).
51. Dirzo, R. et al. Defaunation in the Anthropocene. Science 345, 401–406
(2014).
52. Faith, J. T., Rowan, J. & Du, A. Early hominins evolved within non-analog
ecosystems. Proc. Natl Acad. Sci. USA 116, 21478–21483 (2019).
53. Holdo, R. M. et al. A disease-mediated trophic cascade in the Serengeti and
its implications for ecosystem C. PLoS Biol. 7, e1000210 (2009).
54. Janzen, D. H. & Martin, P. S. Neotropical anachronisms: the fruits the
gomphotheres ate. Science 215, 19–27 (1982).
55. Dantas, V. L. & Pausas, J. G. e legacy of the extinct Neotropical megafauna
on plants and biomes. Nat. Commun. 13, 129 (2022).
56. Harris, G., irgood, S., Hopcra, J. G. C., Cromsigt, J. P. G. M. & Berger, J.
Global decline in aggregated migrations of large terrestrial mammals.
Endanger. Species Res. 7, 55–76 (2009).
57. Seersholm, F. V. et al. Rapid range shis and megafaunal extinctions
associated with late Pleistocene climate change. Nat. Commun. 11, 2770
(2020).
58. Alroy, J. A multispecies overkill simulation of the end-Pleistocene megafaunal
mass extinction. Science 292, 1893–1896 (2001).
59. Berger, J. e last mile: how to sustain long-distance migration in mammals.
Conserv. Biol. 18, 320–331 (2004).
60. Faurby, S. & Svenning, J.-C. Resurrection of the island rule: human-driven
extinctions have obscured a basic evolutionary pattern. Am. Nat. 187,
812–820 (2016).
61. Wilman, H. et al. EltonTraits 1.0: species-level foraging attributes of the
world’s birds and mammals. Ecology 95, 2027 (2014).
62. Smith, F. A. et al. Body mass of late Quaternary mammals. Ecology 84,
3403 (2003).
63. IUCN. IUCN Red List of reatened Species 2019 (IUCN, 2019).
64. Toljagić, O., Voje, K. L., Matschiner, M., Liow, L. H. & Hansen, T. F. Millions
of years behind: slow adaptation of ruminants to grasslands. Syst. Biol. 67,
145–157 (2018).
65. Pinzon, J. E. & Tucker, C. J. A non-stationary 1981–2012 AVHRR NDVI3g
time series. Remote Sens. 6, 6929–6960 (2014).
66. R Core Team. R: a Language and Environment for Statistical Computing (R
Foundation for Statistical Computing, 2020).
67. Revell, L. J. phytools: an R package for phylogenetic comparative biology (and
other things). Methods Ecol. Evol. 3, 217–223 (2012).
68. Blomberg, S. P., Garland, T. Jr. & Ives, A. R. Testing for phylogenetic signal in
comparative data: behavioral traits are more labile. Evolution 57, 717–745
(2003).
69. Orme, D. e caper package: Comparative analysis of phylogenetics and
evolution in R. R package version 1.0.1 https://cran.r-project.org/web/
packages/caper/vignettes/caper.pdf (2018).
70. Beaulieu, J. M. & O’Meara, B. OUwie: Analysis of evolutionary rates in an
OU framework. R package version 2.6 https://rdrr.io/cran/OUwie/ (2014).
71. Cressler, C. E., Butler, M. A. & King, A. A. Detecting adaptive evolution in
phylogenetic comparative analysis using the Ornstein–Uhlenbeck model. Syst.
Biol. 64, 953–968 (2015).
72. Ho, L. S. & Ané, C. A linear-time algorithm for Gaussian and non-Gaussian
trait evolution models. Syst. Biol. 63, 397–408 (2014).
73. van der Bijl, W. phylopath: easy phylogenetic path analysis in R. PeerJ 6,
e4718 (2018).
74. Chen, L. et al. Large-scale ruminant genome sequencing provides insights
into their evolution and distinct traits. Science 364, eaav6202 (2019).
Acknowledgements
We thank A. C. Staver, E. J. Sargis, J. T. Faith and G. P. Hempson for the many
thought-provoking discussions regarding ungulate migration and mammal evolution
that inspired this project. We also thank the Edwards and Dunn laboratories at Yale
University and Pringle laboratory at Princeton University for providing helpful feedback
on this work. Finally, we thank J. R. Goheen for valuable feedback on the manuscript.
J.O.A. was supported by the United States National Science Foundation (NSF) Graduate
Research Fellowship Program (GRFP 2019256075) and N.S.U. was supported by the NSF
VertLife Terrestrial grant (DEB 1441737) and Arizona State University President’s Special
Initiative Fund.
Author contributions
J.O.A. conceived the study. J.O.A. compiled the underlying ungulate trait data from the
literature and B.R.J. calculated the green wave metrics for all species. J.O.A. and A.D.-S.
designed the analyses, with significant contribution from N.S.U. J.O.A. and B.R.J. wrote
the initial manuscript drafts with significant input from N.S.U. and A.D.-S. All authors
discussed and provided feedback on subsequent manuscript drafts.
Competing interests
The authors declare no competing interests.
Additional information
Extended data is available for this paper at https://doi.org/10.1038/s41559-022-01749-4.
Supplementary information The online version contains supplementary material
available at https://doi.org/10.1038/s41559-022-01749-4.
Correspondence and requests for materials should be addressed to Joel O. Abraham.
Peer review information Nature Ecology & Evolution thanks Nic Bone and the other,
anonymous, reviewer(s) for their contribution to the peer review of this work. Peer
reviewer reports are available.
Reprints and permissions information is available at www.nature.com/reprints.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
© The Author(s), under exclusive licence to Springer Nature Limited 2022
NATURE ECOLOGY & EVOLUTION | www.nature.com/natecolevol

Articles NATURE ECology & EvolUTIoN
Articles NATURE ECology & EvolUTIoN
Extended Data Fig. 1 | The role of green wave tracking in the evolution of migration. Relationships between (a) green wave sill, (b) green wave range,
and (c) green wave seasonality and migration are depicted, as well as between (e) green wave seasonality and latitude and (f) green wave seasonality
and grass dependence. Of the green wave metrics we calculated, only green wave seasonality significantly predicts migration (two-sided PGLM; n = 189
species), with migratory behavior more prevalent amongst taxa exposed to more seasonal green waves. Green wave seasonality is likewise positively
correlated with latitude and dietary grass fraction (two-sided PLMs; n = 189 species). The asterisks (*) in (c) and solid regression lines in (e, d) denote a
significant relationship (P < 0.05), whereas the ‘N.S’ in (a,b) denotes the lack of a clear relationship (P ≥ 0.05), corrected for multiple comparisons. White
bands in (a-c) represent median values, the colored bars represent the interquartile range (IQR), and white whiskers extend to ±1.5 × IQR. Grey shaded
regions in (d,e) represent 95% confidence intervals on the regression. Full model details are available in Supplementary table 1.
NATURE ECOLOGY & EVOLUTION | www.nature.com/natecolevol

Articles
NATURE ECology & EvolUTIoN Articles
NATURE ECology & EvolUTIoN
Extended Data Fig. 2 | Measuring landscape suitability for migration. A simulated (a) perfect resource wave, (b) heterogeneous landscape
with no resource wave, and (c) landscape intermediate to (a) and (b). Brown pixels represent areas where the date of peak NDVI occurred early, whereas
green pixels represent relatively late peaks NDVI. (a-c) The x-axis represents the distance travelled by resource waves (distance lag in km) and y-axis
represents magnitude of the green wave (semivariance). Dashed lines illustrate maximum semivariance (horizontal) and maximum distance lag (vertical).
(d) Empirical variograms for mule deer (Odocoileus hemionus) and white-tailed deer (Odocoileus virginianus), depicted in purple and black respectively.
Vertical and horizontal dashed lines represent maximum semivariance (horizontal) and maximum distance lag (vertical) just as in panels (a-c).
(e) Illustration of how seasonality in resource waves varied among the geographical ranges of mule deer (O. hemionus) and white-tailed deer (O.
virginianus). Horizontal dashed lines depict the minimum and maximum magnitude of resource waves throughout the annual cycle. Note that the distance
between purple dashed lines for mule deer (O. hemionus) is much larger than the distance between black dashed lines for white-tailed deer (O. virginianus),
indicating greater seasonality in resource waves across the geographic range of mule deer (O. hemionus).
NATURE ECOLOGY & EVOLUTION | www.nature.com/natecolevol

1
nature portfolio | reporting summary March 2021
Corresponding author(s): Joel O. Abraham, NATECOLEVOL-211014922A
Last updated by author(s): Mar 8, 2022
Reporting Summary
Nature Portfolio wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency
in reporting. For further information on Nature Portfolio policies, see our Editorial Policies and the Editorial Policy Checklist.
Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.
n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement
A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly
The statistical test(s) used AND whether they are one- or two-sided
Only common tests should be described solely by name; describe more complex techniques in the Methods section.
A description of all covariates tested
A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons
A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient)
AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals)
For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted
Give P values as exact values whenever suitable.
For Bayesian analysis, information on the choice of priors and Markov chain Monte Carlo settings
For hierarchical and complex designs, identification of the appropriate level for tests and full reporting of outcomes
Estimates of effect sizes (e.g. Cohen's d, Pearson's r), indicating how they were calculated
Our web collection on statistics for biologists contains articles on many of the points above.
Software and code
Policy information about availability of computer code
Data collection No software was used for data collection.
Data analysis Data were analyzed in R statistical software version 3.6.1
For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors and
reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Portfolio guidelines for submitting code & software for further information.
Data
Policy information about availability of data
All manuscripts must include a data availability statement. This statement should provide the following information, where applicable:
- Accession codes, unique identifiers, or web links for publicly available datasets
- A description of any restrictions on data availability
- For clinical datasets or third party data, please ensure that the statement adheres to our policy
All data generated and analyzed during this study are included in the supplementary material files of the published article as a supplementary data file and are also
available through Dryad Data Repository (DOI: doi:10.5061/dryad.g79cnp5rj)

2
nature portfolio | reporting summary March 2021
Field-specific reporting
Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.
Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf
Ecological, evolutionary & environmental sciences study design
All studies must disclose on these points even when the disclosure is negative.
Study description We compiled existing data on the incidence of migratory behavior within ungulates to evaluate how migratory behavior evolved.
Research sample Our study focused on species-level variation in migratory behavior. As such, we sought to include all extant species of ungulates and
to characterize migratory phenotype at the species level.
Sampling strategy We included all 207 extant ungulate species that were included in the species-level mammal phylogeny of Upham et al. (2020) in our
analyses.
Data collection To curate a list of migratory behavior in ungulates, we first compiled published syntheses of migratory species and performed an
exhaustive literature review, searching Web of Science and Google Scholar for any records of migratory behavior for each ungulate
species. For the purposes of this study, we reduced migration to a binary characteristic; ungulates were considered migratory if any
population exhibited seasonal round-trip movements between discrete areas and/or if they were explicitly described as migratory in
published literature.
Timing and spatial scale We coded species as migratory if there was any record of any population within the range of the species having ever exhibited
migratory behavior in the past or present. This approach enabled us to capture any evolutionary history of migration, and therefore
control against the confounding influence of anthropogenic impacts. As such, our study is global in scale and extends as far back as
there are publicly accessible published records documenting ungulate behavior.
Data exclusions No ungulate species present in the species-level mammal phylogeny of Upham et al. (2020) were excluded from the study. We were
not able to derive NDVI data for some species, however, as their species ranges smaller than the spatial scale of NDVI data and/or
species range maps did not exist for these species. These species (n = 18) were excluded from analyses that incorporated NDVI data,
but we indicate this in the text where applicable.
Reproducibility All underlying data are provided in a supplementary data file and are also available via Dryad Data Repository, and we cited the R
packages (and the specific functions) used for all analyses.
Randomization We incorporated all extant ungulate for which phylogenetic data were available in our study, precluding the need for a randomized
sampling design.
Blinding Though it was impossible to truly blind our data collection (we employed the species name of each ungulate in our literature search),
our data collection protocol was designed to control against canonized notions of what defines a migratory ungulate. We were
systematic in our literature review and had objective criteria for classifying an ungulate as migratory or not.
Did the study involve field work? Yes No
Reporting for specific materials, systems and methods
We require information from authors about some types of materials, experimental systems and methods used in many studies. Here, indicate whether each material,
system or method listed is relevant to your study. If you are not sure if a list item applies to your research, read the appropriate section before selecting a response.
Materials & experimental systems
n/a Involved in the study
Antibodies
Eukaryotic cell lines
Palaeontology and archaeology
Animals and other organisms
Human research participants
Clinical data
Dual use research of concern
Methods
n/a Involved in the study
ChIP-seq
Flow cytometry
MRI-based neuroimaging